De-twitching an electronic tachometer

Unfortunately, I don't think that debounce circuit will work as drawn. There are several issues, actually. I suspect some/most 555 chips will output a continuous low signal, or nearly continuous low signal. I can elaborate if anyone is interested.

weeZee wrote: Going to put a quick note here. It's also possible to get a beautifully clean signal from the alternator output, but the voltage can go up to 70V on a kz. I would guess that an hefty Zener and resistor crowbar circuit should provide enough for a few mA without too much load/heat. This note is awaiting further wisdom rather than providing sound advice.

The raw alternator output signals are basically square-wave and will regularly drop pulses due to shunting so I'm not sure how beautiful the signals will be. I have screenshots of the raw alternator signals if anyone wants to see them. The rectified output is full of all sorts of noise.

s3.amazonaws.com/gpzweb/GPZAltntrWaves/AlternatorWaveforms.html

There is no "Edit" option on my post above (778376), so I'll add here:

For a situation where the rest duration of the Dyna S signal is too short etc. I would build a discrete SR latch (aka flip-flop) circuit instead of trying to make an artificial timed-pulse signal. That way you can get a nice wide 50% duty cycle across all RPMs while still getting a 3msec on and 3msec off pulse at 10,000 RPMs...much like you would get with points. I can elaborate on that as well, if needed.

But since I would suspect, as Doctorot stated, it is more likely due to noise, I would probably try a simple filter first.

Thanks for the reply and your explanations weeZee and loudhvx.
I will add an input low pass filter to the circuit board .
Do testing the circuit board on Saturday and give you a feedback.

The alternator output is a smooth sinusoidal 3-phase output, it's sometimes used in tach applications in diesels.
The issue with a 50% duty cycle is that it does not work with the full-range of the Emgo aftermarket tach, it's explained in the video.

weeZee wrote: The alternator output is a smooth sinusoidal 3-phase output, it's sometimes used in tach applications in diesels.
The issue with a 50% duty cycle is that it does not work with the full-range of the Emgo aftermarket tach, it's explained in the video.

One need only look at the signal to see it is not a smooth sinusoid when in actual normal operation. The link, above,shows what it looks like and why. It's a constantly changing square wave. The battery and regulator clip and shunt the altenator output, so it becomes closer to a square wave.

This is the actual signal from a Kz stator.



In order to be a smooth sinusoid, it would have to be open-load, or going into a passive/linear, unregulated, unchanging load. Obviously it does not.

Also, any tach that can't take a 50% duty cycle tach signal would be a very poor design since that is a very common signal from points-based ignitions. I suspect it may be some other issue preventing the tach from functioning.

The signal off an unloaded alternator is a sinewave, with a loaded shunt R/R it is closer to a partial sinewave depending on the shunt/switching design.
The point I made is that the signal is pretty clean (as your scope trace shows), you don't get switching ringing from the transistors like you do in an ignition circuit and as a result you don't get false triggering. I believe I've mentioned earlier that this tach was designed for the original kettering ignition from a kz900 and is sensitive to the ringing from an aftermarket transistorised ignition system.
The reason a latch flip-flop is not used is because there is a minimum duration signal from a monostable that prevents false re-triggering from spikes. You can build the circuit I've detailed and test it for yourself. It works.

weeZee wrote: The signal off an unloaded alternator is a sinewave, with a loaded shunt R/R it is closer to a partial sinewave depending on the shunt/switching design. The point I made is that the signal is pretty clean (as your scope trace shows), you don't get switching ringing from the transistors like you do in an ignition circuit and as a result you don't get false triggering.

I think it's a stretch to call that signal a partial sinewave, but ok. Unfortunately, it is still not very useful since the regulator will be periodically shorting that signal to ground.

weeZee wrote: I believe I've mentioned earlier that this tach was designed for the original kettering ignition from a kz900 and is sensitive to the ringing from an aftermarket transistorized ignition system.

If it's designed for a Kz900 then it should be able to accept a 50% duty cycle signal, as that is approximately what a Kz900 dwell consists of...probably more like 55%, if I recall.

The ringing is due to residual coil energy after the spark ceases. This happens with both transistors and points. A filter should be able to clean that up.


Just to clarify:
I'm using the term "false-triggers", here, to mean a signal from the pickup or points that comes at an unexpected, undesired time that causes a spark to occur. The signal it produces at the coil is similar/same as a correctly timed, desired signal. The tach cannot discern the two, and thus registers an incorrect RPM.

This is opposed to spikes or other noise on the coil that "triggers" the tach to incorrectly register a spark event.

weeZee wrote: The reason a latch flip-flop is not used is because there is a minimum duration signal from a monostable that prevents false re-triggering from spikes.

If implemented correctly, the debounce circuit can work to reject spikes. But it won't necessarily reject false-triggers as defined above.

A filter can handle the spikes and noise, but also will not likely reject all false-triggers.

An SR latch would be better at rejecting the type of false-trigger described above. But the exact implementation would require an analysis of the false-trigger's timing. It's not random. The false-triggers may come and go during different periods, seeming random, but the exact timing is usually not random. False-triggers are usually in sync with the crankshaft, as can be seen by a timing light. But as I mentioned, I was only addressing the possibility of the Dyna S rest period being too short, not necessarily false-triggers or noise, which would be better addressed other ways.

weeZee wrote: You can build the circuit I've detailed and test it for yourself. It works.

The drawing appears to have a mistake in component value(s) so I don't think the circuit you show to be the one you intended. And possibly there is a wiring error in the drawing as well.


I've built that circuit many times and can tell you it won't work with a Dyna S (which is the ignition in question). A bare debounce circuit is not meant for that type of situation.

A debounce input signal (trigger, pin 2) should float high most of the time and briefly pull low. The Dyna S signal is opposite of that. It pulls low most of the time and briefly floats high. This results in the output staying high until the trigger returns to high. To correct this, an inverter of some type is needed at the trigger input. It can be a simple transistor with resistor.


But there are other problems as well, which may just be errors in the drawing.

First, the pin 4 hard-reset is tied to the input so the circuit will be forced to reset whenever the coil signal is low, which is most of the time. This will hold the output low. Even if an inverter is used, as mentioned above, the circuit will be forced into reset, which produces a low signal right when the input signal should be triggering a high signal. When the input signal goes high to pull the 555 out of reset, the trigger (pin 2) will also be high, which means the 555 never gets triggered, so the output will likely stay low all the time. Typically, pin 4 is just tied to pin 8 to disable the hard reset function.

It's possible not all 555's react to the reset-tied-to-trigger the same way, as that does create a slightly ambiguous timing condition, but the output on the ones I use would just stay low in this condition.

Second, the drawing appears to show 22K for the resistor ("22.K"). I think you meant 220K as that's what you mentioned in the earlier post. 22k gives about a .2msec pulse, whereas a 220k would give the aforementioned 2msec pulse (assuming the other issues are resolved). Maybe the capacitors are swapped? I often see .01 for the pin5 filter, and using .1 for the timing cap would also correct the pulse duration discrepancy. So maybe the .1 and the .01 were inadvertently swapped?



Another problem related to the input signal is that a coil primary produces spikes in the hundreds of volts. Obviously a 555 timer won't like that. You can use a simple blocking diode, like a 1N4007, but then that will affect the operation of the input signal with respect to the inverter requirement. So a couple more parts would be needed.

Yes, the alternator circuit is periodically switched to ground with a shunt regulator. This is not an issue for a high impedance signal.

The resistor in the circuit is 220K, my zero looks like a dot in the drawing. The reset is triggered at a lower voltage (0.7V) than the trigger input so the chip will act as a non-retriggerable monostable multivibrator for the duration of the RC constant on a rising input signal. Saves a few parts. Caps across supply and on pin 5 are not too critical.

The input for aftermarket tach are usually the ignition primary, so it's probably best to use a fast schottky diode to take the edge off spikes. The application I built the circuit for takes a 12v signal from an aftermarket electronic ignition. This won't work with a dyna, but stock igniters have a 12v tach output. The other thing with Dyna's is that they have much higher ringing than from a standard Kettering ignition with condensers.
This issue is with the Emgo/Hi-Level units, it appears mine is fussy about pulse width. The circuit shown takes an input signal pulse and conditions it to a uniform width, but retains the frequency. TBH, I should just pry the unit open and check the circuit, but this digital filter circuit seems to work fine.

Now a kz kettering ignition has a 55% dwell, but the absolute width of the high signal is going to change with RPM, I found the aftermarket tach had a minimum signal width for responsive behaviour. Dyna and electronic ignitions have shorter dwell (10% & 30% ish), so my guess is that at about 5k-6k rpm, the unit started aliasing input signals, causing the needle to start oscillating.

The alternator signal dropping to ground means the tach will not be able to count revs reliably regardless of impedance because pulses will be missing randomly.

The 555 circuit shown simply does not work as drawn for the dyna s for the reasons explained. With a long dwell ignition like the dyna s, the 555's output just stays low.

Even with the reset wired conventionally, the output will not be as claimed due to the extended low input. The dwell is not 10% on the dyna s, it is over 80%. The dwell is over 300 deg, leaving less than 60 deg for the rest period. 300/360 is 83%. This means the primary signal is low most of the time, so it is not useful as a debounce input signal. It will force the output to look like the input, thus defeating the purpose of the debounce circuit.

I tested it. And I've used debounce circuits many times for other things, and have designed mods to make them work better. The standard textbook version is really only for learning. With afew extra parts some of the problems can be fixed on the textbook version.

1. All that a tach has to do is detect zero crossing for the alternator whether a shunt or switch mode (as in your scope signal) is used, perhaps with a few volts of + bias. I don't see why a ground signal is an issue for a high-Z input, or why there would be missing pulses if the engine were running. A resistor divider on the input might be a good idea, an op-amp or 70v signal transistor should work.

2. Works with an NE555N that I had handy, though it seems that I'd used a 0.1uF cap for the RC part and the 0.01uF as a supply cap.
I think I said monostable re-triggerable in the video, it's not a re-triggerable.
It would be interesting if you could post a 555 it doesn't work with. I think I also tested it with a LM555CN. Cheers.

Most tachs I've dealt with seem to work on various voltage threshold crossings, rather than simple zero crossings. So a converter might need to be made for detecting current-polarity or even simple zero-voltage crossings. But there is still the problem of shunting. On the page I linked to, I show how one phase wire, when referenced against ground, can have multiple zero-crossings when other phases get shunted. This will cause additional pulses to the tachometer.

Here there is little to no shunting on any phases, so this one phase has two zero-crossings per cycle as would be expected.


Here there is much more shunting. Notice there are spikes on this phase when another phases gets shunted. these spikes cross zero, albeit very briefly. So then you would be back to making essentially another noise filter.



Then of course there is still the conversion from 12 pulses per revolution to 4 or whatever the tachometer requires. It just seems easier to use the ignition for off-the-shelf automotive tachometers. I certainly don't mean to say it can't be done, I just don't think it would be easier than the currently available options.

I don't seem to be able replicate the low signal output I thought I had earlier. But I don't really know which chip I was using. It may also have been due to the 22.k / 220k confusion producing what looked like a low signal or it may have been the signal shown below that is nearly straight-line, but closer to the supply voltage.

However, with a Dyna S type signal I also don't get the same results as yours unless I tweak the input signal to a very idealized one, one that would not really exist on the bike. The problem in this case, still, is with using the reset function for input. The circuit is dependent on the internal construction of the of the chip, which may not be the same for all chips.

The Dyna S signal, like most/all igniter driver signals, does not go to zero volts. It gets down to the typical 1 volt or so and is current-dependent. This leads to some erratic results when driving the reset with the input signal, as seen below.

Here is the replication of your results using what I consider to be an overly-idealized input signal.
Input trace is on top. Output trace is on bottom. Both are on 5v scale. The voltage reference lines are there to show 0v reference for each signal.


Here is the output with a slower RPM.


Here I've added one diode to model the Dyna S signal a little more realistically.
Note the input signal never goes to zero, but is just slightly above zero volts. Note the output of the chip is very different than in the previous trial.


Here I've added one more diode to go slightly beyond the what the model for a Dyna S may look like, but could still exist depending on connections etc.
Note the output is now completely inverted into the textbook version of a debounce circuit.


So with just slight alterations to the input, the output goes through radical changes. This means the circuit could conceivably change modes during operation. These problems can likely be overcome, but again, at the cost of more parts etc.

And the other issue of pulse-width changes is still present. In either of the methods (positive or inverted), the Dyna S signal alters the pulse width at higher RPMs or all RPMs, respectively. This is evident in your video as well. I think you noted it. So in any case, there needs to be more parts to get the circuit to be an edge detector, and be immune to varying pulse widths at higher RPMs.

This also doesn't address the high-voltage issue which will require a few additional components as well.