Tuning AVCS without a dyno (AVCS for Gruppe-S header post 12 and 24)
 

I still didn't mess up my AVCS, so am thinking about making this my next project. :D

In order to simplify it and reduce the number of flashes I so far came up with this idea. This is different intake valve settings that I plan to use (excuse my poor excel skills):

https://s1.postimage.org/827466de9b/...t_00.48.03.png

I would modify my current table first and flash that as a baseline, and make 1 tune with -4 deg and 2 tunes with +4 and +8 more advance. (The dip at 4200 in 2 of the variations is to see if I can get rid of a, in all likelihood fake, lean hump in my AFR curve).

I assume that following the original curve is not all that bad (OTS OFT AVCS) and that I, in any case, would cover most settings the would be realistic to potentially offer gains?

E.g I suppose it would hardly be realistic to expect +25 advance at 7400 to be reasonable, or +20 at 4800? So by staying around the original curve, I suppose I should have the best chances of actually finding any improvement?


Next, I would drive off to a flat road and do 4 pulls on each tune. I'm not entirely sure how to see which version is better. Virtual dyno is out of the question, due to the inaccuracy, unless it would surprise me with huge gains.

Here an example of 4 pulls done on the flattest road I have available, the fluctuations are too big (also you can see my lean hump in the AFR and that I tried to make VD show load in the boost graph without much success):
https://s1.postimage.org/83aoaib41b/leaner.jpg

Instead, I was considering if mass airflow or load (or calculated volumetric efficiency) would be a good indicator of what makes more power? I would suppose if the engine is pumping more air that it would mean that it's making more power?

Here the loads of the same 4 pulls overlaid in Datazap (the blue one is slightly ahead in rpm). Of course, it wouldn't be necessary to overlay it like this, it just to visualize that they are indeed fairly similar compared to what VD outputs:

https://s1.postimage.org/9rwslutr7z/...t_01.10.18.png
https://datazap.me/u/tor/test-2?log=...-4-21-22-23-24

What do you think about my idea? Any ideas or input appreciated. Or any fallacies in my reasoning or proposed method?

Thanks!

Not used the VVT tab in VGI's tool? It's designed to look at VE, but load values could also be an indicator if you don't change any MAF scaling. The helpful thing of the tool is that it will use the actual logged values of AVCS rather than the target in the table.

Don't forget that it might be worth taking some timing out the ignition tables as AVCS can have a significant impact.

You'll probably find that the intake has more effect than the exhaust cam too.

I only looked a little bit at it, but it seems to do the same as your formula? And then it's just comparing automatically instead of having to go over the curves manually, is that correct?

As far as I read ignition timing also affects the optimum AVCS timing. If I keed the changes fairly small, like not deviating from the shape of the original curve, wouldn't it be better to stay with the timing I intend to run and let the ECU pull timing? I understand if I opted to make flat tables and tried e.g. 40 deg intake advance at full throttle it might have a more profound impact on knock? This was in part my reasoning for sticking around Shivs curve.

My original plan was to make a flat curve from 5200 rpm to 7400 with 0 to 25 deg intake advance in intervals, which would equate to 6 tunes plus a baseline tune. With 4 pulls per tune, 28 pulls - bit much! Whereas sticking around Shiv's curve, seems to not only test out the most realistic option in the least amount of pulls but also offer the most safety without having to reduce timing?

Another thing, if I go down this route. Do the pulls even have to be made on a flat road or for that matter the same road? I would greatly reduce the chances of curious men in cars with blue lights on the roofs if I could do the pulls at different locations.

I looked at it a bit more... and now I get it. I can throw in all the runs I make and then let Vgi's tool do the comparison for me of all logs at the same time and it will automatically snake it's way between the curves as required. I suppose the more runs I make for each setting the more accurate it will be.

Thanks for the tip!

First off caveat, I've done some studying with this and toyed*a bit* with AVCS timing, but novice here so take my advice with a grain of salt :D

From a string empirical approach, even on a dyno AVCS optimization takes time. Doing runs on a road takes even longer obviously. Create a handful of tunes, load them to your tablet, load one to your car (wait), log a run, load the next one (wait), make a run, and repeat for as many tunes as your table can hold. Analyze results, make new tunes, do it again. It's a fairly iterative process that can be shorted if you have some experience with this kind of tuning before (i.e. make educated guesses on next adjustments instead of the blanket +/- approach). I dabbled but gave up as I just don't have the time, and my wife thinks I've gone crazy. Cam timing and ignition timing do affect each other, although my typical approach is to tweak AVCS intake/exhaust first... get that right.... then go after timing. Ideally you'd probably make another AVCS pass after that, and then timing again... but that's an unrealistic amount of work for me. Also keep in mind that we're not just talking about pulling timing - AVCS changes can actually reduce compression in certain areas and allow advancing of timing - so you can just rely on tuning from watching where the ECU pulls timing via the knock sensor.

Again, I'm a newbie, so Kodename or anyone else feel free to correct me.

Yes, no doubt a long process! I might be pulling my timing back a little bit anyway since the the last bit I added didn't really show any noticeable gains. As you said I can always revisit that part for fine adjustments.

I was curious about decreasing the intake advance where I have the lean hump in my AFR curve, so I flashed the yellow tune from above.

Completely unscientifically without any same day comparative runs, I suppose I can imagine why it won't work (optimally) with looking at load or MAF.

When comparing with older log, I can see load and MAF going down, while VE remains about the same. If it's a fake lean and air is rushing straight through the cylinder, it would make sense it's reading a higher MAF/load without a difference in VE? If that theory is correct, I suppose it would only apply to when there is too much overlap?

Anyway, I'll stick with VE and use the Vgi tool for further tests. It will probably be a while (looking for better weather and I have too much other stuff to do at the moment) but I'll update here on my progress when I get that far.

To be fair, if you did 4 pulls on an actual dyno the variation between traces would not look too far off from what you'd see in your virtual dyno. but yeah.. you'd have to expect the virtual dyno app to be a bit off in terms of accuracy.

Carry on, good sir.

In any case I will stick with Vgi's tool and use VE as I suppose it will give the best consistency.

Unfortunately, I have to go on a long business trip starting tomorrow. But if the weather still plays along I might get it done at the end of November - main thing is no salt on the roads. Otherwise my car will be in winter sleep and it will be a project for the spring in that case.

You: Yep, just going for a drive...to do some tuning...again...
Wife: O......K.......


:D:D:D

Yes, you guys with wifes are heroes fighting for the survival of the human race. You should all be given medals for putting up with all that BS! :D

Back to AVCS... I did my 4 logs today under great trouble (traffic is horrible here). So doing 3rd gear pull to redline is quite a challenge. Especially trying to keep somewhat consistent IATs.

I drove prepared with my "base tune" to my favorite road. Once I arrived I chickened out with the timing and decided to pull -1.4 deg. So since I had to flash again, I thought I might as well start out with the most aggressive advance. The "plus 8" purple line:

https://s1.postimage.org/827466de9b/...t_00.48.03.png

It changed the sound of the engine completely. Like the OFT stage 1 groooar around 4000 rpm, just the grooar is there all the way up through the rpm. From the butt dyno it felt faster, but I discounted it as placebo.
Here the log +8:
https://datazap.me/u/tor/tor-236-8base?log=0

Next up was -4, I didn't really expect much from this as my gut feeling was that this would be a step in the wrong direction. But just for verification. It did indeed feel meeeh, especially around 4200 rpm where I made the largest dip in the advance. It didn't really help on the lean hump either, so I'll either leave it or go back to fix it with Engine Load Compensation:
-4 log:
https://datazap.me/u/tor/tor-236-44?log=0

Next +4, and then the "base tune" again with -1.4 deg timing. Both felt kind of similar.
+4 log:
https://datazap.me/u/tor/tor-236-44?log=1
Base log:
https://datazap.me/u/tor/tor-236-8base?log=1

On the way home, I prayed the +8 tune would be the one that had the best VE, because it sounds soo cool! :D

Back home I throw all the logs into Vgi's too, and it comes up with this: :thumbsup:
https://s8.postimage.org/hiehqrh45/2...12221974_n.png

Values copied into my excel sheet:
https://s8.postimage.org/w6pe1vrut/S...t_18.50.47.png

Now I wonder if you can tune AVCS by sound. Haha :D

Next up, it to make one or two more tunes with even more advance above 5200 rpm. Maybe two more with smaller steps, like +3 and +6. Once done I'll try to increase the exhaust retard and see if that brings anything.

Looking over my logs more thoroughly today.

First, I don't like that the cams appear not to follow each other around 6000 rpm with the tune most advance.

https://s2.postimage.org/7j695tubt/S...t_13.54.57.png

So regardless of what the Vgi's tool says, I will dial the cams a bit back there. It doesn't appear in the other tunes.

If I compare the logs individually, I get other results.

E.g. when taking the "+8 tune" out of the Vgi tool: It follows the +4 cams up to 5200 rpm, then jumps to the -4 tune from 5200 to 5600 rpm, then back to +4 up 7200 rpm and finally the -4 at 7400.

When comparing +8 and +4: +8 wins in every rpm.

When comparing +8 and -4: +8 wins in every rpm, except suprisingly in 4200 when I made the strongest dip.

When comparing +8 and base: +8 wins in every rpm, except again 4200 and 5200.

Given the above, I think this methode still leaves some guess work. But another factor to consider may be consideration for the hardware. I will disregard what comes out at 4200 so the valves don't have to cycle up and down and in the area from 5200 to 6400 rpm, I will reduce the advance a bit.

What is shown conclusively is that above 6400 rpm to 7200 rpm it benefits from more advance. At 7400 I think the results may be affected by running into the revlimiter where the timing reduces. I could probably check this by editing the logs to exclude parts above 7450 rpm. For future versions of the Vgi tool, a max rpm option might be useful.

I decided to combine using the Vgi tool with another method.

What I did was using Kodename47's VE formula in my logs and using log stats. I would open 4 instances (one for each tune) side by side like this:

Here shown counts in each cell

https://s17.postimage.org/f9c2djmun/avcs_count.jpg

I then printed them out and by hand looked at each rpm cell to see which made the highest VE for each statistical option:

https://s17.postimage.org/6r2m9nygv/IMG_7657.jpg

After that, I checked which tune got the most hits for each rpm. This method of elimination gave some clear winners in specific rpm, especially above 6400 rpm. From 3200 to 5600 it was a wash between +4 (38 deg) and +8 (40 deg), some cells favoring 38 and some 40. The cell with the highest uncertainty was 6000 rpm.

Given the above, I decided to keep 39 from 3200 to 5600 rpm. I simply don't think neither the cams nor my logs are accurate enough to justify letting it cycle between 38 and 40 according to rpm. So I simply averaged it out.

6000 was the rpm where the cams seemed to lose control with the tune with most advance, so my plan was to reduce them a bit here regardless. Statistically less advance was favored here, but as I don't think the cams can cycle fast enough I decided to make a blend between 6000 and 6800.

The changes compared to the +8 tune makes a slightly less aggressive value noise (unfortunately :D) and keeps the valves synchronous through 6000 rpm:

If it matters, the curve ended up looking pretty nice and smooth:

https://s17.postimage.org/oxpigr9z3/...t_13.40.06.png

I test drove it extensively, including highspeed autobahn, and it doesn't cause any knock or unwanted side effects. It actually didn't even FLKC more than -0.65 at high rpm in 5th gear with my -2.46 A/B and -1.05 C/D TCPC disabled but with -1.41 timing from my normal timing at high load. So actually, with a slight net timing increase.

I couldn't rev it out in 5th due to traffic. But given that it doesn't FLKC anything noteworthy in the first 4th gear pull, it shouldn't in 5th either.

https://s17.postimage.org/9g20d9lkf/...t_14.07.53.png

What is missing now is Virtual dyno pulls from before/after. But since it's raining all time and even snowing occasionally it won't be anytime soon. I did try making pulls in the wet, and the numbers show lower as expected but with really nice looking curves. It's probably going to remain wet here for a foreseeable future and when they start to salt the roads the car won't come out of the garage. So don't hold your breath. The wet curves:

https://s17.postimage.org/efzgl7brj/237_wet.jpg

The settings were tested out on another car with Gruppe-S header and the owner said it felt like an improvement as well and ran without problems. A 3rd car is being flashed today too.

So if anyone else with a Gruppe-S wants to try it out, here is the intake cam, low loads is stock/OFH values:

If anyone decided to use these intake cam setting, I would appreciate a comment with your results/impressions in this thread. Thanks.

https://s17.postimage.org/3zjcj2fdr/...t_13.36.24.png



Edit:
Future plans for editing exhaust in 3 flashes:

https://s17.postimage.org/mx85np2yn/...t_16.38.20.png

I'm going to test these cam settings with my OpenFlash Header.
Vincent is quite happy with these on his Gruppe-S
Going to try to get some results in the weekend and will post a log here, together with my buttdyno results :D

I'm not sure this will be better on an OFH, after all Shiv tuned the cams specifically for that header.

But sure you can try. Maybe the OFT cams are set conservative because it needs to run over a broad range of cars. Only Shiv will know how he decided on the OFT cams. I'm just saying this might not be an improvement for you.

If you run a lot of ignition timing, it might be a good idea to take out -1.4 deg from load 0.9 and 4000 rpm up. On my car, Vincent's and a 3rd one (all Gruppe-S) it doesn't cause FLKC. On my car it infact looks like there is room for more timing now.

Regarding gains. I did some messurements of the stock tires today vs. the 18 inch I have been running over the summer. As it turns out the circumfrence of the tires are practically the same, give 2 or 3 mm.

So the actual gains look like this:
Highest and lowest of 6 pulls today, and 2 highest and lowest out of 6 with my pre AVCS tune:

https://s33.postimage.org/ksop1xse7/2371_vs_235.jpg