1997 F-Body ECM

[Tom H]

I would like to gain an understanding of the spark knock system used in the LT1 and other engines of the time. A while back I purchased a few GM ECMs just to have a look at common parts and do testing that might destroy things. One point of interest is the spark knock filter. Here is a picture of the filter with the elastomeric goo removed.

IMG_1757.jpg

The module contains a bunch of laser trimmed resistors, capacitors and two chips. The first chip is easy to identify, it is a Philips Semiconductor NE532D. This is a very common dual op-amp. The second part is not so easy. It is a "flip chip" design with no package. The part is mounted directly to the board using solder bumps. This can only be done where the board expansion is similar to the semi-conductor expansion. This is the reason it is mounted on a ceramic hybrid.

I started looking at this after reading the thread at: http://www.gearhead-efi.com/Fuel-Inj...-knock-sensors
This encouraged me to look at the system a bit and ask for some help.

--> If anyone knows anything about the flip chip part OR earlier systems that use a similar part in the detection of knock please post what you know. I can only guess that this part is involved in the filtering of signal from the sensor
--> If anyone with an LT1 could give me an accurate measure of voltage on the blue connector pin 13 engine not running/key on. I know there are different modules used (at least two) so if possible post the number of your filter module

The second ask is because I think this voltage is used to level slice the filtered knock signal. The voltage is created using a divider composed of two laser trimmed resistors. These resistors are the ones between pins 2-3 and between 3-4. I have yet to set things up and start measuring signals. To be honest, the whole system is a bit confusing. In any case this is where I am starting out.

-Tom

[Tom H]

I should have mentioned... the knock module from a v6 of the time has exactly the same pattern of resistors/caps/chips. The hybrid module has different laser marks and I think the only difference is in the trimming. Some LT1s with factory installed roller rockers had de-tuned modules to reject the mechanical noise. Quite interested in those modules, if you have one and are willing to do some measurement it will be interesting.

The voltage on Blue 13 (which I believe is the slice level) for a stock LT should be about 2.5v. The V6 module I have sets it at about 3v.

-Tom

[kur4o]

Blue 13 on 94-95 pcm is not used but it is connected internally to the knock filter. I was thinking it could be for second knock sensor but according to my measurements it has 1.3kohm pull up resistor and there is no voltage[d13 to ground]. There is 9.45volts if measured: d13 to b+, So it should be some form of grounded that can accept either some voltage or can be used to control something.

Some of the knock modules I have looked at don`t have the extra philips chip.

I did mentioned that at high rpm the airflow measured from VE tables is crippled at some rpms above threshold. I suspect maf frequency can also be affected by this saving cpu time.

It is because at 96-97 pcm the airflow is measured from both MAF and VE and at the end the pcm decides which value to use. On earlier pcm the ve is measured only if the maf fails.

[Tom H]

Blue 13 on 94-95 pcm is not used but it is connected internally to the knock filter. I was thinking it could be for second knock sensor but according to my measurements it has 1.3kohm pull up resistor and there is no voltage[d13 to ground]. There is 9.45volts if measured: d13 to b+, So it should be some form of grounded that can accept either some voltage or can be used to control something.

Some of the knock modules I have looked at don`t have the extra philips chip.

I

The knock modules I have both have a voltage divider between the internal +5V and ground. LT1 & V6 have different trims on the resistors. The output reference analog is used [edit not internally] externally to the knock filter module connected on pin 3 (of my numbering...left to right with connectors up). There is a 100 ohm buffer resistor and a cap on the line before reaching the blue connector pin 13. There is some possibility that this voltage is only an identifier. That is, manufacturing wants to know if the correct module is installed without opening the unit.

I have traced the configuration of the op amp... one inverting amp one non-inverting amp. Still no clue regarding the 18pin filp chip part. Starting to look through patents to see if anything looks to match up...

-Tom

[kur4o]

There is a chance that when the measurements were taken the knock module was disconnected from the board explaining the lack of voltage.

One way of insight how it works is to wire a knock sensor and simulate knock while measuring inputs and outputs of the knock module.

[NomakeWan]

The knock modules I have both have a voltage divider between the internal +5V and ground. LT1 & V6 have different trims on the resistors. The output reference analog is used [edit not internally] externally to the knock filter module connected on pin 3 (of my numbering...left to right with connectors up). There is a 100 ohm buffer resistor and a cap on the line before reaching the blue connector pin 13. There is some possibility that this voltage is only an identifier. That is, manufacturing wants to know if the correct module is installed without opening the unit.

I have traced the configuration of the op amp... one inverting amp one non-inverting amp. Still no clue regarding the 18pin filp chip part. Starting to look through patents to see if anything looks to match up...

-Tom

This is likely correct. 69427, one of the engineers of this system, recalls from memory that the 5-pin knock modules from the L98 engines had five pins but only four were used--power, ground, knock sensor, output. The fifth pin was an ID pin to tell which module was which at the factory. This jives with what you just said. In addition, he stated that the PCB layout of the 5-pin module was almost identical to the modules used in the later $DA2 and $DA3 MEMCAL-based cars, and that those were likewise almost identical to the modules used in the LT1 cars. I do have a dismantled knock filter from an LT1 on my desk (though you did a much better job with yours than I mine), and I do have an unopened 5-pin knock module from an L98 on my workbench. However, unlike the LT1 module, it is not obvious how to pry it open without destroying the internals. If I can figure out how to open it, it may give insight into which pins do what on the later module, since we already know the correct pinout for the 5-pin L98 one.

In addition, here is a spreadsheet listing certain values from various knock modules GM used. 69427 was at least able to confirm that it was legitimate, and that he likely knew which engineer had leaked the data. He would not comment on the accuracy of the data, however.

UPDATE: While it would appear that reverse-engineering the 5-pin module is a lesson in futility (it appears to be sealed such that you would have to slice it open, at which point you would likely find the internals potted with goo, making visual confirmation of the circuitry nigh impossible), I do have a partial reverse-engineer of the $DA3 MEMCAL-based module, which appears to be very close to the $EE module. I will attach the photo of that module with the pins I was able to identify to this post.

I was able to ID the pins thanks to "How to connnect ESC module inplace of a memcal knock filter" and a memcal layout I found. The former document says that MEMCAL pin 29 is the knock sensor input into the knock module, and MEMCAL pin 32 is the knock filter output to the ECM. The latter document not only shows where those pins are physically located, but additionally shows that pin 31 and 62 are both grounds. This leaves only four 'unknown' pins. We know one must be 12V, and likely one is that 'calibration' value. UPDATE 2: The middle two pins, 47 and 48, are marked on the 1227730 schematic as N/C. Thus we can assume that these pins are not actually connected to the ECM. So in fact that leaves pin 30 and pin 61. Pin 30 does not have an external connection, while pin 61 does. This could potentially suggest that pin 30 is VCC, and pin 61 is calibration. This is further backed up by the schematic for the 1227730, which shows CAL 30 being connected to VIGN through a 30ohm 1/4W resistor.

I have updated the attached image accordingly.

In any case, perhaps this can help in some way to figure out the layout of the LT1/LT4 module as well.

[Tom H]

In any case, perhaps this can help in some way to figure out the layout of the LT1/LT4 module as well.

Thank you! This is a great help. -Tom

[Tom H]

Some more work on the knock module...

KNOCK_MODULE.jpg
Clean picture of knock module

KNOCK_REFDES_C.jpg
Reference designator for capacitors

KNOCK_REFDES_R.jpg
Reference designator for resistors

Schematic of 8pin knock module is in the attached PDF.

Kur4o: Is it possible that the modules that don`t have the extra philips chip have a flip chip varient? I see other semiconductors mounted like this in the ESC modules. Can you photo the module?

Working to complete analysis of the module now. Almost for sure the two resistors I wrote about earlier are just for module ID.

-Tom

[kur4o]

Here are some pics. The module with 8 pin chip is from 2000 blackbox pcm, the other one is probably from 94 v6 pcm.

[Tom H]

Here are some pics. The module with 8 pin chip is from 2000 blackbox pcm, the other one is probably from 94 v6 pcm.

Thanks! One picture confirms the layout of a resistor I was un-sure of, on the underside of the op-amp.

Are any of these modules scrap? The filter chip seems the same in all. Should one be a scrap module, we could heat the ceramic substrate and remove the filter chip. This would let us look at the "ball side" of the chip. Perhaps some microscopic markings. Just like the TPU we need to find out details and with luck a part number. I don't think the filter chip is a switch capacitor filter... just not sure what it is. Hmmm

-Tom

[kur4o]

This one is from stock 94 lt1 pcm fbody. It also misses the 8 pin chip. I suspect they switched to 8pin chip in 96, might be obd2 related. lt4 knock module have the extra chip too. Could the 8 pin philips chip can be some preamp or a filter.

I doubt there will be any markings. Some schematics of the board + some knock simulations and minitor the output with a scope. The rc network could be fine tuned for a specific frequence of the engine knock. A real art of science.

The main chip is 18pin 3x6 layout[ [It could be also 4x6 or 5x6 or 6x6]. It is stripped down to save space. ??some philips chip with similar layout or custom delco unit??

What is the signal that goes from knock module to tpu and how it is processed by the tpu. I guess something like the maf but it measures some delta knock events for predefined time. It also have some dtc set but it is not clear how they are processed either.

[NomakeWan]

>What is the signal that goes from the module to TPU
The knock filter pulls the signal line low when it believes that 'real knock' is occuring and pulls it high when the knock event ends. This has been true since at least 1985 and did not change up through at least 1997. My assumption is the same as yours--that the TPU looks at delta knock events over a fixed time period and then decides whether that's enough to warrant retarding timing.

The DTC for the 96-97 is most curious for me. It's much more complex than the 94-95 which relied on the knock sensor itself to have a specific resistance value. That's no longer the case for the 96-97 so I'm incredibly curious as to how that works. Not as curious as knowing the full pinout for the knock module, but you know, pretty close. :)

Great work so far!

[Tom H]

Thanks for the input, there is still a lot to uncover in this design.

I looked at the power pin (pin 4) it seems to be driven from R520 (see my Tside schematic). There are a couple of zener diodes that form a very basic regulator at 10V. This type of supply is typically used for analog because it is quiet. My thinking is that this supply may be used for both the analog knock section and the knock module. This will also differ for 94/5 as the board layout is different here.

I need to get hold of a frequency generator or hook my pc's sound output up somehow. Any suggestions? I understand that knock is in the 5-15khz range depending on the cylinder size and so on. I need first to get a setup where I can generate knock and regular running engine noise. Perhaps I did read somewhere that the system looks for "non-knock" noise to verify that the sensor is in place. I would have expected the knock module ID to go to the a/d such that the system could verify the module was plugged and the right one. So far, I can't see this

-Tom

[Tom H]

PS: Interested to know what voltage the '94/5 provides on pin 4. Ground is pin 6

[NomakeWan]

I can't get you that reading right this second, but this weekend I need to tear into the car again, so I should be able to get that reading for you at that time assuming no one beats me to it.