Hi All

I have been comparing the knock voltage on the ECU+ to the knock count on Evoscan and it seems that they relate like this:

The knock voltage threshold for the ECU seems to be around 1V (0 counts). After that it looks like you get 1 knock count per 100mV of knock voltage. Any input/comments?

Example 2.5V = 15 counts

 

I've seen the same thing.. Only the knock filters in the ECU have an affect on what the ECU determines to be knock.. You will occasionally see that at high RPM where there is alot more background noise (and the knock voltage is higher) but the ECU registers no knock..

I use the ECU+ knock voltage as a sanity check, and then go back to look at the mitsulogger data in the same area..

Different RPM's will register knock differently and at different quantity and levels in the ECU thanks to those filters.. Its kinda like a spectrum analyzer of sorts..

 

If you have a TurboxS tuner pro, you can do what I was doing not too long ago and using the audio recorder to record the knock sensor output from the tuner pro's sensor.. since the recordings is done in seconds from start, you can actually listen to the playback later, or watch it in an audio analysis tool at the same point you look at your logs.. I think you can do the same thing with the stock knock sensor assuming how you tap the signal doesnt degrade it to the ECU..

You can clearly hear knock and suspension/exhaust system clunks as unique and different sounds, both of which can trigger the ECU's knock correction...

 

Hey Tom.. Since you tap into the knock sensor, is it possible to record that data as an audio output that can be listened to later?

 

Good info Malibujack, I have not really looked at the knock filter yet but that defiantly makes sense. Its knida funny I work with a lot of telemetry instrumentation at my job then I come home and I am doing the same thing on my car. Logging the actual frequency at the knock sensor would be great. It would give more information to characterize actual knock. I wonder how much filtering the ECU does for high frequency aliasing.

 

hmm.... I havent noticed much of a correlation. I think the knock filters play a huge roll, making the voltage no more than a "sanity check" as you said Jack.

 

FWIW there are 12 filters, each obviously represents a frequency range and threshold.. The ECU isn't really all that fast or complicated to do anything elaborate, its obviously some sort of simple threshold for different frequencies.. I have to assume if you combine the different filters together, you get a signature that mitsubishi determined to be cumulative frequencies that constitute knock...

 

Are you guys logging both ECU+ and Evoscan/Mitsu logger at the same time? I have been doing two seporate runs to get both sets of data (not idea at all). I was only looking at one rpm range at the time so the numbers work out pretty well at that given point.

 

yep.. logging from both simultaneously

 

To actually record the knock signal, I'd need around 30,000 samples per second, which isn't feasible to record. Just retrieving that data (16-bit samples) would be 480,000 baud on the serial port.

The knock signal is on the high end of the range of human hearing, and if you could hear it (older geezers like me probably can't), it'd be a reeealy high pitched squeal, and you wouldn't be able to distinguish intensity changes. I'm guessing that what the Turbo XS device gave you was an aliased (distorted in frequency) version, which is of no real use to anyone. BTW, suspension klunks and the like are way outside the range of the knock sensor - if those really do influence the stock ECU's knock sensing, then the stock ECU must have a really lousy (er, inexpensive) hunk of hardware to process it, with the knock tables just a big band aid to work around the bad hardware.

Tom

 

Metal to metal contact typically creates a plink in a similar frequency range as some types of knock.. So its certainly possible.. And the ECU definitely has very rudimentary (Band pass) filters so I'm sure it just attenuates out the crap.. But if you clank something loud enough, you will meet all the criteria sees as knock and it'll pull timing..

I kinda figured it was a bit optimistic to think it could record the audio, I keep forgetting that its connected through a serial interface.

It didn't really distort the frequency, but you really had to listen for the "rattle can" carefully to hear anything, but it could be heard.. But I can hear flourescent lights squealing and the sound of the power supplies hissing when data is being transferred through my network port.. Its hard to explain, only that my computer resonates and I can hear it.

 

so what is the considered 'safe' knock voltage? i see somewhere between 1-1.8, peak 1.9V once. i don't have evoscan so i can't record exact count. but anything above 1V on ECU is considered knock count?

then...my question is what is the 'safe' knock voltage with ECU+? should I tune my ECU+ so that i have knock voltage close to 1V all across the RPM range?

Chong.

 

From what I have seen using both the ECU+ and mitsulogger. The knock filters really effect what voltage the ecu considers knock at different RPMs. The knock counts are a better gauge of when the ECU thinks there is knock.

 

With the ECU+, you really shouldn't be looking at the knock voltage per-se. Rather, look at the graph - it'll show a gradually-increasing voltage with RPM. When there's real knock, you'll see spikes in the voltage. Think: changes from the background noise. The ECU+ really does a much better job of showing you what the knock looks like, and you'll see spikes with the ECU+ that MUT dataloggers won't show, and this indicates that you're right on the verge of getting a knock event that the stock ECU will see.

Tom

 

Hi

I've noticed that the ECU will pull timing when the knock voltage spikes in two peaks very close together, without a deep valley . This seemed to be a repeatable trend.

MB