Knock module for MS3 (beta test)

[Reverant]

http://e2e.ti.com/support/applications/ ... spx#553334

Post #4 by a TI employee:

"The TPIC8101 only works with the resonant type knock sensor. We do not offer a device that works with the non-resonant type."

[gross polluter]

http://e2e.ti.com/support/applications/automotive/f/26/t/118763.aspx#553334

Post #4 by a TI employee:

"The TPIC8101 only works with the resonant type knock sensor. We do not offer a device that works with the non-resonant type."

Seems to be an error in nomenclature. If it was truly the case that the TPIC8101 only worked with resonant type sensors then there would be no need for the device to have a bandpass filter.

[GintsK]

TI page says TPIC8101 "drop-in" for HIP9011 http://e2e.ti.com/support/applications/ ... 84859.aspx
I knew OEM ECU where HIP9011 is used together with "wide band" sensor.

And what else is strange: in their training document scopeshots are not accordant. In real life both sensors gives signal all the time and difference is visible (except amplitude) just in spectral analysis. Flat response scopeshoot comes from picoscope webpage where signal is gotten by clicking engine block with wrench. It does not comes from operating engine!

I suppose they simply do not want support designs with flat-response sensors because of higher complexity.

TIP's band pass filter is not so "sharp". Device still "sees" side frequencies very well. So main software should be very sophisticated to catch just knock as knock and noise as noise.

Gints

[savagerocco]

I would doubt that James would even be releasing it if there weren't some kind of success with this. Matt hasn't said that there had been any issues with the units they have tested. I would hazard a guess that this will work with maybe a little tweaking. That being said, I am sure that we will overcome.
According to the engineers, swapping an AC system from r12 to r134 was theoretically impossible, new lines, bigger condesor, blah blah...... Turns out they were wrong.... I guess argue all you want, but there are more than one way to skin a cat....How about a little trust in James, not like he is the habbit of letting us down.

[dontz125]

Electrically the card would work with MS2, but then you'd have so little I/O left it would be pointless.

JBPerf IO/Extender?

There's also plenty of firmware required to support the comms and windowing. I have no desire to support this on MS2.

Separate topic, but ... le sigh ...

[racingmini_mtl]

Electrically the card would work with MS2, but then you'd have so little I/O left it would be pointless.

JBPerf IO/Extender?

The problem is that this would take the spark outputs and you can't move those to a CAN device.

What could be done is for James to have another board with a CPU to handle all the code needed for the SPI comm and possibly using CAN for the knock data. This would then require a knock window output on the MS2. The code for configuring parameters and dealing with knock would need to be split between the MS2 and the new CPU. It would likely be possible to reuse some or most of the MS3 code but that would still not be trivial.

Jean

[Dookie454]

TI page says TPIC8101 "drop-in" for HIP9011 http://e2e.ti.com/support/applications/ ... 84859.aspx
Flat response scopeshoot comes from picoscope webpage where signal is gotten by clicking engine block with wrench. It does not comes from operating engine!

I suppose they simply do not want support designs with flat-response sensors because of higher complexity.

For what it's worth, I remember tapping my block with a wrench and triggering knock. 1995 + 1996 GM LT1 OBD 1 and 2. Knock sensor in bottom right side of iron block.

My setup is noisy, ATI D1SC and some noisy, headers, lifters/rockers/pushrods, and only real knock sets it off.

[masterx81]

TI page says TPIC8101 "drop-in" for HIP9011 http://e2e.ti.com/support/applications/ ... 84859.aspx
I knew OEM ECU where HIP9011 is used together with "wide band" sensor.

And what else is strange: in their training document scopeshots are not accordant. In real life both sensors gives signal all the time and difference is visible (except amplitude) just in spectral analysis. Flat response scopeshoot comes from picoscope webpage where signal is gotten by clicking engine block with wrench. It does not comes from operating engine!

I suppose they simply do not want support designs with flat-response sensors because of higher complexity.

TIP's band pass filter is not so "sharp". Device still "sees" side frequencies very well. So main software should be very sophisticated to catch just knock as knock and noise as noise.

Gints

it's a 3rd order band pass filter. Used with windowing and with the table un change sensibility with rpm. After a good calibration must be quite accurate. Plus, the freq of det change a bit with load (due to different in chamber temp, that change the speed of sound), so a little bit of margin is useful

[jsmcortina]

Electrically the card would work with MS2, but then you'd have so little I/O left it would be pointless.

JBPerf IO/Extender?

The problem is that this would take the spark outputs and you can't move those to a CAN device.

Absolutely, but also, the present code is sending data to the knock chip before each spark event and the window signal is synchronised to engine position. That's a lot of information to attempt to relay to a second CPU. It would be necessary to us a minimum of two pins (even if the device was CAN connected) plus develop a new CPU board and write new firmware on both devices. That's a lot of work when the "fix" already exists with the MS3 card.

James

[S.Bretz]

I need to stop by this section more often. Good show on this one James.

$60 for something like this seems very reasonable; even if you had run off a large batch.

I can't really tell from the pics, but is the card supported on the other side opposite of the dip-40 pins on the ms3 card? If not supported on the other side; how stable is it? I would worry about it shaking around from vibration ....then again the Mapdaddy is just soldered in and held by pins too.

[jsmcortina]

There are two pins on the other side to solder on for a rigid mounting. They are mentioned (but not shown) in the manual.

James

[gross polluter]

Received my board today. Hopefully this week I can start providing some feedback.

[masterx81]

attaching the knock sensor at the mic input of the pc (with some diodes of protection) allow to record an induced knock and lately do an fft to find the right frequency of det of a particular engine. Use the right frequency is quite important...

[gross polluter]

attaching the knock sensor at the mic input of the pc (with some diodes of protection) allow to record an induced knock and lately do an fft to find the right frequency of det of a particular engine. Use the right frequency is quite important...

Based on the FFT analysis I've done on three engines (B230F, 2JZGTE, and LS2) I've found that the formula 900/(Pi * r) gets you within 200hz of the center frequency, at least on the three engines I've tested. James mentions this in the manual for the knock board, which I believe is sound advise based on my own research. Since you have to select pre-defined frequencies on the board itself it looks like "roundin up" to the highest matching frequency would work best.

One thing I'm curious to try is using the strategy that mitsubishi uses on the older DSM ECU's where the bandpass frequency was the second octave and not the primary knock frequency. On the iron block engines I've done FFT analysis on, the primary knock frequency gets noisy at higher engine RPM's where the second octave intensity only increases during knock conditions. The LS2 is the only aluminum block engine I've done analysis on and it does not exhibit a second octave resonance like the iron block engines I've tested.

Looking forward to getting the knock board installed on my ECU this week.

[HeinrichG]

Hello guys,

it's a really good thing.

It is possible to use them on V12 full sequential fuel and coil-on-plug ignition?

Regards,
Heinrich

[jsmcortina]

I don't think there are enough pins on the MS3 to be able to do that. Certainly the current design doesn't permit that.

James

[elaw]

Based on the FFT analysis I've done on three engines (B230F, 2JZGTE, and LS2) I've found that the formula 900/(Pi * r) gets you within 200hz of the center frequency, at least on the three engines I've tested.

Just a hint for anyone like me who has struggled with that formula: "r" has to be in meters! Millimeters, centimeters, and inches do _not_ work.

[nuvolarossa]

enter half bore in meters and you'll have the result in Hz, enter it in mm and you'll have it in kHz

Please report how good it works/engine noise/etc. and datalog of it in action when tuned

BTW it's two channel, so two sensors can be connected with one module, right?

[Matt Cramer]

enter half bore in meters and you'll have the result in Hz, enter it in mm and you'll have it in kHz

Please report how good it works/engine noise/etc. and datalog of it in action when tuned

BTW it's two channel, so two sensors can be connected with one module, right?

Right, one module can work for up to 2 sensors.

[gross polluter]

Not having much luck with mine. Seems like there needs to me more attenuation before the signal gets to the ic, at least with the factory 2JZ knock sensors. With the gain set to 1 the knock gauge always reads 101% with the gain at minimum the knock gauge bounces between 50% and 70% at idle and reads 101% off idle. Haven't had much time this week to tinker beyond installing the module and running it as-is.

Any advice on setting integrator time constant? Can't really figure that one out.