Sale ends todayGet 30% off any course (excluding packages)

Ends in --- --- ---

High knock level reading on certain cylinders

Practical Standalone Tuning

Forum Posts



Tech Articles

Discussion and questions related to the course Practical Standalone Tuning

= Resolved threads


Had my first attempt at tuning on a dyno today. The shop owner was nice enough to babysit me the whole way to make sure I don't mess up anything although it does limit the way I tune. We were tuning a SW20 MR2 with a gen 3 3SGTE engine with a Precision 5558 turbo and high lift 262 cams on a link G4+ plugin.

One of the major problem we encountered with the tune was high knock level reading from cylinder 2 and 3. During peak load and RPM, knock level on cylinder 1 and 4 were at around 600 but cylinder 2 and 3 would be reading much higher at roughly 900. According to my tuner, it was scaringly high. What was most worrying is that it was only cylinder 2 and 3 giving out such high readings. If all 4 cylinders were high, we may attribute that to a noisy engine but a noisy engine will always be noisy no matter which cylinder is firing and not only when cylinder 2 and 3 are firing.

Most of the local tuners here are not confident at using audio knock detection devices and trust the ECU knock detection more than their ears. I tried listening to the knock with a knock block anyway but don't seem to hear any knock (but I only have experience on hearing knock from listing to sound clips online and I am not even sure whether I can properly hear knock in real life situations).

We tried retarding the timing by 4 degrees but there seem to be no change in the knock level detected so I guess the high level should just be other noise and not knock. However, given the high knock level detected, the tuner is reluctant to let me go for higher timing.

Should I be worried about the high knock level on the specific cylinders? Should I try using the 2nd harmonics as knock frequency? Will making an audio recording from the knock block help?

The knock sensor used by the ECU for knock detection is the same sensor included with the knock block.

Attached is the current map used and log file pulled from the ECU. I have foolishly discarded the PC logs I made during the dyno session but after betting home, I managed to pull the logs from the ECU which contained the last two final WOT pulls at the very beginning of the logs and followed by some steady state fuel tuning and my drive home from the tuner.

Sadly we did not do any steady state ignition tuning as the tuner suggested me to leave it for next time as somehow torque is falling flat on its face after 6500 rpm and he thinks the cams had not been degreed properly. We were only making 220PS at the wheels while he thinks my setup should easily make 300+.

Thank you!

Attached Files

Without knowing all the details behind your engine and configuration I would say that what you are experiencing is not knock. You've only got 6 deg advance with 220 kPa boost which even for a low grade pump gas is incredibly conservative. Couple this with a mixture that is probably much richer than it needs to be and this adds up to an engine that is probably not running too happily which would explain your low power.

The knock levels alone are meaningless as they need to actually be validated to ensure the ECU is detecting real knock. I recommend doing this with audio knock detection. It is also very common to have 1-2 cylinders with much higher noise profiles than others. This can be related to the knock sensor location for example. This is why it's essential to use the individual cylinder gain function to equalise the non-knock noise profiles.

My definition of scary knock is knock you can hear with your ears without using a microphone or sensor.

I don't know where the knock sensor is on the 3SGTE but remember that it is going to pick up vibration from the cylinders it is closest to. If it is between cylinder 2 and 3 it's going to get a stronger signal from those two.

Thank you Andre and Raymond.

Yes, I am inclined to use audio detection for knock confirmation too. The problem is, there isn't much information on how to listen to knock within the HPA modules. I have tried listening to different sound clips with knocks, some sounds very obvious to me while for some I am not so sure. I guess it all varies slightly between different engine and audio devices used and that is why there isn't any demonatration on how to listen to knock in the HPA courses given how abstract it may be.

Hiring an experianced tuner to teach me hands on might help but I don't think it is easy to find one locally as most tuners here rely on ECU knock detetion and a few even claim that if they can hear knock on the headphones, it is already be knocking badly.

I guess this would mean I am on my own to get myself trained listening to knock. However, I do not want to blow up my engine in the process which took a lot of time and effort over the past 3 months to build.

Will using the current noise profile be sufficient to provide a basic safety net for me to tune and learn how to listen for knock. I will be adjusting the individual gain levels to even out the noise profile between the cylinders and set a trashold a little above the detected noise levels once all cylinders have been evened out. I am just worried the high amount of engine noise may be masking the knock on cyl 2 and 3 should knock happens.

Wow yes, the AFR had drifted very rich in the logs and I didn't even notice it. There was way to many things to handle and being my first dyno tune ever I was overwhelmed and missed it. IAT may have a play in this or maybe it simply was not tuned correctly. I will correct it on the street in the next couple of days. Anyway, the way torque have fallen off in the dyno plot suggest something more than the over rich mixture and retarded timing. I do not have a picture of the dyno plot with me right now but I will upload one a little later.

Many thanks!

Attached is the dyno plot.

I have started another thread in the general discussion to discuss the drop in torque.

Attached Files

I guess I can start building up my experience on listening for knock by purposely inducing knock at the light load area. But what I am worried about is, knock might creep up without me being able to identify it and I will continue to increase timing more and more. Will that eventually be sufficient to cause damage even when load is light? Under what RPM/load should it be done at as the definition of light load seem to be quite subjective. What is the maximum advance I should try at specific RPMs before i should back off?

In a practical sense a lot of times you take your existing spark table (or some area of it) and increase spark by 1 or 2 degrees at a time and look for the power/torque benefit on the dyno. As you start hitting diminishing returns (1 degree gives 1 horsepower or something) you stop. Stock ECUs use this torque vs spark relationship in their engine torque model. See the attached image for a stock Bosch ECU. Don't pay attention to the exact number of spark degrees on the axis, just try to understand the basic relationship of diminishing returns of advancing spark.

Attached Files

Yes, but what you mention is the method for finding minimum best torqure correct?

I do not know how knock limited my engine is but I suspect it might not be knock limited at the lower load areas. If I want to induce knock, I am going to have to advance past MBT? At higher loads where it would knock before MBT, I am afraid the load may be too high to be safe (especially if I miss the knock and continue to advance the timing).

I would adjust the individual cylinder gains right now until the noise profiles are relatively equal. I would also try the second harmonic frequency as this is likely to show a better signal to noise ratio. I would teach yourself what knock sounds like at low rpm and high load by purposely making the engine knock. This will also let you confirm the knock control system is responding as you'd expect.

In terms of your torque I actually wouldn't be too concerned with this. You are going to naturally see the engine torque drop away as you move higher in the rev range and you don't have cams that are really suited to ultra high rpm airflow. the power level is currently lower than I'd expect but then this may be more related to your conservative timing and rich AFR.

Although my tune is not ready I had already signed up and paid for a track day beforehand and went to the track last weekend. I was able to rent a dyno at the track (dynojet with no load control) to carry out some last minute fixes to the tune and corrected most of the overly rich fueling before I actually hit the track. I have also advanced ignition by roughly two degrees at higher rpms which i think should still be conservative and safe but it made a HUGE difference on the dyno.

I have set the knock sensor to the harmonic frequency of 13kHz and although the overall signal volume had decreased (I've added in more gain to bring it back up), the sound signature still seems to be a little erratic where minor spikes would occur on random spots from different cylinder between each run. With only one knock sensor available (I've since ordered another one and it is now on its way), I was not able to confirm this with audio.

The log attached contained the last few dyno pulls and the subsequent logging at the track for the rest of the day until a hose on the wastegate popped which had caused a huge boost spike.

Looking at the logs seems like there might be some knock between 6000 and 6500 rpm where cylinder 1 is spiking quite a bit at such RPMs. Does it really look like knock?

Attached Files

I have touched up the map a little more for the second day where I have decreased the gain on cylinder one, retarded the timing slightly as I am worried about the knock reading and also upped the wastegate duty cycle as I was still below my boost target (no time to setup closed loop yet and boost level seems to be varying greatly depending on the ambient conditions for open loop).

The ambient temperature was quite a bit hotter on the second day and I was suffering seriously from overheating problems. RPM limit actually kicked in due to high ECT temps which I had to reluctantly raise in order to run more than a lap at the track. What scared me was the knock level I was getting from cylinder 4 once the temperature have risen. It seems to be knocking all over the place!! I have attempted to pull further timing from the igniton base map but it is still knocking eventhough significant timing had also been pulled by the ECT and IAT ignition corrections at the same time. The engine had so much igniton pulled that it had no power at all at this point.

Looking at the logs now, I am actually very worried about the engine given how much noise we are getting from cylinder 4. The knck sensor is placed on cylinder 2 so cylinder 4 is actually the furthest away but it is still giving out such a high output! Seems like I have been pushing the engine hard under serious knock!!

Also, I think the HUGE spike at 10:32 and 10:53 in the log file labelled Day 2-5 and 10:58 in the log file labelled Day 2-7 is most definitely knock??

Attached Files

One interesting thing to note. The track is in mainland china where we need to apply for temporary permits to in order drive our car up from Hong Kong. We run on 98 RON normally but have started to run low on fuel on the second day. With no gas station anywhere nearby that sells petrol with octane higher than 95 RON we are forced to refuel at the track paying more than double for their "100" octane fuel in tin cans. Knowing not to trust anything from china, I've even added a bottle of octane booster to the tank.

Given I am on 100+ octane compared to the 98 RON that I had been using for the tune, my engine should be much less likely to knck. However, The HUGE spikes I mentioned in my last post above only started appearing after I have refueled at the track which suggests the fuel being sold is actually of an inferior quality with a much lower octane than advertised.

For those of us without the Link software, can you post a screenshot of the concerning area of your log?

Thank you.

Here you go!

Attached Files

ignition maps

Attached Files

Honestly, after seeing the logs I have my doubts that that is real knock. First, it's at really high rpm. Engines get exponentially more noisy with engine speed. Over 5000rpm is definitely sketchy to judge. Second, knocking can have an exponential sensitivity to spark. 3 or 4 degrees should be way more than enough to quiet it down.

Ok so the spikes that you've got showing in the chinese fuel screen shot does indeed have the look of real knock. I'd still recommend confirming with audio knock detection gear but that is the sort of trace you'd expect with a few reasonable knock events - The signal level spikes way above the background level and then returns when the knock goes away. I wouldn't read much into this though since you really don't know what that fuel was.

The trace you have with a high level from cyl 4 looks more like a gain problem since the signal is basically high everywhere.

My bosch KS4 knock sensor came in today so I should be able to hook up my G4+ knock block to listen for knock at the same time with the ECU detection.

What confused me with the cyl 4 trace though, was that the noise level of cyl 4 is normally in-line with other cylinders until the IAT and ECT rises to a very high level. Then all of a sudden, the cyl 4 trace gets high everywhere. I am taking measures to solve the overheating problem now so I probably won't be hitting those areas again (and therefore gain should not be adjusted to maintain sensitivity for cylinder 4 when temps are at normal level) but I just fear the high noise level when hot might indicate other problems.

If you look at the actual logs for which I took the screen shot from, you can see the "background" noise would appear suddenly when the ECT reaches 100C and there was no noise before the temperature rises.

Here is a zoomed out screen shot of the log to show what I mean for cyl 4.

Attached Files

I agree that trend is a little weird. I don't actually have an explanation for that as it isn't a correlation I've experienced myself. I would be suggesting you find a way to combat both IAT and ECT however for both the reliability of your engine as well as the potential to make more power. One compensation I use heavily in knock limited engines is the ignition vs IAT trim. High IAT can easily result in the engine becoming more prone to knock.

Hi Andre,

How much timing do you pull for each degree of temp? Is there a way to calibrate this?


Hey Matt, I don't have a solid rule to work from as it will depend how knock sensitive the engine I'm tuning is. On most turbocharged engines I'd leave this table zeroed out to around 40 degrees as that would be a typical IAT to see under our ambient conditions. It's above this that I may start pulling timing of maybe 2-4 degrees as I move from 50 deg and above. You can test this on the dyno by blocking airflow to the intercooler and artificially create a higher IAT and see how sensitive the engine becomes. You do need to be careful though as if you end up pulling a lot of timing in this table and you have other ignition comps active (ign vs road speed or gear for example) then the cumulative retard can end up larger than you think and can result in poor running.

We usually reply within 12hrs (often sooner)

Need Help?

Need help choosing a course?

Experiencing website difficulties?

Or need to contact us for any other reason?