Hi guys,
From the first time I drove a pre-production STi press car several months ago, I've been very interested in the detonation issue. Or more specifically, why the detonation was so localized and serious. And why, at times, it wasn't there. I think I've isolated one of the causes (if not the cause) for this random detonation problem.

First a little history on the Subaru ECU and how it works as far as ignition advance and knock sensing is concerned:

The Subaru ECU relies on both a base ignition advance table as well as a knock correction table for ignition control. The base ign. advance table is just that, a base table. That is, it is a very conservative foundation on which to add additional ignition advance.

This additional advance comes from the knock correction table. The knock correction table dictates the maximum amount of ignition advance that can be added to the base advance table. In the Sti, the values in this table range from 0 to as high as 11 (or more) degrees. The bigger the number, the greater authority range the ECU has to add in timing. If the value is 0, there is no ability for the ECU to add in any timing. The ECU decides how much of this authority range to use by listening to the knock sensor. When minor knock is determined, the ECU stops adding in timing. It does this during all load/rpm points (with the exception of extreme loads and engine speeds where extraneous engine noise tends to make real knock hard to pick-up). Over the course of a drive, the ECU "learns" how much advance it add in through th table.

What's more, the ECU looks the learned knock correction values (in certain yet-to-be-determined load/rpm points) to determine a general state of knock tendancy. When situations are deemed to be "knock-full", it essentially clamps down up on the knock correction table, only allowing a fraction of the total authority range to be used. As things become more "knock-less", more of the knock correction authority range is dolled out. This greater or lesser amount of knock correction authority range is still subject to the feedback system based upon knock noise from the knock sensor.

As I see it, I think the problem with the original ECU code is that the ECU is learning its AIM during off-boost conditions when knock will never occur. For example, at 4000rpm at 10 inches of vacuum, I seen knock correction values anywhere from 2 to 7 degrees. This is unusual. In the WRX, for example, one would never see positive knock correction values until transitioning on to boost (as one would expect given the tendancy to knock at higher, boost-induced cylinder pressures. Learning knock correction from off-boost conditions seems, from what I've seen at least, to give the ECU a false sense of security with respect to knock tendancies.

This false security results in the AIM climbing and climbing during cruise conditions. If it climbs enough, the engine will likely detonation the first time you go onto boost. Once detonation begins, it is very difficult to quell it. And the STi ECU does try, sending knock correction values from the initial 10 or more degrees to 5 to 2 to sometimes -2 or lower. Go enough times under boost (without giving it enough time at cruise to get aggressive again) and the ECU will eventually learn and pull back timing during the knock-prone on-boost zones. But cruise some more and the AIM will climbs up again. Then go back on boost and the detonation occurs once again. This said, the problem with the original STi ECU isn't that the ignition and knock correction tables were mapped aggressive. It, IMHO, was caused this glitch in the learning system. It is learning poorly. This poor learning meant that the ECU had to react to knock instead of anticipating it (which the 2.0 WRX ECU does so well). This doesn't work very well.

Okay, so that's my take on the situation. It may or may not be accurate. There may be other issues involved as well. But it's the best I've been able to come up so far. So pls take everything with a grain of salt or go running to your dealers.

Also, I haven't seen or tested any of the factory reflashed STi ECUs. I hope to in the next couple of weeks as we do more Stage 1 tuning. I'd be interested to know if the new ECU code has the same or different learning characteristics. If anyone has the new factory reflash and Delta Dash, shoot me an email at shiv@vishnutuning.com and I can give you some tests to do and some logs to take. This would go a long way in coming to some sort of conclusion with respect to this whole STi knock issue.

Best Regards to all,
Shiv

Very interesting...

Hopefully, the Subaru Reflash only fixes this possible bug and doesn't detune the car any.

I wonder why some people seldom hear knock/pinging? I've heard it twice in 3,000+ miles.

Maybe those that drive agressively enough are on boost often enough to keep the ECU's false feeling of safety at bay! :)

Russ

Thank you Shiv :)

so based on this "poor" method of learning for the sti.......do u think after the break in period, u should reset the ecu, and go hard on the boost right away....this will give it a chance to "learn" a more aggressive map

btw.....u didn't mention anything about it.....but do u think the shiv reset on the wrx would work on the sti? hmmmm

Shiv, while this may be the case it leaves some big questions unanswered.

1. Why do only SOME of the STi's ping?

2. Why does the pinging have NO direct correlation to the age of the car. IE there are cars newer than mine that knock bad where as mine doesn't.

3. Why do the pinging cars tend to run 2psi more than normal. I only saw 14.2psi peak on mine, and they see 16.x psi.

4. Why does forcing a det cars boost back in line with stock clean up most if not all detonation... while mine has no problem running 17psi det free

Originally posted by Vishnu Performance
As I see it, I think the problem with the original ECU code is that the ECU is learning its AIM during off-boost conditions when knock will never occur. For example, at 4000rpm at 10 inches of vacuum, I seen knock correction values anywhere from 2 to 7 degrees. This is unusual. In the WRX, for example, one would never see positive knock correction values until transitioning on to boost (as one would expect given the tendancy to knock at higher, boost-induced cylinder pressures. Learning knock correction from off-boost conditions seems, from what I've seen at least, to give the ECU a false sense of security with respect to knock tendancies.

This false security results in the AIM climbing and climbing during cruise conditions. If it climbs enough, the engine will likely detonation the first time you go onto boost. Once detonation begins, it is very difficult to quell it. And the STi ECU does try, sending knock correction values from the initial 10 or more degrees to 5 to 2 to sometimes -2 or lower. Go enough times under boost (without giving it enough time at cruise to get aggressive again) and the ECU will eventually learn and pull back timing during the knock-prone on-boost zones. But cruise some more and the AIM will climbs up again. Then go back on boost and the detonation occurs once again. This said, the problem with the original STi ECU isn't that the ignition and knock correction tables were mapped aggressive. It, IMHO, was caused this glitch in the learning system. It is learning poorly. This poor learning meant that the ECU had to react to knock instead of anticipating it (which the 2.0 WRX ECU does so well). This doesn't work very well.I guess this is possible, but I think the problem is due to a bad ignition correction map, perhaps accompanied with a bad main ignition map. Let's consider your example where you observed 2 to 7 degrees of knock correction at 4000 rpm and 10 inches of vacuum. This would fall into a low load area of the ignition maps. If you examine the ignition correction map from a base WRX, you'll find zeroes in the low load areas, which indicates that the WRX can run knock-free MBT timing in those areas on 87 octane and there's no need or desire to increase timing when running 93 octane. In the WRX case, regardless of what the IAM is, knock correction will be 0 degrees in those low-load areas because the ignition correction map entries are zeroes. The fact that you observed positive knock correction during low loads suggests that the ignition correction map is screwed up.

EcuTeK says that positive values of knock correction indicate the degrees of advance added to the appropriate main ignition map entry, but negative values are not relative to the main ignition map entry. Instead, negative knock correction represent degrees of ignition advance the ECU has just retarded due to detection of knock. So when you observe negative knock correction log entries, you don't know how many degrees was added to the main ignition map but you do know that the ECU was detecting knock and actively retarding. Presumably, the larger the magnitude of a negative knock correction log entry, the worse the knock.

Originally posted by XT6Wagon
1. Why do only SOME of the STi's ping?
I would guess because some drivers like a nice cruse when others drive like mad. Funny, cause driving like mad helps to stay knock free. So everybody go mad! :banana:

I drive both ways... trust me saint or sinner you have to WORK to get my car to knock.

Shiv,
I have this exact problem in my 04 wrx(not STI). If I go on a long Hwy cruise and then get on it it will ping. If I continue to beat on it the ping goes away but as soon as I take a long cruise or just baby it for a few days and then beat of it again the ping comes back. The only mods I have UP/DP and resonator removed. Going to the dealer with a catless exhaust is not going to get me to far, could an Xede or a reflash possibly help?

Originally posted by XT6Wagon
I drive both ways... trust me saint or sinner you have to WORK to get my car to knock.

reset the ECU, baby it during shifting, and go on long crusies for 1000 miles. then drive like mad and i will put money on it that your car will ping

I put 2,600 miles on the car in 6 days... 2 of which were in vegas (100degrees on 91 octane). Never a ping. I drove it 240 miles off the lot in one sitting in high temps and only one mildly agreessive pull about 100 miles in.

FACE IT, SHIV has found PART of it. I don't think for a second that this is the whole story of "why knock happens to STi's".

Originally posted by XT6Wagon
I put 2,600 miles on the car in 6 days... 2 of which were in vegas (100degrees on 91 octane). Never a ping. I drove it 240 miles off the lot in one sitting in high temps and only one mildly agreessive pull about 100 miles in.

FACE IT, SHIV has found PART of it. I don't think for a second that this is the whole story of "why knock happens to STi's".

what do you use for knock detection? i think you just don't hear the pinging. i know my utec has lit up my CEL several times to let me know that the engine is knocking.....but i didn't hear a single weird "pinging" sound from the engine bay.

jason

Originally posted by XT6Wagon

FACE IT, SHIV has found PART of it. I don't think for a second that this is the whole story of "why knock happens to STi's".

So what's your take on it then? Where's your data?

Armchair engineering doesn't solve anything :rolleyes:

What roll if any, does the fuel mixture play here?

The reason I ask is this...my car is not a pinger, but when put on the dyno it went uber rich as we turned the boost up.

I am curious if these things are related or are seperate.

Originally posted by XT6Wagon
Shiv, while this may be the case it leaves some big questions unanswered.

1. Why do only SOME of the STi's ping?

2. Why does the pinging have NO direct correlation to the age of the car. IE there are cars newer than mine that knock bad where as mine doesn't.

3. Why do the pinging cars tend to run 2psi more than normal. I only saw 14.2psi peak on mine, and they see 16.x psi.

4. Why does forcing a det cars boost back in line with stock clean up most if not all detonation... while mine has no problem running 17psi det free

1.) Possibly variations in Octane, temperature and driving styles all of which would effect the above variables and either exacerbate or ameliorate any ECU learning issues.

2.) Doesn't seem like Shiv's theory requires that the car be old or
new....but rather that the ECU either has not learned...or has learned
poorly.

3.) This is a good question, since the car isn't supposed to be
producing that much (16+ psi) boost anyway.

4.) If an engine is detonating then a lot of different things might reduce the det. higher octane, lower IC temps, lower boost, retarded timing. This is without regard to the actual cause of the det.

Just some opinions on your questions. What I am interested in seeing from tuners like Shiv, is whether or not they can consistently tune out the knock while adding power, which is what Shiv has reportedly done. So far, I'm impressed.

Originally posted by Jon [in CT]
I The fact that you observed positive knock correction during low loads suggests that the ignition correction map is screwed up.


It's very likely from what I've seen. Another possibility is that the knock learning load/MAF breakpoints (in the WRX, there are only 8 of them) are too widely spread out and overlap off-boost regions. To me, this would be an easier mistake to let slip by than having bad data in the knock correction table which can be easily seen during a quick glance.

Just my 2c,
Shiv

PS. The STis that I've tested all ran 12-14psi of boost at WOT and pinged audibly.

Good to know that the ones you've seen don't show the extra boost.

My guess is that its more than just one thing screwed up, since most noticable failures are the result of multiple "oops" events.


Next I'd not take a Utec's word about "detonation" as the absolute truth. I've had mine flash when there is NO reason for any detonation. Basicly this is why I ignore 1 "knock event", and only start to worry if its 2+ in one consistant condition or 3+ at any time.


On the learning, could it be that only certain ECU's got the bum program? I "woke up" my STi driving in conditions moderately likely to cause it to toss in tons of timing.

Shiv,
Thanks for the info, very interesting.

To others FWIW
I have a had a knock link on my STI for over 3 months now. I have tested all mixes of fuel from 91 to 96. Using both toluene to get upto 96 and upto 94 pump gas.

I have never heard Knock "BUT" the Knock Link on 91oct lit up ALL the lights. On 92oct it lit up the 2nd orange caution and on 93oct lit up the second green.

When I had the UTEC on my WRX I also had Knock link installed. When the knock link lit up the second GREEN the UTEC had already started to pull timing.:eek:

IMHO
If you hear knock its too late.

If you don't hear knock with your ears it doesn't mean it ain't happening.

If I recall from a post last year Shiv said its the little knocks over time that will do damage. (Shiv please correct if this statement is incorrect).

Originally posted by Vishnu Performance
Hi guys,

What's more, the ECU looks the learned knock correction values (in certain yet-to-be-determined load/rpm points) to determine a general state of knock tendancy.
Best Regards to all,
Shiv

Assuming the 8x8 grid used by the WRX still holds true on an STI, theoretically the solution sounds like rescaling the MAF load references higher up for ignition and knock correction. That way, if you find favorable conditions down load (lower in the matrix) your ok, but if it learns knock up high, that will only impact that cell in the matrix.

Jorge (RiftsWRX)
www.ProjectWRX.com

Shiv: Any update on this? I had my ECU reflashed by Subaru and noticed a big drop in power (seat of the pants). I figured that it would have to go through the learning process again, and have given it 2 weeks now. No improvement, and in fact it still pings a little bit (audibly) on the new reflash, especially when hot out.

I would like to remedy the problem and NOT lose any power in the process. Your XEDE piggyback sounds like a good solution to me, but I want to remain in AS when I autocross, and that would mean shifting pins constantly, which would be a pain. Plus, when I get to the track, the stock ECU (reflashed) will cost me 20HP and 20TQ. Before the reflash, I always added a gallon of toulene and never had any problem with det. :mad:

WOW!

i just read this thread.

a learning disability? where have i heard that before?

shivmeisters analysis and conclusions are very close to what i am experiencing with my vehicle.

i'm friggin' imprezed!

thank you shiv and jon, and all the rest too.


wow!

Originally posted by bunot


what do you use for knock detection? i think you just don't hear the pinging. i know my utec has lit up my CEL several times to let me know that the engine is knocking.....but i didn't hear a single weird "pinging" sound from the engine bay.

jason

I don't know anything about the UTEC and it's ability to discern knock from other vibrations from the engine block, but I'd venture a guess that not everything the UTEC is calling 'knock' is knock.

There is a very fine line between perfect combustion and knock. I can't help but wonder whether 'inaudible knock' has any meaning at all.

I've experienced knock in my STi twice. Both events lasted less than 2 seconds.

The first time was pulling out to pass a car after a 10 minute light-throttle cruise.

The second and most significant time was after a long cruise, a 5 minute stop and an almost full throttle acceleration.

The second event left me wondering about intake charge temperatures. No doubt in my mind that the IC was well-heated by the 5 minutes at standstill after a long cruise.

Shiv and others... correct me if I'm wrong (and I usually am), but doesn't the STi monitor charge temperature PRE-turbo, PRE-IC?? Is it possible that the ECU is reading air temps from the airbox and not accounting for heat soak of the IC in its 'predictions' of knock risk?

Again, this is based on my VERY limited experience with pinging over several thousand miles of hard driving.

Nice input from shiv...

I made some testing on my 2.0 STi 8 (type euro) and if we have the same overall principe, we definately don't have any correction during close loop drivng: KC remains at 0 until ECU switches to open loop and active timing control.

To Konekiller,
As I don't expect you're getting any additional probe on this side of the ocean, the ECU monitors the air temp at Mass Air Flow location... not very usefull :huh:

Originally posted by FRSTi

To Konekiller,
As I don't expect you're getting any additional probe on this side of the ocean, the ECU monitors the air temp at Mass Air Flow location... not very usefull :huh:

Yes, I thought that quite odd when I first read it. On a normally aspirated motor, the Mass Air Sensor makes sense, but on an intercooled motor, it would seem an obvious problem.

<shrug>

Originally posted by KoneKiller


Yes, I thought that quite odd when I first read it. On a normally aspirated motor, the Mass Air Sensor makes sense, but on an intercooled motor, it would seem an obvious problem.

<shrug>

When measuring mass airflow, you need to bring temperature into the calculation. If all the sensors that determine mass airflow are not all in the same spot, then the calculation would be off.

So, technically, you can measure mass airflow before or after the turbo, and the readings should be the same. PV=nRT.

Originally posted by ShapeGSX


When measuring mass airflow, you need to bring temperature into the calculation. If all the sensors that determine mass airflow are not all in the same spot, then the calculation would be off.

So, technically, you can measure mass airflow before or after the turbo, and the readings should be the same. PV=nRT.

Sorry Shape... I was a bit unclear there. The issue is not where you detect the mass of gas, but where you detect the termperature of the mass of gas. If you do it at the airbox (where the Mass Flow Sensor is), you can't detect Intercooler heatsinking and a subsequent rise in post-IC temperatures.

It would seem that a temperature sensor on the post IC side of the intake system would help prevent pinging due to IC soak (which is ALL I've had).... assuming the ECU was so inclined.

Sorry about the message that was unclear in the absence of the context... my bad!

Shiv - Any update on this after tuning some reflashed STi's? Have you been able to confirm/disprove your original theory?

TIA if you have the time!

-st