Ms3 knock control hardware!
 

This looks pretty awesome, apparently he's doing a small beta run... If I already had my NB motor installed I'd be all over it, but its going to be a while. Someone help test!

http://msextra.com/forums/viewtopic....ee23ec829a96e6

We've been testing an older version on a beater Nissan already. It's already showing quite an improvement over older MS knock sensing systems.

Sounds interesting.

Hard to be absolutely certain from the picture, but it looks like he's using the TPIC8101 chip: http://www.ti.com/lit/ds/slis110a/slis110a.pdf

I'm trying to guess as to what the various pins from the MS3 he's called out are doing. Some combination of INT/HOLD (for windowing) and implementing a bit-bang SPI interface?

Correct, that's what he is using, and it is a SPI interface.

I'm probably in for a beta board. anyone want to donate $5 to me?

Very cool.

Something I have long wondered, being by no means an expert in the science of engine knock:

My own personal experience with knock sensing is limited to one attempt, using the KnockSenseMS. I found that the ambient noise from the engine, which increased as a function of RPM and load, tended to limit the usefulness of the device. I could adjust it so that I got accurate knock readings at lower RPM/load but constant false indications at higher RPM/load, or I could set it up to mask out the high RPM/load noise while sacrificing sensitivity at the low end.

I have seen different companies attempt different solutions to this problem.

TurboXS, for instance, takes an RPM sample and scales the sensitivity of their KnockLite based on RPM.

J&S uses individual-cylinder windowing, and also takes both RPM and MAP into account somehow, although I can find no technical detail on this.


Where I get confused with the knock-windowing approach is this: If ambient mechanical noise is more or less constant (or at least pseudo-randomly distributed), then isn't there just as good a chance for noise to induce a false-knock indication during the knock window of a system so equipped as there would be for a system that does not use knock-windowing?

i'm in.

So I bail on the MS, and they finally get serious about this. It will probably end up being well-sorted by the time I get my AEM ems installed...

Knock control has always been something Ken and James have been serious about adding to MS3. It is a time thing more then anything else. There have just been other things that they want done before they add in knock support.

Y8s, if you serious and need funds, send me paypal. Im in for $5 and would send more if need be. I see this as a fundamental tuning aid.

Several months back, I spent 2 days or so playing with one of these (well a handbuilt pre-cursor to it) on the dyno. It is way cool. You set the bandpass knock frequency range for your engine and a table of "knock threshold" vs RPM. This allows you to compensate for engine noise with RPM. There is also a "knock count" adjustment that is how many times the knock threshold is breached before anti-knock action will be taken. You can specify how much timing is pulled at a knock event, and you can specify how quickly you want the MS3 to add the timing back. Also, it does window knock events as knock can only occur in a narrow range of crank angles, so if there is a noise spike but it occurs outside of a knock window, then it's not knock.

James did a wonderful job on this.

There's no reason I need any outside funding. I'm happy to own one at that price actually.

Plus I'm pretty good getting my MS3x feedback to James and Ken so I'm sure they appreciate that.

Ah ha! When I was looking through the datasheet I completely overlooked the fact that gain was dynamically adjustable in real time. (The more I read about this chip, the more I understand why, at first glance, the parts count on that board seems so amazingly low.)

Knock windowing plus gain-vs-RPM does indeed sound like the recipe for a robust and effective product.

Now you've got me curious as hell. Does the MS3 have on-board hardware support for an SPI bus, or do they really have enough surplus CPU cycles that they can just bit-bang it on GPIO pins?

The MS3 CPU has a real SPI bus (actually more than one and the SD card uses another one). The MS2 CPU also has one but the pins are used for the LEDs and idle so using the SPI bus would make MS2 quite short of I/Os.

Jean

knock module ordered. very exciting.

and probably long overdue in a 10:1 turbo car.

So much win.

y8s: you're running upgraded rods with stock pistons on 93 octane? I switched to e85 to prepare for the same setup, though I think with this winning knock control I might switch back.
Since they dropped the ethanol subsidies, e85 is only 10c less than 87, the price benefit isn't there anymore, and I kinda feel silly since you damn sure don't need e85 to pretzel stock rods. The mileage drop and limited number of e85 stations is a PITA.

Y8s is still on a completely stock block. He parted out all his stuff for his engine build.

Would you recommend buying module now, or wait for further development?

It's likely that any further development will come from the firmware side and not the hardware side.

http://msextra.com/forums/viewtopic....301071#p300994

Because knock only occurs inside a specific range of crank angle (window) after spark occurs, I suggested that the algorithm compare noise levels between in-window and out-of-window. This difference should be compared to a table that has RPM as its x axis. When the difference exceeds a certain amount then there is knock.

This would improve knock discrimination vs. noise that rises with RPM.

The reason for the table is that even without knock, the crank angle during which knock can occur is also the period of combustion and flame propagation - and thus there is a noise difference. Said normal signal delta between in and out of window *also* increases with timing and power production, even in the absence of knock.

The characteristics of the bandpass filter and actual knock noise should be considered, in that a high-Q bandpass filter will stretch out the knock signal in time and can leak outside of the accepted knock window. So the valid knock window should be stretched out a little bit, with a fixed time adder (i.e. not angle adder). Either that or apply the windowed signals into 2 separate bandpass filters. However the windowing function itself has to be designed not to excite the bandpass filter with its edge rise and fall times (e.g. use a Hamming window http://en.wikipedia.org/wiki/Window_function#Windowing )

The "signal size measurer" after the filter and window might work best as a quasi-peak, or RMS detector. (This is the part that comes up with a single number that reflects the size of the signal)
http://en.wikipedia.org/wiki/Quasi-peak
Something up Joe's alley.

An even more sophisticated approach would use DSP that spots the characteristic "descending bell tone" of knock. (Slow knock down 20x and listen with headphones, very distinct)

To come up with a simplified version of the algorithms, and to test its effectiveness, one can start by examining and simulating it with a Math package (e.g. Matlab, Scilab, Mathcad) with a nice datalog of various knock and no knock signals, which includes crank angle information (e.g. CKP signal). I do this shite professionally...

Obviously, before embarking on such investigations, one should always check the current literature - patents, SAE papers, and the web:
https://www.google.com/search?ix=heb...ock+detection+