944 Turbo Knock Counter

[Tom]

Here is a downloadable DIY guide showing how to make and install a knock counter on a 944 Turbo. A knock counter (and wideband) are invaluable tools for modified cars!

Porsche 944 Turbo Knock Counter DIY.pdf

knock counter pdf pic.jpg

[whalenlg]

Hi Tom - I put one in and has been working for the past year or so. I used a Trumeter 7110DIN that had a reset already. Wired for the fast count line (min 50us pulse width) since the slow one was spec'd for 15ms pulse minimum.

Only issue is that I think I get a lot of false counts with low boost.

I added in a Hobbs like pressure switch (verified normally open and closes about 4psi):
https://www.amazon.com/dp/B08N4FDX38?ps ... ct_details
between the KLR diagnostic connector under the glove box and the counter fast input pin.

At low to 0 boost no counts.

Once I get to the active range (5+ PSI), way too many counts (like 80 counts for 1 2nd gear pull).

Has anyone seen this before? A Res + Cap RC delay might filter out the spikes, but I really couldn't find what normal pulse widths are for this signal.

[Tom]

Hi Tom - I put one in and has been working for the past year or so. I used a Trumeter 7110DIN that had a reset already. Wired for the fast count line (min 50us pulse width) since the slow one was spec'd for 15ms pulse minimum.

Only issue is that I think I get a lot of false counts with low boost.

I added in a Hobbs like pressure switch (verified normally open and closes about 4psi):
https://www.amazon.com/dp/B08N4FDX38?ps ... ct_details
between the KLR diagnostic connector under the glove box and the counter fast input pin.

At low to 0 boost no counts.

Once I get to the active range (5+ PSI), way too many counts (like 80 counts for 1 2nd gear pull).

Has anyone seen this before? A Res + Cap RC delay might filter out the spikes, but I really couldn't find what normal pulse widths are for this signal.

You are most likely picking up electrical interference from the ignition that is lasting longer than 50us and triggering the sensor repeatedly. I can go back and check my o-scope shots of the KLR signal, but when I build my own using an arduino, I learned quickly that it's very hard to distinguish noise from knock. So, what I did was hook up both a Curtis counter and my Arduino counter to the same line, and worked (via software) to ensure the Arduino had the same count as the Curtis. I figured it was most important to have a counter that was comparable to every one else's counter -- so that I could compare my knock counts to what others are seeing. The Curtis counter looks for something in the 500us range, and can't trigger more than once per millisecond. As a result, it filters out most or all of the noise. You could try putting an rc filter in front of your sensor to stretch out the pulse, but ultimately I think it would be way easier to just get the Curtis counter for $30 or whatever it costs... If you then want to filter out no-boost knocks, you could use your pressure switch, or use an arduino to serve as the gatekeeper. More than happy to help if I can....

Also, if you want to stick with your sensor, you might also look at a Schmitt trigger like the 74C14 chip. When I build the SpeedoBooster, the speedometer signal had tons of electrical noise that was causing the speedometer to go whacky just like your knock counter. The 74c14 removed 100% of the noise....

Here was my knock counter, which replicated the count on the Curtis counter (and had other features....).

[whalenlg]

Thanks Tom - I don't think it's ignition noise since it doesn't get so many counts under boost without the Hobbs switch.
I think it makes more sense to do an AND circuit with the Knock signal and pull the Hobbs input to 12V with some RC to remove any switch bounce or noise.

Any recollection on how wide the Knock pulses coming from the KLR are?

BTW - I had a speedo calibration issue as well - trial and error'd my way to the right resistor value on the speedo circuit board after replacing burnt caps.

[Tom]

Gotcha. If that pressure switch is mechanical and activated by air pressure, then the 'bounce' on that switch could indeed be what you are seeing (and maybe a little noise). An RC circuit with Schmitt trigger is the way to go if that's the case. An RC circuit alone might mask the bounce, but it then creates an analog signal rather than the digital (on/off) signal your counter is looking for. You can still get it to work, but you'd be relying on the counter's threshold voltages and most likely slow down your effective sample rate waiting for the voltage to build and decay. Adding a Schmitt trigger snaps the voltage up and down to create a true digital pulse. See video below. You can then scale the RC circuit to create (say) 1ms pulses, which will negate the 50us sample rate of the counter and make it act comparable to the Curtis counter.

I was mistaken about having o-scope shots of the KLR signal. Random pulses like that, in a running car, are tough to catch on the scope (for me anyway), so what I did was add an RC circuit to see if I could data log them. The KLR pulse isn't documented as far as I know, so I'm not able to say exactly how long the pulses are, or if they are always the same length (or height for that matter). I can tell you that the Zeitronix data logger samples 65 times a second, and that is too slow to capture the pulse. 65 times a second is about 15ms, so I'd be pretty confident saying the pulse is shorter than that. Then speculating/extrapolating a bit, at 6500 rpms, there are 13,000 combustion events in a minute, or about 217 every second. Assuming Bosch intended to send out a pulse for every knock event, then that means the pulse would need to be shorter than about 4.5ms (at least at redline -- maybe longer at lower rpms?), or more likely no longer than 2.25ms in order to have a 50% duty cycle during rapid-fire knock counts. While it is possible the pulses are shorter still when coming out of the KLR, I wouldn't see any reason to trigger the counter any faster than the car is able to actually knock. If you aimed for a 500hz pulse frequency (50% duty cycle) with minimum of 1ms between consecutive pulses, you should be good to go -- that's the spec on the Curtis counter.

[whalenlg]

Thanks - those Hobb switches are a diaphragm pushing against a spring, so definitely mechanical.

I did the same mental math on the theoretical KLR knock signal pulse width and took a guess at the settling time of the switch of ~0.5 sec. Counter threshold voltage is 0.7v.

I drew up and Spiced a 1 transistor circuit last night and put it on a breadboard - will try it this evening and post results and schematic if it does what I want.

[Tom]

Thanks - those Hobb switches are a diaphragm pushing against a spring, so definitely mechanical.

I did the same mental math on the theoretical KLR knock signal pulse width and took a guess at the settling time of the switch of ~0.5 sec. Counter threshold voltage is 0.7v.

I drew up and Spiced a 1 transistor circuit last night and put it on a breadboard - will try it this evening and post results and schematic if it does what I want.

.5 seconds, as in 500ms? The KLR pulse isn't that long, and I think you'd miss counts at .5 seconds....? Lots of ways to clean up the bounce, so not suggesting this is the only way, but I forgot to include the video I mentioned above...

[whalenlg]

This seems to work for me - I did get count a few knocks at 10+ psi and no counts for less then 2 psi.
This version doesn't simulate quite right in LTSpice for the MYSW pressure switch model, but my other spice runs that simulated switch bounce differently were fine.
Spice shows 2.5V pulses coming out which is a reasonable margin above the 0.7v counter input threshold.

Screen Shot 2022-07-18 at 7.50.47 PM.png

[whalenlg]

BTW - I re ran the Spice simulation and was seeing too much bleed through of the input pulses to the output with the pressure switch open, so changed R3 to 100k and R4 to 270k. Will test in the car later this week.

[Tom]

This seems to work for me - I did get count a few knocks at 10+ psi and no counts for less then 2 psi.
This version doesn't simulate quite right in LTSpice for the MYSW pressure switch model, but my other spice runs that simulated switch bounce differently were fine.
Spice shows 2.5V pulses coming out which is a reasonable margin above the 0.7v counter input threshold.
Screen Shot 2022-07-18 at 7.50.47 PM.png

Let us know how it does! I'll have to study that schematic a bit. What do the pulse times mean next to V1 and V2 -- are those LTSpice inputs? It seems like the RC circuit (c1 and r2) will keep the output high for longer than you want, no? -- like hundreds of milliseconds? Or am I not understanding the circuit correctly?

Edit: looking again, maybe I see it now? You are debouncing the Hobbs switch, which is resulting in the long RC time I was stuck on. But a few hundred ms isn't important when you are talking about whether the boost is over or under 4psi... I was thinking that you'd want to debounce the KLR pulse to filter out noise, but if I am understanding this circuit, you are passing through the KLR pulse in its original length without any noise filter.... ?