Knock Sensor, Knock Filter, ESC and swaps!

[EagleMark]

and he has a SWEET camaro! But I forget his forum name, and forgot I was supposed to add this to Write Ups... my bad, I fail...

[1project2many]

"Zaut"
Used to build wiring harnesses for Packard Electric / GM / Delphi.
http://www.gearhead-efi.com/Fuel-Inj...-Z-(From-Intro)
http://www.gearhead-efi.com/Fuel-Inj...n-from-1226870
http://www.gearhead-efi.com/Fuel-Inj...ning-questions

[RobertISaar]

been looking into alternative knock detection/filtering methods....

as much as i'd love to be able to do the ionic sensing type of stuff(i believe SAAB used it for a long time), i'm pulling a blank when it comes to figuring out a way to adapt sensing and filtering into a system that can produce somewhere in the range of 60,000 volts...

so, piezo-sensor based sensing is what will have to be stuck with for now. googling various terms has come up with some interesting solutions, some of which are dedicated knock sensing ICs. texas instruments makes one(that GM probably used once they got rid of the hardware filters), the TPIC8101. you can program it over a SPI dataline and change the characteristics of it as much as necessary. the only issue with it is that it NEEDS to be programmed over the SPI dataline.... so you pretty much have to have another IC communicate with it(smaller AT units have been suggested) to send its settings.

the other only real option i can see is more hardware based filtering, just obviously we would want to be able to adjust it to the frequency and sensitivity needed for a wide range of engines. googled a LOT for this, but i couldn't really find anything simple/better than one of the first results found, from the megasquirt guys, no less.

http://www.msextra.com/doc/ms1extra/...nual.htm#knock

that basic design SEEMS like it would be compatible with any C3/P4 unit that allows using an external filter. it uses REALLY simple, cheap and common components, there's only a couple of problems with it:

as displayed, it only uses a high-pass filter(set for 1.6KHz), not a band-pass filter. not difficult at all to fix by adding a low-pass filter and tweaking the high-pass for something a lot more usable.

as displayed, a 0-5V signal is produced.... 5V when no knock, 0V when knock. this is very similar to how GM's external knock filters worked, but those gave ~10V or so(according to some more MS documents) instead of 5V... this may need to be corrected.

400mS of knock signal is a LOT of time(circuit actually analyzes out to 429mS)... from what i'm told, since holding the knock signal high(since there is an inverter inside the ECM causing the low pulse to become a high pulse) causes the counter to increment continously, 400mS would be treated as a LOT of knock counts.... i've seen references of the filter requiring to be in the "knock" condition for 5mS before the processor will recognize it(start incrementing the knock counter)... seems kind of long to me? may have to see if i can verify this on the testbench...

otherwise, it looks fine.... i just need to figure out what that pot is being used for... i assume level matching or possibly sensitivity adjustment?

[JeepsAndGuns]

You are definately way smarter than me. I think I understood a few of the words....

Keep us updated on any big breaks you have. I offer to help however I can, though I dont think it will be in any usefull way.

[EagleMark]

Me either, except maybe another idea?

What does a 0411 use? Did you see all the adjustments? The code is already there, could write an XDF for it? Now how could we get that part and program it? Just thinking...

[RobertISaar]

0411 uses some type of "programmable" filter.... probably the TI unit, but you would have to look pretty closely inside to confirm.

as much as i'd like to use it, for the OBD1 and early OBD2 applications, would need to use another microprocessor and download code from a laptop/etc into it, which will then push settings into the knock filter IC. maybe go really nuts and use DIP switches or a rotary encoder or something and have it switch between multiple knock sensor profiles based on that?

[EagleMark]

I'm glad you have it figured out! So when can we expect one?

[34blazer]

iirc there was an aftermarket adjustable knock detection system available. I cant seem to remember the brand, MSD maybe?

[RobertISaar]

i've seen them before.... just skeptical of how well they can be suited for any random application.

this should allow for nearly anything to have accurate knock sensing, regardless of bore size. background noise should also be able to be accounted for. yet another benefit would to be able to hookup any knock sensor, regardless of impedance (3.9K vs 100K) or broadband vs resonant.

BTW: the resonant sensors? they have a bandwidth of roughly 1000Hz around their target frequency.

[EagleMark]

Sounds like a cure all!

I'm familar with 3.9 and 100k, whats the broadband vs resonant and 1000hz refer to?

[RobertISaar]

broadband = the newer 2-wire sensors, they have a "flat response" over a VERY wide range of frequencies... i think i saw 1,000-19,000Hz? anyways, using those, you pretty much capture a LOT of "sound" from everything and the filter has a lot of work to do to try and determine knock from other sounds. resonant = the 1-wire sensors used with all of OBD1 and some OBD2 applications. they only create a signal when a "sound" is close to the frequency they resonate at. say you have a sensor setup for 5,000Hz, since it has a bandwidth of roughly 1,000Hz, then it will only ever make a signal when something in the 4,500-5,500Hz range happens. the filter has less work to do since the sensor is a rough estimation of where knock will happen.

[EagleMark]

Ahhh! So that is what all this work in PCM is doing? Not the table shown but the other parameters available like frequency and sensitivity?

[RobertISaar]

yeah, the Knock-Hardware looks to be what it's referring to.

[daleulan]

Memcal.jpg

Functional sketch, I may have a couple of errors but basically there is a filter, precision rectifier, fast filter, buffer with gain and offset adjustment, slow filter, and comparator.