Knock?

[Howard Coleman]

i ran one of my best 3rd gear pulls yesterday. approx 530 rwhp at only 15.8 psi. two rotor with a Garrett G40-1150.

all my metrics were great... 80 F IAT measured with a thermocouple, mid 1500 F EGTs etc.

my G4X Extreme was measuring knock around 50-80. (Gain at 1)

68% 93 octane pump, 32% ethanol along with 1400 CC of 100% methanol as AI. 

knock was normal except for two data points.

at 6:23.604 i recorded 236. knock before the data point was 31 and after 22.

at 6:24.925 i recorded 389. previous was 29 and following was 41

AFR, EGT, IAT, Exhaust manifold pressure, fuel pressure and AI pressure showed zero impact from these two data points.

AFR was 11.2 during the 236 and 10.42 during the 389.

i am especially concerned as to these two data points because in March of last year i had a similar one data point of knock that registered 1178. adjoining readings were around 60. same RPM area and the same front rotor. 24 PSI, 37% TPS, 9.72 AFR, 9.4 IGL around 500 rwhp. given all the metrics were fine there was a thought proffered that the knock reading could have been my front iron breaking. (due to an issue with the thru bolts.)

i also had my Link Knock ears on and heard nothing amiss. 

i successfully ran my V88 for 10 years with essentially the same setup and never had a single data point knock reading. i have a log where i had bad gas and it showed a series of knock readings... as i would expect. i realise the G4X knock system is more advanced than the V88.  

is this one data point outlier knock reading a normal generation for the G4X? can this be knock? 

map: https://drive.google.com/file/d/1vaIbH-YN69ffDT8lRhaldydv0BAaDw2C/view?usp=sharing

log: https://drive.google.com/file/d/1zHrgH0DV6K8CdOcXBYu1zpAXQUNCXlo_/view?usp=sharing

thank you

   

[Adamw]

I would struggle to believe that is knock with that amount of alcohol in there, relatively low boost, and your ECT at 65°C, I think you would have to try very hard to make that thing knock.  

I do have a log from a similar run in our FD where I purposely made it knock, I did have 2-4deg less advance than you at that point but I also had 20kpa higher boost, 20°C higher ECT, leaner lambda, no alcohol and similar pump gas.  

Are there any mechanical noises - exhaust hitting the floor or anything like that?

[Howard Coleman]

have you seen a significantly higher knock reading that was one data point that was knock? all my V88 real knock had other surrounding knock data points.

i will be doing a similar run w a touch less timing

[koracing]

If it were on dyno I would say pull 3 degrees of timing and see if power drops evenly and that data point goes away.  On the street, just see if the outlier is still there or goes away.  It will tell you something anyway.  You can try across the board change or just at those particular rpm points.

[Howard Coleman]

Thanks Adam and Kris. i will reset timing a bit and that may supply an answer

[Adamw]

7 hours ago, Howard Coleman said:

have you seen a significantly higher knock reading that was one data point that was knock? all my V88 real knock had other surrounding knock data points.

Yes, there are many explanations for that, a knock sensor is just a microphone, any sharp noise or impact that occurred within the window will make it through.  Gravel hitting the sump guard in a rally car for example makes knock detection almost impossible.  You can only confirm what is knock Vs other noise with audio connected to a good human ear or in cyl press measurement.  Compared to a V88, you are using a different band pass frequency for a start, the G4X engine position and knock window is more accurate and there is less hardware filtering so you can see smaller/faster spikes than wouldnt have been possible in older ECU's.  

[Pete_89t2]

Howard - I'm in agreement with Adam that those two data points were most likely noise induced outliers for all of the reasons he mentioned. And I think Adam's statement that the knock sensor is "just a microphone" might confirm what I've suspected regarding how the G4+/G4X processes raw sound data from the knock sensor to decide on what a knock "detection" is - it is simply an amplitude only strategy. So if there's enough audio energy in the sensor's output to cross your set threshold within the bandwidth selected, and it occurs within the knock window, the G4+/G4X declares knock and takes action as programmed.

Adam - this is probably something for the engineers to look into for the G5 generation & beyond ECUs wish list, but have you guys considered combining frequency domain & time domain processing of the knock sensor's raw data with the amplitude only/threshold based processing already employed to better discriminate actual knock against background/spurious noises? In essence, this would be an AI that would act like a skilled pair of human ears listening to the knock audio.  As implied, frequency domain processing looks at the tonal qualities of the sound, and time domain processing looks at the rise/fall times of the sound signal - both of these when used together with an amplitude based measurement can help discriminate the actual sound of interest (knock) against all the other noise.

[Howard Coleman]

FWIW, i had my Link Knock Ears on and didn't hear anything unusual.