Harlan

I'm building a prototype ion-sensing ignition mod for my Rx8. It's providing good pressure traces, and I have a microprocessor set for signal analysis to detect knock at 3.5+- .3khz and it also checks out on a signal generator. The problem is, I can't know for sure if it can detect knock unless I can induce some knock to detect.

So, what is the safest way to cause knock? I'm thinking advance spark at idle until I get a reading, while also standing by with det phones. Any better ideas?

RotaryEvolution

iTrader: (30)

best way would be with a lower compression engine like a series 4 non turbo. advance the timing about 30 degrees and with quick throttle inputs you will hear it audibly.

but i hate doing it and still do not recommend it, knock always has a chance of causing damage.

Harlan

All I really need is to advance enough to get a reading significantly above baseline, I don't need audible knock. Yeah, it's a risk, but I'd rather know it works for sure before I start doing anything too crazy. Thanks!

Vierte

iTrader: (2)

This is interesting, I have some input.. maybe some of it will be helpful. I will say that I know close to nothing about knock.

Sooo.. stop me if I'm wrong but knock is measured in a count right? Some number of discrete occurrences of knock over some period of time. There are acceptable levels of knock in any engine.. so you should be able to run the engine and observe knock without attempting to introduce it (unsafely). Then the question becomes how can I tune my system to read knock when there is actual knock and not when there is noise. You could find a aftermarket knock system that is tuned for the car and check the knock count against your system. I've also heard of people building diy knock listening stethoscopes and actually listening for knock to occur... and my favorite.. Doesn't Rx-8's have a stock knock sensor that is able to be monitored using an obd scan tool? If that's the case you could probably talk autozone into letting you borrow theirs for a ride around the block while you compare knock levels with your own system.

Another thought... would advancing the ignition at idle and low loads really cause knock or would it be more similar to the characteristics (frequency) of preignition? My feeble mind seems to think there is a distinct difference in the frequency of preignition and detonation because of the process of combustion is different.

Harlan

Stock knock sensor on the RX8 is worthless, or nearly so. So checking knock count with an accessport may or may not save me. Knock is measured by the ecu as a count, because an acoustic knock sensor is hearing all sorts of things other than combustion. Having a knock count with no actual knock is pretty common. Direct pressure measurement with ion sensing or with a spark plug pressure sensor will only detect actual knock and that's what I want to test.

Could buy a J&S system just to check, but part of my problem is that I don't know for sure the knock frequency of a 13BMSP is the same as a 13BREW which I have the data for, in which case the J&S system could be just as deaf. Knock cans make more sense, and I've already tapped into the stock knock sensor to listen to it.

I'm not sure if preignition really has an associated sound. Seems like it's more a loss of power and silent killer of engines often associated with knock.

Thanks for the ideas!

Howard Coleman

iTrader: (6)

a very interesting project...

where did you come up w the 3500 Hz number? i recall the 13 b series around 4200. i am not question the veracity of your number, just trying to better zero in on it. i run a ViPEC V88 ECU and it has a pretty fair internal/active knock system. i am also seriously looking at their Knock Block system as it employs earphones. Richard Green (NZ) seems to be tuning in a whole 'nuther world w the earphones. hearing advance modest knock and being able to take action.

i realize earphone tuning isn't new but it is to me and has caught my interest.

BTW, your comment about RX* knock sensors being worthless also caught my attention. when i installed my V88 i decided to "upgrade" from the single wire FD sensor to the double wire Bosch RX8 item. i thought the Bosch RX8 sensor would be tuned right for the rotary's resonance. i have great respect for the FD sensor BTW.



howard

Harlan

I got the 3500 +-300 from
https://www.rx7club.com/rotary-car-p...engine-839059/

Since the number is based on combustion chamber shape and size it shouldn't be that far off, piston engines of similar design have very similar knock frequencies. I'm just afraid of the frequency being outside my detection band.

Listening to a stock knock sensor is pretty easy. You just use the piezo element as a microphone and plug it into an amplifier. You can also use a 100k resistor and a capacitor and tie into the knock sensor with it still attached to the ecu. Use a cheap laptop running a spectrograph and it should be pretty sensitive.

I kinda misspoke, the stock Rx8 sensor is fine. It's not specifically a tuned sensor, although they all have a resonance(not sure what it's resonance is). It can hear everything, but the ecu has some sort of issue detecting the knock. Someday I'll actually test it to find out if it's a frequency issue or what, but I haven't done that yet.

RotaryEvolution

iTrader: (30)

it's only worthless because in rotary engines the knock level is so discrete that even transmission gear noise will mask lighter levels of knock.

knock earphones can be built rather cheaply but it is difficult to mask the ignition noise since on these engines you have to mount the mic right near the ignition system. i put one together and it works fine but i do need some better shielded wire to cut out the electrical noise induced.

Vierte

iTrader: (2)

what about using the analog version of knock headphones... aka automotive stethoscope. No more noise from electromagnetic interference. I seem to recall seeing an automotive steth with a long tube and a "sensor" that bolted to the engine block for tuning from inside the car.

Popular Mechanics - Google Books

RotaryEvolution

iTrader: (30)

analog does pick up electrical noise, quite a bit of it unless you have very well shielded leads to the microphone.

what you're referring to is a mechanical stethescope, which of course can't pick up electrical noise. i'm sure that method also works but not quite as amplified as an electrical microphone can pick up and amplify. i can hear knock with my electrical amp that i could never hear with my ears from the driver seat, it also is helpful that the speakers double as sound deadeners to kill exhaust noise(regular full size stereo headphones).

madbouncy

Since what you're worried about is where you can't reach MBT, you might as well test there. From what I remember seeing on the dynos, the knock at low load and high load showed the some breakup on the pressure trace but I wasn't the one running the dyno so I can't say if they sounded the same or would be seen the same from a knock sensor perspective. I've only ever tuned for knock while using a microphone on the block with headphones but we heard stories of some companies that would run a stethoscope outside of the cell as you mentioned.

Personally I would go with rotaryevolution's suggestions and shield the living hell out of the wires and run a mic.

Harlan

Jumped the gun a little bit when I said I had good pressure traces. I can get good pressure traces, but I also have a phantom signal which comes and goes. Still a work in progress.

Between a laptop running a spectrograph and det phones I should be able to do this safely. I'll just advance timing slightly until I start getting a reading. Then I'll know what the frequency is for sure, and I'll be able to test the digital filter.

The end goal of this is to both have a knock sensor and have peak combustion pressure point to get MBT. Going more advanced than MBT will quickly cause knock too, so knock is most telling.

I'll post more if I get it working. Wish me luck.

j9fd3s

iTrader: (3)

i'm not sure you will be able to get it to knock it idle, and at slightly higher rpm and load, it might be tricky to figure out what is knock and what is preignition, as even a stock Rx8 will do a little preignition at 1500-2000rpm, that's what that chirping noise is

RotaryEvolution

iTrader: (30)

and the frequency of knock in an older rotary may not be all that similar to a renesis due to the differences between the engines.

i have heard the renesis "chirp" but i can't even say for sure that it is knock, compared to say a series 4 non turbo the chip turns to a "ping ping ping". much more definitive even from a basic standpoint.

madbouncy

I talked to my buddy at work who has been running the dyno the longest about the knock at low load and he said that it's very very faint when it happens but anytime you can see it on the pressure trace you'll be able to hear it in the head phones. However, most of the time they actually just go to the next point once they hit 50 degrees advanced and just record the ignition timing for MBT.

Edit: The point I was trying to make is that it be a lot easier to go to a higher load region where you only have to advance a few degrees to find knock opposed to low load where you spend 30 times flashing and are still never sure you correctly heard knock.

arghx

iTrader: (3)

Knock is almost always audible before it can be easily distinguished on a pressure trace, at least according to the guys I know who do that for a living. That's why, depending on the engine, knock is usually set by a subjective rating by somebody who's trained to listen for it. Basically, when you can start seeing knock on a pressure trace you're already risking damaging something. Once the tech decides the engine is knocking, he will record the engine's onboard knock sensor reading and submit that information to be used as the knock threshold.

So in the case of the Rx-8, somebody at Mazda set a pretty insensitive knocking threshold OR the knock control was set to not be very active, most likely for fuel economy & catalyst protection. You see, what probably happened is that somebody at Mazda realized an overactive knock retard system would raise the EGT, and burn up the cat. Then you'd have warranty claims and possible fines from the government if the vehicle start failing government in-use emissions tests. I'm sure all the software for aggressive knock control is built into the stock Mazda ECU, just nobody set it for aggressive retard at Mazda and nobody in the aftermarket really figured out how it all works.

I'm not sure what your engine management is, but without realtime adjustment (not just flashing with a Cobb AP) it's going to be very difficult. You have to be able to hold the load point and sweep your spark timing and listen. If you ever watch engine dyno technicians do it, it can be a pretty intense process. They're watching the monitors trying to hold the engine together, while listening to the knock microphone & adjusting load + spark.

Howard,

the biggest thing about knock control is setting the proper knock/noise threshold & then understanding the actual control logic used in the ECU. As I was alluding to, during engine development usually the knock threshold is set by mapping out a series of load points on an engine dyno. Basically, sweep RPM & torque points and then try different timing at each point. The knock sensor reading at borderline detonation point is, generally speaking, considered the trigger point for a knock count as the ECU software sees it. You can map out knock trigger points by rpm & load. The AEM EMS for example lets you set some baseline noise threshold, and modern stock ECUs are the same way.

Once the knock count has been incremented, it's up to the ECU to figure out what to do with it. The simplest and oldest way is to immediately pull timing according to strength of the knock (feedback correction). Then a timer counts while the timing slowly returns to normal. Newer stock ECU's though will calculate a fuel octane rating or some other kind of knock sensitivity factor and use it for timing correction in the future. It can get pretty complicated.

It all starts with mapping the knock threshold/trigger point though. And for that, usually it takes a good setup and somebody's trained ear.

madbouncy

I completely agree with that, that's what I was try to say when I talked about that if you can see it on a pressure trace that you'll be able to hear it. So far I've only ever seen it done the same way with somebody listening with headphones to set the values, hell even the automated mapping software has a long calibration time where the person basically trains the software to know which events are knock and it all comes from what he hears.

RotaryEvolution

iTrader: (30)

guess i'm just hard headed and prefer the old technique of hard mapping the timing versus having an overactive auto compensating ECU. if the engine is going to knock at one point, it most likely is going to knock when it gets to that point again.

i've never experienced decent enough gains in excessive advance with these engines to spend all that much time figuring out threshholds for each setup. knock is usually caused by severe lean conditions anyways(in turbo engines).

Howard Coleman

iTrader: (6)

Richard Green's tuning process w headphones sold me.

i find this comment important:

"Knock is almost always audible before it can be easily distinguished on a pressure trace" /arghx


a Vi-PEC Knock Amplifier Kit is in the air from New Zealand and will be a part of our forward tuning process. i expect it within a week and we will be on the dyno w it immediately thereafter.

one of the reasons knock reading is important to me is i have the ability to accurately deliver my AI. so the question becomes... what should be the proper amount be everywhere on the map.

only the best knock reading ability will provide the answer.

Howard

Harlan

What I know about the stock Rx8 knock system is this:
The ecu generates a tone and sends it to the knock sensor. (Haven't checked what the frequency is yet.)
The amplitude of the tone the ecu generates changes with load. Probably controlled by a table accessport doesn't show.
The ecu does have coarse and fine knock adjustment, the max retard is 10deg, but I don't know if that's leading only, or if it retards both.

What I think:
The ecu hears the tone and finds the frequency. If it hears same frequency it's generating all is good. If not it increments the knock number. Somewhere in there is a filter that may be tossing out the actual frequency the engine knocks at.

Or the threshold is so high it can't hear anything.

And several tuners on the 8 forum have claimed that the chirping sound is knock. I get it sometimes when starting from cold, so I may not have to find knock after all, it may have already found me. Frequency analysis should tell the tale.

RotaryEvolution

iTrader: (30)

knock sensors are a basic piezoelectric generator, they generate very small amounts of voltage that the ECU translates into a herts signal to compare with on a knock table to determine if the knock level is acceptable or not.

the ECU does not send any type of signal to the sensor.

sensors can be innacurate because the signal could be generated from noise vibrations carried out through the engine from various places. exhaust droning is an example of a hertz signal caused by resonance and even the knock sensor could translate that into a knock reading.

this is why many tuners are aiming more at their own senses, because a human can easily differentiate a "pop" from a droning sound or clattering from the transmission under light decelration. most of the highest knock readings i see are usually caused by gear lash in the drivetrain under light loads. sometimes you have to think of the engine and drivetrain as a bar of metal, you can tap anywhere on the bar and you will hear it at the opposite end.

Harlan

Yes it does. I'm not saying that all systems operate this way, but this system for whatever reason does. It really confused me at first because I was hearing a tone from the knock sensor that changed with engine rpm/load. When I disconnected the sensor and ran my own leads to it the tone disappeared and engine nose is all that remained.

It still could be interference from the coils or such, but that didn't make sense with how it changed either. I'll test more and post results.

RotaryEvolution

iTrader: (30)

it is probably just feedback from the input circuit to the ECU.

Harlan

Did more testing, even more confused. Signal the ecu is generating is not as clean as I remember it, it sounds more like interference than a generated tone. It persists with the knock sensor unplugged, and still changes with rpm. The sound from the knock sensor alone is very clean. Also used a tone generator to try and induce knock as indicated on my accessport. Couldn't get indication of knock retard, but it only indicates one kind of knock retard out of 3... Don't know if I failed or if I can't read the knock I was causing, either way the engine didn't change at all so if it did retard it wasn't significant.


Next step is to disconnect the knock sensor wire and see if the ecu is generating the signal or if it's interference from the rest of the wiring harness.

I'd share the sound files as MP3, but can't get them small enough to upload.

RotaryEvolution

iTrader: (30)

how are you incuding knock? a simple method is rapping on an engine lift bracket with a ballpeen hammer.