Glad it is working for more people!

This morning I logged (DMA, every spark event up to 6000 RPM) my commute, and there were a few times I could go WOT through the gears, and I got 2 episodes of knock sum of 1. Because the logging was fast, every time there was a new knock sum I could see the triggering event of knockadc crossing knockbase. So the slow (MUT) logs are missing out the peaks as we suspected.

So with MUT logging the best thing to do is to look for the new or increases in knocksum and assume knockbase has been breached there (either due to real or false knock).

I'll post the graphs later.

 

Hmm, cool findings.

 

RPM scaling on the right hand axis, all the rest on the left.

ks3_partthrottleboost: here you can see knocksums of 3 built up during part throttle boost (about 140kPa absolute, or 6 PSI)

spoolupknocksum: during spool up

4500rpmknocksum: AVC-R spools up quickly to 250kPa and then on towards 280kPa (it must have some sort of differential gain that kills it when boost is arriving fast). You hardly notice this in the car, the log is very fast. See the spikes that breach knockbase and result in knocksums. This is the best contender of all for real knock I think.

 

5.75 second full throttle pull in IIRC 3rd gear from 2200 to 7300 RPM.

Knocksum of 1 starting at 4688 RPM, switchover of gain at ~140 knockbase and 4812 RPM.

 

Very cool data, John. I would be interested to see how some logs like this look on cars known to have phantom knock., especially bad 36 count phantom knock.

Finally understanding the knock control is great, though. Even with the slower MUT logging, this can give us a good enough picture to see how close we are to the knockbase line, to see how safe our tune is or how much room for improvement there may be at certain RPM ranges for more timing, etc.

This is definitely a great breakthrough and we appreciate all of the effort that you have put in.


Eric

 

wow, yeah. Grahping knockbase would be very helpful. I know MUT is slower but what would be the MUT addresses to log both "knockfiltadc" and "knockbase"?

 

I know the latest version of EvoScan has those built in already, just check the box.

 

I wanted to figure out where the transitions occured for the Low, Mid, and High RPM tables on the knock Multipliers. So I set them to 5, 25, and 50 respectively. Below is the log from a simple rev up to 5.4K at idle. Looking at the steps in the knock_base you can see the transition points.

Looks like:
LOW = 0- 3800 RPM
MID = 3800 - 4800 RPM
HIGH = 4800 to redline.

I've got things adjusted and have been doing a little light cruising with NO Knock flashes from the CEL. Where my first day out with the car I got all sorts of knock CEL flashing on just easy driving - THIS STUFF IS AWESOME! Since I'm not really getting into high load cells and haven't been tuning on pulls I'll save the details until I'm fully tuned.

But I think that most engines with false knock should be able to raise the LOW multiplier and get rid of the 3000-3500 RPM problems. And maybe bump the MID a little to make a smooth transition.

 

Great. This solved all my problems.
Stock ecu soon will be better than any ems.

 

Are any of you guys pulling plugs to look for detonation? The reason I ask is that I know a bunch of people who have had "false" knock. You know, it wasn't real knock but when they increased the octane it went away LOL. If the knock is truly false due to abnormal background noise, then higher octane shouldn't do squat to reduce background noise. All higher octane would do is decrease detonation. In turn you have just found the problem, but some people just don't accept the fact that the knock system is doing what it is supposed to.

Also real knock will happen nearly every pull at nearly the same RPM. Whereas false knock is usually completely random.

I'm just sayin................

 

My testing has been done with Det Cans and microphones on the engine, there are ZERO signs/sounds of detonation. Not to mention my knock has shown up during lightly revving at idle under hardly any load and during no boost driving.

I'll check the plugs just for kicks and giggles though.

 

ive changed my plugs probably 5 times on this engine, the plugs always look perfect, except for an orange tinge after the torco lol.

and my car had the same false knock issue on 93oct, 93 + torco (104 oct), sunoco 100 oct, and sunoco 116 leaded

its definitely false.

although you bring a valid point up, everyone modifying there knock tables should make sure that they need to

 

Sweet, thanks for figuring out the transition points. Time to readjust the values.

 

This is by far the most informative area on this forum, or any other forum I have ever been on. Wow just Wow.

 

jcsbanks can you comment on those crispy details like MUT_C9, MUT_CB, MUT_CC,MUT_CD, MUT_CE and MUT_72?

I have lately read in service manual that there is some knock learning procedure after replacement of knock sensor. At least we know there is something like knock learning via MUT.

Do we know how actuator commands work (I mean, Hamish knows but is it public/open knowledge available?) We might find that learning stuff, if we know how it works.