Model T Ford Coil Tester

In my quest to set my coils so they all fire at the same time (they run on DC in my Model T), I designed this instrument to observe coil firing points. I had been thinking of using the current measuring method, and when I discovered a Hall effect current converter IC mid 2012, (ACS712-TELC-20A) the idea of this instrument was born. While one could sample coil current with a low value resistor, I didn't like that idea as any resistance would cause the coil magnetic field to build up more slowly.
When current from a battery or magneto is applied to a Ford coil via the timer, there is a finite time (a few milliseconds), for the magnetic field in the iron core to build up before the points open and the coil fires. In the interest of perfect timing, it is obvious that each coil should take the same time for fire as the others. This way, advanced or retarded firing between cylinders is avoided.

The completed unit set up and operating.

The unit powers the coil at 6, 9, or 12V at a rate equivalent to 2000rpm, to simulate in car use at the highest likely speed.  The internal timebase that switches the coil is also available to drive the horizontal axis of the CRO in X-Y mode, so that there is no CRO triggering difficulty. In fact, there is no need for a timebase in the CRO; it merely functions as an X-Y display.
The ACS712-TELC-20A provides 100mV output for every 1A that flows through its internal current shunt. Thus, setting the CRO to 100mV/DIV shows coil current in amps per division.
The coil is provided with supply for 6ms out of every 60ms. 6ms was chosen because it shows one complete fire. 60ms between fires is equivalent to running the coil in an engine at 2000rpm. Note that the timer rotational speed is half the crankshaft speed. So, for 2000rpm at the crankshaft, the timer is rotating at 1000rpm, or 16.7 revs per second. The period per revolution is thus 60ms. Although there are four coils, each is fired only once per revolution.

What the instrument displays.
Apart from showing firing time vs. timer contact, observing coil current over time also shows multiple firing on maladjusted coils. So, this instrument can also be used to set coils for magneto powered cars.
Initial coil set up is done by means of the usual setting of current to 1.3A at 6V. The tester allows for continuous coil operation, "buzz box" style, and includes an ammeter for this. Then, the firing point is checked and adjusted, and if necessary multiple sparking removed. It can also be seen that average coil current drops when powered from 12V.
I included a 9V supply as I also have the facility to run the coils in my own car at this voltage. This is provided by a switchmode converter described here.

Many Model T owners simply set coil current after adjusting the points, either using a homemade or commercially available unit with nothing more than an ammeter and a box to hold the coil.  While a spark is produced, this "buzz box" method of testing does not show firing time, or multiple sparking. The reason why this is undesirable is that multiple sparking does not have a defined firing point, and as can be seen from the CRO waveforms, the actual peak current when the coil fires is less. Multiple sparking is more of a problem with magneto operation because  the coil fires on the rising voltage of a ramp waveform, as opposed to being suddenly presented with 6 or 12V. As the speed of the magneto varies, so does its output voltage, and thus the firing point. This is normal provided the firing points are always consistent. However, the degree of multiple sparking on a misadjusted coil varies with input voltage, so it can be seen how erratic firing can occur with a varying engine speed.

We can also see the difference in firing time between 6 and 12V operation, and why lower coil current on 6V gives improved high speed performance.

Dual trace Mode.
Here we see the timer vs. coil firing with the CRO in dual trace mode:

The upper trace shows when the coil is being applied with 6V DC (equivalent to when the timer makes contact), and the lower trace is the current waveform of the coil. Each horizontal division is 1ms. Each vertical division is 100mV on the CRO, which translates to 1A of coil current. So, this coil takes 3.5ms to fire, at a peak of 4A. The smaller secondary waveform is simply the coil current building up for the next fire. Why 3.5ms? This is because 3.5ms corresponds to greatest coil efficiency. Less than 3.5ms results in a lower peak current, and thus a weaker spark, while applying 6V for more than 3.5ms causes the iron core to saturate, wasting energy, as well as unnecessarily retarding the firing time.
From this, it could be assumed that other kinds of igntion coil, such as used with stationary engines could be set up for efficient operation on this tester.

X-Y mode:

A ramp generator driven off the timebase in the instrument drives the horizontal axis of the CRO. It is set so that each division in 1ms. Thus, we see here the full 6ms when current is applied. Using the instrument this way avoids any triggering problems. Ideal firing point is just before the waveform flattens off, as shown here. This is when the coil will produce the strongest spark for least power input.

6V operation of coils.
For 6V operation it can be advantageous to advance the firing by reducing the current, in cars where the initial timing has been set for an initial 15 degree retard. This seems to be common with left hand drive cars in the U.S., where an accessory timing gauge is often used to set the initial timing. RHD drive cars tend to have their ignition set to only about 2 or 3 degrees ATDC, so 6V operation of coils in these cars tends to work well by default, as there is sufficient advance available with normal coil settings, for good performance.

12V operation of coils.
When the coil supply is switched to 12V, we see this:

Note that the firing period is now 2.2ms. Peak current is increased to 4.5A. This clearly demonstrates why coils perform differently on 12V than on 6V. It means that for successful 6V coil operation at high engine speed, the timing must be advanced as much as possible. 9V operation causes the coil to fire in between where it does on 6 and 12V. Note that the top of the rising waveform is flattening off. The flat period is indicative of wasted energy which does nothing but heat the coil. It shows the current is set too high. If this particular coil was to be used off a 12V battery, current should be reduced to the top of the curve (about 2ms) before it flattens out.

Multiple sparking is shown.

Note that peak current is now only about 3.2A. The sparking energy is now being dissipated over at least 6ms instead of being concentrated within 3.5ms. Thus, we see multiple sparking robs the initial spark of energy it could have otherwise had. The coil under test was deliberately made to multiple spark by sliding a paper clip over the upper point cushion spring. The average current shown on the meter does not change when a coil is misadjusted this way, and proves that the "buzz box" method of coil adjustment is a fallacy.

Erratic Firing:
If evident, this shows up as a jitter in the waveform, with the firing time fluctuating. It is also sometimes seen that the coil does not actually produce a spark, even though a firing pulse has been applied to it.

The Circuit.

Timebase.
This is the heart of the unit. Initially, I did the easiest thing and tried various IC circuits. However, radiation from the coil caused erratic operation, even if placed several metres away. Bypassing etc., did nothing to improve the situation. I even tried a 2N2646 UJT which again was troubled. I went so far as to try a relay in astable mode, but the problem was anything with mechanical contacts has contact bounce, and this gave an erroneous display. What I needed was a low gain oscillator. Indeed, using a couple of audio type germanium transistors solved the problem. I used AC128's. Not only are they low gain (compared to modern silicon types), but bandwidth is limited. So, I figured the circuit should be less susceptible to triggering. Indeed it was, and there was rock steady oscillation regardless of what the coil was doing. The oscillator is a simple cross coupled multivibrator with different R-C combinations in the two base circuits. This provides the asymmetrical duty cycle required. For calibration, trimpots are provided.
The output of this drives a BD139 emitter follower, and then a 2N3055, which is used to switch the coil. Given the few hundred volts that may appear across the 2N3055 when the coil is switched off, a diode is used to prevent any negative voltage appearing, and a neon bulb prevents the Vce rating being exceeded. It clamps the voltage to about 70V. The base is only 150 ohms away from earth which enables the transistor to be used to its full Vce rating of 100V.
A LED shows the timing pulses, and a timebase output is provided for the CRO second channel, for the X axis, when used in dual trace mode. Output voltage is attenuated by the 270R so that the X axis CRO amplifier can always be left on 100mV/div. The vertical amplifier can be set to DC coupling to avoid showing a tilted timing waveform. The CRO is triggered from this pulse when in dual trace mode. An NE2 neon bulb prevents excessive voltage from damaging the circuit should it somehow be fed back in from an operating coil.

Current converter.
Based on a ACS712-TELC-20A, this converts 1A of current flowing through its shunt to 100mV output. Thus, by setting the vertical channel to 100mV/division, we can read off 1A of coil current per division. Note that there are several different versions of this IC rated for different maximum currents. However, only the 20A version has the direct 1A to 100mV ratio.
The IC is surface mount. While I was prepared to make a PCB for it, I discovered it was cheaper to buy the whole thing already on a small PCB on eBay. This was about a third of the price of buying just the IC locally! So that one side of the current shunt is always earthed, it is placed between emitter of the 2N3055 and earth. This is done to again minimise high voltages getting in the wrong places. It is true that the base current is added to the total current, but as it is DC, present for the whole time, and of insignificant amplitude, it is simply ignored.
There is a DC offset on the output which means the display may be off scale when set to 100mV/div. So, I AC coupled it with a .1uF. We don't need the DC component for this application. Protection from any high voltage nasties is afforded by the 2x 1N914's and limited by the 10K resistor. It is all too easy with a buzzing Ford coil nearby for the wrong thing to accidentally happen!

X-Y Mode.
While the instrument works well in dual trace mode, an improvement is available whereby the timebase that switches the coil also drives the CRO X or horizontal axis. The advantages of this are that a single trace CRO can be used, and there is no need to set up the timebase or triggering. Simply, the CRO becomes an X-Y display device. Even the cheapest CRO's allow for external horizontal input. The coil tester output is calibrated so that each horizontal division is 1mS when used this way.
The 6ms pulse is converted to a ramp waveform by means of the 10K and 10uF. A 2K trimpot allows for calibration of output level such that 100mV per division when fed into the X amplifier, corresponds with 1ms per division. In other words, the horizontal trace is set for 6 divisions, showing the 6ms period.
Because of the DC offset, a .1uF is in series with the output. Being a simple waveshaping circuit, linearity is not perfect, but quite sufficient for the purpose. Dual trace mode can always be used if greater accuracy for the waveform needs to be observed.

Accuracy.
Like any analog CRO display, the accuracy of this instrument depends not only on the calibration of the CRO, but also how the operator reads the waveform. Parallax error, and the accuracy of the X and Y shift controls have a bearing on this. Also, it must be remembered that both the tester and CRO use an RC timebase which is susceptible to drift. Added to the current reading, as previously mentioned, is the 2N3055 base current. This is visible as a slight offset in the Y axis amplitude during the firing period. As the firing period is displayed in less than half of the eight divisions of the CRO screen, the resolution is less than it could be if the timebase was expanded.  Because the output of the current converter IC is AC coupled, it means the Y axis 0V level changes slightly depending on the waveform. Thus, it is necessary to check this on each coil test, if an accurate current reading is required. Possible improvements would be to use a crystal locked, or mains frequency derived timebase, with it expandable over more of the screen. Also, DC coupling into the CRO from the current converter IC, with DC offset correction would improve current measuring accuracy. For X-Y mode, the linearity of the ramp generator would be improved with a constant current source to charge the 10uF capacitor. Nevertheless, the instrument in its present form is quite sufficient for the task.

Power Supply.
At the outset it was decided to run the unit from the mains given that 12V is required for some coil tests and Model T's normally have only 6V batteries.
A standard 15V transformer, rectifier, and filter circuit provides around 23V unregulated. A large capacitor of 10,000uF is used to help cope with the spiky high peak current drawn by the coil. Because a faulty coil may present an excessive load, a 1.8A circuit breaker is connected in series with the rectifier. It is on the AC side to provide protection should the rectifier short out. Further protection is via a 500mA fuse for the transformer primary.
A 7812 provides 12V for the timebase. This also feeds the input of a 7805 which powers the current converter circuit. Note the 100mA bulb in series with the input of the 7805.
This was included to limit the current should the ACS712-TELC-20A lock up. When this happens it draws about 250mA and gets hot - not good! The first time it happened I thought the IC was dead, but surprisingly was totally OK at next power up. I figured it would be wise to limit supply current. This "locking up" problem sometimes occurred with the breadboarded prototype but has never occurred with the finished built up model. I found that if the coil was run continuously the radiation from the coil could cause this to happen, and was no doubt due to less than optimum earthing and circuit layout.
The coil under test is powered from a regulated 6, 9, or 12V via a standard zener diode and emitter follower circuit. The zener voltages chosen allow for the base emitter voltage drop of the regulator transistors. For improved regulation, the series transistors are wired as a Darlington pair. This means less fluctuation in current drawn from the zener regulator. Some may wonder why I've used another 2N3055 to drive the 2N3055 series pass transistor, when the base current required is so small. I've learnt years ago that if this type of power supply suffers from an overload that destroys the series pass transistor, it often takes the driving transistor with it. With the driving transistor rated at 15A, that won't happen here.
Final filtering is done again by a 10,000uF capacitor to present a low impedance, high peak current capability, supply to the coil. A 100R 5W resistor is across the supply output to prevent any high voltage nasties getting in (the EHT output of a Ford coil would drop to virtually zero if presented with a 100R load). It also ensures a load outside of the 6ms period.

Construction.

Unit under test during construction. Note the simple aluminium chassis and tagboard construction.

The circuit was built on an aluminium chassis which slides into a wooden box. On this is mounted the meter and coil connections.  I use springy brass strips to make contact with the coil terminals, while on the opposite sides angle aluminium provides a backstop for the coil. To do the job quickly and simply, everything was assembled on tagboards and connected point to point. All the 2N3055's were mounted in chassis mount sockets. The front panel is attached to the chassis and contains the user controls.

Calibration.
Simply connect the timebase output to the vertical channel of a CRO, set to 100mV/div. vertical, and 1ms/div. horizontal, and the switch in 'dual trace' mode. The two timebase pots are set to provide a total time of 60ms with a 6ms duty cycle. It is quite possible there might not be enough adjustment given use of different transistors, or normal component spreads. If so, changing the values of the two base capacitors will fix this.
Next, switch both the tester and CRO to 'X-Y' mode and set the 2K pot for 6 divisions horizontally.

Using the tester.
The coil is set up in the usual way with points at .032" and upper cushion spring at .005". Then with the unit on 6V and continuous, the coil is set to draw an average of 1.3A by means of adjusting the vibrator tension. This provides an initial starting point to get the coil going.
Next, connect the CRO and select X-Y mode, with both channels set to 100mV/div. Set the mode switch to "2000rpm". A waveform like one of the above will appear.
It can be examined to see at what time the coil fires, and then the others compared. Slight adjustment in vibrator tension can adjust them so they all fire at the same time, which from observation should be 3.5mS at 6V. Note that when coils are set up for equal firing time, the average current read by the panel meter may now differ coil to coil. Here, we see one of the limitations of the "buzz box" tester.
If there is multiple sparking, the cushion spring tension needs to be adjusted. Note that the chances of multiple sparking increase with supply voltage, so this test should also be done on 12V. The lover vibrator contact and upper cushion spring can be interactive, so it may be necessary to readjust one after the other.
If the coils are to be optimised for 12V, then the current should be checked so that firing occurs before core saturation occurs. For 12V, 2ms has been found to be optimum, and this actually corresponds to 3.5ms at 6V, meaning the coils can be set up on either voltage regardless of the voltage they will be ultimately used on.
Obviously, a set of coils for a particular car need to be adjusted for equal firing time, and not necessarily equal current.

Limitations of non-electronic testers.
Originally, the best method of setting Model T coils was to use a hand cranked coil tester. The technology at the time was well before electronic methods were practical.
The hand cranked coil tester shows multiple sparking, evident by more than one spark close together on the sparking ring as the tester handle is cranked.
However, the simple ammeter only gives a broad idea of coil current, and as the electronic tester shows, optimimum coil current for one coil is not necessarily the same as another.
It is not possible to show what this is, and as the tester is not calibrated, it is not known what the actual firing times of the coils are. Thus, it is hard to create a matched set of coils with such a tester. The mechanical tester does not  contain a timer, so one could argue it provides a dissimilar operation to that of a car ignition system. The difference is the coil is presented with a slowly rising voltage, rather than being suddenly hit with a higher voltage as when a timer is used.
Popular amongst Model T owners are the simple "buzz box" testers - so called because they usually consist of a wooden box, into which the coil is placed, and powered up from 6V.
A continuous spark occurs, and the rms or average primary coil current is measured by an analog meter. Users are instructed to set to 1.3A at 6V. Unfortunately, as each individual spark cannot be observed, there is no way of testing for multiple sparking. As with mechanical testers, the meter reading cannot display firing times.
Nevertheless, this kind of tester will get a set of coils going in an emergency. Just don't expect optimum operation from them.

So, you'd like your own electronic tester?
For those who are technically inclined, the tester described here is easy to build with mostly common parts. Those not into vintage electronics may be unfamiliar with germanium transistors, but they are still available with a bit of searching. Similar types include AC188, 2N185, OC74, 2SB56, 2SB172, and so on, which should all work. Basically, they are a type that was used in the audio output stage of many portable transistor devices in the 1960's and 70's. If you already have a buzz box tester, the circuitry described here can be added on, saving some construction.
Unfortunately, I am not in a position to manufacture the testers or supply parts. However, an excellent alternative which is available for sale is the ECCT.

The ECCT.
Those who follow the MTFCA forum would have seen mention of the "Electronically Cranked Coil Tester". It was, coincidentally, developed at around the same time as my design, but completely indepedently. The principle of operation is the same, but that's where the similarity ends. Unlike my design, the ECCT is designed for use by Model T owners who are more comfortable with mechanical things, than electronic. Instead of a CRO, the display is via a set of LED's which show any coil setting errors. For the technical user, an optional computer interface is available which shows a more detailed graphical result. It shows all the coil operating parameters while being fired. Furthermore, there is a capacitor tester and an optional magneto tester. It is basically a comprehensive Model T ignition system analyser.
I've had opportunity to try one out, and to come to the conclusion that this is the "must have" Model T instrument of the decade is an understatement.
This is a truly revolutionary design, which for the first time allows any Model T owner to set their coils correctly. It is compact, elegantly designed, and can be run off a 12V SLA battery for portable use. See my review of the ECCT here.
For further details of the ECCT, please visit its official website here.