Pulse multiplier for tachometer - opinions wanted on method

[Paul Moir]

Hello,

One of my various hobbies is old outboard motors.  One problem with them is that analog tachometers for them are now very hard to find since their ignition systems are not compatible with any sort of modern marine tachometer.  There is a manufacturer of digital ones that work well.   
Years ago I came up with a circuit that conditions the signal seen by the points in the magneto and feeds that to a tachometer.  This works for the two cylinder models since I can get a pulse from each magneto and feed them to a marine tach setup for a 4 cylinder inboard (4 stroke car type) engine, which expects 2 pulses per revolution.

I stuck it up on the web and surprisingly get a few emails about it a year despite it being pretty hard to find.  Unfortunately it's function is limited to a particular brand of outboard and only their two cylinder magneto ignition models.  Also accessing the point circuit is difficult on some of the models.  What I'd like to do is a second effort that uses a capacitive pickup on a spark plug lead, takes that and multiplies it by 2 4 or 6 (modern outboard tachs use 6 pulses per revolution, which are derived from the CDI charging stator).

I tried doing it with a 4046 PLL setup as a multiplier but found it tricky to get it to lock well and quickly over the required frequency range, which is 5-1200Hz 5-100Hz (300-6000RPM).  I had squared up the input signal duty cycle by dividing by 2 which didn't help matters.

It would be trivial to do this with a MCU like an ATTiny, and I could add some helpful features.  However my target audience likely doesn't have a programmer, and I'm not interested in supplying programmed parts.    I doubt there's enough volume to get any 3rd parties interested.

I was thinking today about it and thought maybe a F->V coupled with a V->F might be the solution.  I wouldn't have tried it a couple years ago because I would be afraid of the inaccuracies, but I think my analog is getting a little better.

For the detector, I've rigged up a functional one using a FET.  I saw another design somewhere that used a sensitive gate SCR which seemed like a better idea.  For one thing I'm not too interested in having a FET that near an ignition circuit, but the other was that pulse stretching with it was trivial.

So what do you think? Is there something seriously wrong with the PLL multiplier that I can't get it to lock quickly over that range?  Am I an idiot for not doing it with an MCU, perhaps an easily programmed one?  Can I get better than 1% (50 rpm) accuracy with a F->V->V->F converter without twiddling or exotic components?  Got an awesome capacitive pickup front end?  Oh, and where do people publish schematics these days?

Thanks for any advice!

[Fank1]

Use the pulses to gate a fixed oscillator.
ie : 100 KHZ oscillator and .01 sec between the pulses will give you a reading of 1000

[Paul Moir]

Ah, I think that gets me the wrong way round. ie the higher the RPM the lower the number of output pulses.  Might be some way of getting it back around the right way with even pulse spacing (which I suspect the tach will need).

Let me think about it some more.

[Andreas]

Hello,

which 4046 did you use?
The 74HC(T)4046 is the only which can be used for a larger frequency range.
There is a additional phase comparator (PC2) which compares not only the phase
difference but also wether the frequency is higher or lower.
But I have not calculated if the internal VCO of the 4046 can be used for such a large range.

With best regards

Andreas

[SeanB]

I saw an old method that used a counter fed from a high frequency oscillator ( or at least high for the range concerned, but still within the range of the counter) which was held in reset by a RS flip flop, and when the input set the RS the counter would count up, and on TC would reset the RS thus resetting itself. The gate for the counter would also enable an AND gate to output clock pulses so the external circuit would operate. IIRC it was implemented in 3 CMOS IC's, a 4017 as a counter, and a NAND chip as oscillator and gating, with the third NAND being the RS and assorted inverters.

It gave in this application 5 pulses out per input pulse, but you could set it from 2 to 9 pulses just by using different outputs of the counter.

[rob77]

Hello,

which 4046 did you use?
The 74HC(T)4046 is the only which can be used for a larger frequency range.
There is a additional phase comparator (PC2) which compares not only the phase
difference but also wether the frequency is higher or lower.
But I have not calculated if the internal VCO of the 4046 can be used for such a large range.

With best regards

Andreas

such a range is not a problem even for the original ancient cmos 4046 - i was using it in a similar range( in terms of span) , the VCO was running  1kHz to 1MHz without any issues and the phase comparator PC2 did the job well in the loop so a bigger cap for the VCO and tuning the control voltage to middle of the range would do the trick for the 5Hz to 1200Hz.
however the  PLL lock might take some time to lock on those low frequencies (especially the 5Hz) during fast RPM changes - but that should be acceptable, because even the mechanical tachometers are not responding to quick RPM changes quickly - so it will be kind of an imitation of that mechanical tachometer behavior

[Paul Moir]

Andreas,
I was using the old CMOS version.  MC14046BCP with an '88 date code to be precise.  I'll have a look at the improved 74HC version.  If I remember right, range was the primary problem:  to keep the loop stable, I had to dampen it a lot, which hurt response time.  This is going back a year though.
I've been rereading some stuff on this and I ran across some of Bob Pease's application notes which suggest 3 decades of lock is possible.    Since I'm only going 5-100Hz (Sorry, error in original post! Don't know how I got that.) that's only 20.  But I'm not 1/5th a Bob Pease!
Using 74 series logic will mean I have to regulate down from 12v and maybe have to put on an output driver, but that's not such a bad thing.  At least I won't have to account for supply voltage as well.

SeanB - I have to think about what you wrote some more, but I really like the concept.  If I understand right this will give N*x output, but the pulses will not be evenly spaced since they come from the high frequency oscillator.  I have to make sure the tach will work OK like that.  I suspect it will as long as it's not filtering out signals above it's range.  Time for some testing.  I don't mind saying "I've tested it with this brand tach and it works fine, YMMV with other tachs"

[Paul Moir]

rob77 - that might be a good point.  It seemed to me I was laggy on the analog tach, but I was feeding from a signal generator rather than an actual motor.  I was thinking of revving in neutral, but that's a bad idea on an outboard and of course the throttle response isn't exactly snappy in gear.

Note I somehow calculated 6000RPM at 1200Hz rather than 100Hz, so the correct value is well within the relative bandwidth you were operating.

I'll try getting a schematic up of where I was.

[Rerouter]

a module to investigate would be a general automotive ratio box, there are many designs around, most with up to 3.5 times multiplication of the signal. e.g. vdo/continental sell one as part number : 411.110, as a last resort option (this model would happily take whatever funky signal you have)

the next best option is look into the tachometer, near every model i have encountered in aus has had some form of 1 pulse per rev option, most of the digital ones involve holding down a button or a pin at a logic level while turning the ignition on to enter the calibration menu, (The latest vdo ocean line tach-hourmeters  have 2 buttons and a rather convoluted setup sheet, equally they can only be set as low as 1 through the button setup, not the programmer)

as for your range of frequency i know the vdo ratio box works that low but not the circuit it uses, at 16.66Hz at 1000RPM, i have only had to make up a automotive gauge that low before, it was an analog veiwline tachometer but you had to dramatically reduce the capacitance in the circuit to get the time constant low enough (think we had to swap out the 110n to a 55pF cap, ) if your gauges are mounted in a cabin one of those might do the trick, as they are quite cheap, but not water proof in any way and are based on the CS8190-D chip in-case you want to investigate the idea,


I do work for a vdo reseller in aus,

[motocoder]

You could implement it as follows:

Use a binary adder / accumulator. For each pulse you add a constant value to the accumulator. The exact amount depends on the number of pulses per rotation (1, 2, 4, or 6) and how frequently you will sample values to update your RPM display. You could set this with a dip switch or a few jumpers.

So let's say you sample the RPM every 1 seconds, the value to add to the accumulator for 1, 2, 4, and 6 pulse engine cycles would be:

(60 sec/min / 1 sec/sample / N pulses/rotation)
1 pulse/rotation -> 60
2 pulse/rotation -> 30
4 pulse/rotation -> 15
6 pulse/rotation -> 10

Then you have an oscillator that drives a circuit (counter) to generate RPM sampling pulses at some fixed interval. On every pulse, you grab the value in the accumulator and use it to update a register that drives your RPM display. On the falling edge of the sampling pulse, you clear the accumulator to zero.

Technically you need to synchronize the incoming pulses to the clock that is driving the oscillator, or you could see strange metastability issues.

If you put all that together, it's quite a bit of discrete logic and will definitely cost more than using a programmed part (a micro, or a PLD/CPLD).

[motocoder]

After I posted I realized what I am describing is basically a programmable frequency divider, of which there are many available options (TI makes some, for example).

http://www.ti.com/lsds/ti/logic/rate-multiplier-frequency-divider-timer-products.page?paramCriteria=no

[tautech]

I was without a tach for setting the max no-load RPM for a chainsaw, made an inductive pickup with a few turns on a ferrite rod with a 150 Ohm resistor in parallel to adjust sensitivity, grabbed my DSO and read the frequency x 60 = RPM. Easy for a single cylinder and non-contact.
Many small motor repair shops have very small LCD devices for this, some are cylinder number selectable and some are POS. The best are from their brand suppliers,(Stihl, Husqvarna etc)
I have always thought that a simple frequency counter was the way to go, just divide /multiply by for cylinder number and 2/4 stroke.

[dannyf]

which is 5-1200Hz

That's very low and a wide band vs. the center frequency.

The easiest is to use a mcu.

[Joule Thief]

Paul,

I did something similar many years back to allow an automotive tach  to be used on a single cylinder 4 cycle garden tractor engine.

The circuit was more "created" than "engineered" but as I recall I triggered a 558 quad timer off the signal conditioned pulse from the engine points. Each of the four 555 timers was increasingly delayed by using individually tailored RC delay circuits. This delay scheme allowed the one input pulse from the points to create a four pulse train. The four outputs were then OR'd into the tach.

My requirements may have been more modest than your boat engine. I was only worried about a span of 1200 to 3600 RPM. The four packet pulse train did vary in spacing over the RPM but individual trigger delays were optimized for 3600 RPM (max allowable RPM for garden tractor engines). The mechanical dampening of the tach needle covered many sins created the lack of design elegance.

[Paul Price]

This is a simple problem for a microcontroller to solve and at the same time it is simple to provide a very accurate and precise digital or analog meter reading( with the help PWM circuitry.)  All that is needed to interface with a MCU is a simple circuit to convert the capacitive pickup spikes to trigger an interrupt, (a single pickup using capacitive coupling to a single spark plug wire for input will do.) A small amount of MCU intelligence (programming) does the rest for a rock stable visible output at any RPM.

All a MCU has to do here is count tach pulses/time interval and then present the numerical result onto a digital or analog display. Any MCU can do this with just a few lines of programming.

With a single Microchip PIC or a tiny beginner's Arduino board, this ain't so hard to accomplish.

[eetech00]

Hello,

One of my various hobbies is old outboard motors.  One problem with them is that analog tachometers for them are now very hard to find since their ignition systems are not compatible with any sort of modern marine tachometer.  There is a manufacturer of digital ones that work well.   
Years ago I came up with a circuit that conditions the signal seen by the points in the magneto and feeds that to a tachometer.  This works for the two cylinder models since I can get a pulse from each magneto and feed them to a marine tach setup for a 4 cylinder inboard (4 stroke car type) engine, which expects 2 pulses per revolution.

I stuck it up on the web and surprisingly get a few emails about it a year despite it being pretty hard to find.  Unfortunately it's function is limited to a particular brand of outboard and only their two cylinder magneto ignition models.  Also accessing the point circuit is difficult on some of the models.  What I'd like to do is a second effort that uses a capacitive pickup on a spark plug lead, takes that and multiplies it by 2 4 or 6 (modern outboard tachs use 6 pulses per revolution, which are derived from the CDI charging stator).

I tried doing it with a 4046 PLL setup as a multiplier but found it tricky to get it to lock well and quickly over the required frequency range, which is 5-1200Hz (300-6000RPM).  I had squared up the input signal duty cycle by dividing by 2 which didn't help matters.

It would be trivial to do this with a MCU like an ATTiny, and I could add some helpful features.  However my target audience likely doesn't have a programmer, and I'm not interested in supplying programmed parts.    I doubt there's enough volume to get any 3rd parties interested.

I was thinking today about it and thought maybe a F->V coupled with a V->F might be the solution.  I wouldn't have tried it a couple years ago because I would be afraid of the inaccuracies, but I think my analog is getting a little better.

For the detector, I've rigged up a functional one using a FET.  I saw another design somewhere that used a sensitive gate SCR which seemed like a better idea.  For one thing I'm not too interested in having a FET that near an ignition circuit, but the other was that pulse stretching with it was trivial.

So what do you think? Is there something seriously wrong with the PLL multiplier that I can't get it to lock quickly over that range?  Am I an idiot for not doing it with an MCU, perhaps an easily programmed one?  Can I get better than 1% (50 rpm) accuracy with a F->V->V->F converter without twiddling or exotic components?  Got an awesome capacitive pickup front end?  Oh, and where do people publish schematics these days?

Thanks for any advice!

Hi

Have you looked at the LM2917/LM2907 IC? It's designed for tachometer applications...

http://www.ti.com/lit/ds/symlink/lm2907-n.pdf

eT

[Paul Moir]

Rerouter - I'll check that out.  I do want to use a marine tach since you can get matching speedometers & water pressure gauges and what-not.  Since these are for outboards they'll almost invariably be out in the environment.  I haven't looked at many but the ones I've seen (and the Faria I have sitting here) only go down to 2 pulses per revolution, and that's one designed for inboards.  The outboard ones (typically read to 7000RPM rather than 6, but I don't care much about that) are sometimes fixed at 6 since this is what they all use nowadays.  I'd go for hacking the tach but they're usually sealed.

Motocoder  - I was beginning to look at those last night but didn't get too far and gave up.  Thanks for fleshing out the idea for me.  I agree a MCU would make *way* more sense if I could get easily programmed parts, but it's an interesting challenge to do it that way without too many parts.  One of the Pease circuits I was looking at was a ultrafast response F->V that worked kind of like that with the output going to a DAC to get V.  Got it in just a few chips.

Tautech - yes, there is already a manufacture of digital tachs that work just fine (tinytach).

dannyf & Paul Price - I agree completely, a MCU is the way to go if it wasn't for the other considerations.  It's just that I'm not building this just for myself, and those that I expect also want to build it won't have access to a programmer.  If I could point them to a source of getting programmed parts for say ~$10-$20, I would go this route.  If I can't solve it without one (and without a handful of chips) maybe I'll have to resort to one of those MCUs on a stick with USB.  Also I somehow got that wrong (I'll fix the OP) - the range is only 100Hz-5Hz.  What's worse is I think I have to divide the signal by 2 to get an appropriate duty cycle for the PLL?

Joule Thief - sometimes you can't beat brute force and ignorance when it fits your specs.  The wrong way can be right enough!

eetech00 - yes I have, as a F->V solution.  And as what I'm guessing is used in most of these tachs.  It's also why I think having evenly spaced pulses might not be important since all you're doing is pushing up a charge pump.

I'll try doing some real work on the PLL and see if I can get anywhere.  It's looking like F->V->V->F gets expensive if you want accuracy and/or quick response.  The chips themselves are kind of expensive.

[theatrus]

You can also consider getting a batch of PICs preprogrammed from Microchip. You'd have to handle distribution yourself, but the minimums are surprisingly small.

[Paul Price]

Cheap MCU starter kits come with a PC USB or serial port interface programmer. In the case of PIC chips, you can buy a suitable  PIC device for about $2.25 each (even in small quantities) and the programmer for them (for about $7,  shipping free) on ebay that can program any number of them you want in a few seconds each.

http://www.ebay.com/itm/PIC-USB-K150-Automatic-Programming-Microcontroller-Programmer-ICSP-cable-/141317097545?pt=LH_DefaultDomain_0&hash=item20e727c849

http://www.newark.com/microchip/pic16f886-i-sp/mcu-8bit-pic16-20mhz-dip-28/dp/74K8577

[Paul Moir]

I really don't have the wherewithal or time to get into distribution.  It's not me at all.  And I'm not doing this professionally.

Paul Price - I'm worried it'll be above my target audience.  I think they tend to be older, have some soldering skills and can figure out where to buy parts.  Like a kit builder.  Personally I have AVR & PIC programmers.  PS - I have one of the KIT149 - and while it gets the job done there are *serious* software issues with them.  Something like the original designer encrypted the communications protocol, and then they lost the source code, some soap opera and yadda yadda yadda.  Then the software doesn't work with knockoffs which causes all sorts of problems.  I've abandoned mine even though it's not a knockoff and sorta works.

[dannyf]

A pickit2/3 would be ideal here. forget about the others.

[eetech00]

eetech00 - yes I have, as a F->V solution.  And as what I'm guessing is used in most of these tachs.  It's also why I think having evenly spaced pulses might not be important since all you're doing is pushing up a charge pump.

I'll try doing some real work on the PLL and see if I can get anywhere.  It's looking like F->V->V->F gets expensive if you want accuracy and/or quick response.  The chips themselves are kind of expensive.

Not really. The LM2917 chip is about $1.75 USD and the remaining parts less than $10.
The display would probably cost the most...

eT

[Paul Moir]

The "Display" is intended to be an actual purchased tachometer.  ~ $100 for a marine one.

I was eyeballing the LM2917 in an effort to keep the error under 1%. 

At the moment I am getting somewhere with the PLL.  Actually I have good lock across the range, down under 2Hz.  There's some settling flicker (like ringing) which I'm working on.  Right now it's been a bit too experimental so I'm trying to get it resolved with the datasheet to see where I'm actually operating.  I had to open R2 which I just learned is OK from the Fairchild datasheet.  Didn't know that before!  Makes the values a *lot* more reasonable.

PS - the weird counter is for a 2-in-one package.  I was thinking the PLL needed a decent duty cycle on the input.  I think I'm wrong about this?

[TerminalJack505]

I played around with the attached circuit in the simulator.  It needs only three ICs to produce 6 pulses or 4 pulses.  It needs only two ICs to produce 2 pulses.  It also needs a bunch of passives and probably a bit of twiddling.  All of the components are extremely common.

The pulses wouldn't be evenly spaced.  I don't know anything about tachometers so I don't know if that matters or not.

[Paul Moir]

You know what, it doesn't seem to matter at all!  At least the one I have.  It'll integrate a bunch of quick pulses fine.  That's pretty cool, it's sort of like Joule Theif's 558s.  I'll try breadboarding it while I'm waiting for parts.  I think we will be able to get around the 555 because we ought to get a nice decaying slope out of the front end.  Two identical ICs are hard to beat for simple.

I think I have the PLL low pass tuned pretty good for very nice response over 1000 RPM with no hunting, and only a little bit at the lower frequencies, and that settles out pretty quick.  Locks in at about 3Hz.  I also verified that there's no problem sending the PLL a short pulse signal.  Of course one you understand how they work, it's fairly obvious it really doesn't matter.  It just fights to align it's local oscillator's rising edge (or the 1/4 signal derived from it) with yours.  But I wanted to make sure.
 
I had the 58K resistor in the last schematic in the wrong place.  Won't work like that!  (At least not well..)

So here's where I'm at.  Much easier than I thought it would be.  Should be only 2 chips, no regulator, and I'll replace that 4520 with a 4017 I guess to make N=4 or N=6 selection simple. 

Now I've got to get in an order for some parts for the frontend, anyone suggest a fairly standard TO92ish sized sensitive gate SCR?  I haven't used them since I was a kid.