Making the stock knock sensor reliable and accurate...

[cbrfrenzie]

I am currently in the process of developing an electronic knock sensor conditioning unit. This will be able to tune the center frequency, bandwidth and sensitivity of the knock signal before it goes into the stock ECU. It will also have the capability to run 2 knock sensors into this module, so that you can see knock from each cylinder pair (1,4) and
(2,3).

There would be to re-locatable LED outputs, one for each sensor that you could mount anywhere... i.e. one on the left side of the instrument cluster for cylinders (1,4) and the other on the right side of the instrument cluster for cylinders (2,3).

The technology built into this device would be very sophisticated... where there are blanking times where knock is completely ignored... except when inside a crank angle "window" where knock CAN occur, such as the compression stoke... this would greaty reduce the effects of valve train noise, being the valve would be closed and not slamming open or shut.

The reason for this post is to see if anyone would be interested in a product like this, specific for our DSM's. There would be a model for the 2.0L, 2.3L stroker, and the 2.4L 4g63 engines, given they have different engine harmonics. I know how much it sucks ass to loose 20+ WHP due to some valve train noise on some of our higher horsepower setups... especially those with stiffer valves prings.

Please clobber me with your inputs and suggestion on what would be desired as far as a user interface for knock indication.... i.e. the LED idea blows, or the LED idea sounds sweet...

Extensive R&D will be done on my personal car to ensure correct operation for ALL ranges... as I have 2 AWD dynos at my workplace and can simulate ALL engine loading conditions with the load control features...

Please and thank you!
Chad

 

[cbrfrenzie]

Alright fellas,

After doing a bunch of research last night I found out alot on the factory knock sensing hardware located on the ECU. It turns out that the hardware on the ECU consists of two bandpass filters, centered around two distinct frequencies based around the harmonics the 4G63 engine. The 4G63 Block with a 80mm bore resonates anywhere between 4kHz - 5Khz, where the knock sensor's main harmonic is centered at 6.1 Khz. The stock ECU uses the fundemental frequency as well as the second harmonic in its filtering. The explanation I found for the factorys maddness is that when pre-ignition occurs, the second harmonic is dominant enough to have to be filtered out (12.2kHz). Other detonation events create a frequency at approx 6.1 Khz. I have more info on this so if anyone is interested, please PM me.

It also appears the stock ECU does provide a short blanking pulse on the input just before the ignition coil fires for a period of time porportional to RPM.

This now makes me wonder why so much flase knock is making it thought to the processor causing timing is being pulled.

Since the factory ECU is already doing partially what I started off to do, I will change my focus on providing a nice conditioned signal to the ECU's input hardware.
My theory is that becuase every engine is different (a 0.020" bored out block is 80.5mm which equates to a different frequency), therefore frequencies present in the knock signal are not a constant value. Since the ECU's hardware filtering is set at a constant center frequency, I'm going to try and develop a unit that "learns" your engines harmonics and dynamically slides a front end bandpass filter across a determined (by testing) frequency window to elimate more noise, as well as provide the user with information on "what cylinder knocked" and "how bad" thourough my software application.

Please give me feedback on this new plan of attack.

 

[kenamond]

Do you have any gut feel if things like ARP headstuds, non-OEM HG or other common upgrades would change the harmonics at detonation or pre-ignition in a non-negligible way?

 

[cbrfrenzie]

Do you have any gut feel if things like ARP headstuds, non-OEM HG or other common upgrades would change the harmonics at detonation or pre-ignition in a non-negligible way?

That is a very good question. The only way I can think of to truly test that would be to find a single stock block, do some testing on it to determine the resonante frequency. Then repeat the test on the SAME block with different parts installed.

Going off of gut feeling, I would say yes, it would change it, but the amount would "probobly" be so little that the width of the frequency filter would still reject the changed frequency. This is why I think I'm going to experiment with a "learning" type filter, so that it will optimize itself for each engine it is to work with. This would also tune out distorted frequencys that might occur if the engine were ever to be upgraded... ie ARP studs and so on.

I like the thought of electronics that "learn" about their enviroment and make adjustments real-time to optimize its effectiveness.

 

[turboglenn]

Damn, I wish i had a 5th of the electronics knowledge you appear to have. I'd be building my own stand alones On another note, maybe you can find some people near you that have motors that you cna test for different frequencies in Hz to see what kind of noise a motor makes with "X" parts on it and so forth. Hell if you lived near me, you could use my car for testing since i don't use a knock sensor anyway, there's a hole right there for something to go in.

 

[Spoolin98]

subscribing...

 

[cbrfrenzie]

On another note, maybe you can find some people near you that have motors that you cna test for different frequencies in Hz to see what kind of noise a motor makes with "X" parts on it and so forth. Hell if you lived near me, you could use my car for testing since i don't use a knock sensor anyway, there's a hole right there for something to go in.

I have a guy at my place of work that owns a 90 talon TSI with stock head gasket and stock head studs running an evo3 16g on a tubular manifold. He is ordering ARP studs and a MLS or Cometic head gasket, so I will do a test on a before and after sort of deal. I will also check out my car, (mods in profile) and use that as my first three sets of data.

Both engines at 0.020" overbore -> 85.5MM Bore

 

[GVR4592]

I have a better understanding of what your goal is now after reading the rest of your posts, I think it's a great idea.

With the ecu you're designing, are you planning on having it interface with the dyno so it could essentially tune itself? Of course there would be some user adjustable parameters that would need to be set up, but something like that would be pretty sweet. I heard Autronic is working on something like that, or they had in the past.

This now makes me wonder why so much flase knock is making it thought to the processor causing timing is being pulled.

Are you observing false knock or are you just going by what other people say about phantom knock?

 

[GVR4592]

I know with my AEM EMS, I didn't have to make any changes to the knock threshold when I put in the ARP studs and mls gasket. I also didn't have to make changes when I installed dual valve springs and cams. Not sure how that applies to a factory ecu though.

 

[cbrfrenzie]

Are you observing false knock or are you just going by what other people say about phantom knock?

Both. Other DSM's that I have tuned have shown knock in regions were the AFR, timing and engine load tell me there should be no knock. I have also heard people on these forums complain about the same thing.

On some of the Audi's I've tuned I have also experienced phantom knock, more so than on the DSM's, but still it is present.

I did some dyno testing yesterday with a friends talon and got some really nice sets of data to start my analysis.
These are the remaining problem "areas" I see and their respective actions:

(1) Changing engine displacements (overbore) -> Adjustable bandpass center frequency.
(2) Changing engine noise -> Adjustable bandpass frequency width ( or window) as well as (1).
(3) Auto tune OR User tune feature to set the above two parameters atomatically, or by using software (higher resolution for specific applications, since every engine is different). The auto tune will get you really close.
(4) Software adjust ability with individual cylinder knock feedback and data-logging.

I've also located some 1G ECU code in assembly, so I may also be "tweaking" the knock input code as well for some more testing.

 

[cbrfrenzie]

Alright, I've done some VERY thourough testing on the ECU's knock circuitry and found the exact frequency range that the filter is centered at as well as the 3dB bandwidth.

It turns out that there is a very large frequency "window" that the stock circuitry is allowing through as a potential "knock" signal, however engine noise sweeps right through that frequency range every time you rev past 3000 and 6000 RPM.

What I need from you guys (The DSM community) is as many datalogs of both knock and RPM on the same graph, so I can make a plot of knock frequency vs RPM. I have two local buddies with logging setps that are currently logging every time they drive their cars to get me as much data as possible. I would appreciate if anyone else with those loggin capabilites could do the same

After this final step of testing, I will be finishing my circuit design and testing it out on the 2local cars, (3 including my own) to confirm my theoretical calculations.

I think the biggest interest in this part is the ability to tune out noise and dial in kock more closely on a "per vehicle basis" , as well as log knock vs RPM for each cylinder in engine.

Any thoughts, questions, comments?

 

[GVR4592]

Would an AEM datalog work or would it need to be a stock ecu?

 

[cbrfrenzie]

Would an AEM datalog work or would it need to be a stock ecu?

It would log through its own software interface. In order to calculate a histogram on what cylinder knocked at what RPM there is some post processing that will need to be done. Being that is the case, the only way an existing datalogging software would work is if the vendor for that product supported my product. This is for a bunch of reasons, including logging resolutions, logging frequencies, data format..... there are just too many variables between different systems.

Being this is very un-likely, I plan on making my own datalogging software that would communicate directly with my device, as well as adjust the "tunable" parameters of the system.

 

[cbrfrenzie]

Would an AEM datalog work or would it need to be a stock ecu?

I'm a retard, I miss understood your question.

If your question was in regards to the data-logs I requested, any logging setup will be fine, as long as you have RPM and knock on the same graph so the data is overlaid with respect to time.

 

[cbrfrenzie]

Hello all. An update on my progress:

I've established the frequency ranges my device will operate in as well as the Bandwidth of those ranges.

My next question is a user interface specific one. In the logging software, do people prefer tables, graphs or histograms for reviewing their cylinder specific knock datalogs? I think i' going to put both tables and histograms for referencing knock counts, intensity and where the knock occurred with relation to that cylinders TDC and also with relation to RPM.

I am STILL in need of peoples data logs showing knock, timing and RPM on the same log. I would LOVE it if I could get some logs from different tuning interfaces, (i.e. AEM ( thats you GVR4592 ) DSMLink, Haltech, Motronic, Pocket Logger, ECU+, MegaSquirt (sorry for the swearing...haha) and so on..... the more the better.

I think once I get this done it will be a great analyzing tool for us DSM'ers. Not only will it show you knock in very high detail, but it will be a great diagnostic aid, in the fact you can tell if you're getting pre-ignition, post-ignition, no ignition..... etc.

I'll make up a sample picture of the interface I'm envisioning and you all can tell me what you think.

Chad

 

[kenamond]

I think once I get this done it will be a great analyzing tool for us DSM'ers. Not only will it show you knock in very high detail, but it will be a great diagnostic aid, in the fact you can tell if you're getting pre-ignition, post-ignition, no ignition..... etc.

That's true...you could log misses. That'd be something interesting to see...percent miss (misses per 200 revs). Then you could probably more accurately tune your plug gap or probably other things related to ignition and have quantitative feedback. Heh...may give credence or shoot down indexing plugs on our motors.

 

[cbrfrenzie]

That's true...you could log misses. That'd be something interesting to see...percent miss (misses per 200 revs). Then you could probably more accurately tune your plug gap or probably other things related to ignition and have quantitative feedback. Heh...may give credence or shoot down indexing plugs on our motors.

BINGO! This would open up a completely new can of worms for people tuning DSM engines. You're only as good as your tools and resources.

 

[crensinivon]

Is there a way to use a noise canceling circuit for this.

 

[cbrfrenzie]

Is there a way to use a noise canceling circuit for this.

Yes there is. The stock system actually does use a noise canceling circuit, however the bandwidth of the band pass filter utilized in the stock ECU is quite large. My system will have a "tunable" filter width and center frequency to help aid in the surpression of un-wanted noise.

 

[soldave]

It turns out that there is a very large frequency "window" that the stock circuitry is allowing through as a potential "knock" signal, however engine noise sweeps right through that frequency range every time you rev past 3000 and 6000 RPM.

Very interesting. I'm having a problem with what I'm 99% certain is engine noise causing my knock counts to hit about 25 counts as soon as it hits 6,000 rpm and this could be why.

Are you still wanting log information? I've done some logs on my Evo 1 trying to figure out if this knock was real engine knock or not, and could send them through to you if they'd help

Any ideas when this item might be ready for sale?

 

[cbrfrenzie]

Very interesting. I'm having a problem with what I'm 99% certain is engine noise causing my knock counts to hit about 25 counts as soon as it hits 6,000 rpm and this could be why.

Are you still wanting log information? I've done some logs on my Evo 1 trying to figure out if this knock was real engine knock or not, and could send them through to you if they'd help

Any ideas when this item might be ready for sale?

Yes!! Please send me your logs before and after your cam install! (I'm guessig you're the one who posted about knock showing up after the FP1x cam install). If I could get those logs, then my product will be that much closer to production. I'm trying to finish up my 3D mappable boost controller vs RPM vs IAT, so as soon as that is finished up I'm jumping on the hardware circuit for this knock "tuner". I've got all the conceptual and math worked out, its just more logs would help prove OR disprove my current concept.

What did you log with?

I'll PM you my email for the logs. Thanks, and I'll keep this thread updated as my progress continues!

 

[soldave]

Yeah - I'm the same guy. I have logs before and after cam-install and can send you a bunch of them actually. I log with Pocketlogger and can send you the data as Pocketlogger files or CSV files.

 

[cbrfrenzie]

Yeah - I'm the same guy. I have logs before and after cam-install and can send you a bunch of them actually. I log with Pocketlogger and can send you the data as Pocketlogger files or CSV files.

CSV files please

 

[Black_Bullet]

Any updates to what you were doing here?

 

[sniver]

project dead?

 

[EGuenther]

x2, Did all the theories turn out to be just theories? I was getting all excited as I was reading through this post and then bam! nothing. Updates?