created 6/08
My related essay on the telephone is here
Go to homepage

Telegraph dates
Reference telegraph values
Telegraph essay
Can the telegraph work with only one wire?
       Earth Return
       Modern high current, low resistance earth ground
Multiplexing introduction
       Bridge duplex
       Bridge and differential duplex (side-by-side)
       Edison's duplex
       Edison's quadruplex
Tuning the line
Very bref history of telegraph
       Did Morse invent the telegraph?
Relay types
Submarine cables
      Edison's telegraph patents
      Telegraph books
      Telegraph links
      Beautiful telegraph keys
      Restored AT&T 1930's telegraph key

        This essay is about the 19th century telegraph as seen by a circuit designer. While I lay out a time line of historical development of the telegraph and provide a little history, my real interest and focus is on how the early telegraph designers solved some difficult and tricky circuit-like problems like multiplexing, earth return, and high capacitance lines. Included are figures from key patents, from Prescott's 1800's telegraph books, and figures I've drawn myself.

Telegraph dates
         1820                                                         Orsted discovers that current in a wire deflects
                                                                                magnetized needles
         1825                                                         Sturgeon invents electromagnet
         1830                                                         Henry improves electromagnet and shows it can
                                                                                be operated one mile from battery
         1831                                                         Faraday discovers mutual inductance
         1835                                                         Henry invents the electromagnetic relay and
                                                                                shows how it allows weak currents to control
                                                                                big electromagnets (local loop)
         1836-37                                                   first practical batteries developed by Daniell and
         1837-39                                                   first short telegraph lines (few miles in Europe)
         1839                                                         Steinheil in Germany demonstrates 'earth
         1843                                                         Wheatstone develops the bridge circuit
         1844                                                         first experimental line built by Morse (Wash,
                                                                                Balt) and first major US telegraph line
                                                                                (Wash, NY, Boston). (Early receivers scratch
                                                                                dash/dots on tape which provides both speed
                                                                                 and message accuracy)
         1854                                                        sounders largely replace tape as good operators
                                                                                become fast and accurate
         1860's                                                     gravity cell, an improved wet style Daniell type battery
                                                                                developed by Callaud in France

Multiplex telegraph
         1872  Joseph Stearns                             Stearns duplex telegraph adopted by Western
                                                                                Union (both directions simultaneously)
         1874  Edison                                           Edison quadruplex telegraph adopted by
                                                                                Western Union.  (Edison's two messages
                                                                                in same direction using polarity and
                                                                                amplitude combined with Stearns duplex
                                                                                resulted in quadruplex.)
         1876  Elisha Gray                                  Octaplex (Electro-harmonic) demonstrated at 1876
                                                                                Philadelphia Centennial. (used commercially
                                                                                for a few years on a high performance line
                                                                                NY, Chicago)

Reference telegraph values
          Line resistance                                            20 ohm/mile   (#16 AWG wire)
          Line inductance                                           1,600 uh/mile  (1uh/meter rule)
                                                                                      (8 to 14 mh/mile for iron wire)
          Line L/R time constant                               80 usec
                                                                                      (400 to 700 usec for iron wire)
          Line capacitance (air line)                          5 to 20 nf/mile (est)
          Voltage (sending)                                        1 V/mile (up to 150V ?)
                                                                                        eq to 50 ma line current
          Distance (no repeater)                               300 miles (typ)
                sounder in local loop
         Sounder coils                                                4 ohm,   100 ma pull-in       local loop
                                                                               20 ohm ,   50 ma pull-in        up to 15 miles
                                                                             150 ohm ,   20 ma pull-in        very long lines

         Transmission rate (manual)                        40 wpm (40 x 5 letters = 200 letters/min,
               land line, good operator                                or 3.3 let/sec) Pretty fast! If av Morse
                                                                                        letter is three signals, equiv to 100 msec
                                                                                        for av dot/dash (+ separation space).
      Transmission rate (automatic)                      130 wpm   (later 300 to 400 wpm on good line)
         Wheatstone 1855
              sender --- perforated tape
              receiver --- inking receiver
              (+/- current pulses)
          Atlantic submarine cable                          8 wpm (40 letters/min)
                Kelvin's mirror galvanometer
               (+/- current pulses)

        In the 1840's, only about 20 years or so after it was discovered that current in a wire produced a (weak) magnetic field, a practical device emerged, the telegraph. It was the beginning of the electrical revolution, and it quickly revolutionized communications shrinking message time from days to minutes.

Telegraph essay
        For much of the 19th century the telegraph was the state of the art in electrical engineering. The Scottish physicist Lord Kelvin, the thermodynamics expert and one of the most famous physicists of the 19th century, was also a telegraph engineer. He designed, and figured out how to make work, the long submarine telegraph line under the Atlantic. Edison did a ton of work on the telegraph with over 100 telegraph patents (!), most famously developing the quadruplex telegraph. As the first major, and world altering, electrical technology, the telegraph probably did much to create the field of electrical engineering. You don't see much written about the telegraph these days, and I learned nothing about it in engineering school.

 Basic telegraph
        The basic (Morse) telegraph is pretty simple. It's just a battery, a key (a switch to open and close the circuit), and a sounder (an electromagnet that makes a metallic click when energized and deenergized). It turns out that you can put many miles of wire in the circuit and it still works. This means you can switch the current on/off in one city and listen to the electromagnetic clicks, which arrive almost instantly, in another city. Combine this with a long/short code for each letter (invented by Morse) and messages can be sent long distances. Presto -- basic telegraph
        But still there were things about the telegraph that I never understood and have long been curious about. Two things in particular always bothered me. One, how did the 'earth' work as a 'return', i.e. how do you model it, and two, how in the 19th century, well before electronics, were they able to send more than one message at a time over a line (multiplex telegraph). When I began researching the telegraph for this essay, I was unable to draw even a simple circuit model of the telegraph because I had no baseline numbers. I didn't know if at the receiving end the line looked like a current source, a voltage source, or somewhere in between.

How did they....?
Earth Return
       How did they figure out that the telegraph needed only one wire to work? The answer to this turned out to be good old trial and error. When equipment at both ends of the line was properly 'earthed', it was found that not only did one wire work, but one wire worked better than two wires doubling the current in the sounder. But why and how 'earth return' worked was kind of a mystery and was the subject of study and debate by 19th century electricians. I certainly didn't understand it. It don't ever remember it being discussed in engineering school, and it's not the kind of thing you can investigate in the lab, because for experiments you need wires that stretch for miles.

       How were they able to send more than one message at the same time over a wire (multiplex)? Consider, pretty much all they had to work with in the 19th century was coils, switches, magnets, resistors and batteries. The answer turned out to be good old clever circuit design. The early telegraph designers like Wheatstone, Steinheil, Stearns, Edison, and Gray were really the first generation of circuit designers.

       How did it work over such long distances? Didn't the build up of resistance from hundreds of miles of wire make the current too weak to get a good click from the sounder?  The answer was yes, but a work around was found in the form of an 'electromechanical amplifier' (local loop) invented by Joseph Henry.

        I read a telegraph battery 'rule of thumb' was one battery cell per mile of line. At 10 to 20 ohms/mile (most refs' say 20 ohm/mile, eq to #16 AWG wire), this yields a current of about 1V/(10 to 20 ohms) = 50  to 100 ma. But it's very likely for safety, insulation, or economic consideration that the maximum sending voltage was limited. I have not found a reference, but my guess is that sending voltage was limited to somewhere between 50 and 150 volts. This means of course that on line longer than 50 to 150 miles the receiving currents starts to fall.

Local loop
        The trick for getting good distance out of the telegraph, which I suspect was pretty obvious to early researchers, was at the receiving end to have sensitive/light relay switch current from a local battery into a heavy/strong relay. (Henry in the USA is generally credited with inventing this in the late 1820's.) The relay at the end of the line was constructed and wound so that it could be driven by a small current (20 ma). The magnetic flux in a coil depends on NI, meaning the number of turns (N) times current (I), so with low I you want high N. This means winding the receiving coil with lots of turns of fine wire. It was also made sensitive by having the movable (switch) lever be light and with limited travel. This relay then switched on/off a much larger current from a small local battery (couple of volts) that ran a high current (250 ma), low impedance sounder (4 ohms) or recording apparatus. Loosely speaking this structure, called a 'local loop' by the telegraph guys, could be considered an electromechanical amplifier, amplifying the incoming current by x10 to x20.

Can the telegraph work with only one wire?
        Originally telegraph lines had two wires, out and back. But people soon figured out only one wire would work if they grounded both sending and receiving stations. But how does this work? Is the resistance of the earth between cities really that low? How can it be low for stations that are thousands of miles apart, for example telegraph stations communicating across the Atlantic using an undersea submarine telegraph cable? There is absolutely no doubt that it does work, because early undersea submarine telegraph cables had only one copper wire.

       The first US telegraph line, Washington to Baltimore, built by Morse in 1844 was built with two wires for out and back, because although by 1844 ground return, invented by Steinheil, had been in use in Europe for a few years, Morse did not know about it.  (Prescott, p272, 428)

        Well, that's Prescott's version, but it seems to me that this may just have been a (good) conservative engineering decision. After all, when the line was built earth return was new and not well understood. Just because it had worked in a few trials in Europe didn't mean it would work everywhere, so it was safer to build the line with two wires. The line could easily be converted later to earth return if tests were successful, which they were, the two wires then allowed simultaneous transmission in both directions.

It's a mystery
       One of the (big) mysteries of the telegraph, at least to an electrical engineer, is how can the telegraph work with only one wire. All electrical engineers know that current flows out & back, in loops, in closed paths. In fact one of the fundamental theorems of circuit theory is that current always flows in loops. The electrical wiring to outlets in your house have two (current carrying) wires. When an appliance is drawing power from the outlet, the current (at every instant) flows out from one wire, through the appliance, and back into the other wire.
        In AC (60 hz) power the direction of current flow reverses every 8.33 msec, but at any instant of time the current flow in two wires to an appliance (as engineers say) is the same in amplitude but opposite in direction. In other words if the current is +5.34A (meaning out or right) in one wire at the same instant the current will be -5.34A (meaning in or left) in the other wire, so the net current (sum of the two currents) is zero at all times. {With a modern appliance that includes a 3rd (differently shaped) 'ground' terminal and 3rd 'ground' wire to the appliance a small (capacitive) current may flow in the 3rd wire, but it is still true that the sum of the three currents in the wires is zero at all times.}
        In a horizontal system like the telegraph is it not true that the requirement that current flow in loops means that current flow, say left, in the telegraph wire must be balanced (nulled) by current flow right in the ground (earth) at the same time? In other words it would appear that a current similar to the on/off (dots/dashes) current in the telegraph wire must be flowing miles through the earth, but in the opposite direction, between the telegraph station grounding poles. For example, does current sent through a submarine telegraph cable from Ireland to New Foundland, really flow all the way back from North America to Europe through either the earth or ocean?

Why is this a mystery?
        So if the telegraph only has one wire, where is the loop? It must be include an earth return path, right? In other words if the current goes into the telegraph wire at Boston and comes out (of the wire) at Washington DC, then it must be returning to Boston from Washington DC through some path in the earth, right? The problem is that when you measure the resistance (so I've been told) between a couple of pipes driven into the earth (simple grounds) that are only a few feet apart it is not all that low and the resistance goes up as distance between the pipes increases. Ohms laws tells us that (for DC) current equals voltage divided by resistance (I = V/R), so for a given voltage the higher the resistance the smaller the current that will flow.

        So basically the question is,

        How (the hell), with the modest battery voltages used in 19th century telegraph systems and the need for fairly substantial currents to drive 19th century mechanical telegraph sounders, is (enough) ground current able to flow tens, or hundreds (or thousands in the case of the undersea Atlantic submarine cables) of miles through the earth between stations to form a (substantial) loop of current?
        It is a puzzlement!

        In modern terms the issue is whether an earth (in an earth return) 'looks like' resistor or a capacitor. On page 282 Prescott says for the moment he is going to reserve comment on whether the earth is a resistor or capacitor, but that Steinheil, who invented 'earth return', thought that the earth was acting as a resistor, meaning the current was actually flowing between stations. Steinheil knew that poor conductors (like earth) could in fact have low resistance if the conductioning cross sectional area was large enough, and this was done using large metallic plates to ground the telegraph. Prescott has the following sketch implying that this is either from Steinheil or represents Steinheil's view about the current path.

Steinheil's view of telegraph earth return current.
Note large (20 square feet) grounding plates, P)
Prescott's 1888 'Electricity and Electric Telegraph' Vol I (p 283)

History of Earth Return (from Prescott)
        In 1838 Steinheil is testing whether he can use the two rails of a railroad  as the out and back conductors for a telegraph. He finds no he can't because the current 'leaks' from one rail to the other, and (for some unspecified reason) he thinks it's flowing through the ground not the ties. So from this he gets the idea that the ground might work as a return. According to Prescott Steinheil tries running his telegraph with ground returns and it works. In 1858 Steinheil wins a European telegraph cash prize for the invention/discovery of earth return (Morse is the other telegraph prize winner in 1858.)  Prescott comments about this as follows:

        "The discovery by Steinheil that the earth may serve as a conductor for the galvanic current is justly regarded as one of the most important discoveries in the art of electric telegraphy ever made."  (p282)
       The first telegraph lines of a few miles were built 1837 - 1839 in Europe (some by Steinheil). The first US line of any length (Washington, New York, Boston) is built 1844. So earth ground was used in telegraph lines (of all length, land and water) almost from the very beginning. The very first US telegraph line, an experimental line from Washington to Baltimore (44 miles), built in 1844 was built with two wires, because Morse who built it was unaware of earth return being used in Europe, but it was soon converted to earth return.

Using the earth as ground
        Using a computer in the New York Public library main reading room on 5th Ave (of all places!) in summer 2008 I found a very interesting discussion of the earth as telegraph ground in a 19th century telegraph book. I'm pretty sure it was Prescott's book on Electricity and  Telegraph, since I had discovered it before the trip to NYC, but I am not sure what edition of the book came up. I now find a similar discussion in the first edition, Prescott's Electricity and the Electric Telegraph, 1860 edition, page 70.

Curious resistance vs distance relationship
        Presott quotes an experimenter (Matteucci) who measured resistance of earth vs distance. He found that current (for a fixed voltage) decreased with distance initially, but when the distance reached 60-100 yds the current stopped decreasing and for longer distances began to increase! At a distance of ten miles or more it is found that the resistance of the earth return is small compared with the resistance of the wire.
       In other words (according to Prescott) something very interesting starts to happen at a distance of 60 to 100  yards. For distances less than 60-100 yards earth resistance is (pretty much) proportional to distance in the classic sense, but at 60-100 yards something very curious happens. At this distance is approached earth resistance stops increasing with distance, and it plateaus at a maximum value (unfortunately no ohmic value is given). For distances longer than 60-100 yards resistance starts to decrease as distance increases! Prescott says that at a few miles or so it decreases to what the 19th century electricians generally refer to as a 'null' value. Again unfortunately references never give (that I can find) an ohmic value.

       What 19th century telegraph electricians generally measure, in other words what they really know, is when earth return is substituted for wire return, the current (as best they can measure) doubles. If we make a guessimate that they could only measure the 'doubling' to 10% or so accuracy, then their 'doubling' claim is equivalent to saying that earth resistance is less 1/10th of line resistance. Of course, (effective) earth resistance between cities may really be close to zero, but from the 19th century 'current doubling' claims, I think all we can say is that (effective) earth resistance is probably less than 2 ohm/mile or so (=  1/10 x 20 ohm/mile). {I  did find one reference that said the resistance of earth was about the same as a seven miles of telegraph line. This gives us a nominal resistance of 140 ohms (=7 miles x 20 ohm/mile), which is consistent with the my guessimate of 2 ohm/mile (max) for a line of 70 miles or more.}

Earth return's big advantages
        Note the property of the earth of providing an easy 'grounding' together of circuits hundreds of miles apart has two huge advantages for the telegraph. One, only one wire rather than two need be run between stations (obviously a big reduction in construction cost), and two, for the same battery the current is doubled as the resistance between stations is (pretty much) halved.
        * Matteucccci runs an 4 terminal experiment at a distance of 160 yd. Digs four wells in a row and connects a battery to the outer wells (160 yd apart). A galvanometer connected between the two inner well shows a deflection. A voltage in the ground path is proof that over a distance of 160 yd (at least some) of the current is returning in the ground. (Prescott Vol I, p403)

        * Baumgartner runs an experiment measuring the resistance between Vienna and cities of different distance when the line was all metallic and was metallic with earth return. Prescott gives the following data, but unfortunately he is not clear as to what is being ratioed. This data makes some sort of sense (to me) if the ratio is assumed to be the ratio of the earth path to some (unspecified) reference. Then the interpretation would be that earth resistance goes up approx with distance between 20 and 50 miles, but something happens above 50 miles. A distance increase from 52 to 134 miles results not in a resistance increase of x2.5 but a small decrease.
                               19.44 miles                        ratio 3.14
                               52.36 miles                        ratio 6.98
                             134.06 miles                        ratio 4.7

        *  One analogy is a constant current flowing into an infinitely large reservoir with no apparent accumulation. Prescott, who was the chief engineer of Western Union, thinks the earth acts like a reservoir for current,  meaning it acts like a very large capacitor. Prescott says at one point current entering the earth plates is "diffused in every direction without producing around them any appreciable state of potential."  He describes a couple of experiments where double earth connections are shorted and unshorted by a wire and a difference is found. Prescott says this is a very strong indication that the earth is acting as a reservoir not as a conductor, but the experiments are describes so poorly (for example, the distance between the plates is not given) that I can't review it.

        --- On 'very short lines' the resistance of the earth is has 'an appreciable value', but for lines of '75 miles or more' the resistance of the earth return becomes 'too small to be taken into account.' (p 411) {This is also consistent with the previous estimate that the resistance of earth is 2 ohm/mile (or less).}

How does the current flow in earth?
        A subject that has long bothered me (How does the current flow in earth?) I see from Prescott's book that it bothered Prescott and the other electricians of the time. Prescott says:

        "It is not easy to determine whether the earth really conveys the current in the manner of an ordinary conductor (resistor) from one station to another, or whether it should be regarded merely as a reservoir (capacitor) into which the electricities of the battery pass." (p282)
        ----  Telegraph operators found the telegraph would (sometimes) work even with the batteries removed (replaced by a short).  Proves that over many miles the earth can have several volts or more.
an aside  --- iron telegraph wire
        Prescott says the first telegraph lines were built with copper, but copper was abandoned very quickly (within three years in US) in favor of iron. He says (in 1888) that virtually all telegraph lines in the world are made of iron wire. Iron's big advantage was that it was 12 times stronger than copper, so it could be run longer distances between poles. To compensate for its higher resistance (x6 higher than copper) an iron wire needed to be sqrt{6} = 2.4 larger diameter than an equivalent copper wire. The cost of the two wires was about the same.
Telephone use of earth ground
        Why did 'ground return' not work for the telephone as it did for the telegraph? noise voltages in audio band? ring problems? The current output of the early Bell designed electromagnetic telephones of 1877 and 1878 was tiny and his receiver sensitivity was low. In fact the earth was used in the first years of the telephone (late 1870's) as the return, but its use as a return was soon abandoned. Why? My first guess was there was a  problem with pickup of 'hum' from nearby power lines, but this can't be right since the first power plants weren't built in the US until five years or so after the telephone begins commercial operation. When lightning storms were around, I bet you could hear some crackling and popping.

        In 1878, only a year or so after the first telephones begin to be used, Bell is finding that a wire return works better than earth and files for a patent on it. The next year in an 1879 patent filing, issued as patent #220,791, he says "much difficulty has been encountered when (telephone lines) are employed in the neighborhood of other lines." In other words use of a single wire with earth return was leading to cross talk between circuits, you could hear other conversations. It's hard to say how much the resistance of the ground contributed to the crosstalk, but Bell in the patent says the crosstalk is due to "induction" coupling between adjacent wires, and this is probably correct. Two long adjacent wires form sort of a transformer with a fraction of the magnetic flux of one wire encircling the other. Bell's 1878 solution is to run telephone signals out and back on a "twisted pair" of wires abandoning earth return. I know from experience that twisted pairs work great for isolating signals, and telephone signals to this day are run on twisted pairs.

        In his 1879 filing Bell shows that what works best for telephone signals is metallic out and back (on a twisted pair) with one wire of the pair also earthed. The earthing allows (he says) metallic return and single wire with ground return circuits to be combined. I believe the telephone circuit in my family house was (crudely) earthed by a connection to a water pipe. So in effect modern telephone currents have two possible returns paths, through a  wire in the twisted pair or earth. What happens (I'm pretty sure) is that due to the inductive effect between the two wires of the pair nearly all the telephone audio and higher frequency currents return in the wire. The grounding serves only to prevent DC potential from developing between circuits. This provides safety in handling telephone circuits and allows them to be interconnected in any which way without DC currents flowing in the wires.

an aside --- power distribution with one wire
        I read that Tesla had patented means for (supposedly) delivering power with one wire. The figure in his patent (593,138) does in fact show only one distrubution wire with earth return. But the patent is really about the design of a of a transformer with a huge step up/dn ratio built with a sophisticated winding. Running the line at very high voltage reduces both the line and ground current and makes the system tolerant of substantial resistance in the earth path. Tesla in his patent gives no step p/dn ratio, no voltage and no power numbers, but he does say the transformer could be built with 50 miles of wire! This is clearly a very high voltage transformer, maybe tens of KV?

        I read somewhere that power distribution in rural areas is sometimes done using only one wire with earth return, but I did not dig out a reference on this, so no details..

Modern high current, low resistance earth ground
            How does the earth work as a ground? What are the electrical characteristics of the earth if you want to sent current hundreds or thousands of miles? In forty years as an electrical engineer I have never come across any information like this. A colleague made me aware of Wikipedia entry (below) about a high power DC power link from Oregon to Los Angles that uses earth grounding. How the grounding was done is described in detail and from the electrical data given I can (roughly) figure out how well it works.

        The power grounds on this DC line are really massive. "The grounding system at Celilo consists of 1,067 cast iron anodes buried in a two foot trench of petroleum coke, which behaves as an electrode, arranged in a ring 2 miles in circumference.  The Sylmar grounding system is 24 silicon-iron alloy electrodes submerged in the Pacific Ocean suspended in concrete enclosures about one meter above the ocean floor."

         The line consists of two wires each about 1.5 inches in dia. The max power of the line is 3,100 MW (million watts), roughly equivalent to the power output of three (1,000 MW) large power plants. The max voltage is given as 500,000 V = 1/2 million V. Since this is DC, the current is just the power divided by the voltage.

                           (3.1 ^ 10^9 W)/(0.5 x 10^6 V) = 6,200 A

    The Bonneville Power Administration, which runs the Pacific Intertie, gives the spec of the line this way:

                    Monopolar            2,000 Mw   at 500 kV        3,1000 A/pole
                    Bipolar                   1,550 Mw  with Earth Return or 1,000 Mw with metallic return
                    Conductors            2 x 1,171 mm^2   ACSR/pole  (cross sectional area does not
                                                                                          include the steel reinforced core)
                    Length                    1,161 km (850 mile)

       The slightly cryptic Bonneville specs are consistent with Wikipedia if we assume that the 1,550 Mw (with earth return) applies to each wire of the bipolar system, giving us a total power of 3,100 Mw. The conductor areas (exclusive of steel core) are consistent with 1.5 dia copper wire [1.52 in dia = sqrt{1,171 mm/pi} x 2 x (1in/25.4 mm)].

        Monopolar means one voltage     +V0
        Bipolar means two voltages         +V0,  -V0 (with respect to ground)
        Bipolar is a preferred mode of a two wire system because while it nominally uses an earth return, the earth return currents of the +V0 and -V0 lines (nominally) cancel.  (It's reasonable to assume the plant operators always keep the power balanced in the two lines). This makes bipolar the most efficient configuration for the same reason that three phase AC system is most efficient. In both case there is (sort of) a phantom return path (earth return) with little to no current in it and hence no losses in it. However, it does look like a disadvantage of bipolar (assuming the voltage is 500 kV to earth on each wire) is that the isolation between wires is doubled to 1,000 kV.

        From the reduction is the line's rating between bipolar (3,100 Mw) and monopolar (2,000 Mw) with earth return we can get an estimate of the earth's resistance over the roughly 850 mile length of the line. Lets assume the reduction in delivered power is due to ohmic losses in the earth path. (Of course, it's possible, but unlikely, that the reduction in power is due to some limitation in transformers or the conversion plants.) A simple estimate can be obtained by assuming the current remains at max (6,200 A) and about 35% of the original 500,000 volts delivered is now lost in the earth path.

    Bipolar delivered                             500 kV x 3,100 A x 2 = 3,100 Mv
    Monopolar delivered                      (2,000 Mw/3100 Mw) x 500 Kv x 6,200 A = 2,000 Mw
    Earth voltage @ 6,200 A                (1100 Mw/3100 Mw) x 500 Kv = 177 Kv
    Earth resistance @ 850 mile          177 Kv/6,200 A = 28.5 ohm
    Earth resistance per mile                28.5 ohm/850 mile = 33.5 mohm/mile

    total power lost in earth path          (6,200 A)^2 x 28.5 ohm = 1,100 Mw    (checks)
   power dissipated per mile                1,100 mw/850 = 1.29 Mw/mile
   power dissipated per foot                 1,290,000 w/5,280 ft = 244 w/ft
            check                                          (39 mv/ft x 6,200 A =244 watt)

How much current can wire carry?
        Here is the check of the current handling ability of the wire. A good way to check this is to figure how many watts are dissipated by each foot of wire (I^2 x R) when running the line at full current. I will ignore the steel core and assume the wire is all copper. Scaling from #10 AWG wire that has a resistance of 1.00 milleohm/ft, we find the resistance per foot for the 1.5 in dia wire is 4.6 microohms (see below).

                        #10 AWG wire                   0.102 inch dia                   1 mohm/ft
                        R/ft (1.5 in dia wire)   (0.102 in/1.5 in) ^2  x  10^-3 ohm = 4.6 x 10^-6 ohm/ft

        Each wire carries half the total current, or 3,100A  The equation for power dissipation in the wire is P = I x V, which since V = I x R (ohms law), then P = I^2 x R.

                     P/ft (1.5 in dia wire)   =     (3.1 x 10^3 A) ^2  x 4.6 x 10^-6 ohm/ft
                                                           =     44 W/ft

        We all know how hot a 100W light bulb runs. A 1.5 in dia wire one foot long (probably) has more surface area than a 100W bulb and is more spread out, so at 44W it will run pretty warm but for copper that's OK. The wire is not insulated, and one reason for that is probably so it can dissipate the heat.

              R (each wire, 841 miles) = 8.41 x 10^2 miles x 5.28 x 10^3 ft/mile x 4.6 x 10^-6 ohm/ft
                                                         = 20 ohm

Earth resistance estimate --- 28.5 ohms for 850 miles
            In this 850 mile DC system with two 1.5 in dia wires, with massive 2/3 mile diameter earth grounds, the wire resistance (each wire) is about 20 ohms. From the reduction in delivered power when monopolar operation forces high current (6,200 A) to flow in the earth return, we estimate the DC earth resistance (total) for the 850 miles to be about 28.5 ohms, or only 33 mohm/mile.

Footnote -- In normal bipolar operation the power lost is all in the wires @ 3,100 A/wire is  (2 x 44 w/ft = 88 w/ft) or 395 Mw total or 12.7% of delivered 3,100 Mw . In monopolar operation the wire loss (@ 3,100 A/wire) remains the same (395 Mw total) and the earth ground, even though earth resistance is only 33 mohm, adds another 244 w/ft @ 6,200 A, so the total monopolar loss comes out to be (1,495 Mw = 1,100 Mw + 395 Mw total), which means that total loss comes out to be 75% of delivered 2,000 Mw. Pretty bad.
Multiplexing introduction
        The telegraph (for many years) was limited to one message at a time in one direction over one wire. If there was only wire between stations, the two stations would alternate between sending and receiving. To send more messages per hour between stations you either had to click more rapidly, hence speed of keying and 'reading' code was valued among telegraph operators, or you needed to string more wires between the stations.

        It was clear that there was lots of money to be made if anyone could invent a practical way to send more than one message over a telegraph wire, so lots of inventors worked on it, including Bell, Edison, and Gray. (I think there may have been a prize offered too.)  A telegraph wire (or any communication channel) that is shared is referred to as 'multiplexed'.

        There were two general approaches:
                    * Send in both directions at the same time
                   * Send multiple messages in the same direction

        The first required figuring out how to send and receive at the same time without sending interfering with receiving. The second required figuring out how separate multiple, overlapping, messages at the receiving end.

        There were three general approaches as to how multiple, overlapping, messages might be structured so that they could be separated at the receiving end:

                    * positive and negative current separated with a 'polarizing' relay
                    * switching high frequencies on/off with tuning fork like resonant sounders
                                    (frequency multiplexing)
                    * high and low voltage (amplitude multiplexing)

        Duplex telegraph is (generally) defined as a system that can send both ways simultaneously, and diplex is used to mean sending two signals simultaneously in the same direction. However, Edison appears to use the word duplex in his patents to mean both.

        The equipment in a duplex (both direction) telegraph must prevent the (stronger) outgoing (sending) current from tripping the local sounding relay while at the same time having a (weaker) incoming current from a distant station trip it. A seemingly impossible requirement! In fact this is a non-trivial problem. The telegraph had been in widespread use around the world for 30-40 years before anyone figured out how (in a practical way) to do it!

Differential duplex
        Differential duplex is an interesting and clever scheme. Invented by Stearns in 1872. (In spite of a lot of searching I can't find the original Stearn's patent #.)

        It depends on the fact that the magnetic field developed in the magnetic core of a coil depends on the net amp-turns (Ni) of current in all its windings. This means sending the same current around the coil in two similar winding, but in opposite directions, cancels the magnetic effect. This was exploited by Edison (among others?) to prevent the sending current from tripping the sounding relay while still allowing the distant current to trip it. The sending current is inserted into the center of the sounding relay coil arranged so the current splits equally, half going out to the Line and the other half (in the opposite direction) to a local resistor and coiled trimmed up to have the same R & L as the line.

Differential Duplex Circuits
Telegraph key is SPDT and must be 'Make before Break'. Outside adjustable resistors are the dummy line.
Notice anything interesting? Ref left says batteries must be opposite, but 1922 book right shows same polarity.

Operates with no line current!
        In virtually all telegraphs, of course, a depressed key causes a current to run down the line (between stations) and to trip a relay at a distant receiving station, but this telegraph is different. The circuit on right with (equal) positive batteries at both ends operates in a very curious (counterintuitive) way when both keys are depressed. The voltage being equal at both ends of the telegraph line (effectively a resistor) results in no current flows in the line, yet the sounders at both stations are tripped! What? Yup it works. In this case the local battery in each station is sending current into the dummy line, and this is what activates the local sounders. Interesting.

Line 'twiddle' box
        To set up a differential duplex an RC dummy line was needed, and every telegraph line was different because the impedance depended on the length (weather too). To make it easy to match the dummy line to the real line one telegraph company manufacture a real twiddle box (maybe the first?). A 'twiddle' box is an box filled with resistors and/or capacitors that allows you to 'dial in' the impedance that you need. It's a an electronic lab staple. The telegraph box had three resistance sections (10's, 100's and 1k's) and a parallel series RC section (0.1uf to 2.5 uf and 0 to 400 ohms), which, of course, bounds the range of telegraph line impedances.

19th century 'Twiddle Box' used for RC dummy lines in duplex telegraphs

Stearns duplex
        The idea of using a sounder with a split coil and a dummy line to make a duplex telegraph had been in the air (and experimented with by Gintl & others in Europe) for 15 to 20 years before Stearns (of Boston) in 1872 solved a problem it had. The problem was that the line had capacitance, and when it charged and discharged, it created an (uncancelled) pulse of current in the sending coil. The duplex would work for a short line (50 miles), got shaky for lines of 100 miles, and was unusable for lines of 400 miles. It was just not robust.
        The electricians knew the line had capacitance because when they quickly switched a line from the battery to earth through a galvanometer they saw the galvanometer pulse (transient deflection). This was the line capacitance discharging.
        Stearns solved the duplex problem in 1872 by adding a (parallel) capacitor across the dummy load resistor. This capacitor when adjusted properly caused the dummy line to draw a transient current pulse (at the edges of the pulses) just line the line. [We know today that the area (charge) of the pulse was the same as the line, but not its shape.] This neatly solved the false trip problem robusting the duplex differential telegraph causing it to be immediately adopted by Western Union, which put it into widespread use.

        It's interesting that today to an electrical engineer adding this capacitor is trivial, but capacitors (at least of this size, around 1 uf) were (apparently) in the 1870's exotic animals, not (widely) manufactured and (perhaps as a consequence) not well understood.

Stearn's differential duplex circuit showing his 1872 addition of a parallel capacitor, marked 'C', to the dummy line (on left)
(from Prescott's 'Electricity and Electric Telegraph' 1888)

        The key element in the figure above is the split coil, wound on the top receiver. Note the sending current (circuit below and right) goes into the center of the split coil, half going right to the line (L) and half to the left to the dummy line (X C, where X is a resistor and C a capacitor). The left end of the split coil is shown controlling a lever that would operate a sounder in a local loop (not shown). The magnetic flux generated in the split coil iron, hence the force on the lever, depends on the net NI (amp-turns) in the coil. The polarity of the two coil halfs would be chosen so that the sending current flow is clockwise in half the coil and counterclockwise in the other half. This causes the magnetic effects of the two terms to cancel resulting in no net flux in the bar, thus the receiving sounder does not respond to the pules being sent out.

        The complication shown in the sender circuit (bot, rt) is not relevant to the invention. It looks complicated only because the sending key (K) is driving the switch that drives the line through a local sounder. The local sounder is there to allow the sending operator to hear what he is keying.

Bridge Duplex
        There was anther clever way to make a duplex telegraph. This method used a bridge configuration and was also used by Western Union. As with the differential duplex, it needed a key that was 'make before break' so that there was always a path to ground for the receiving current. The bridge configuration was developed in the 1840's (by Wheatstone) to be a supremely sensitive and accurate way to measure an unknown impedance. Essentially a bridge circuit is two impedance dividers connected to the same source with a differential sensor between the outputs. A nulled sensor means the dividing ratio of the two impedance dividers is the same. [Even today precision impedance bridges are still in use.]

        A bridge in electrical engineering is a four impedances arranged in two legs (usually drawn as a triangle) that is typically excited at the top with the bottom grounded and with a sensor hung differentially. When the ratio of impedances of the two legs are the same, the center voltages of the two legs (to ground) are also the same resulting in no (net) voltage across a sensor hung differentially. The bridge is said to be balanced.
        An unknown impedance is measured by putting it into a bridge with three known impedances one of which is adjustable. The adjustable impedance is changed until the bridge nulls, then with the ratios of the right and left legs the same, the value of the unknown impedance can be read off as a combination of the three known impedances. For example, if A/B = C/D, then A = (C/D) x B.
         Below is a figure from Prescott's 1888 book showing the principle of the bridge duplex telegraph. (Prescott labels this as Stearn's Bridge Duplex telegraph not because Stearns was the first to conceive of a bridge duplex, it was invented by Maron in Germany in 1863, but because it includes the shunt capacitor that Stearn's first added making it practical.) The bridges are a little hard to see because they are drawn rotated 90 degrees with the excitation (keyed battery) drawn on the side (rather than the top as is conventional). The left bridge is leg [A + L (line)] and leg [B + RC], but note L (line) is the line 'as seen from' the transmitting side, so it includes not just the resistance and capacitance of the true line but also a complex impedance of all the elements on the right including the sounder.

Stearn's bridge duplex telegraph (1872)
(from Prescott's 'Electricity and Electric Telegraph' 1888)

        How it works is that the ratio of B to RC (dummy line) in left bottom leg is adjusted to match the ratio of A to L (line as seen from left) in top left leg. This prevents the keying at left from exciting the left sounder, which is hung differentially across a nulled bridge. From a superposition viewpoint (short all voltage sources except one) the current in each sounder can be seen to come only from the excitation at the other end. In the simplest case:

        Set A=B and dummy line (RC) = (Line + termination). Excitation current splits evenly at node H with half in A and half in B, so each excitation makes no voltage across the sounder branch at its end.
        On the receiving side (right) as seen from the line there is no bridge configuration. In essence the transmitting side is driving a bridge in the conventional way whereas on the receiving side the line current excitation enters at a node rotated 90 degrees, so from the perspective of the current coming in from the line no bridge is visible.

        The current from the line wants to get to ground (directly or via the battery). It sees two paths in parallel, one is resistance k and the other is the sounder in series with B,R,C in parallel. If all the elements are roughly in the same ohmic range, then half (or a little more than half) of the line current flows through the sounder. The loss of some of the line current at the sounder probably means that the bridge duplex technique is less sensitive than the differential duplex technique. However, an advantage it has over the differential is that there is more freedom is picking component values.

        Note the batteries are shown reversed. Line currents (pointing left) are 0, +i0, +i0, +2i0 (approx), but I think it works with batteries of same polarity, then currents are 0 +i0, -i0 and 0 With battery polarity and both keys closed superposition gives 0 for line current and says opposite end drives sounder, but, of course, the equivalent (and common sense) viewpoint is that in this case the sounders are driven by the local batteries.

Bridge and differential duplex (side-by-side)
        Here's my sketch comparing the bridge and differential duplex circuit architectures.

        The principal of operation of the Bridge Duplex is the Wheatstone bridge. In this circuit the sounder is placed in the center of a bridge that is balanced from the sending side, but not from the receiving side. Technically, voltage from the sending key creates a common mode voltage across the sounder leg of the bridge, but no differential voltage (or current) because the bridge is balanced.

        The principal of operation of the Differential Duplex is that the flux in a solenoid (sounder) depends on the net 'Ni' of its windings. Technically, this is superposition. The net flux in a solenoid is sum of the fluxes created by the currents in each winding (taking polarity and the number of turns into account). The trick in differential duplex is to add a second winding to the sounder and flow some of the sending current though the added winding backwards. With the right scaling the two flux terms cancel resulting in no (net) flux in the sounder when sending.

        This is most easily done by center tapping the sounder coil and injecting the current from the sending key into the center with the line (to ground) at one end and a matched dummy line (to ground) at the other end. Note the dots near the right hand ends of the two coil windings in the figure. This is the modern method of indicating winding polarity of a coil in a schematic. The number of turns of each winding is also marked (N). Notice the current to the line (i0 through N turns) flows out of the dot, while the current to the dummy load (also i0 through N turns) flows into the dot. This means that the Ni's, or fluxes, of the two windings cancel (Ni0 - Ni0 = 0).

two duplex telagraph configurations

        Below is basically the same thing (but less clear I think) from a site maintained by R. Victor, Jones (Research Professor of Applied Physics in the Division of Engineering and Applied Sciences, Harvard University) with bridge duplex on top and differential on the bottom. In his sketch the sender is flipping the battery polarity and the receiver is a polarized relay, but the architectures work also with battery voltage just switched on/off and a normal (non-polarized) sounder used as receiver.


Edison's duplex
        The story in the Rudger's archives is that after Western Union adopted the Stearns duplex method in 1872, it then hired Edison to invent and patent other ways of duplexing so Western Union could hold a monopoly. When Edison was doing this duplex work for Western Union (so the story goes), he realized that his amplitude/polarity duplex (also diplex), filed in 1873 and issued as patent 162,633, could be combined with existing (Stearns) duplex circuits to make a robust a quadruplex telegraph. The Edison quadruplex was adopted by Western Union in 1874 just two years after the Stearns duplex was adopted. The Rudger's archive says the Edison quadruplex was the most significant of his many telegraphic invention.

Amplitude/polarity duplex & diplex
        Below is my sketch of a basic amplitude/polarity telegraph of the type invented by Edison. Top shows it configured as duplex, where messages are sent simultaneously in opposite directions, and bottom as diplex, where two messages are sent simultaneously in the same directions. Notice all it takes to go from duplex to diplex is to switch a couple of boxes. The operating principle and waveforms are unchanged. The text of Edison's key patent on the amplitude/polarity telegraph (162,633) says it can do both duplex and diplex, but he only sketched duplex operation in his patent figure.

        I drew my duplex/diplex sketch to help explain what Edison is doing and as a key to 'reading' his patent figure. There are several ways to implement the building blocks shown in my sketch, which makes for lots of minor variants. I show the 'polarity flip' (left) implemented with a rather complex switch (double pole, double throw) and a single battery. Edison uses a simpler switch (single pole, double throw), but this forces him to double up on the batteries (marked MB and MB prime) and on the sounder coils. His normal, double coil, high threshold sounder is rather tricky. It has a center pivoted bar (pulled right by a spring) which is pulled (left) to 'click' by powering either coil A or coil B.

duplex and diplex telegraph configurations

Figure from Edison's amplitde/polarity duplex telegraph (patent 162,633), filed April 1873.
This became the basis of the Edison quadruplex telegraph.
Double circle symbols (marked MB, MB prime, SB) are batteries. The top CC device is a polarized relay.
The current amplitude (at the far end of the line) is varied by just shorting out the line grounding resistor (R).

Edison's quadruplex
        Prescott says the Edison's quadruplex (as shown below) was the first quadruplex put into production at Western Union (in 1874), though several improved versions followed it into production. Note this is basically the bridge duplex configuration (above) combined with Edison 162,633 amplitude/polarity diplex (above). It's two messages in both directions at the same time, so it has double excitations and double sounders at each end. The two ends are the same (except I think for reversed battery polarity.)

Amplitude/polarity quadruplex
        Below is my sketch showing how Edison's amplitude/polarity diplex was combined with Stearn's bridge duplex to make the commercially successful quadruplex telegraph. The key trick here is that anything placed in the center leg (vert leg on right) of a balanced bridge is unable to 'see' the sending voltage. So what is done is that a polarized sounder and high threshold, normal sounder are put (in series) in the bridge center leg. In this location they get little to no current from the sending circuit (left), which is basically the same as in Edison's diplex (above), but they do see (a fraction) of the current coming down the line, so they respond basically the same as in Edison's diplex. By using the same configuration at both ends of the line two messages can now be sent in both directions all simultaneous, hence a quadruplex telegraph.

edison-stearns quadraplex telegraph configuration
(Line current is shown increased by 'k' account for current losses in a bridge configuration)

        The bridge configuration produces the duplex action (excitations are prevented from driving the local sounders), and diplex (two messages in same direction simultaneously) is accomplished by sending one message with current polarity and the other with current amplitude (small or big). In Edison 1873 polarity/amplitude duplex/diplex patent the amplitude variation was provided by shorting a resistor. In the figure below from Prescott's book, supposedly showing the commercial quadruplex configuration, the line current amplitude variation is accomplished by switching a extra battery (3 V0, the long stack) in/out of the circuit.

Edison's bridge quadruplex telegraph (1874)
(from Prescott's 'Electricity and Electric Telegraph' 1888)

        The lower section of the keyed excitation controls amplitude [by selecting V0 or (V0 +3V0 = 4V0) battery voltage], and the upper section of the keyed excitation controls the polarity (by flipping which end of the floating battery stack is earthed).  There are two sounders in series placed in the center legs of each bridge. The upper is a polarized relay that detects polarity, the lower is a neutral relay that only responds to high amplitude currents (positive or negative). The receivers look complicated partly because they are shown as relays driving sounders in local loops. Both ends of the line are the same except (I think) the battery polarities are reversed.

Ride-thu circuit
        Prescott points out that one of the difficulties in this circuit that had to be managed, and which got worse as the lines lengthened, was the tendency for the amplitude relay to briefly drop out when the polarity reversed. This is the +3i0 to -3i0 voltage transition that I sketch in my Duplex/Diplex figure above. The general approach was a second relay that responded more slowly coupled to the first to provide ride through. This is (apparently) the function of the additional extra business in the lower sounder section (though I haven't worked out the details.)

        Prescott also points out that a real strength of the configuration was the large separations in the four signals. The large battery could basically be any ratio of the smaller battery, typically the ratio was set at 3 (or 4) : 1.

Automatic telegraph ---1,000 wpm
        An alternative to multiplexing, used by some telegraph companies, was to crank up the send/receive speed using automated equipment. This involved preparing a tape with the message translated into dot and dashes so that it could be dragged under the send key by some sort of clock mechanism. Receivers of various types existed that would ink, or scratch out, the dots and dashes, to be read later offline. But as pulses got shorter a problem arose, especially with longer lines. The inductance of the line causes the dots and dashes to have a finite rise time. As speeds pushed 1,000 wpm the dots and dashes at the end of the line were overlapping and receivers stopped working.

Baseline numbers
        Assume for the moment that the line can be modeled by lumped circuit elements. Using the 1 uh/meter rule of thumb for straight wire in free space, which I know is accurate in the lab for copper wire, we can estimate the inductance of the line as 1,600 uh/mile (approx). Since nominal line resistance is 20 ohm/mile, that makes the line L/R time constant = (1,600 uh/20 ohm) =  80 usec.

Baseline numbers don't work out
        But somehow the baseline numbers don't show the problem. 1,000 wpm @ 5 letter/word = 5,000 letters/min = 83 letters/sec. This is 12 msec/letter, which at an av of 3 char/letter is 4 msec/char. With a 3:1 dash/dot ration this is (3 msec + 1 msec space) for a dash and (1 msec +1 msec space) for a dot. A 1 msec pulse is only slightly softened by an 80 usec rise and fall time.

What can be wrong here?
        I can think of three possibilities. One possibility is that some lines may have been 10 ohms/mile. This doubles the L/R time constant from 80 usec to 160 usec, but it's still not large enough to cause pulses to overlap. A second possibility is that the inductance causing the overlap is not in the line its mostly in the sounder. Who knows what the inductance is in the sounders, all I ever see listed is the resistance. But it would have to be awfully large. If the line was say 300 miles, the line inductance (@ 1 uh/m) would be 480 mh (= 300 mile x 1.6 mh/mile), so for overlap the sounder inductance would have to be probably x10 this value or larger, which is 5 henry. May be possible, but not likely.

        The third possibility is that the culprit is the iron wire. Copper wire was found to sag and break so much that it was very quickly replaced iron wire. In any wire with current some of the magnetic field is actually inside the wire. It's likely that high-u ferromagnetic iron telegraph wire had substantially more inductance than non-ferromagnetic copper wire. The 1 uh/m rule of thumb is for a non-ferromagnetic material. The L/R dynamics of iron telegraph wire may also be further complicated by what is known as the skin effect. When current changes rapidly in a conducting wire, it is pushed to the outside of the wire. (This happens in copper wire as well, but the increase in wire resistance with frequency is significant only at frequencies much higher than those used in the telegraph.)

(update 5/09) Iron wire inductance
       I did a little searching  to try and find the inductance of iron wire. I found an 1893 Google Book reference (Transactions of the American Institue of Electrical Engineers) that says the permeability (u) of iron wire at telegraph current is about 150. This means the fraction of the magnetic flux within the wire stores x150 energy than in copper wire. The inductance for a telegraph wire is said to be about 12 mh/mile higher than with copper wire. Using the 1 uh/meter rule copper wire has an estimated inductance of 1.6 mh/mile, so these values would indicate that iron telegraph wire probably had about x8.5 times higher inductance than copper telegraph wire. The reference (p211) says that copper wire (with ground return) has inductance of about 3 mh/mile vs 15 mh/mile for iron wire or iron is x5 times higher.

        Ignoring the complications of skin effect with iron wire sized to have an equivalent resistance to copper wire the L/R response time would be not 80 usec, as it is for copper, but (5 to 8.5) x 80 usec = 400 to 680 usec. This would have a major effect effect on 1 msec pulses.

Prescott describe the overlap problem and its fix this way
       "Due to inductive action of the line the dots and dashes ran into each other. Little improved thing somewhat by putting a shunt around the receiving instrument with an adjustable rheostat in the shunt. (A parallel resistor.) Edison added an electromagnet in the shunt and this vastly improved the response. This effect is due to the opposing induced currents set up by the magneto-electric action within the short circuit formed by the shunt and the receiving instrument." (Prescott page 725)
Isn't the line a distributed system?
        But of course the line is a distributed system and for accurate modeling the speed of light needs to be taken into account. We can estimate the speed of propagation of signals on the line as between 1 nsec/ft (speed of light in free space) and 2 nsec/ft (typical coax cable). This translates into a delay of 5 to 10 usec per mile. Since this is about an order of magnitude less than our estimated L/R time constant (80 usec), we probably can with reasonable (90%) accuracy model the line with lumped components ignoring the fact that it is a distributed system.

Tuning the line
        This initially seems like an unsolvable problem. After all if the line inductance limits the rise time of the receiving current, isn't this a fundamental limitation?  But it turns out there is a fix for this, which in the 1870's was found by Little and Edison. The fix can be dated, because at the 1876 Centennial Lord Kelvin was a judge and in recommending an award for Edison's automated telegraph, he described the way Edison had tuned the line to fix the problem.

Lord Kelvin on Edision's line tuning
        William Thompson (Lord Kelvin) reviewed Edison automatic (telegraph) at "Centennial Exposition of 1876" and wrote a report recommending it for an award. His official report closes thus: "The electromagnetic shunt with soft iron core, invented by Mr. Edison, utilizing Professor Henry's discovery of electromagnetic induction in a single circuit (an inductor) to produce a momentary reversal of the line current at the instant when the battery is thrown off and so cut off the chemical marks sharply at the proper instant, is the electrical secret of the great speed he has achieved.
        I have been scanning Edison's patents of this era looking for a more complete description, but have yet to find it.

How tuning the line works
        The descriptions by Kelvin and Prescott are not described in terms of circuit theory or equation, so to the modern ear they are not very clear, but the tuning concept is based on a very simple equation. Assume the line has some impedance Z, typically this would be (R + Ls), but it really doesn't matter. If the line is loaded (terminated) with the same (matched) impedance Z, then the receiving voltage is

                      V receiver/V send = Z receiver/(Z receiver + Z line)

Set Z receiver = Z line

                      V receiver/V send = 1/2

        Wow! Sure you get only half the amplitude, but almost like magic the mathematics say the dynamics of Z, which is the L/R time constant, is gone.  But how does this really work? Consider that the line is modeled as a series L+R, and the line is terminated by a shunt series L+R adjusted to match the line.

Caveat --- This model is calculating the 'open circuit' voltage. This would apply (approx) if the impedance of the load, in this case the mechanism driving the inking pen, is high relative to the shunting tuning element. Whether this is realistic or not, I don't know, since I have not found any info on how much current it took in automated telegraphs to drive the pen.
        At the beginning of the pulse, when the current in the line is low, the impedance it is driving (terminating L+R) is high, so the open circuit receiving voltage can 'step' up. The rise time of the current in the line is matched by the rise time of current in the terminating shunt inductance. When the transmitting key opens dropping the sending voltage to zero, the current in the line only slowly decreases due to its inductance. This slow decay of line current is kept out of the (high R) load because it is 'stolen'  by the inductance of the shunt terminating L+R, allowing the open circuit voltage at the receiver to 'step' down. In other words most of the slowly ramping up/down line current is the line is absorbed by the shunt load that ramps in the same way, leaving a (small) unbalance current to (approx) step up and down, replicating the rectangular shape of the sent pulse.

        If this explanation and line model is right, then we should see a series L+ R shunt (to earth) added in parallel with (a higher impedance) receiver. This is in fact what we see. Below is a sketch of Edison's tuning from Prescott's book. M1 is across the receiver (on right) as the main shunt tuning element and a short rC ladder (on right) is also included to compensate for the (distributed) line capacitance.

Prescott on line tuning in Edison's Automated (high speed) Telegraph (p727)

Very bref history of telegraph
Dozens of wires
       In 1820 Ampere's idea for using electricity to communicate was to run dozens of wires (one for each letter of the alphabet) and to sense the presence of current in the wire by deflection of a needle.

Five wires
        The so-called 'needle telegraph' of Wheatstone and Crooke was first installed in 1839 and continued to be widely used in England. It used five wires (plus return) and five needles. A pair of needles would point to the letter in a grid.

Two wires
        Morse in 1844 on his first line (Washington-Baltimore) used two wires with the current actuating a steel point to scratch a dash or dot in a tape pulled by a clock mechanism.

One wire
        Later that same year (1844) a Boston-NY-Washington telegraph was built with only a single wire and earth return. Later it was found the tape could be dispensed with as good operators could 'read' the message directly off the sound of the relays clicking.

Did Morse invent the telegraph?
        Prescott says Morse sketched out his ideas for a fast, automated, recording telegraph on a cruise to Europe in 1832, but did not build a working table top model until 1835. There is a sketch of Morse's 1835 tabletop apparatus in the book. The cruise sketch included (says Prescott) his soon to be famous dot/dash letter code. The recording was done by a clock mechanism pulling paper through the receiver. The pen moved back and forth driven by current in the line leaving up/down squiggle on the paper to be read later by an operator.

        There was no telegraph key. Precision and speed were to be obtained by setting the message in movable type, each letter having a series of raised lines for its code. To send the message the type was smoothly cranked under a level that dipped into a dish of mercury completing the circuit and driving the line as the raised sections passed under.

       Something happened with Morse code (in the US) that is similar to the QWERTY keyboard. It was found that sloppy sending could make it difficult to identify some letter combination, because the space between letters tended to get mixed up with the space between dots and dashes of a letter. When the code was introduced into Germany, the code was improved to minimize this problem. Morse then tried to introduce the improved code into the US, but all the telegraph operators objected! So the US (as of 1888 anyway) continued to use the old (buggy) code while everyone else in the world used the improved code.
        So was the telegraph invented in the US by Samuel Morse? Prescott says Morse that in the sense that he pulled together a  lot of ideas (of others), made improvements, and made it all work reliably, then yes he invented the telegraph. The proof being that with within a few years other telegraph systems (like the needle telegraph) all over the world were replaced with the Morse (recording) electromagnetic system using his dot and dash code.

        Morse worked with Alfred Vail from 1837 to 1844 commercializing the telegraph. From the Wikipedia entry on Vail it appears that Vail provided engineering and Morse was the boss and entrepreneur. One reference (Calvert) credits Vail with the design of the electromagnets on the first Morse telegraphs, saying Vail wound his receivers with 3,000 turns of wire, and they weighed up to 75 lbs. Morse and Vail were the two telegraph operators on the first 1844 (Morse) Baltimore-Washington telegraph line. Wikipedia says the two of them co-invented the dot and dash code, but there is a dispute about the code. While some scholars give credit to Vail, most scholars say Morse developed the code over a period of several months referencing a contemperaneous Vail letter in which he wrote that Morse had come up with a letter code.

        Below is a great photo of Morse, which is perhaps not too surprising, since he had met Daguerre in Paris and later taught daguerreotype photography in NYC. Mathew Brady, the famous civil war photographer, learned the daguerreotype process from Morse. This photo is not dated, but he looks about the same as an 1845 Library of Congress daguerreotype, which would make him age 54. (Notice how the face is in sharp focus yet most of the hair is blurred. Either the lens had an amazing narrow depth of field, or else the blurring has been done in the printing process.)

Samuel F. B. Morse, -- artist, photographer & 'inventor' of the telegraph

Morse and Bell
        Although Morse (born 1791) preceded Bell (born 1847) by more than fifty years their engineering bio's are somewhat similar.

        Bell was professor at BU in Boston who specialty was speech and teaching of the deaf. He became interested in the telegraph and taught himself about electricity and the telegraph by doing lots of experiments. Bell's work on the harmonic telegraph led to his design of an electromagnetic transmitter and receiver that he and partners put into production in 1876. The telephone business then took off like gang busters.

        Morse was a painter, ran a daguerreotype portrait studio, and was a professor of painting and sculpture at the University of the City of New York. Morse was introduced to the wonders of electromagnets in 1832 on cruise to Europe by a Dr. Charles Jackson of Boston, and on that cruise the possibility of a telegraph excited his entrepreneurial sprite. Morse cobbled together some equipment to do telegraphic demonstrations and worked for twelve years to build interest and get the funding to build a demonstration line. But Morse unlike Bell was not much of an electrical experimenter, witnesses say he knew relatively little about electricity. He consulted with a science professor at his school, and he developed the first telegraph with an engineering associate (Alfred Vail).

        Morse's personal contribution appears to have been engineering the whole system (historically many different approaches to the telegraph had been tried), a key component of which was his coming up with a letter code that allowed the the use of an electromagnetic telegraphic receiver. He ran tests on short lines and gave demonstrations including to the president, and in 1843 US Congress gave him 30k for construction of a Washington-Baltimore test line (1844). The telegraph business then took off like gang busters with twelve thousand miles of telegraph lines installed in the US in the next six years (1850).

Morse's first telegraph
        Morse's idea for the telegraph (formulated in the 1830's) is that the receiver have a steel tip that could make an indentation on paper pulled smoothly under it by a clock mechanism. A long current pulse (dash) caused a long trace and a short pulse (dot) a short one. And this is how early telegraphs were built and operated.

        Only later (when ?) was it discovered that the tape and clockwork mechanism could (often) be dispensed with because some operators got so good that they were able to 'read' the incoming message just by listening to the relay clicking without looking at the tape. (Very likely this transition was accompanied by a change in the sounder design to emphasize the sound.)

Sensing weak line currents
        On a long line it was not necessary for the line current to drive the sounder directly. To get a nice strong click sound takes some power. The trick was to use a local loop at the receiving station. The line switched a sensitive, high impedance (lots of turns) relay that in turn controlled the sounder with power coming from a local battery. A telegraph local loop was (sort of) an early 'amplifier'.

        Prescott credits Joseph Henry (of inductance fame and first president of the Smithsonian) with first figuring out this amplifier concept in 1836, when he used it to ring bells at a distance. Henry also developed the electromagnetic relay in 1835, so clearly Henry was a key inventor in laying the ground work for the telegraph.

Telegraph batteries
        Another requirement for a practical telegraph was a practical battery. In the early telegraph days you basically had to build your own battery from a few known recipes, and these batteries had lots of problems. But as the telegraph developed so did the battery with new chemistry and engineering. Very likely the huge demand for batteries for the telegraph greatly spurred battery development, in essence the telegraph and the battery co-developed. Prescott says a 'good' (a relative term) battery for telegraphic use was not developed until 1836-1837 (by Daniell and Grove), which is just a little before commercial telegraph service began. (Prescott p410)

        The very first battery, invented by Volta in 1800, used a single (acid) electrolyte and suffered from self discharge and degraded with use. The degradation, known as polarization (an odd name), was caused by tiny hydrogen bubbles that formed on the cathode and slowly blocked it. The hydrogen came from H+ ions of the acid electrolyte being reduced at the cathode [2H+ +  2e- = H2 (g)]. Even though the Volta 'pile' used zinc and copper only the zinc was chemically active (zinc dissolved into the acid as it oxidized), so the cell voltage was low at 0.75V.

        Daniell's 1836 battery was also a zinc-copper battery, but it was a huge improvement over Volta's 1800 cell. Daniell was the first to put a separator in the battery (initially a partially conducting clay pot and later linen). Nearly all modern batteries, with exception of the car lead-acid battery, have two electrolytes kept from mixing by a separator that allows some ion flow. When the anode and cathode metals are able to react with different electrolytes, a much wider range of chemical reactions can be used in a battery. In Daniell's cell the zinc anode dissolves into zinc sulfate (salt) electrolyte, and the copper ions in the copper sulfate (salt) electrolyte plate out onto the copper cathode. With no H+ ions in the electrolytes Daniell solved both the degradation and self discharge problems. With a chemical reaction at both electrodes his cell voltage was higher [0.75V (zinc oxidize at anode) + 0.35V (copper reduce at cathode) = 1.1V].

        Thirty years later (in 1860's) Callaud of France comes up with an improved (wet) Daniell cell called the gravity cell. It worked better than the Daniell's original battery because the separator was gone, density differences kept the electrolytes separate. No separator means lower ESR and current higher. Perhaps more importantly the battery was so simple, two liquids floated on top of one another covering metal electrodes (see figure below), that it could be built on site. Its state of charge could be checked visually and new zinc could be added as needed. It quickly became the standard telegraph battery and remained in service for nearly the next 100 years.

Gravity cell telegraph battery with zinc-copper Daniell chemistry.
(top) zinc anode dissolves into zinc sulfate (0.75V)
(bot) copper ions in copper sulfate plate onto the copper cathode (0.35V)
 (figure from wikipedia - Daniell cell)

Relay types
Neutral relay
        When current flows in the coil,  it induces a magnetic field to flow in the iron bar that runs down the center of the coil. A nearly continuous iron path outside the coil then carries this loop of magnetic flux around the coil into the movable sounder where the flux is forced to jump across a small air gap to get back to bar. Counterintuitively most of the magnetic energy in a system like this is stored not in the iron, but in the small air gap at the sounder. Any system left to its own devices (technically a closed system) tries to arange inself in a lower energy state. The result is that when current starts to flow in the relay coil it pulls the movable sounder inward (toward the coil) narrowing the air gap. When the sounder hits a metal stop, it produces the classic telegraph click sound.

        From a duplex point of view the important thing about normal (neutral) relays is that they operate the same with positive or negative current. True the direction of the current controls the direction of the magnetic flux, i.e. whether is circulates clockwise or counterclockwise around the coil, but the forces in a normal relay come from changes in magnetic energy and the magnetic energy depends on the flux squared. The formula for energy in the gap is (Energy = 1/2 u0 H^2 x gap volume), where H is proportional to the current.

Polarized Relay
       However, there is a type of relay that is sensitive to the direction (sign) of current. This is known as the polarized relay, and it proved useful in constructing duplex and quadruplex telegraphs. One reference credits Henry with its invention, and a well known polarized relay was manufactured quite early by Siemens. It had another useful properties besides being sensitive to the direction of current. It could be wound to be very sensitive (10-20 ma), making it a good receiver for long line where the current was low, and it could be constructed so that it would hold (remember) its last state when the current went to zero. This type of relay has a PM (permanent magnet) in it, which provides the latching, and two gaps, one of which is open and one closed. Electrically it functions as a current controlled, polarity sensitive, latching DPST (double pole, single throw) switch.

        This relay was used by Edison to detect the polarity of current in his amplitude/polarity duplex (& later quadruplex) scheme, and the fact that it was sensitive was an advantage because it allowed for a large ratio between small and large pulses.

How a polarized relay works
        A polarized relay is shown below in a realistic geometry. It's a twist on a standard relay which would normally be two coils that when energized pull down a horizontal lever against a spring to close both gaps. But in a polarized relay a PM is positioned between the coils, and the lever is pivoted in the center at the top of the PM giving the lever has two stable positions: tilt right or tilt left. The force of the PM holds the lever at its last position when the current in the coils is turned off, so it's a latching relay.

Polarized relay operation
(from Stan Schreier of Antique Telephone Collectors Association
describing the operation of a pay phone coin polarized relay)

        You can think of the magnetic fluxes from the coils and PM this way. The PM flux rises up in the center (think of it squirting out of S) and splits to return on the two outside paths, paths which it sees in parallel. The flux in a PM is (pretty much) hard fixed, so the flux created by current in the coils (mostly) circulates around the outside path, going either clockwise or counterclockwise depending on the polarity of the current.

        The left figure shows the lever perfectly horizontal with no current in the coils. Here the PM flux divides equally between right and left paths, so the flux is the same in both gaps. With the energy in both gaps the same the force pulling downward on the lever is the same at both ends, so there is no net force to rotate the lever. It's in an (unstable) null position. Now put some current in the coils. If we think of flux coming out of S and into N, then in the right figure the net flux in the left gap is increased, because the coils and PM fluxes add, and it is decreased in the right gap, because the coils and PM fluxes subtract. The gap with the higher flux will pull down harder on its end of the lever (because force depends on the change in magnetic energy for an incremental change in position). Hence if the lever is free to move, it will tilt to close the gap with the higher net flux.

Simpler view
        A simpler view of a polarized relay is this: An electromagnet has NS poles just like a PM magnet. If a PM is brought near an electromagnet it will be attracted or repelled depending on which pole is presented (opposite or same).  The polarized relay is just a magnetically combined electromagnet/PM structure. There are two electromagnet coils wound such that the poles at the top always have opposite polarity. Both top electromagnetic poles 'look' at (are close to) a south PM pole that can move (a little) because it is pivoted. As shown in the figure (above right), the electromagnets tilt the pivoted PM south pole lever by attracting it with one coil and repelling it with the other. Of course, when the coils current reverses, the poles of the electromagnets flip, so the PM lever flips too.
        Bottom line: In a polarized relay the polarity of the current by controls whether the lever (which of course is coupled to contacts) tilts right or left. And because much (half?) of the force to move the lever can come from the PM, the relay can be designed to operate on a relatively weak current.

Submarine cables
        In all long telegraph lines the current amplitude at the receiving station is low because of the high resistance of the wire and current losses in the insulation system. However, with land lines the shape of the current pulse, i.e. its on/off time (current rise and fall time) is pretty close to what is sent. This is not true for submarine cables, in long submarine cables the shape of the current pulse is all stretched out. Since one pulse has to (pretty much) die out before the next pulse can be sent, the transmission speeds on submarine cables are quite low.

High capacitance problem
        The reason for the pulse 'stretch out' is that submarine cables have much higher capacitance per unit length than land lines.  When voltage is applied to the cable (meaning a pulse is being sent), an electric field (usually designated E field) radiates out radially from the cable wire into the surroundings. While it's unlikely that most of the electricians of the day thought in these terms (Kelvin may be the exception), in fact this picture of the E field had been developed my Michael Faraday in the 1830's.

Submarine mystery
       Why submarine lines were so slow, and acted so weirdly (distorting the pulse) compared to land lines, was initially a mystery. It was the one of the world's greatest physicists/engineers, Lord Kelvin, who figured it out. The problem turned out to be high capacitance combined with high resistance.

Atomic picture of capacitance
        When a voltage is applied to a cable wire, here's a picture of what happens at the atomic level to electrons outside the wire (technically in the capacitor dielectric). Voltage has units of volt/meter, so a voltage on the cable wire means that an electric field (E field) is projecting radially out of the cable all around. The electric field exerts a force (F = qE) on the nearby electrons (outside the wire), which acting like little springs move a little. The equation for mechanical energy is {Energy = Force x Distance}, so when the voltage on a little section of cable wire increases it radiates out an increasing E field that does work (puts in energy) on the electrons surrounding the wire.
        The only place for this energy to come from is the source driving the cable. The result is rise time of the voltage is slowed as the energy is pulled from the cable to move all the adjacent electrons a little. Most of this energy (1/2 C V^2 = 1/2 u E^2 x volume) that is stored in the capacitance, which at the atomic level is stored in the displaced spring-like electrons, is recovered when the voltage goes to zero killing the E field. In circuit terms (ideal) capacitors store energy, but do not dissipate energy like resistors do.
        Cables under water have much higher capacitance per mile than land lines. The electric field in a submarine cable is projecting into sea water not air (see caveat below), and water has a huge dielectric constant (capacitance is proportional to the dielectric constant). The dielectric constant of water is x80 higher than air! The reason it's so high is that water is a polar molecule, its charge is spacially unbalanced giving the E field has something to 'grab onto'. Of course, a key problem was what to do about the high capacitance.
Caveat --- Above argument is valid for a simple insulated wire in water, but a submarine cable has an outer steel conducting protective sheath, and it may very well be that the E field does not penetrate (to any significant extent) the steel sheath. It depends on how 'grounded' the steel sheath appears to the wire.

        For the E field to project into the water then voltage applied to the cable wire has to also increase voltage of the steel sheath above earth, but there's a good chance that the high conductivity of sea water prevents this. In this case all the E field would arise between the wire and the outer steel sheath across the insulation. This would still increase the capacitance of the cable (above an air line of equivalent length) because the dielectric constant of most insulating materials is x3 to x5, but not nearly as much as if the E field penetrates into the water.

Resistance is a problem too
        How do you send code when the received current pules of a long submarine cable are so weak and spread out that they overlap each other? On the first Atlantic cable (1858) the chief electrician, who according to a biographer of Kelvin was a hack, attacked the problem by cranking up the sending voltage with coils and promptly shorted out the insulation in the cable, killing it.

        Again it was Kelvin who figured out an answer. Brought in as the chief electrician on the second Atlantic cable (1866), he invented the supremely sensitive mirror galvanometer to use as the receiver (it was used to test the cable too). This allowed the high resistance of the cable to be lived with because on the other side of the Atlantic only a tiny current (ua) was needed to deflect the mirror over a small angle. Its response speed, which is somewhat slow, could be adjusted and made compatible with the cable response time too. I've never seen specs on a 19th century mirror galvanometer, but from what I know of the physics and modern instruments it was probably at least x10 times, and probably more like x100 times, more sensitive than the best electromechanical relay.

 Using a mirror galvanometer
        I remember using a mirror galvanometer in engineering school. Even into the 1960's mirror galvanometer's survived, being used as the detectors in sensitive, high precision bridges. A mirror galvanometer is a tortional balance. It consists of small mirror connected to a fine wire coil, driven by the current to be sensed, free to rotate in a strong magnetic field. Tiny deflections of the mirror/coil are sensed without disturbing it by shining light on the mirror and reflecting the light spot onto a distant calibrated scale.
Modeling a high capacitance cable
        The capacitance of a long cable is not really like a normal capacitor. The reason is that the capacitance is spread out all along the length of the cable and basically due to the speed of light it takes some time (about 2 nsec/foot) for the edges of the on/off pulses to propagate down the cable. In technical terms the capacitance of a cable is not a 'lumped' capacitor, it's a distributed capacitor.

        This distributed nature of the submarine cable capacitor was a real problem when attempts were made to duplex the cable. Duplexing (both differential and bridge) only works if a dummy line similar to the real line can be constructed. Sterns had solved dummy line problem and made duplexing work robustly in 1872 for land lines by just adding a single capacitor across a resistor. But it was found that this didn't work well for submarine cables. The capacitance in land lines is distributed too, but the capacitive correction for land lines is a 2nd order term and fast, so a single (lumped) capacitor while a little off, does the job.

Kelvin has a model
       A long cable in technical terms is basically a transmission line. Transmission line theory is a relatively advanced (undergraduate) concept in electrical engineering. (It's a little tricky visualizing the how the pulses run up and down the cable and reflected off the ends.) It turns out that the distributed nature of a transmission line can be modeled by (an infinite) string of (lumped) inductors (L) and/or resistors (R) and capacitors (C) in what is called an LC or RC ladder. I found it interesting that the RC ladder model for a submarine cable is discussed in Presott's book (1888 edition). On page 905 he says,

        "According to the opinion of Sir William Thomson (Lord Kelvin), it would require, in fact, an infinite number of resistance and of condensers to obtain a perfect assimilation of the artificial line with the actual cable."  (Right on, Kelvin nailed it.)
        Below from Prescott's book (p904) is the bridge duplex circuit for a submarine cable. It's similar to that for a land line, except the lower bot corner (dummy cable impedance) has a short RC ladder. Units: the modern convention is that when no units are marked on the schematic the value of a capacitor is to be interpreted as microfarads (uf). I have no idea what the convention was 120 years ago, but uf could be right. 30 uf (or so) against the 1k ohm's make the natural response time of the cable (sqrt{# of RC's} x RC approx) about 1/20 of a second. This is basically consistent with 8 wpm, the initial rate on the Atlantic cable (8 wpm x 5 letter/word x 2.5 pulses/letter = 100 pulses/min). If the cable used # 16 wire (20 ohm/mile) (I don't know what size wire was typical in a submarine cable), then each 1K section of the ladder represents 50 miles.

A bridge duplex circuits for submarine cables by De Sauty (after Kelvin) showing cable modeled by RC ladder (bot, rt)
Assume: cable wire R = 20 ohm/mile, then each 1K is 50 miles,
and propagation time per 50 mile is RC = 1K x 32 uf = 32 msec/50 miles.
Response time of whole cable is approx sqrt{3} x 32 msec = 55 msec (consistent with 8 wpm)

Edison's RC ladder
        I found the same RC ladder for an improved dummy line in one of Edison's duplex telegraph patents (207,724) filed Dec 1874. In the figure below 'b' is a series of resistors and 'a' a series of capacitors (summing to the line capacitance). In the text Edison makes it clear he understands the line is a distributed RC system and that modeling it this way improves the match of the dummy line to the real line.

        So, who was first with an RC ladder model of the line, Kelvin or Edison? Edison in 1874 is claiming it as an invention, but very likely Kelvin is first. Prescott credits Kelvin, and the second (& successful) Atlantic telegraph cable, for which Kelvin was chief electrician, was laid in 1866 eight years before this Edison patent was filed.

(RC ladder dummy line model from Edison Duplex Telegraph patent 207,724, filed Dec 1874)

Theoretical understanding
        Prescott (in 1860) speaks of two electricities (positive and negative), which when they meet neutralize themselves. It turns out that this is not correct for normal current in a wire (it's all electrons), but it is equivalent. But interestingly within semiconductors Prescott description is correct, there are hole (positive charges) and electrons (negative charges) that both flow and when they meet the annihilate each other.

Test instruments
        Galvanometer was a current meter. It consisted of a magnetized needle free to rotate, suspended by a fine thread, in the magnetic field of a large coil of fine wire. Essentially it was a very sensitive torque balance. The direction of deflection of the needle indicated the direction of the current and the amount of deflection the strength of the current.


Edison's telegraph patents
        Edison receive an astounding 186 telegraph and telephone patents! In the Edison section of Rudger's Univ site it says the quadruplex was Edison's most important telegraph invention. What appears to have happened is that Edison's idea for a duplex (& diplex) shown in patent 162,633 (filed 1873), which used polarity for one message and amplitude for the other, was quickly combined with existing with Stearns duplex circuits (with dummy lines and bridges) to make a quadruplex. It was adapted by Western Union almost immediately and the quadruplex came into widespread use by 1874.

        I've reviewed the names of Edison patents. Seven times in years 1873 to 1885 Edison filed telegraph 'duplex' patents. This produced a total of 12 duplex telegraph patents, six resultiing from the filing on just one day in 1874. However, the key duplex patent (162,633) came from his very first filing on 4/26/1873 (it's figure is shown above). This is the amplitide/polarity diplex idea that was combined with the known bridged duplex scheme to make the widely used 'Edison quadruplex' telegraph.

        In 1877 he files several patents on a six way (sextuplex) telegraph, but I'm not sure if his sextuplex went into use. I've been downloading and reading these patents. Most are just one figure and one page of text. The dates below are from the Rudger's site, but I'm not sure they always agree with the patent office filing dates.

filed     4/26/73          162,633  (amplitude/polarity diplex that evolved into Edison quadruplex)
              6/27/73         147,917
             8/19/74          178,221  (split inductor duplex) , 178,222, 178,223, 180,858,  207,723, 480,567
            12/14/74          207,724
             1/18/75           168,385
            11/11/78          217,782
              4/30/85          333,290

Edison patent site
        The Edison patent list maintained at Rudgers University provides a quick way to scan through Edison patents. This site has all Edison patents in .pdf format. Click on a patent name in the patent list and it opens the pdf patent at the first page, hit return and you're back to the patent list. The patents are organized by 'executed' date, which just slightly precedes filing date and is probably the signature date in the patent.

Patent 162,633 polarity & amplitude
        This (see above) is the first of Edison's dozen or so duplex patents. According to Prescott (1888) the Edison duplex scheme of patent 162,633 (filed April 1873) became the basis of the famous and successful Edison quadruplex telegraph. Patent 162,633 describes a duplex scheme that is unusual, as Prescott notes, because it is asymmetrical.

        Transmission one way is done by varying the polarity of the current, while the other way is done by varying the amplitude of the current. While the patent text says this scheme can also be used either to send two messages simultaneously in different directions (duplex) or in the same direction (diplex), only the duplex scheme is detailed and shown in the figure. But it appears that it is the diplex implementation of polarity & amplitude that evolves into the quadruplex.

        In 162,633 left varies the polarity by having both a positive and negative battery (of equal voltage) and switching between them with a 'make before break' switch. This is a binary system. If the left key is open, then the line voltage is (continuously) held at the negative battery potential. The current amplitude is set by the resistance of the line and a resistor to earth at the distant right station. The +/- current pulses are sent to the line (at left) via a pair of sounders that do not respond to this level of current.

        At right the current polarity is detected by running the current through a polarized relay, which is only sensitive to the polarity of the current. Right sends in an amazingly simple way. Right has a key that just shorts the line terminating resistor (to earth). (No battery needed at right!) This varies the amplitude of the +/- currents on the line, modulating it above and below the current threshold of the two left sounders (one for positive and one for negative), which are mechanically coupled together.

Edison duplex patent 147,917 (50 ma, 100 ma, 150 ma)
        Two months after 162,633 Edison filed on a totally different duplex scheme. This is an amplitude approach with three thresholds. Each station has three cascaded sounders with thresholds of 50 ma, 100 ma, and 150 me and a resistor to earth shorted by a key.

Edison differential duplex patent 178,221
        Here is Edison's version of the Differential Duplex telegraph (patent 178,221, filed Sept 1, 1874). Not sure what his contribution here is. From my reading this does not appear to be much different (or even as good, no capacitor across dummy line) as the the Stearns duplex that had been in widespread use for two years before this was filed.

        What is new in patents of this vintage is hard to figure out because, unlike modern patents, there is (generally) no reference in the patent to previous work or patents. The claims are also often vague, saying something like I claim the configuration as drawn.
        Edison claims the polarized relay connected to differential coil in local loop and also the dummy line. This is probably more sensitive than sketch above. Instead of directly using the flux in the coil (as in the simplified text book sketch above), he uses a sense voltage from the coil to trip a local polarized relay (no details), which then in a local loop with battery operates the sounder.

Figure from Edison's duplex patent 178,221, filed 1874 -- His version of the (split coil) differential duplex telegraph.
Polarized relay (top) changes state depending on polarity of di/dt inductive pules and drives local sounder (presumably) equipment including battery polarization is same at both end of line. Telegraph key switches line to battery or earth with 'make before break' action
(Details are more easily seen in a full size version of this figure included at the end of the essay.)

My days as a telegraph 'operator'
        In my teen years I learned to use a telegraph key and 'talked' over the 40 meter radio band with it for a couple of years as a novice ham radio operator and later (I think) a technician licence. (Maybe  KN1GN or K1GNx. Surprisingly there appears to be no old call sign list online). To get a general radio operator's license in those days you needed to demonstrate (to FCC) a fairly high level of proficiency to send and read morse code (13 words a minute) whereas the novice and technician licenses required only 5 word per minute.  I remember taking the FCC general exam three different times and coming close, but failing each time. (I never did pass.) The test was a bitch because they always includes some oddball punctuation  symbols that you never heard on "the air". The FCC test was given way up in the tiny tower of the Boston Custom House, which in recent years has been converted into a Boston hotel.

         My transmitter, which I built myself from plans, had only one tube. It was only 7 watts (I think). And for a receiver I pulled apart and modified our old family floor-standing RCA radio (which had shortwave bands). I think all that I did was to pull plates from a variable tuning capacitor to make turning the dial less sensitive. For an antenna I initially ran copper wire from my bedroom to a trees on the far side of the backyard, and later I put a 40 meter dipole in the eves of our unfinished attic. I had to bend the dipole at the corners of the attic because it was longer than our house. The wire was fished out a hole near the gutter and ran down the side of the house and into my bedroom window.

        This equipment was virtually unusable during the day, because the receiver selectivity was terrible, but if I got up in the middle of the night, and the ionic conditions were good, I was able to communicate about 1,500 miles with this primitive equipment. I remember contacting Florida a couple of times and lots of places east of the Mississippi. I think one of the Florida contacts was on a night of one of the early (exciting) rocket launches from Cape Canaveral.

        Like most novice amateurs I had cards printed up with my call sign (KN1GN ??), and when a distant contact was made, we traded addresses and mailed cards to each other. I had a collections of call sign cards from my 'distant' contacts. (I've searched the FCC database for my call sign, but I'm not in it, apparently it doesn't go back to the 60's)
Telegraph books
        The best telegraph reference (concurrent and comprehensive) I found turned out to be a huge (1000 page, two volume) book published published in the 1888 by the chief electrician of Western Union, George B. Prescott. It turns out this was a later edition of Prescott's telegraph book, which went through many editions over 28 years. An 1860 version (460 pages) titled: "History, Theory and Practice of the Electric Telegraph is available on Google books'. This looks like it is probably the first edition, since Prescott, born 1830, is only age 30 when its published. In its preface Prescott says he has been working in telegraph for 13 years, meaning he started at age 17 (1847), so he starts in telegraph pretty early, the first telegraph line going up in the US in 1844 when he is age 14. A paperback version of the first edition is sold by Amazon for $30. The 7th edition, published in 1888 as two volumes, totaling over 1,000 pages, titled: 'Electricity and the Electric Telegraph' is also available on Google books. Prescott also wrote "Bell's Electric Speaking Telephone: Its Invention, Construction, Application, Modification, and History" 1884.

        These two editions of Prescott's book have been photocopied by Google books and can be downloaded in .pdf format free, and I took advantage of this. However, I found these downloaded books are not searchable, they are only page images without underlying text. On the one hand I read it is Google's intention to make its scanned books searchable, but from my reading it appears that Google is crippling all (free) download versions by making them unsearchable. So to find information in my downloaded Google books I had to use the table of contents and then start reading. On the Amazon's site these same books are searchable. Another frustration is that Prescott doesn't give patent references for the inventions he describes.

        Doing a Google search for "differential duplex and bridge duplex telegraph" up popped pages from a 1922 engineering book on the telegraph. At first I thought it was an 'inside' book look from Amazon but I noticed the entire 420 page book seemed to be available (free) as an 8M pdf file.

        Telegraph Engineering: A Manual for Practicing... by Erich Hausmann - 1922 - Telegraph 406 pages (scanned by Google Books) Google says this 1992 book is now public domain because its copyright has expired.   Link
        This book has a chapter on duplex telegraph and another chapter on quadruplex  telegraph. Reading 1870's patents on duplex and quadruplex (& higher) you find many newly hatched schemes. They are often hard to understand because they are just sketched, use notation and conventions that are 130 years old, and usually have no numeric values. This book written 40-50 years later gives perspective, because you can see what actually got used, and there is also basic data on telegraph lines, sounders, etc.
Google books
        The Prescott book is the first Google scan book I have come across. Google started in 2004 with plans to robotically scan (robot turns pages and camera takes photo @1,000 pages/hr) tens of millions of books. As of Mar 2007 according to NYT they have digitized 1M books. If the copyright has expired, they are available free for download in .pdf format. At least two editions of George Prescott's massive book on the telegraph have been digitized by Google.

        (update) I read in a 2006 blog a complaint that all Google book downloads are crippled by not being searchable, whereas other digitizing projects routinely have searchable downloads. Certainly the two Prescott Google books that I downloaded were not searchable.
Telegraph links
        A large telegraph online 'museum' site maintained by retired Montclair professor Tom Perera. He has collected 3,000 items including a huge number of pictures of historic telegraph keys. (A lot of his links don't work.)

        A good, modern detailed 50 page technical history of the telegraph has been written by Dr. James B. Calvert, Associate Professor Emeritus of Engineering, University of Denver, April 2000. It's posted on his home page along with technical essays on other topics. His home page is somewhat similar to mine, though his bent is more mathematical and academic.

Wheatstone bridge
        History of Wheatstone bridge and pictures of commercial bridges.

Telegraph sound
        Here is a recording of a telegraph sounder from Calvert @ 20 wpm (click 'What Hath God Wrought'). Note the clicks of a sounder are very different (& strange) compared to long/short tones used when Morse code is sent by radio. The keying is the same in both cases, but learning to 'read' the sound is very different.

Telegraph keys
        New telegraph keys for sale from 20 manufacturers, including these beautiful models below.

 Straight Key Century Club (SKCC) telegraph key, made by LTA in Spain, $110

GHD telegraph key, model GT501MIL, made by Toshihiko Ujiie in Japan, $449

LTA telegraph key, model AMCC Antique European Camelback,
made by Llaves Telegraphicas Artesanas in the Balearic Islands of Spain, $90

Restored AT&T 1930's telegraph key
        After working on this essay, I had to have a telegraph key, so I bought an old AT&T brass telegraph key (probably 1930's) on Ebay and restored it. This key has a fairly heavy brass base and nice action. I fully disassembled it and cleaned all the parts. I bought several types of metal/brass cleaner, but the type I found worked the best was a wadding material impregnated with a cleaning oil called Never Dull, George Basch Co, Freeport NY.

        After I took it apart, it sat for more than six months, so it took me a while to figure out how to put it back together again. When I took the photo below, I didn't have it quite right. In the photo the electrical terminals are shorted, because I hadn't yet figured out the silvery looking bar needed to be isolated from the brass base with the mica washers. Also the spring is mounted to the base incorrectly. It should be inserted into a tiny hole in the brass base, not be squeezed by the silvery terminal.

30s brass telegraph key that I restored
Telegraph key I restored (here not quite assembed correctly)
[To get the best depth of field I used aperature priority
stopping down all the way (F8)
Exposure was fine tweaked by adjusting the flash intensity]

cleaned up brass base

Disassembled (mica washers not shown)

Edison patents full size
        Below are the two Edison patent figures in the text above, but here at full size so details are more clearly seen.


Raw notes
       -- The bridge duplex, typically used with submarine cables, fed the line and earth through a pair of resistances, and a rheostat in series with the earth connection that could be adjusted to the line resistance. When the key at one end was closed, the potential between the ends of the pair of resistances was the same, so the receiving instrument was connected here. Therefore, the key did not affect the local receiver, but the current sent over the line operated the distant receiver. It was easy to accommodate the double-current method of working always used with cables with this system.

         -- Duplexing was very often used, since the capacity of a busy line was practically doubled by the addition of operators and a little apparatus at the terminals only, with no change in the line. By 1872, it was quite common on Western Union. In 1874, Edison showed how to double the capacity of the line again, by quadruplexing. The key to this was a way to send two independent messages in the same direction at the same time, a method called diplexing. Diplexing was never used by itself, but always in connection with quadruplexing. Edison's idea was to send one message by varying the strength of the signal (as in off and on) and the other by polarity, positive and negative. There were two sounders or relays, one responding to strength, the other to polarity. The diplex circuit now only had to be duplexed by one of the existing methods, and two messages could be sent in each direction at the same time, or four in all. Soon, most important lines were quadruplexed. By 1878, Western Union had 13,000 miles of quadruplex. There were four operators at each end, two sending and two receiving, all at the same time. Edison made many contributions to telegraphy, of which quadruplexing and the stock ticker are the most famous. The principle of impressing two different kinds of signals at one end of a circuit and untangling them at the other was used in later inventions, such as the Phonoplex where audio frequency and direct current signals were superimposed.

**  -- (Edison telling how Prescott's name came to be connected with the quadraplex)
        About this time I invented the quadruplex.I wanted to interest the Western Union Telegraph Company in it, with a view of selling it, but was unsuccessful until I made an arrangement with the chief electrician of the company (not named, but probably Prescott), so that he could be known as a joint inventor and receive a portion of the money. At that time I was very short of money, and needed it more than glory. This electrician appeared to want glory more than money, so it was an easy trade.

        -- Little had a fast automated system (pumched tape for transmission), but it didn't work well on loinger lines.

        -- Edison demonstrated transmission of 1,000 words per min (vs 40-50 wpm manually keying).

        -- To achieve such speeds required improving the pulse shape which due to self inductance was "sluggish")

        -- Stated very briefly, Edison's principal contribution to the commercial development of the
automatic was based on the observation that in a line of considerable length electrical impulses become enormously extended, or sluggish, due to a phenomenon known as self-induction, which with ordinary Morse work is in a measure corrected by condensers.

        -- Edison discovered that by utilizing a shunt around the receiving instrument, with a soft iron core, the
self-induction would produce a momentary and instantaneous reversal of the current at the end of each impulse, and thereby give an absolutely sharp definition to each signal. This discovery did away entirely with sluggishness, and made it possible to secure high speeds over lines of comparatively great lengths.
        -- At each end of a duplex wire, there were two operators, one at the key and the other at the sounder. The key at each end only operated the distant sounder, not the local one, which seemed strange to the operators, used to hearing their own sounder. There are two ways to accomplish this, called differential duplex and bridge duplex. Differential duplex, the most commonly used system, used a relay with double windings, as shown in the diagram on the left. Current from the key went through one winding to the line, and through the other to a dummy line of the same resistance and capacitance. The two currents caused opposing ampere-turns, and so did not operate the relay. The line current, arriving at the distant relay, passed through only one of the relay coils and operated it. It was always necessary to adjust the dummy line with a rheostat to achieve a balance, at both ends of the circuit. Note that a battery is required at each end, and that the polarities are opposite. You may find it interesting to determine the current in the line with both keys up, one or the other key up, and both keys down. The key must have both front and back contacts. The differential duplex works equally well with polarized currents. (Duplexing, diplexing, and quadruplexing, from 50 page telegraph reference)

Rudgers Univ Edison papers
        "The common approach to diplex was the use of weak and strong batteries to produce signals of different strengths, with relays at the receiving end designed to respond to one or the other signal. However, it proved difficult in practice to prevent the sensitive weak-signal relay from responding to the stronger signal current. Edison tried a different approach that used a common element in many of his designs -- the polarized relay. He continued to use one receiver with a common, or neutral, relay that responded only to changes in current strength but employed a second receiver with a polarized relay to respond to changes in the polarity of the current."

        "The use of current reversals, however, presented a new problem. At the moment of reversal, a momentary drop in current strength caused the common relay to lose its magnetism just as it was supposed to act, thus mutilating the signal and causing false breaks. Rather than trying to prevent the moment of no magnetism, Edison decided to isolate the effect electromechanically so that it would not interfere with the signal. He did this by means of what he called a "bug trap." Instead of preventing the neutral relay from releasing when the current fell to zero during the moment of reversal, he used it to work a local relay interposed between it and the sounder relay. This local relay was adjusted so that it responded sluggishly to the signal, and thus failed to act before the neutral relay regained its magnetism and acted on the sounder relay. In essence, Edison used a cascade of electromagnets to bridge over the time during which the reversed current regenerated the magnetic field in the main relay magnet.

        --- discuss terminology of Edison's patents (voltage is vague).