So I know knock does... what else? And a follow up, how much "learning" does the computer do with signals that retard the ignition (such as knock for instance)? How long does this event last for?

(If "I" were an ecu and I detected knock, I'd dial back ignition alot to get the signal to go away, and then slowly elevate to the stock map while checking for knock. Once I detected a significant amount id retard ignition a bit and then leave it there, checking every so often if i could advance ignition. So basically learning where is safe and where is not... but that logic may not be built into the ECU... that's what im wondering basically)

As far as knock goes, does the amount of ignition retard vary with the amount of knock? i.e. more knock -> more retarded.

Ok, lets hear how much you guys know about being retarded

-Heath

 

Knock sensor frequencies and basic principles of piezo-electric sensors:



sample knock sensor design 1 (mass type?):



sample knock sensor design 2 (resonance type?):



basic FD knock control logic:



If I had to guess, the FD sensor is a mass type, meaning it tends to produce a wide range of voltages which peak at a frequency of about 7 kHz. This may explain why they tend to read a lot of activity if you monitor them with a Power FC. The stock FD ECU must have some kind of logic to filter through the signal. I know in the Subaru ECU's the knock sensor reading is ignored under certain conditions (very high rpm for example).

I was just tuning a Subaru Legacy GT (the turbo model) and I got a few small knock events, so I've had knock sensors on the brain today. The Subaru knock sensors don't have anywhere near the activity of the Mazda knock sensors from what I can see, which makes me think that they are mass type.

 

^ but what im wondering is it instantaneous... meaning, when the knock disappears, does all of the ignition retard? or does the ecu LEARN to keep it slightly retarded after an even with say max knock?

 

speaking of retards, i just realized my last post was somewhat unintelligible and its too late to edit

what im trying to say is: does the ecu return to its previous state once the knock signal is gone?

 

the ecu is stupid it does not save knock any changes cause of knock. some new new cars can do this but not back in 93.

 

Given that the series 6 8 bit FD ECU doesn't even have any kind of A/F learning (fuel trims), it's unlikely that it memorizes any learned knock over time like newer cars. So I think the answer would be yes, and it probably returns to that state pretty quickly.

 

thanks guys, exactly what i was wondering. -Heath

 

From Service Highlights book

 

from the same document - sorry I should have copied this entirely the first time...



don't understand how (5) and (6) can be different...

 

yeah i was wondering myself how it would know the difference ....

 

Hmm. I forgot about that part of the service highlights manual, especially #5.

There may be at least two types of knock control strategies employed. The calculated ignition retard from each strategy may be added together, or the ECU may switch between the two. I say this based on my experience with modern Subaru knock control systems. The FD may have a primitive version of what the STi's etc are doing.

The way it may work is this:



The ECU has one set of knock control tables for responding to knock based on rpm and MAP. So more timing would be pulled under high levels of boost for example. There must be a set of conditions to determine when this type of correction is used. I'm kind of curious how this type of logic could be used given the hardware constraints they had at the time. In Subaru terms they would call this Fine Learning Knock Correction, and there are tables learned and stored within the ECU that you can view if you have the right software.



This would be a fast acting knock retard strategy that would apply ignition retard almost across-the-board based on sudden increase in knock sensor activity. Some kind of timer logic would be used, or the ECU would just continually monitor the sensor until the the knock activity falls below some threshold. This may be used in addition to the load based logic above, or they may be used in mutually exclusive fashion. The part about fuel injection exceeding a particular level may just mean that this type of feedback is not triggered under really low loads.

Now the question is, does the ECU actually remember knock sensor activity over a significant period of time? The newer Subaru ECU's do, as long as you don't disconnect the battery. Again we have no indication that the FD has any kind of long term fuel learning so long term knock learning seems unlikely.

Here is an excerpt from an article about Subaru's knock control system that was originally posted on Rom Raider, a Subaru tuning forum.

 

Beyond Knock control The PCM also retards timing at boost, and as a torque reduction for automatic transmission shifting. I believe also the O2 sensor, air temperature sensor, TPS, and water temperature sensor have a roll in the ign. timing but don't have the data handy.

The TPS acceleration is also measured to advance ign. for off line performance and turbo spool; again don't have the data here but the dyno pulls seem to confirm.