I ended last race season with holes in 2 pistons from detonation, as also shown by the pitting in the head.
I have backed off the timing using @curly 's suggestions in some other threads.
And I have added a Bosch knock sensor, and am trying to get it to work, again using some threads from on here.

But I don't really know what I am looking at in the data, is it noise, or is it knock?
8.7:1 Wiseco pistons, Garret 2560 turbo, 8 to 10 psi boost depending on creep from day to day.
Megasquirt MS3PnP for 2001 Miata with VVT.
Here is a recent track day log and tune, and below are some pics of the knock in"" signal (set for a single read right now, not all 4 pistons)
The knock sensor controls are set for it to do very little right now, (-3 deg, high threshold) until I understand if the signals observed are meaningful.

 

Turn monitor and control per cylinder on, you can compare then adjust individual cylinders.

Start a log with a warmed up engine, and slowly increase RPM while in neutral, all the way to redline. Change your "knock sensor settings" RPM break points to 2000rpm, then every ~500rpm until redline. No point in having less than 2000, because you've told it to monitor only above 2000. Plug in all the numbers you're seeing in the neutral log, adding just a bit for margin of error. Then every time you exceed this threshold, it'll pull 3 degrees. I'd increase max timing retard to 6.

Id love to see that same log with the above changes, dunno how often you're doing track days.

 

OK, will set to work on that. Many thanks for responding quickly.

 

How much timing were you running before? Have you been running on 93 octane (or your equivalent)?

 

91 or 93 depending on what is at the pump.
Shown are last season's timing values, basically a mashup of Trubokitty base map and DIY MS3 PnP base map (interpolated to my x-y axis ranges), choosing the more conservative of the two across the range.
Followed by a heat map of the changes made, negative numbers meaning a decrease from last year.
(You can see the current tune in the files attached to the original post, or the images in the same post, if you can read that small of print.)

 

Did some of my homework. Would not have thought of this baseline measurement Curly, thanks for the suggestion.
Tested 6.64, 6.94 and 13.8 Hz (or whaetver the exact choices are in Tuner Studio) as the frequency, settled on 6.94 as the most linear. The other two gave a big peak in response around 3500-4500 rpm. Calculation suggests best value is 6.8 Hz. Still not totally linear, as seen in the image below.
Raw data traces for knock sensor noise at various rpm, no load, and the revised setting choices that came from that. Set threshold at 15 % higher than max signals seen at each rpm.
Have to do some pulls next.

 

So I did my homework @curly , many thanks for the advice.
Attached is the tune and data log for a series of pulls at relatively constant load values. Not a track day data set, but a lot of low and high load pulls.

Some images from the log are in a post below.
I can see the knock retard is being triggered at higher load, and low rpm. A very few occasions show it triggered at high rpm, where the threshold is set higher.
It looks to me like the threshold at lower rpm may be too sensitive.

But to quote the Tragically Hip, Boots or Hearts, "what the hell is happening here?" When it comes to the off throttle, overrun, fuel cut portion of the sensor signal, there are digital spikes from 0 to 40 across the various cylinders.

 

And now for some images, showing the weird digital looking noise in overrun with fuel cut, and individual pulls at increasing load.

 

Regarding the knock values going nuts in over-run-fuel cut, I experienced the same issue with the stock knock sensor. I posted here and at msextra forums and could not find the answer why. I suspect the megasquirt firmware.

 

The old timing table was the main cause, but there are issues. The mat timing retard is not good, the temps are too high. I would start at 120F and increase in 10deg increments. The VE/fueling table seems low.

Below is a timing table for a similar setup for a street car. There is a lean cruise section for economy from 3-4k below 60kpa, if you don't care about economy run less timing is this area. It also matches a corresponding increased AFR.

 

I agree that the old timing table seems aggressive for 93, much less 91. I've never had the car on the dyno, but FWIW these are the timing maps I'm running on E85 and 91. I'm not saying these are good or efficient, but I don't see any knock activity - that I think is real or concerning - with these.




When I had ~3 degrees more timing in the table for 91 (prior to flex fuel) I'd see some spikes on the knock sensor. Pulled this old photo from my build thread to illustrate. FWIW I wasn't as diligent with setting the threshold value for the knock sensor, but now I consider any normalized knock over 2.5-3 something worth looking at. This was also a single, hot (probably around 80-90 F/~27-32 C ) day of autocross with minimal tuning to the knock control settings, but I still consider it to be a valid data point. My current map might be a bit overly retarded on 91, but considering we'll see ambient temps around 100 F (~38 C) here and I run E85 as much as possible, it's not a huge concern.

 

Not enough timing will result in high EGT’s. There should be a balance between timing and AFR when controlling the combustion temps. A high AFR will delay combustion, but will also pull out some of the accumulated energy. Reducing the timing does not have that effect. It will only delay the process which moves some of the energy from the cylinder to the exhaust. Melted piston vs melted exhaust valve.

With e85 you can run a lot of timing. It is not just the higher octane or stability characteristics of the ethanol, but the endothermic properties and slower burn rate.

 

Thanks for the input and the timing tables guys.
I max out at about 155, with WOT initially running in 145 range and creeping up to 155 toward the end of a pull on some occasions, not others.
My table is scaled to 190, not 250.
Anyhow, when I look at these you have posted in the 145-155 range they seem very similar to the values I am using to give the results I am showing in the above images. In some cases about 2 degrees lower at most. (Agreed it look's more aggressive in my current table at 190, but my highest values are at 155, and I can correct above that ).
This comes from eyeballing your charts, I will have to transcribe them into Excel to get a true comparison.

Highest IAT I have run was about 125, on a stinking hot day for Alberta, but I can certainly adjust the MAT timing retard table to be active.

From LeoNA
Not enough timing will result in high EGT’s. There should be a balance between timing and AFR when controlling the combustion temps. A high AFR will delay combustion, but will also pull out some of the accumulated energy.
Indeed, the current spark table is running ~50-60 degrees hotter, at 1340-50, compared to the more advanced table that blew up last year. And that is after enriching the AFR target to 11.7 at 130 kPa and 11.2 at 150 kPa.

No e85 in Alberta.

 

I would setup the MAT timing so that it will pull some timing at a temp that is relative to your max average usage. Before it was set so it would not activate because the temps were too high. I have on occasion mixed in a few gallons of VP 100oct to a tank of 91 on hot days at the track for an added safeguard. The e85 comment was a reply to SimBa’s post.

 

Update: Used 100 octane fuel and saw a big reduction in apparent knock
Thanks for all the input and information. I used a blend of LeoNA (street) and Simba's (91 octane) spark settings, plus another from some other thread.
The threshold settings were the ones from image 9/16 in this thread, and I was still getting apparent knock spikes on street testing, as suggested in image 13/16 on octane 93.

So for the next race I gave in and bought 100 octane race gas (unleaded), from a local blender selling Fury race gas. It smells rather strongly of toluene or xylene I have to say.
First two times on track it was ~ 50/50 blend of 93 in the tank and 100 octane added, and I saw many (but not a huge number) of apparent knock spikes above the threshold I set, often in the high load mid rpm range right after throttle tip in.
Each race after the 93 was diluted out with more 100 octane added, and by the time it was ~ 95 % 100 octane I would see only 1 spike or so per race that triggered threshold.
The following race weekend I ran the race gas again, with the same low number of knock triggers.

So, I take that to mean the signals I am seeing are apparently due to knock, or else the 100 octane would not have shown such a difference.
But I still use the word apparent, because the signal traces I see are really quite noisy compared to the ones you guys have shown above.

Anyhow, the car's parked away awaiting next season, so no more testing for now.