Nothing needs to be done to benefit individual cylinder knock retard. It will only pull timing from the cylinder that is knocking while allowing the remaining cylinders to run at full timing. If only one cylinder is knocking more than the others, then you'll get 75% of the possible benefit just by using the individual cylinder knock retard. You'll only gain another 25% of the possible total benefit by fixing the issue with the cylinder that is knocking. Its actually an awesome feature. I'm still surprised that it can be done. I didn't think the ECU was fast enough to do it.

 

What about swapping spark plugs along with injectors? Is it possible 2 and 3 cylinders get taken out more because they run inherently hotter being in the middle of the block or maybe the airflow to those cylinders is not as good? Any way just random thoughts I guess.


PS Nice work as always Banks and thanks for sharing

 

That has always been the theory, that the middle two run hotter. However what is odd is how you would think it would be the cylinder furthest from the water pump, but on the case of the DSM its the middle cylinder closest to the water pump. On a DSM the cylinder closest to the knock sensor is the one that lets go first usually. Mitsu knew that it was the most vulnerable, and they did the best to protect it.

That brings me to another point. Is the detonation wave propagation through the block even across all the cylinders? IE: does the cylinder furthest from the knock sensor sound the same as the one closest to the knock sensor when both ping the same amount? I know I'm getting nit picky, but I like to ask the loaded questions.

I do like individual cylinder retard, I'd like to increase the amount of timing pull if I'm only pulling timing on a single cylinder.

 

As with a sound wave travelling through air, a knock vibration will get attenuated as it travels through the iron block. It is interesting to consider that there's no accounting for that in the knock code. Perhaps the background noise is way above any difference in knock signal attenuation between cylinders.

 

I agree in priciple but think that sound would not get attenuatd as much as in air. Isnt the speed of sound faster in a solid like metal and attenuated less?

I agree that the background noise of combustion , valves opening and pistons going up and down would be very high. Maybe though the frequency and wave patterns suggested for the knock are what Mitsu has identified and is what is quantified as JSCbanks has suggested above...so identified by CAs and filters app-lied, if a knock then added to KS.

What jscbanks has done is trluy great here. Specific knock on specific cylinder. If a tiny snip of code will individually tune for timing / cylinder then we have a huge improvement in tuning to the safe edge. Couple that with individual injector control and there will be standalone control on the stock ecu

Wait..doesnt the ECU fire the plugs in pairs ?? Will this be a problem ?

 

Holy hell I didn't realize it was also controlling ignition timing on a per cylinder basis. I thought it was just to log the knock the car was seeing. This is truly amazing.

 

WRC, the wasted spark can still be wasted, it will be wasted later than "normal"

 

I don't think you are really going to get 75% of the benefit without subjecting one cylinder to a lot of abuse. If you tune the maps so that the 1 cylinder is always getting knock retard, that cylinder is still going to fail in the long run. Being able to pull a little timing out of that one cylinder separate from the knock control is still going to be beneficial.

Does the knock sum still look at all 4 channels to count up? Or is there a per cylinder knock sum with this change?

 

I'm speaking from the situation of a car that does not have any serious knock issues. If one cylinder is creating 3-4 knock counts while the others are 0 knock counts, I wouldn't worry about that. If it were to turn out that one cylinder is consistently creating 10+ knock counts while the others are 0 knock counts, then it would be a good idea to find the cause of the high knock counts in the one cylinder.

 

I think we underestimate the abilities of an interrupt driven microcontroller and its ability to respond in real time, they absolutely eat this sort of thing. It has considerably more processing power than some Bosch ECUs that run individual cylinder knock control, and way more than the DSM whilst doing little more. Whilst the face that we are presented with of slow 8 bit logging and coarse 8 bit maps looks clonky, under the hood it is quite different. Then length of the interrupts that are processed on each CAS pulse (both rising and falling) is astonishing, and is really the last bit to really understand. Here I'm scratching the surface. It does not look like individual cylinder fuelling will be anything like as easy, appealing as it might be to have it.

I would prefer to place a trim (perhaps with a load threshold, not implemented yet) on individual cylinders as I showed in the code on the last page rather than change the knock control. This has the benefit of reducing possible bugs, simplifying code and not breaking something that works. To rewrite the knock control for individual cylinder would need **lots** more code, a new set of variables, work on the decay properties etc. I do think if the cylinders are balanced then it would be good to apply correction across the cylinders as this will mean you only knock once rather than four times.

03whitegsr, knock sum still counts up for all four cylinders. I propose to leave it alone.

 

I've collected a variety of logs with each cylinder retarded 3 degrees, and two with no retard. It is difficult to interpret with lift off knock on this lower octane fuel, and some randomness about it. However, I think I've narrowed it down to knock3 being reduced by the ignition calculation that happens when the cylindex is also on 3 or possibly 0. Needs more testing to confirm.

 

Further logs suggest that the non-lift off knock is reduced in knock3 by using 1 degree of retard when cylindex is 3, but I'm not certain yet. Will need to test further with usual fuel which doesn't do this lift off knock which confuses things.

 

I added 5% methanol which has got rid of most of the lift off knock so it behaves more like my usual fuel, but still has knock sums under load. With 1 degree off the timing when cylindex is 3, it does seem to equalize things nicely. I'll leave it like that and test further.

 

Are you running individual cylinder knock retard right now?

 

I'm running individual cylinder retard, but it is from a lookup table in RAM that I can alter, the original knock control system is working as usual.

So if we have zero knock sum and 10 degrees in the ignition map, we have 10,10,10,9,10,10,10,9 etc. It runs smoothly, you wouldn't know. I'm not going to claim that it feels smoother under power, it feels just the same. On my usual 99 RON fuel I would be dropping the boost back by 1.5 PSI at this time of year, but I'm not going to do that for now.