Software knock processing

[Hmelik]

https://rusefi.com/online/view.php?log=1217

[neongreen]

I'd really like to use knock logic in my project, but I have Frankenso. I understand it's currently only on some boards (Proteus and hellen?). Is there a reason we can't activate the logic on other boards? I have no issue repinning the harness to put knock on ADC03.

[AndreyB]

it's currently only on some boards (Proteus and hellen?)

Frankenso has totally different electronics for knock, the not-software but ASIC approach which did not work out for us.

This thread saying "Software knock" means "not ASIC like on Frankenso"

[neongreen]

it's currently only on some boards (Proteus and hellen?)

Frankenso has totally different electronics for knock, the not-software but ASIC approach which did not work out for us.

This thread saying "Software knock" means "not ASIC like on Frankenso"

Understood, but what I'm asking is why can software knock not be used on Frankenso? It has an ADC03.

I also realize that the frankenso doesn't have a knock amp, which is why I copied the proteus amp and had it built as an add on board.



There is no access to the software knock in the F407 Discovery firmware version. Could that be activated?

[AndreyB]

Ohhh now I understand the ask.

It could be possible, it could be pretty small effort but I am very overloaded on software side I can really use help from other software people.

https://github.com/rusefi/rusefi/issues/5254

First like of progress https://github.com/rusefi/rusefi/commit/75d14a0cea44c4aa325dd61058353eb8e85946fe

[neongreen]

Thanks Andrey. I wish I were more help on software, but looking at the code I feel like a six year old attempting to figure out how an Airbus A320 works.

[AndreyB]

vag0dr04er is teasing me so let me tease everyone else

WhatsApp Image 2024-03-23 at 06.08.18_348ef01d.jpg (347.64 KiB) Viewed 12455 times

[AndreyB]

[tmbryhn]

Great demo video

I've been playing around with knock in RusEFi for some time, and come across a limitation in flexibility that sets it apart from eg. MS3 that uses the dedicated TI knock DSP chip.

Particularly on setups that utilises one knock sensor, there's sometimes a non-trivial delta in peak noise level generated by the individual cylinders. On systems using a global knock threshold (like RusEFi and MS3), this can be problematic to fine tune unless a sound gain can be defined on a per-cylinder basis. MS3 has this feature as a part of the DSP stage which allows for a more or less equal amplitude between the different cylinders.

As I understand it, RusEFi operates with a fixed gain analog filter input feeding directly into the dedicated knock DSP ADC. To implement a variable gain feature; could instead the digital value generated by the ADC be scaled by a variable factor associated with each cylinder as a type of "software gain" feature?

[AndreyB]

https://github.com/rusefi/rusefi/issues/6413

[AndreyB]

Alexey @vag0dr04er has contributed _something_ - right now it's available in Proteus firmware, let me see if I can also enable it for 4chan and uaefi

https://github.com/rusefi/rusefi/pull/6467

image.png (753.07 KiB) Viewed 10532 times

[vag0dr04er]

While studying detonation on a real car, I found some discrepancy between theory and practice.

I recorded real detonation on a sound card and analyzed it with spectral analysis.
1. light detonation
2. strong detonation

photo_2024-09-08_15-04-47.jpg (207.46 KiB) Viewed 5973 times

VID-20240902-WA0001.mp4

(7.95 MiB) Downloaded 169 times

Given:
4g63 turbo, garrett2871, boost 1 bar, fuel euro RON 95.
Cylinder diameter 85 mm.

Using a simple but widely used formula for calculating detonation frequency, we calculate the estimated frequency:

f (Hz) = 900 / (PI * r)
6744 Hz = 900 / (3.14 * 0.0425)

The estimated detonation frequency is expected at 6.7 kHz, but in a real car it is best heard at about 5.6 kHz.

The worst thing is that detonation is least audible at the estimated frequency!

knock_real_freq.jpg (292.63 KiB) Viewed 5973 times

Perhaps this is specific to this car.
Perhaps the calculation formula is too simple or incorrect, I have not found another.
This requires research.

I'll try to set up software detonation detection better with this reference.

[vag0dr04er]

rusefi plugin knock analyzer

2024-11-05-172216_1920x1200_scrot.png (219.56 KiB) Viewed 3282 times

I transferred the detonation spectrum analysis from the console to TunerStudio.
An analysis of each sensor or cylinder was also added.

The plugin is available for bundle proteus f7, f4. Technically, it will compile on uaefi and Alphax-4chan, but the bundle is not available yet.

knock_parameters.png (106.85 KiB) Viewed 3282 times

1. Enabling analysis, sensitivity and
display of a signal filtered at a given frequency.

2. As well as correction of fuel for detonation.

3. Suppression of all corrections by knock (angle retard, fuel) when theTPS is released. (When changing gears, the engine may make noise like detonation or just when coasting)

[mk e]

Given:
4g63 turbo, garrett2871, boost 1 bar, fuel euro RON 95.
Cylinder diameter 85 mm.

Using a simple but widely used formula for calculating detonation frequency, we calculate the estimated frequency:

f (Hz) = 900 / (PI * r)
6744 Hz = 900 / (3.14 * 0.0425)

The estimated detonation frequency is expected at 6.7 kHz, but in a real car it is best heard at about 5.6 kHz.

The worst thing is that detonation is least audible at the estimated frequency!


Perhaps this is specific to this car.
Perhaps the calculation formula is too simple or incorrect, I have not found another.
This requires research.

The formula really only applies to engines with open combustion chamber heads found on older engine designs. Simple shape, simple math.

Newer combustion chambers (and pistons) are specifically designed to prevent and disrupt detonation. Also any way to reduce engine noise is a concern so castings and mounts designed to damp the loudest frequencies like the primary detonation frequency) which makes the math to predict where you'll get the best indication of detonation a starting point at best so A+ for measuring it.

Forgot, the harmonics tend to be more predictable so you might want to take a listen around 13.5khz

[kneelo]

I have been slowly working my way through the back catalogue of HPA podcasts and came across this interview with a guy from plex tuning. It almost 2 hrs long but theres a lot of really interesting info regarding DSP strategies to detect knock to target frequencies and how to validate on the car.

https://www.hpacademy.com/blog/081-properly-understand-knock-and-become-a-better-tuner-podcast/