Knock sampling Settings (Knock Chan) for 2 Knock Sensors?

[black]

I went trough the docu but there is still a lack of information

5 cylinder S2 engines are working with 2 knock sensors

Sensor 1 - cylinder 1,2,3 and sensor 2 - cylinder 4,5

So what will be the right channel number in Knock Sampling Settings (Knock Chan) for 2 Knock Sensors?

From Defaults I guess Enabled_Alternating_Ch ??  So it changes to 85??

And when I Change the channel rotation order from 1 2 1 2 1 to 1 1 1 2 2 (like it should be for S2) the channel number  in Knock Sampling Settings changes to 29

My result: the channel number is some binary code... 

so 85 = 01010101 and 1 = channel 2 and 0 = channel 1 aka 1 2 1 2 1 2 1 2

So 170 = 10101010 = 2 1 2 1 2 1 2 1

29 = 00011101 (because in knock channels we can only change the order of the first 5 cylinders (in my case) so the last 3 stays 101--> result = 29

So to get the right channel order for a 5 cylinder the following channel numbers should work:

00011111=31
00011101=29
00011110=30
00011001=25
00011000=24

best regards

[VEMS]

Hello Black,

There is a separate dialog for knock channel setup, see inputs->knock channels you can select the channel per ignition output line. Take care the channel selection is per index in the ignition output table in same order, do not look at the number in front of knock channel, but the actual line number -> like fist line knock channel corresponds with first line ignition outputs visual (and the cylinder attached to that in front of it there).

Best regards, Dave

[black]

Hello Black,

There is a separate dialog for knock channel setup, see inputs->knock channels you can select the channel per ignition output line. Take care the channel selection is per index in the ignition output table in same order, do not look at the number in front of knock channel, but the actual line number -> like fist line knock channel corresponds with first line ignition outputs visual (and the cylinder attached to that in front of it there).

Best regards, Dave

Hi Dave, yes I know the knock channel Setup - but I wanted to understand the number calculations ;-)

But you are right, I didnt knew the numbers correct..

so from your description:

5 cylinder engine from top to down on ignition Outputs visual = 1 2 4 5 3

So it should be

0 = 1
1 = 2
2 = 4
3 = 5
4 = 3

aka 11221
aka 00110
aka 53

[VEMS]

Hello Black,

Using your picture as reference:

real cyl nr	knock channel nr
1	4
2	3
4	2
5	1
3	0

Legend:

Real cyl nr = real cylinder number aka the number before the output in ignition outputs visual
Knock channel nr = the imaginary number before the line in the knock channel selection dialog

I am not sure where the knock sensors are located on the 5cyl AAN/3B (would have to look up) but assuming there are two on the engine i suspect cyl 123 and cyl 45 or cyl 12 and cyl 345 are grouped by sensor. Therefore either (from top to bottom in knock channels) 11221 or 22112 would make sense to me. Your current mapping in knock channels (from picture) seems to be shifted, please correct.

There is also a simple way of checking if the right sensor is selected for each cylinder, when the engine is running (at idle) with all gains set to 1. adjust the selected sensor per cylinder, the right one is the one that produces the highest value in view->knock colums gauge (number above the column is cylinder number).

I hope this clears things up; If there are any further questions, feel free to ask.

Best regards, Dave

Edit: Black, i removed your message in reference to my partial post, pressed the post instead of preview button while editing, oops ...

[black]

I am not sure where the knock sensors are located on the 5cyl AAN/3B (would have to look up) but assuming there are two on the engine i suspect cyl 123 and cyl 45 or cyl 12 and cyl 345 are grouped by sensor. Therefore either (from top to bottom in knock channels) 11221 or 22112 would make sense to me. Your current mapping in knock channels (from picture) seems to be shifted, please correct.

Best regards, Dave

Thanks Dave, I will test...

PS as I wrote in my first post: "Sensor 1 - cylinder 1,2,3 and sensor 2 - cylinder 4,5"

best regards

[black]

here is the result...  VEMSTune values shows knock but I know the engine is not knocking (E85)     ??

[MWfire]

You can't use knock sensor above 4500-5000rpm. To much noise from cams, valves ...

[black]

You can't use knock sensor above 4500-5000rpm. To much noise from cams, valves ...

  ok?

the black curve ADC RAW CH0 is 0-5V Output from Phormula Knock Analyzer...  no noise...  max 1.4V

[MWfire]

You can't use knock sensor above 4500-5000rpm. To much noise from cams, valves ...

  ok?

the black curve ADC RAW CH0 is 0-5V Output from Phormula Knock Analyzer...  no noise...  max 1.4V

Only 100% way to know if engines knockes is cylinder pressure sensor or ion current or dyno. Sometime you can see real knock on ind cyl power.

Mapping with knock sensor is good way to kill engine.

You can't trust any box that use standard knock sensor in midle and high rpms.
Because of that all stock car tests fuel at low rpm(usually at 2000-2500rpm) to determinate fuel octane rating. The use map for that fuel made on dyno with pressure sensor.

[Denmark]

Knock control Can be used full rpm , your gain setting might be to high ,
On the subaru engines i tune with forged pistons , i Can use the standard knocksetup setting and dont lower the first setting that the manuel says from 300 to around 150 that the manuel States


 ,Knock Integrator Time Constant (uS): Works as an amplifier, we will use this together with channel gain setting to get our reference measurement setting correct. Larger time constant results in lower knock reading. For the best accuracy try to keep time constant between 100-150uS and use lower gain settings



If i do that the knocksensor show knock when there is No knock

Subaru standard ecu optimise the ign table from the knocksensor to 6000 rpm in standard form

[VEMS]

Hello Black,

here is the result...  VEMSTune values shows knock but I know the engine is not knocking (E85)     ??

Did you follow the full calibration procedure from: http://vems.hu/vt/help/v3/v3_knock_control.html up to (and including) knock detection setup ?
From your log i would say the base reading at idle must be quite high, either gain per cylinder or integrator time constant would need adjustment (or both). If the first frequency picks up a lot of valve-train noise you could opt for the first of second harmonic. e.g. if base freq is 2kHz, 2nd harmonic is 4kHz etc. (but follow the calibration procedure first).

Make sure to enable reatime data option: individual knock readings for more accurate knock data (especially needed during knock integrator and gain setup per cylinder), but keep in mind this disables individual power (ADC channels are still logged).

Yes picking up knock with ion current or cylinder pressure sensor is more accurate, but using the acoustic method is very accurate as well (almost all the oems use it), of course it needs to be tuned properly to function properly (its not a magic box that knows all engines).

Best regards, Dave

[gunni]

You can't use knock sensor above 4500-5000rpm. To much noise from cams, valves ...

  ok?

the black curve ADC RAW CH0 is 0-5V Output from Phormula Knock Analyzer...  no noise...  max 1.4V

Only 100% way to know if engines knockes is cylinder pressure sensor or ion current or dyno. Sometime you can see real knock on ind cyl power.

Mapping with knock sensor is good way to kill engine.

You can't trust any box that use standard knock sensor in midle and high rpms.
Because of that all stock car tests fuel at low rpm(usually at 2000-2500rpm) to determinate fuel octane rating. The use map for that fuel made on dyno with pressure sensor.

This is incorrect, just last week I setup a full knock control on this engine


Before setting up knock control we did some spot checking with and without knock at 5000rpm, 6000rpm and 7000rpm while recording the knock sensor with a datalogger at 250khz
The data is then sampled to create a db level spectrum at each time interval broken down into frequencies. We noted the obviously prominent frequencies that picked up the knock noise and pick them for the ECU

In the ECU I then did alot of running to get the engine noise table filled in, I won´t go into all the details of the knock control but we use a MQ12 from Pectel and its extremely comprehensive.
I can assure you that knock is very very easily detectable at any rpm when knock filtering and sampling is done correctly. Every race engine in GT racing and higher has knock control enabled at all rpm's. You wouldn´t be able to make the power they make without tickling the boundaries of knock.

[black]

thanks mates...

pressure sensor system is well known but not the question here...

I'm calibrating Bosch Motronic for years now, using the OEM knock sensors and OEM knock control.
The OEM Knock control works superb over the full RPM range and after doing a lot of Motronic cars up to 700hp know I had never one single damage with that.

back to topic

I added two picture of my testing. I did not the full Setup, I simply used the default values to see what generally happens. One on boost and one on idle.
I did not lower the Knock Integrator Time Constant (uS)
"Make sure to enable reatime data option: individual knock readings for more accurate knock data" - yes I did that before

What I still dont understand is the cylinder sequence. I used the sequence Dave told me (A) but VEMS is doing a different sequence from that (B) 





plus steady 2500RPM:

[VEMS]

Hello Black,

What I still dont understand is the cylinder sequence. I used the sequence Dave told me (A) but VEMS is doing a different sequence from that (B) 

No need to get confused, it is exactly the same sequence; just the reporting in colums gauge order is shifted by one cylinder. The cylinder number is at the top, for your own piece of mind  its very easy to see when changing gain (big step up or down) for that cylinder, you will see the corresponding bar (and value) change.

I also noticed you have set a substandard window length of 52 degrees -> 40 degree is the standard setting, 52 degrees is way to long, please adjust. Also important to note that integrator time constant is only sent to the knock chip on ecu bootup, when changing this value power cycle the ecu (burn, key-off, restart engine) so it can take effect.

Best regards, Dave

[black]

Hello Black,

What I still dont understand is the cylinder sequence. I used the sequence Dave told me (A) but VEMS is doing a different sequence from that (B) 

No need to get confused, it is exactly the same sequence; just the reporting in colums gauge order is shifted by one cylinder. The cylinder number is at the top, for your own piece of mind  its very easy to see when changing gain (big step up or down) for that cylinder, you will see the corresponding bar (and value) change.

I also noticed you have set a substandard window length of 52 degrees -> 40 degree is the standard setting, 52 degrees is way to long, please adjust. Also important to note that integrator time constant is only sent to the knock chip on ecu bootup, when changing this value power cycle the ecu (burn, key-off, restart engine) so it can take effect.

Best regards, Dave

Mh... I don't see only a shift...  cylinder numbers in knock channels do not match to the numbers in the report numbers. So the knock sensor my not match to the right cylinder. But to get that clear, I will simply disconnect one knock sensor an see what cylinders are effected. Window length will be corrected.

thanks