Re-Mapped ECU Learning Trick 10-20hp in 5 seconds :)

[MadWagon]

I've experienced this behavior on mountain, curvy type driving. Sometimes I'm in 2nd or 3rd for quite some time, using very light throttle and little boost, but holding it for I know more than 5-10 seconds. I would/still experience awesome throttle response after this type of driving.

* Registered users of the site do not see these ads.

[Jduke]

"First: With the car fully warmed up, reset the ECU. This can be done by killing power the ECU or by simply pulling off the neg. battery terminal and pressing the brake pedal for a couple of seconds."

***Do this with the car running or turn if off once its warm? then pull
off the neg. battery terminal? Just curious I have never reset the
ECU before.

[SlideWRX]

Correct me if I'm wrong:

From what I see, this doesn't need to be done on a reset car, unless there is another factor that prevents pulled knock timing from being re-added in this situation. I should be able to go grab my car as is(stock), make a run and produce the same timing changes. Unfortunately, I don't have any logging software, so I can't confirm this.

Are the rates of change in the algorithim same for adding & pulling ignition? Is there a time variable that is virtually non-existent after a reflash that has a lot of control after weeks/months that would prevent rapid timing change?

Thanks for info. I'll be trying this on my un-reset car just for the placebo effect!

Tom

[Jon [in CT]]

Originally posted by AZScoobie
Anyone who has used an SAFC on a WRX while monitoring a Delta dash can clearly see that MAF input effects ignition advance.
...
I found Shivs post interesting because I have been using this trick for about a year now without knowing what I was doing and or why it was working. I discovered this trick while drag racing. I noticed on a reset ECU my times would drop. This was back before the delta dash existed so I was guessing as to what was causing the power loss. While experimenting and pulling a few ideas of posters on this forum I found that reseting the ECU and then running wastegate boost on the way to the track before boosting up high would always net a faster time. Add race fuel, Reset ECU, Turn boost down to 8psi, Drive to track, turn boost up.... I posted that suggestion many times to drag racers on the forums. Now I know why it worked.

The low- and high-octane ignition tables may, indeed by indexed by "Load" and not MAP and "Load" may be based, in part, on MAF. I guess the point is that "Octane Learning Mode" is entered based on MAP, not MAF.

And I could only find one instance where you publicly recommended something along the lines of "Add race fuel, Reset ECU, Turn boost down to 8psi, Drive to track, turn boost up" and it was different in one very important aspect. The post I found was http://forums.nasioc.com/forums/show...81#post2684681 and you said there:

Tip:

Reset ECU
Set boost low like 10psi
Drive 20 miles and 4 or 5 restarts
Toss in race fuel
Turn boost back up to 15-16 psi.
Race....

CT

The significant difference, of course, is that you recommended adding the race fuel after the ECU had learned the octane of some other lower-octane fuel. But maybe that was a typo.

[AZScoobie]

Yes. That is because back then I did not store race fuel in my garage so I was recalling my actual scenario where I would reset, then drive to the track and then add the race fuel. I dont think adding the race fuel before the reset would cause any different effects. I think the act of running the the Reset ECu at low boost causes the same IG multiplier to go up just the same as Shiv is suggesting in his way. Mainly, the lack of knock because of the lack of boost allows the ecu to increase that multiplier. If I have time this week I can test that theory out.

To find more instances of where I suggested this method of ecu reset simply search more. I estimate I posted that about 10 times. As a result many people have used this tactic to gain a few MPH and a tenth or two at the strip and many of thanked me afterword. I doubt I am the first one to use this trick. I did it mainly because of my vast experience with the bosch Motronic ME5 and Me7 ECU's. Peak power and peak advance is obtained on a freshly reset ECU in the Bosch systems. I tried this and found the Denso unit to reduce power output. Back then we (I) had no delta dash and I only had theory's that where posted on these forums. One such theory was that the ECU advances slowly and in steps over a time period after the reset. It made sense to me to reduce boost as to obtain the most advance I could. Then turn up the boost using race fuel. it worked.

CT

Originally posted by Jon [in CT]
The low- and high-octane ignition tables may, indeed by indexed by "Load" and not MAP and "Load" may be based, in part, on MAF. I guess the point is that "Octane Learning Mode" is entered based on MAP, not MAF.

And I could only find one instance where you publicly recommended something along the lines of "Add race fuel, Reset ECU, Turn boost down to 8psi, Drive to track, turn boost up" and it was different in one very important aspect. The post I found was http://forums.nasioc.com/forums/show...81#post2684681 and you said there:The significant difference, of course, is that you recommended adding the race fuel after the ECU had learned the octane of some other lower-octane fuel. But maybe that was a typo.

[Jon [in CT]]

Originally posted by AZScoobie
I dont think adding the race fuel before the reset would cause any different effects. I think the act of running the the Reset ECu at low boost causes the same IG multiplier to go up just the same as Shiv is suggesting in his way. Mainly, the lack of knock because of the lack of boost allows the ecu to increase that multiplier. If I have time this week I can test that theory out.

I guess the results of your test will depend heavily on how Subaru has initialized the two ignition tables. Recall that, according to my research, the ECU will never retard timing below the values represented in the low octane ignition table and will never advance timing beyond the values represented in the high octane ignition table. I think the question comes down to: What octane fuel did Subaru use when it developed the high-octane ignition table? One possibility is that Subaru set up the the high octane ignition table based on the recommended fuel octane, 91. In that case you probably won't see a difference when doing a Vishnu reset while running 91 octane or 112 octane. But what if Subaru had observed that 94 octane was available in many parts of the country and chose to construct the high octane ignition table based on that? Or what if Subaru, because it was easy to do, simply inserted its JDM 100 RON table as the high octane ignition table? In these cases, I expect you would see a difference.

[Jon [in CT]]

I also wanted to mention that it's possible (and perhaps more elegant) to implement the high octane ignition table entries as values to be added to the low octane ignition table entries in order to obtain the maximum advance. In this case, FHI refers to the low octane ignition table as the Base ignition table and refers to the high octane ignition table as the MBT ignition table.

It sure would be nice if Shiv (or any other ECUTEK reflasher) could join in, as they can actually inspect the data.

[AZScoobie]

Originally posted by Jon [in CT]
I guess the results of your test will depend heavily on how Subaru has initialized the two ignition tables. Recall that, according to my research, the ECU will never retard timing below the values represented in the low octane ignition table and will never advance timing beyond the values represented in the high octane ignition table. I think the question comes down to: What octane fuel did Subaru use when it developed the high-octane ignition table? One possibility is that Subaru set up the the high octane ignition table based on the recommended fuel octane, 91. In that case you probably won't see a difference when doing a Vishnu reset while running 91 octane or 112 octane. But what if Subaru had observed that 94 octane was available in many parts of the country and chose to construct the high octane ignition table based on that? Or what if Subaru, because it was easy to do, simply inserted its JDM 100 RON table as the high octane ignition table? In these cases, I expect you would see a difference.

Great thread. What I suggest as to why the IGADV reset works is simple. You take a stock car, you slap on a boost controler and up the boost. This causes the ECU to default down to the low advance map and you loose power. In a sense you have exceeded the knock threashold of the octane you are using based on the conditions (more boost). Now you add in some higher octane fuel and perform the reset as described. Now you have the best of both worlds. High advance and high boost. As long as you add in the required octane you stay this way. But, in the end on sunday when you fill up for the weekly commute and add in 91 your ecu responds with the low advance map and or the ECU trims back troubled areas. You end up with reduced power.

I agree, It would be interesting to see some posts or better yet some graphs of the low and MTB ignition maps and some examples of the fine tune ignition retard map(dont know what to call this). This way we can see the effects of slight knock and the coresponding RPM points. It is also going to be interesting find out what these tuners are going to use as solutions to gaining more power and consistancy. In the end the ECUTEK is just a tool and it will all depend on what the tuner changes/adds. The idea of making the low map match the high map to me with my limited knowledge seems scary. I worry that tuners will end up blind folding the ECU which could cause major damage. In the scenario in the above paragraph would you end up running high advance and high boost with that 91 octane on the way to work monday morning? Is the suplimental retard map(the one based on knock) enough to handle the knock that could be present? Time will tell.

What I hope is that these tuners will take the time to learn how to make the ignition advance scheme faster with more advance for power, yet keep all safetys intact. Teach the ECU that its goal is to run the most advance that it can within reason and if knock happens that it might not happen again.. Something along the lines of the remarkable ME5 or ME7 bosch ignition scheme. Reset the ECU and you start out with a fairly agressive advance map. Run WOT a few times and trims are made per the 4-6 sensitive knock sensors. If knock occurs reduce advance in that area to stop the knock. Then on the next cycle give it right back. In effect, these systems "ride" the knock sensors. Maximum advance and consistant performance is the result along with higher effeciency and better fuel mileage. Maybe the Subie engines cant take slight knock like an Audi engine. Maybe the Subies knock sensors and related systems cant support that kind of sensitivity. Maybe I am dreaming... Its a good dream though.

CT

[Jon [in CT]]

Bump for Shiv to respond to the question he posed.

[Vishnu Performance]

Wow... a lot of info here. But here's what I know (or at least I think I know) from our remapping efforts :

1) There is only one ignition timing table and one knock correction table.
2) The load axis for the fuel, ignition and knock correction tables is MAF, not MAP.
3) The ignition timing map is very conservative and relies on anywhere from 0-12 degrees of additional advance which is provided by the active knock correction system.
4) The knock correction tables resembles the ignition timing table with respect to having MAF and RPM as its X and Y axis. However, in its cells is maximum knock correction authority, not absolute base timing value. At low loads (off boost), knock correction authority is 0, meaning that no additional advance is allowed to be added upon the base ignition table values. Under boost (and especially around 5000rpm, knock correction reaches its maximum authority range (or around 10-12 degrees depending on ECU type). I can only presume that the engine calibrators at FHI determined this RPM range to be particularly "trouble-prone", requiring the "go ahead" from the knock sensor to add in the full amount of timing (learning to advance timing instead of simply assuming that is okay to do so). Naturally, the underlying timing values in the ignition timing tables are unusually low at this rpm/load range.
5) Knock correction authorty range tapers off as RPM and Load goes up (above 5500rpm). I suspect this is because knock sensors tend to become inaccurate at higher engine speeds (unable to differenentiate between actual knock and normal engine noise).
6) The "vishnu reset" simply keeps in the engine operating in a rpm/load zone where base timing is very conservative and knock correction authority range is relatively large. When the engine assumes this rpm/load, the active knock correction system readily adds in maximum positive authority range due to an absense of knock and (I suspect, engine noise). I could make this special reset not work by simply advancing base timing too aggressively in his rpm/load area. This would induce knock prematurely and never all the ECU to add in the allowable knock correction. Needless to say, I keep these rpm/load zones purposely detuned to facilitate this accelerated learning technique.

Cheers,
Shiv
www.vishnutuning.com

[jsarfaty]

So Shiv, there is no fine knock correction table? There is only the base ignition map (low octane map), the ignition advance offset map (high octane map), and one scalar advance multiplier?

for ignition:
total = base map + advance multiplier * advance offset map

?

So if you register a lot of knock events at an RPM/Load range where the knock correction authority (advance offset map) is 0, it will still lower the advance multiplier? This wouldn't change ignition at the RPM/Load where knock occured (because the advance offset map is 0 at that cell), but would retard it over the rest of the rev range where knock wouldn't necessarily have happened at all?

This situation seems to call for a fine tune knock correction table that basically would contain a smaller learned advance multiplier for each cell in the ignition table.

Anyone agree?
__
Josh

[Jon [in CT]]

Originally posted by Vishnu Performance
1) There is only one ignition timing table and one knock correction table.
2) The load axis for the fuel, ignition and knock correction tables is MAF, not MAP.
3) The ignition timing map is very conservative and relies on anywhere from 0-12 degrees of additional advance which is provided by the active knock correction system.
4) The knock correction tables resembles the ignition timing table with respect to having MAF and RPM as its X and Y axis. However, in its cells is maximum knock correction authority, not absolute base timing value. At low loads (off boost), knock correction authority is 0, meaning that no additional advance is allowed to be added upon the base ignition table values. Under boost (and especially around 5000rpm, knock correction reaches its maximum authority range (or around 10-12 degrees depending on ECU type). I can only presume that the engine calibrators at FHI determined this RPM range to be particularly "trouble-prone", requiring the "go ahead" from the knock sensor to add in the full amount of timing (learning to advance timing instead of simply assuming that is okay to do so). Naturally, the underlying timing values in the ignition timing tables are unusually low at this rpm/load range.
5) Knock correction authorty range tapers off as RPM and Load goes up (above 5500rpm). I suspect this is because knock sensors tend to become inaccurate at higher engine speeds (unable to differenentiate between actual knock and normal engine noise).
6) The "vishnu reset" simply keeps in the engine operating in a rpm/load zone where base timing is very conservative and knock correction authority range is relatively large. When the engine assumes this rpm/load, the active knock correction system readily adds in maximum positive authority range due to an absense of knock and (I suspect, engine noise). I could make this special reset not work by simply advancing base timing too aggressively in his rpm/load area. This would induce knock prematurely and never all the ECU to add in the allowable knock correction. Needless to say, I keep these rpm/load zones purposely detuned to facilitate this accelerated learning technique.

Cheers,
Shiv
www.vishnutuning.com

Thanks for replying. I urge you to print out and read the following US patents, all filed by the same two guys at FHI at around the same time and all titled Method for controlling ignition timing of an internal combustion engine:
5,000,150
5,035,219
5,038,736
The first one lays out the basic strategy while the other two add embelishments to that basic strategy. These patents might alter your interpretation of the ignition and correction maps which, in turn might alter your tuning strategy.

Here are the essentials.

There is a base ignition timing table. Each entry (IGB) in that table represents a basic ignition timing advance and is indexed by load and RPM. The ECU will never retard ignition advance for a load/rpm point below the IGB value. Each value in this table represents the timing advance that produces maximum safe torque (i.e. without knock) when operating on the lowest supported octane fuel. I believe this table corresponds to what you call the "ignition timing table" and might explain why it seems so conservative -- FHI might have devloped this table using 87 octane gas.

There is a maximum advance table. Each entry (MBT) in that table represents a maximum value that can added to the basic ignition timing advance and is also indexed by load and RPM. The ECU will never add, for any load/RPM point, more than the MBT value to the basic ignition (IGB). Each value in this table represents, when added to its corresponding base table entry, the timing advance that produces maximum safe torque (i.e. without knock) when operating on the highest suppported octane fuel. I believe this table corresponds to what you call the "knock correction table" and the reason that its entries are zero in the low-load and low RPM regions is that the low octane fuel can successfully support the ideal (i.e. maximum brake torque) timing in those regions without knock and increasing octane in those regions won't change the ideal timing. And you mentioned that the "knock correction table" entries had the largest values in the high-load 5K RPM area. This is likely the area most prone to knock and therefore the area that benfits most from higher octane fuel.

I believe the "Ignition Advance Multiplier" which you mentioned in your first post in this thread is used to determine what the patents call a "rough" (i.e. coarse) correcting quantity. Think of the IAM as an indicator of percentage of the MBT entry to add to the IGB entry, with its lowest value meaning "don't add any advance to the base table entry and its highest value meaning "add 100 percent of the maximum advance to the base table entry." This concept is described in US patent 5,000,149.

Not mentioned in your description is what the patents call the "fine correction" table. It's a table stored in volatile memory (i.e. it's reset when the ECU is reset) with the same dimensions as the base ignition timing table and the maximum advance table. It contains advance corrections for each load point, to fine tune the advance after the coarse correction has been applied.

[AZScoobie]

Originally posted by Jon [in CT]
Thanks for replying. I urge you to print out and read the following US patents, all filed by the same two guys at FHI at around the same time and all titled Method for controlling ignition timing of an internal combustion engine:
5,000,150
5,035,219
5,038,736
The first one lays out the basic strategy while the other two add embelishments to that basic strategy. These patents might alter your interpretation of the ignition and correction maps which, in turn might alter your tuning strategy.

Here are the essentials.

There is a base ignition timing table. Each entry (IGB) in that table represents a basic ignition timing advance and is indexed by load and RPM. The ECU will never retard ignition advance for a load/rpm point below the IGB value. Each value in this table represents the timing advance that produces maximum safe torque (i.e. without knock) when operating on the lowest supported octane fuel. I believe this table corresponds to what you call the "ignition timing table" and might explain why it seems so conservative -- FHI might have devloped this table using 87 octane gas.

There is a maximum advance table. Each entry (MBT) in that table represents a maximum value that can added to the basic ignition timing advance and is also indexed by load and RPM. The ECU will never add, for any load/RPM point, more than the MBT value to the basic ignition (IGB). Each value in this table represents, when added to its corresponding base table entry, the timing advance that produces maximum safe torque (i.e. without knock) when operating on the highest suppported octane fuel. I believe this table corresponds to what you call the "knock correction table" and the reason that its entries are zero in the low-load and low RPM regions is that the low octane fuel can successfully support the ideal (i.e. maximum brake torque) timing in those regions without knock and increasing octane in those regions won't change the ideal timing. And you mentioned that the "knock correction table" entries had the largest values in the high-load 5K RPM area. This is likely the area most prone to knock and therefore the area that benfits most from higher octane fuel.

I believe the "Ignition Advance Multiplier" which you mentioned in your first post in this thread is used to determine what the patents call a "rough" (i.e. coarse) correcting quantity. Think of the IAM as an indicator of percentage of the MBT entry to add to the IGB entry, with its lowest value meaning "don't add any advance to the base table entry and its highest value meaning "add 100 percent of the maximum advance to the base table entry." This concept is described in US patent 5,000,149.

Not mentioned in your description is what the patents call the "fine correction" table. It's a table stored in volatile memory (i.e. it's reset when the ECU is reset) with the same dimensions as the base ignition timing table and the maximum advance table. It contains advance corrections for each load point, to fine tune the advance after the coarse correction has been applied.

Incredible research Jon. My hats off to you on this post. I learned alot and for that I thank you.

CT

[dug-e-fresh]

If knock occurs reduce advance in that area to stop the knock. Then on the next cycle give it right back. In effect, these systems "ride" the knock sensors.

Essentially what the Utec does...

MAF vs MAP

Seems to me MAP is the evil step-son of MAF.

If you knew the total volume and temperature (averaged pre and post intercooler) of the compressed air the MAP is reading and extrapolated this against RPM, would you not get the MAF value?

Assuming that.... MAF is a function of MAP, and vice versa... similiar the way hp is a function of torque.

I would think our ECU would rely on the "total" package value for tuning purposes rather than extrapolating MAP to MAF, and as such incorporating stacked tolerancing like errors...

def

[Jon [in CT]]

On the question of the "fine correction" table, there may actually be four of them, one for each cylinder. Since the ECU, when it detects knock, can identify which cylinder caused it, it might want to maintain a separate ignition advance for each cylinder. However, there's an interesting FHI patent on knock control for the case where there is insufficient memory to store a separate table for each cylinder. It's US patent 4,726,339 and its main ideas are:

• identify which cylinder is most likely to knock and mount a knock sensor close to its combustion chamber
• only pay attention to that cylinder's knock
• adjust timing advance to control knock in that one cylinder
• apply that same timing advance to the other cylinders

Since the EJ20's knock sensor is obviously mounted closest to Cyl #4's combustion chamber, could that mean that Cyl #4 is ...

[AZScoobie]

Originally posted by Jon [in CT]
On the question of the "fine correction" table, there may actually be four of them, one for each cylinder. Since the ECU, when it detects knock, can identify which cylinder caused it, it might want to maintain a separate ignition advance for each cylinder. However, there's an interesting FHI patent on knock control for the case where there is insufficient memory to store a separate table for each cylinder. It's US patent 4,726,339 and its main ideas are:

• identify which cylinder is most likely to knock and mount a knock sensor close to its combustion chamber
• only pay attention to that cylinder's knock
• adjust timing advance to control knock in that one cylinder
• apply that same timing advance to the other cylinders

Since the EJ20's knock sensor is obviously mounted closest to Cyl #4's combustion chamber, could that mean that Cyl #4 is ...

.... the one that slings the rod out the case....

CT

[hotrod]

Thanks Jon for all that research, it is interesting to see how the empherical work of Shiv and the theoretical research of the patents works out. The big danger of empherical evidence is some times you can come up with a perfectly logical theory that just happens to fit reality over a specific range of circumstances, but may in fact not be correct. (no slight intended to Shiv his is the toughest nut to crack)

Thanks Shiv for the qualification that you realize your learning as you go on this and I would caution everyone to understand the limitations that sort of learning process creates. It is terrific that Shiv is willing to lay out on the table what as he so eloquently put it "or what I think I know".

Now that Jon has layed out a theoretical basis for the timing computation process Shiv should be able to figure out a test case (sometimes easier said than done) that will clearly show which possible interpretation is closer to being correct.

Along that line, Shiv your observation about maximum knock correction authority being in the 5000 rpm range makes perfect sense in that , that is the approximate rpm for max torque, which by definition is the point where the engine develops max VE , and as a consequence max MEP (Mean effective Pressure). There would be a natural reduction in cylinder pressures at higher rpms, do to the drop in VE which results from the shorter intake cycles. I have run across several sources that state the max torque portion of the rpm band it typically the most knock prone.

Keep up the brain storming session guys this is getting fun!!

Larry

[NotoriousWRX]

Great thread guys! I'd like to hear from Shiv once he reads through those patents.

[robtmc]

I have run across several sources that state the max torque portion of the rpm band it typically the most knock prone.

I can vouch that this is a fact on centrifugal supercharged engines not originally designed for forced induction. My Mustang with a 9 lb blower always needs more retard programmed in near 4500 rpm (torque peak) than above or below.

[dave_the_driver]

Shiv and All, I know the UTEC is sold by a competitor, but will Shiv's "trick" work if a TurboXS UTEC is installed? Anyone?

dug-e-fresh mentioned the UTEC in his reply above where he mentioned MAP versus MAF, but, as a newbie to my UTEC and tuning, i don't understand the implications of what he said for UTEC owners who are considering doing Shiv's trick. It seems harmless to give it a try, though.

David

Originally posted by Vishnu Performance
Just a little trick that has been shown to accelerate the factory ECU's ignition timing learning process.

Background: There is something called "Ignition Advance Multiplier". It represents, by some complicated algorithm, the average learned positive knock correction applied to the ignition maps. It's represented in 1/8th degrees increments. 1 being the lowest and 16 being the highest (1/8 to 2 degrees in absolute terms).

[SNIP]

Cheers,
shiv
www.vishnutuning.com

[dug-e-fresh]

Since the Utec completely takes control of ignition advance over the set TPS/MAP threshold... this would do no good for anything above that threshold.... however, below the threshold, typically set @ 60% TPS, this could in fact give better throttle response to the car because below that threshold, the OEM ECU is in control.

TXS- correct me if I am wrong...

def

[AZScoobie]

Originally posted by dug-e-fresh
Since the Utec completely takes control of ignition advance over the set TPS/MAP threshold... this would do no good for anything above that threshold.... however, below the threshold, typically set @ 60% TPS, this could in fact give better throttle response to the car because below that threshold, the OEM ECU is in control.

TXS- correct me if I am wrong...

def

Correct as long as you have "ECU" programmed into those columns the utec lets the stock ECU passthrough ignition advance. The longer you own the utec the more you will start to program below the 60 column. Especialy with big turbos.

Clark

[dave_the_driver]

Originally posted by robtmc


I can vouch that this is a fact on centrifugal supercharged engines not originally designed for forced induction. My Mustang with a 9 lb blower always needs more retard programmed in near 4500 rpm (torque peak) than above or below.

With 91-octane ethanol gas and an MRT turbo-back at 5000 feet elevation and with 15 lbs of max boost, Nathan was getting knock signals from the UTEC almost all of them at the 3650 and 3950 RPM points doing WOT pulls from 2000 RPM in 3rd, 4th, and 5th gear. Maybe that's where my peak torque is ... i have no idea since there's no 4-WD dyno here.

[TurboRex]

Thanks shiv for the shiv reset trick. I am going to have to try that and log what happens.

Is there another trick for setting long term fuel trim. On my car, it can take up to two months to have the car stop changing a/f at wot. This is a real pain when you are trying to tune the vehicle. I have the ecutek so I constantly monitor ltft and keep the car rich until it has completed learning. My car normally takes out about 15% once the ecu is fully learned so you can see why it is a concern to me.

thanks,

Greg

[Juliano]

Ok. Cool thread. Great tip. Can we take it one step further? Would it feasible to modify the ECU's power source by wiring in an interupt switch that would allow the user to reset the ECU by way of a dashboard mounted switch, rather than crawling into the passenger footwell every time? Concern: Would the insertion of additional wire and a switch alter the characteristics of the current in any adverse manner?

How 'bout it Shiv. A new and improved Shiv reset?