The information given here is a combination of info obtained from Steve Coniff in the course of a number of conversations, Gail Rodda's "Model T Ford Parts Identification Guide", Volume 2, input by Ron Patterson, and my own observations, measurements, and conclusions.
My current 1925 T does not have a magneto due to a tramp nut left in the transmission by an earlier owner of the car. In the past I have been involved in making Tachometers for Model T Fords; this has necessitated my collecting and evaluating data on Model T magnetos so that I could design tachs for magneto powered ignitions and for battery powered Ford coil cars.
I continue to be amazed that the Ford people got magnetos and Ford coils going at all considering the primative diagnostic gear they had and marginal theory. They seemed to find that the small magnets and small pole pieces of the very early magnetos gave a wave form that was a series of positive and negative peaks with close to zero output in between - hence the reputed "notchy spark advance" of early Ts. There was no chance of a spark in that dead zone regardless of what one did with the spark advance lever!
As time went by Ford rapidly increased the magnet size and then played with magnet pole pieces, single and double stacked coils, and the coil mounting plates until about 1916. They also went from the double stacked coils, which favored voltage ouput (and possibly a hotter spark), to the single stacked coils with heavier copper windings, which had higher current output capability. This was done in the mid teens when they needed the extra current to power horns and headlights. The weakest magneto appears to have been the 1912, when they had gone to a smaller coil pole and had done nothing to beef up the magnetic coupling. The wider magnet pole pieces, used in 1913 and 1914, and the larger coil pole shape, used from 1915 on, helped to spread the waveform out from positive and negative spikes to the present waveform. This later magneto (about the same from then to 1927 with only variations in the coil mounting plate) had more energy in the waveform and could produce a better, hotter spark. These changes, derived in the main from Gail Rodda "Model T Ford Parts Identification Guide", Volume 2, Pages 10,11,12, 22, and 23, are laid out in the timelines shown in the following figure.
I have digitized the
a scope trace of a 1916 T magneto waveform with the engine in neutral and without any
electrical load on the magneto (the car has a distributor). The waveform has an overall
triangular shape, brought about by the wider coil poles.
Here are the digitized recordings of the waveform at three different RPMs. There are easily recognized irregularities in the waveform, particularly in the 500 RPM view. These are due to variations in the magnet strengths and positions on the flywheel.
Note that if this were an early (particularly 1912) magneto, the irregularities would transform into longer periods of near zero voltage.
The waveform at 1500 RPM displays in excess of a complete engine revolution (eight full cycles of the waveform), as can be seen in the next figure. An item of particular interest here is the visible variation in the peak to peak voltage. This is attributed to a fore and aft movement of the engine crankshaft causing a corresponding variation in the magnet pole piece to magneto coil spacing. Any effects due to magneto geometry variations would be repetative with each revolution of the crankshaft - the voltage magnetude changes seen in this figure extend well over one revolution.
The variation in magneto voltage at 1500 RPM indicates that a magnet to coil space change of about 16 percent was taking place while the engine was running (inversely proportional to the square of the magnet pole to coil pole spacing - normally about 0.025 to 0.040 inches). This was most likely due to forward and backward shifting of the crankshaft. It is interesting to note that the Ford Service Manual, in Section 992, puts an upper limit of 0.015 inches of crankshaft endplay; this would allow even larger variations in the magneto output voltage.
Unfortunately, this car had a distributor, so magneto-timer-coil waveforms and spark advance timing effects could not be explored.
The magneto coils from 1909 to part way through 1911 were on a stamped coil sheet with the poles about the size of a 25 cent coin; then the poles were smaller until the oval poles came into use in 1915. This would indicate that the magneto waveform was more of a spike during 1912; that is until the broader magnet pole piece was used in 1913 and 1914.
The output voltage of this 1916 magneto, data from A. L. Dyke,
"Dyke's Autombile Encyclopedia" 1925, Page 1082, data taken from two 1926 T's, and data from an electrical motor driven magneto test
rig are shown here. All data, other than Dyke's, were taken with a variety of analog multimeters. Some of the data were
obtained via telephone conversations. The 40 volt at 2000 RPM data point is questionable - it is probably a bit too high.
The two 1926 magnetos were in well tuned Model Ts that are frequently driven at high speeds. These two sets of data were obtained while the engines were running with timer-coil ignition loading the magneto.
The 1916 was in an older stock engine, without a magneto load, and it output a slightly lower voltage than is shown in the other data.
These data represent what one should expect from a properly setup and operating Model T Ford magneto. The Ford Service Manual, in Section 995, called for a magneto output of 7 volts at 400 RPM.
I have seen a 1914 that did not yield more than 10 V RMS at any engine speed, yet the car was fully capable of running on the magneto - this is probably a marginal case.
Ron Patterson has pointed out that Ford drawing T-701C was used to define the flywheels and magnet positions in all cars 1919 to 1927. This drawing sets the main magnet mounting bolt position as advanced by 7 degrees from centerline of the two dowels that set the crankshaft - flywheel alignment. The brass magnet pole and pole piece retaining bolts near the rim of the flywheel are offset 11.25 degrees to each side of the main bolts.
This seven degree offset from the dowel centerline has been confirmed by measurement of Gail Roddas flywheel pictures 4a, 5a, and 6a on Page 11 of his second book (after geometry corrects were made to correct for parallax due to the camera angle). These pictures show the 1913/1914, 1915/1916, and 1917/1918 flywheels.
Measurements of 1911, a 1913/1914, and a machined 1927 flywheel, done by Steve Coniff, produced similar results.
The two crank dowels are vertical when number 1 and 4 crank throws are at Top Dead Center (TDC). This is clear in the adjacent figure taken from the rear of a crankshaft; the hump is the number 4 crank throw in the background.
Magneto coils are spaced 22.5 degrees apart and are 11.25 degrees each side of 12 oclock, or TDC. A magnetic pole passing laterally across a coil, which is wound around an iron core pole, causes a voltage peak as it approaches the coil and a reverse polarity peak after it passes and then recedes. Hence, with the coils of a magneto wound in alternating directions, the unloaded voltage peak occurs with the flywheel magnet pole midway between coil poles.
This information would indicate that the output peak voltage waveform timing should be 11.25 + 7. = 18.25 degrees BTDC and 22.5 degree increments to each side of that value.
The values then available via the spark advance control should therefore be 40.75 BTDC, 18.25 BTDC, 4.25 ATDC, and 26.75 ATDC. A good coil appears to fire once near the peak of a continuous sine wave or the peak of a magneto triangular wave. It may fire at a different time if it is connected via the timer at some point earlier in the waveform.
Observations are that a properly tuned coil fires once per waveform peak, while a poorly tuned coil seems to fire more randomly and with less vigor. This would say that tuning using a test machine is probably more valid because it shows the uniform single sparking, which one cannot see if using the simple 6 VAC current draw test, unless one uses an oscilloscope. My observations also indicate that either DC or AC tuning results in about the same current draw and that testing at 6 volts or 12 volts DC yields similar current draws. On 12 VDC, the points closed on time draws more current but the off time tends to be longer - hence about the same average current draw. The DC point vibration frequency seems to be about 500 HZ.
I selected a coil for further testing, from a few old ones on hand, on the basis that it showed a more vigorous spark than the rest. This is an apparently unrestored KW coil that is probably very representative of many in routine service in Model Ts today. The capacitor in this coil was measured at 0.12 microfarads and tests good with no leakage. The current draw was measured at several different DC and AC supply voltages while the coil was continuosly firing as shown in the next figure. Note that the current draw assumptions stated above are well supported. The amperage tends to go down as the supply voltage increases - this is due to the fact that the coil points are open a larger percentage of the time as the coil cycles harder with increasing voltage.
A series of tests were performed with the above KW coil. These were done with the coil in a test fixture that was built up as a part of my prior tachometer development and test work. The coil was held in a 1926-1927 coil box that was fired by a New Day timer which was in turn driven by a variable speed motor. In this case the timer connections for all four coil positions were connected together so that the single coil in the box was fired at each cylinder position of the timer. The coil high voltage spark output was connected to a standard spark plug firing in air. The plug ground return was via a piece of 16 gauge wire that had about 120 turns of 26 gauge wire wrapped over about 1 1/4 inches of its length. This was a current transformer that could detect the firing current as the plug sparked. The signal rise time of a spark breakdown is among the fastest of all electronic signals and was well beyond the 60 MHZ response of the oscilloscope in use here. The signal was also quite small; so the monitoring signal cable could not be properly terminated with its characteristic impedance. As a consequence, the spark current firing signal tended to echo up and down the signal cable. A positive current signal could just as often show a die-away as though it were a negative signal. A sample output signal from the tests to be performed is shown in the next figure. The visible coil current signal is in reality only the die-away RC time constant tail of the real signal, but it serves as a good indicator that the spark plug did fire and at what time. A vital point to be seen in this figure is the fact that the spark plug only fires right at the inital leading edge of the coil points opening - there is no "burn time" as some assume.
This picture, as well as all of the similar data pictures used in the following discussion were taken using a 15 ampere 60 HZ supply transformer with a variable output. This transformer was selected so as to have a very low output impedance to be similar to that of a Model T Ford magneto. The pictures were taken with a tripod mounted digital camera aimed at the four trace scope screen. The camera was operated in a continuous mode at about 2 frames per second at a 1/15 second exposure. The motor driven timer was run at a speed so that the timer contact times precessed across the 60 HZ driving signal voltage waveform. The oscilloscope was synched to the 60 HZ line. Some of the resulting pictures were partially or totally blank when the oscilloscope was in the retrace mode. A total of about 1600 pictures were taken in the process of shaking down the method, and finally recording the meaningful data. The final data consited of about 500 pictures that formed a stable set in terms of scope display of waveforms, of which about 269 had useable indications of timer contact closure and the first resulting spark generation and were used in the data reduction. It should be noted that the data from only four pictures were eliminated from this final set as being of questionable nature - such as a shot at the 4.4 VRMS coil firing threshold, where the coil did not fire until two voltage peaks after the timer contact closure.
All data pictures showed a clean single coil firing at the 4.4 volt RMS threshold driving voltage. Many of the pictures taken at 6 volts RMS displayed a closely timed double firing of the coil. The pictures at 8 and 10 volts RMS showed multiple coil firings - please bear in mind that this test KW spark coil was a run of the mill coil that had not benefitted from a professional rebuild and tuneup - it is just the best of a small litter.
The final data pictures were readout using a digitized computer display in which the data points of interest could be read to an accuracy of about 2 pixels, where 93 pixels = 22.5 degrees. Other factors probably introduced another few pixels of possible variation. All data were recorded in a database, where the final data reduction and plotting was done. It was assumed that a positive AC driving waveform peak represented the 18.25 degree BTDC wave and a negative peak was that next one following TDC. All data reduction therefore assumed that the effective TDC was located following the positive waveform and before the ensuing negative waveform. The position of the timer contact closing, the resulting firing of the coil, and the related AC driving waveform peaks were all recorded in the database. The precessing of the timer contact closure across the AC driving waveform enabled a set of data to be recorded that represented all possible timer advances. These sets of data were recorded at the 4.4 volt coil firing threshold, and at 6, 8, and 10 volts RMS.
The reduced data are plotted in the following figure. Note the scatter in the 4.4 volt data; this is indicative of the threshold nature of this voltage. The higher voltage data all tend to follow a smooth curve. The 8 and 10 volt data points are very close to each other, and it was decided not to extend the data to even higher voltages, which in any case would have been beyond the output ability of the AC transformer in use. A coordinated data set of this type would not be possible without the use of a very stable driving power source. A Model T engine driven magneto could not be held at a sufficiently stable speed for the time required to shoot multiple sets of hundreds of pictures.
A clear effect to be seen in the figure is the abrupt advance in the first spark timing as the timer contact closure advances and causes the firing to move from the voltage peak after TDC to the next one before TDC. This change appears to move to a point of less timer advance as the driving voltage increases.
All of these data show a common characteristic. When the timer contact closure was near to or a bit forward of the peak of the driving waveform, the first spark occured on the backside of the voltage peak. As the contact closure advanced forward from the peak, the first firing also advanced up to the time of and then ahead of the peak voltage. Further advance of the timer contact closure beyond the time of the driving waveform zero voltage crossover caused the timing of the first spark to receed back down the backside of the voltage peak - this additional timer contact advance actually causes a slight retarding of the first spark. This is evidenced in the figure by the dropoff in the first spark advance with increased timer advance at the right end of each curve. The advance effect can be seen in the following sequence of photos taken at 6 VRMS.
Note that here the supply voltage is sufficient that, for this particular coil, a second firing begins to occur as the timer contact advances ahead of the voltage peak.
The above tests were repeated using a 1926/1927 type magneto, driven by a variable speed electric motor, that was set up as a test bench by Don Dechant. Both Tony Bowker and Don Dechant helped run this test. The timer and recording system were the same as used in the sine wave tests above. Runs were made at 500 and 1000 RPM on the magneto.
The electrical environment was noisier, the scope trigger point tended to vary a bit, and the magneto could not be held at exactly the same speed at all times. This combined caused the data reduction to be done in a slightly different manner. As a result, the data tend to have a bit more scatter = of the order of a few degrees as seen in the next figure.
It is significant to note that the coil firing point was always somewhere on the back side of the magneto waveform; it never fired at or before the peak voltage. This meant that the timing of the first spark never reached as much as 18 degrees before TDC. This is probably due to two factors; the magneto appeared to have a higher impedance than the transformer used in the sine wave tests, and the more or less linear ramp of the magneto waveform could not provide as fast a voltage rise as a comparable frequency sine wave.
Both sets of tests show a common characteristic in that each grouping of first sparks is about 22.5 degrees wide. The groupings in each set are also about 22.5 degrees offset from each other and are of similar shape. This indicates that each group is a complete set of data points for firing on a particular positive or negative cycle of the magneto waveform. If the spark advance lever should chance to be positioned so that it was before or after the extreams in the figures here, the first spark would occur at points that are shifted 22.5 dgrees both vertically and horizontally from the points shown in the figures, and they would be so far advanced or retarded that the engine probably would not fire properly.
Another important point is that the data from the tests using the magneto have a flatter distribution of timing of the first spark. That means that a Model T Ford driver is basically only selecting what timing group (advanced or retarded) the engine is firing on, and has little control, beyond a few dgrees of spark timing, within that group.
Any resistance in the magneto to spark coil circuit will have a detrimental effect and cause the timing of the first spark to lag. It is important to keep all wiring in the ignition circuit in top shape. Use good gauge wiring, be sure to have clean, solid connections, and be sure that the ignition switch contacts are in good condition. I have seen a number of old switches that made very poor contact. An ohm meter test is not adequate - one must check the voltage drop across the switch when it is under load. The peak currents can be several amps, and an ohm of resistance can result in volts being lost.