I attached a tone generator to the knock sensor leads and pulsed different frequencies. I was just trying to figure out what the ecu could hear.

 

It should also require a minimum amplitude even at the correct frequency. Not sure if your ecu does it but a lot of them also only listen to the knock sensor around combustion timing.

Rotaryevolution is right in just tapping on the engine with a hammer, it's how you test to see if the sensor works for diagnostics as well.

 

Harlan --you may want to call Racing Beat? They had problems getting the knock sensor pcm relationship to "cooperate" when they were dealing with the rx8 pcm reflash development. It was also inconsistent. They had to disconnect the knock sensor mechanically in order to develop their reflash. I bet they saw some things that may be of use.

 

Found knock. At low rpm it sounds like a chirping/belt squealing sound. Somewhere around 3-4khz. The interference is around 1050hz and has no harmonics and it does change in both intensity and frequency with rpm. Might be related to the ignition coils. I tried giving the knock sensor a solid ground, that didn't make things better. I tried disconnecting the line to the ecu and the interference remained (on the ecu side).

Nope, haven't contacted Racing Beat yet. They probably put the thing through its paces.

Realized today that I was connected to the wrong line when I fed the ecu tones. Feel pretty dumb now, but I found out the o2 sensor line has the same exact interference, and it doesn't respond to music.

It this point I'd love to know where the interference is coming from. I bet it holds the key to why the knock sensor is useless.

 

it is probably an amplifier on the ECU side to increase the range of input, giving it better resolution.

 

Frequency is about 3250hz, can be heard in a band from 2400-4100hz. Thanks for all the help guys.

 

If you don't mind sharing could you post up some info on your setup as far as the ecu you are running the system with and whether you're planning on doing real time changes with the ion sensing or just using it as a tuning tool.

 

Sure, I'm running the stock ECU and trying to use ion sense with an Arduino as a tuning aid. The plan is to get knock detection and eventually PPP detection working as a tuning aid, and then work forward to real time timing changes. Right now I just want a solid knock detection system.

So far I have a Goetzel algorithm working at detecting any frequency/band that I want. I also have piggyback ECU script for the arduino to retard timing in 1/2 degree increments. Hopefully I can get the two working together on the car and have a safety net.

It's all pretty rough, but it shows promise.

 

Definitely sounds like fun. Does the rx8 table for knock allow values for all rpms? Once you get the PPP detection working you could always just have the board ouput a signal to the rx8 when it sees MBT that it's actually at the knock limit so it won't advance any further.

Any chance you have a trace of the signals you have been seeing from the ion sense circuit you have? It seems like a lot of people mention that they had to do a lot of averaging because of inconsistent signals, have you had any issues with that so far?

 

Accessing the stock knock control system is limited because the accessport software is poorly labeled and is missing a lot of tables. RX8 stock ECU tuning is only in a semi-functional state because of this. I think running a piggyback ecu config for ignition is my easiest solution, but there are advancements in tuning being made and that may change.

I'm having a horrible time with 60hz interference, and using a resistor bridge to measure has not helped. Gonna switch to a current mirror and see if it gets better results. I wish I knew more about simple electronics. Anyway, here is the best trace so far and a quick video of the setup, the blue line is the L1 coil dwell signal:

 

Generally at work whenever we have 60hz noise we ground to literally everything we can until it goes away. Definitely a brute force method but it hasn't failed. You said your dc power is actually home ac through diodes right?

 

Also, have you put a trace to your ion circuit voltage to see how its doing?

 

I haven't traced the dc voltage, so it may be part of the problem. Yes I am using diodes to rectify household ac. Originally I had a design that used the ignition coil to charge a diode, but it didn't work right with the coils I'm using, and rectifying ac was a quick and dirty way to get a result. I think the problem has more to with the relatively weak signal strength, if I could get more voltage to the scope then It would have less trouble with interference.

Next step on this front is a better dc supply and a current mirror to give me more output. Also I can add a filter now that I have a good idea what frequencies I need for knock. Idealy for knock sensing you take two samples, one at around 25-35ATDC for baseline and one at 45-55ATDC for knock. Since knock only happens 45-55ATDC per Yamamoto it should give a very clear indication. Should also work with a regular knock sensor.

Oh two more things: The stock knock sensor is actually very sensitive. I fed a range of frequencies to the ECU and it pulled timing from 2khz to well past 6khz. It also pulled timing when tapped with socket extension and it pulled a lot when I slammed the hood shut. This was all at idle. Maybe with a proper filter the stock knock sensor will work well? It would help if I had access to the sensitivity table.

 

What voltage are you feeding the ion circuit, is it just the 120V or do you have a transformer in there? Seems like everybody aims for 300-400V. Do you have an output voltage range you're aiming for? Automotive is generally always 5V for sensors but with your own board you're using you might be able to use a higher voltage to get a better resolution.

Not surprisingly, knock generally occurs around peak pressure and since you're advancing towards MBT when you start to knock, it makes sense to put you at or before it. I don't remember off hand what MBT timing is supposed to be for the 13B but you mentioned that it was 45. For a piston engine that would see MBT around 7-10 degrees the knock event would start at or just after TDC whenever I saw a pressure trace of it. It's not stuff I do on a day to day basis so most of my knowledge of that is just from bullshitting over around the guys that deal with it.

 

It's about 160vdc. I'm aiming for 3-4 volts for the PPP, but getting less than 1, my o-scope and the arduino take inputs up to 5volts. Peak pressure at MBT is 44-45deg.
I found this very informative, and most of the information is still valid:
http://foxed.ca/rx7manual/manuals/RE...amoto-1981.pdf

 

It's likely that the Rx-8 looks at the signal distribution over some set period, and when it exceeds a threshold it knocks. When you continuously feed it certain frequencies, the integral easily exceeds the threshold. Basically, in the real world the knock sensor reading under measurement conditions must be different than what you are feeding the ECU for this sort of bench test.

 

When you switch to a current mirror are you just going to use a resistor to pull down the voltage to what you want?

Thanks for the link, I actually have the same book name by the same author but from 1971, however the 1981 has a lot more specific test results for our engines.

 

I wasn't feeding it continuous tones, I was feeding it pulsed tones. Even then it seemed to respond the most when I was sweeping the frequency. I think it looks for a quick hit followed by the sound ringing out. I may generate some fake knock at different frequencies just to be sure. The biggest thing I wanted to find out was if there was some sort of hardware/software frequency filter, and there doesn't appear to be.

Yeah, using resistors to get a voltage signal from the mirrored current. Also needs a zener diode to keep it under 5 volts when the coil fires.

 

Here's a youtube video of what I found and believe is knock on a Renisis please tell me what you think:

Also attached are the spectrograph pictures that helped me track down the frequency.

Let me know what you guys think.

 

Thanks for the help guys, pretty definitely found knock frequency on the Renisis and am slowly getting a better understanding of the knock control strategy.

The tone from the ecu is probably a continuity check. The knock sensor has a 200kohm resistance that the ecu looks for.

At this point I think the stock knock detection works (maybe quite well) and is just badly misunderstood.

Here's my second video if anyone wants it:

 

Very nicely done, the tone is completely different than the engine I'm use to hearing but when you filter and amplified it, it definitely still has that distinctive metallic twing like sound. I've only ever heard knock through headphones for one specific engine so you'll probably want opinions from a lot more experienced people on this.

 

the chirp noise sounds right to me. just about every Rx8 i've ever driven makes that noise, mostly around 1200-2000rpm, it helps to let the clutch out quickly too.

my peripheral port does exactly the same thing, i'm running similar timing and AFR* to the Rx8, in that Rpm range.

*says the wideband, although all the overlap throws it off

 

i would try to filter your pickup leads better to eliminate the ignition noise so you can get a better feel for it.

hell, even wrapping your wire extensions in the engine bay with aluminum foil would probably clean it up a bit.

even though filtered, your ignition may be harmonically resonating what you're seeing/hearing as knock. if the system works as designed, IMO, it should be retarding timing as soon as knock is encountered and some of the clips it was persistent under load. of course it may be octane related though, since the renesis is designed to work on 91+ octane as denoted inside the gas door. but i have a hard time envisioning detonation due to octane out of boundaries that the ECU could compensate for where in turbo engines they produce much more ICR under boost. lower octane would likely just cause knock to happen more frequently as the ECU deals with it.

 

The second video is much clearer than the first. It's not harmonic resonance of the ignition/engine, although there is some of that present. For those sound clips the sensor was disconnected from the stock ecu so the computer couldn't hear anything, and was pushing full timing.

Everything is back to stock now, waiting for parts for an opamp based filter to use as det phones and to feed a filtered signal to the ECU. I'm also trying to figure out the Accessport tables for knock control (which modes they control and how), but without someone a bit more ECU savvy I'm at a loss. I may start begging on the subaru forums at this rate, since they use the same 32bit processor and a lot of the code/tables are the same.

 

Nice work, Harlan