Pages: 1 ... 13 14 [15] 16 17 ... 26
Author Topic: ME7.9.10 - Understanding the torque model  (Read 177421 times)
IamwhoIam
Hero Member
*****

Karma: +43/-99
Offline Offline

Posts: 1030


« Reply #210 on: May 16, 2018, 10:22:39 AM »

long term trims on VAG are rkat_w and fra_w as far as I can remember
Logged

I have no logs because I have a boost gauge (makes things easier)
woj
Hero Member
*****

Karma: +41/-3
Offline Offline

Posts: 500


« Reply #211 on: May 16, 2018, 11:47:02 AM »

There is no doubt about fra_w.

I looked at both rka_w and rkat_w, and actually, rka_w is the one that is applied at the end of the day. rka_w is either directly rkat_w (if a bit on one of the codewords is set, can't be bothered to check what codeword that actually is, but it is not on in my bin anyhow), or rka_w is derived directly from rkat_w with the help of the current nmot (some sort of rpm dependant scaling, again, can't be bothered to dive into this). So it probably does not really matter, provided one of them is logged.
« Last Edit: May 16, 2018, 11:49:47 AM by woj » Logged
woj
Hero Member
*****

Karma: +41/-3
Offline Offline

Posts: 500


« Reply #212 on: May 17, 2018, 12:42:53 PM »

I wonder if anybody has enough experience with E85 to help me ball park my cranking tables (FKSTT). On the first attempt (the _old file in the attachment) I looked up factory tables in the Saab files I was pointed to elsewhere here. I figured out their meaning / unit (or so I think) and converted them to my FKSTT E85 table that blends with the factory one (top of shots). The values seemed excessive to me, "but then, what do I know about diamonds", I just decided it's the best guess I can do.

Well, it isn't. Warm starts are OK, but "cold" (~20-25*C after long stay, that's the weather we have now) are not. The car starts on the first attempt, but rpms choke down to ~500 for a while and then it comes to life (the symptom happened twice, consistently the same at similar temperatures). Clearly overfueled, not to mention the fuel smell. And that is on 22% content, so the effective factor for these temperatures was ~7-8, so more than twice as much as stock.

For now I arbitrarily rammed the E85 table down (second shot) and will see how it goes tomorrow. But this will be a lot of experimentation, the values still seem high to me, could use some help here with approximation.

This can of course be also due to after start enrichment that I have not touched yet, it is stock for both fuels, perhaps it is not overfueled in cranking, but undefueled in afterstart. What do you think?
Logged
prj
Hero Member
*****

Karma: +903/-420
Offline Offline

Posts: 5787


« Reply #213 on: May 17, 2018, 02:49:48 PM »

You can't blend 1:1 between the tables, they are not linear, you need a table for a blending factor.
Same with ignition timing, need a separate table for factor based on ethanol percentage.

They are not excessive, because ethanol does not evaporate easily, only the gasoline portion evaporates.
So basically you're washing fuel down the bores really hard on cold start with ethanol. On E70 you need 3x as much enrichment as on gasoline on crank when really cold and so on...
Logged

PM's will not be answered, so don't even try.
Log your car properly.
woj
Hero Member
*****

Karma: +41/-3
Offline Offline

Posts: 500


« Reply #214 on: May 17, 2018, 03:17:02 PM »

They (the cranking tables) were lineary blended on the Saab bin I think, but I'd have to double check that. For ignition and boost I do have a kline factor for blending the two tables (following your earlier hint in the other thread Wink), but these are so far just stock for both fuels. The code fix is easy then, but it still does not tell me what the kline for fkstt this should be.

OK, but 3x enrichment for E70 on freezing temperatures does not scale (in my mind) to effective 2+x scaling I got for rather warm start ups on E20, so at least in the upper part of the table that is now applicable for me, the values are too high, kline or uniform blending, no matter.
Logged
jcsbanks
Full Member
***

Karma: +15/-3
Offline Offline

Posts: 125


« Reply #215 on: May 17, 2018, 06:52:17 PM »

If you have available flash for the tables and code, and RAM for a few variables, if you are not doing and instead using linear equations, like prj recommends I would make tables for blending. I use three for fuel, ignition and load. You could share ignition and load perhaps, but on my target engines they have different needs where you may have very gentle ignition retard after reaching MBT with increasing ethanol content due to the faster flame speed of ethanol, but steeper drop in max engine load due to running out of pump on engines with a single DI pump.
Logged
prj
Hero Member
*****

Karma: +903/-420
Offline Offline

Posts: 5787


« Reply #216 on: May 18, 2018, 12:34:36 AM »

Linear blending does not work on Ethanol for cold start, end of story.
I highly doubt it is done that way in any OEM bin.
Logged

PM's will not be answered, so don't even try.
Log your car properly.
woj
Hero Member
*****

Karma: +41/-3
Offline Offline

Posts: 500


« Reply #217 on: May 18, 2018, 01:07:06 AM »

Well, actually I have been just now rechecking the Saab maps, it is far more complex than just non-linear blending. For E85 it seems the cranking injection times are done in varying pulses depending on the temperature and number of combustion cycles. Eg. when very cold it actually first does less injection (50%) for 4 combustions, then scales it up to 200%, while on warm first two injections are at 200% and the rest at 100%. I still do not see a scaling map for blending there, but the whole stuff is hardly penetrable, and I am guessing this blending is hidden in all the maps like the one I have seen. All E85 treatment there seems to be much more complex than one would expect.

I still do not have any answer to my initial question though - put blending method aside - the actual E85 cranking factors or the scaling distribution between 95 and E85 ("3x at very cold" does not cut it). For now I looked at my first map (and even the second one I made) and realised it is totally inconsistent, adds 3.8 times more fuel at -15, 7+x more fuel at 0, and then 4x more fuel at 20. That is essentially crap, should have noticed it earlier. Today's morning start with tuned down table was better, but still choked. Will sort it out experimentally for my E22, this will give me at least one scaling point for the non-linear blending.

Logged
woj
Hero Member
*****

Karma: +41/-3
Offline Offline

Posts: 500


« Reply #218 on: May 18, 2018, 04:15:55 AM »

This is going to be a nightmare to get right under these conditions. My first map was overfueled, but the problem lies elsewhere, the choke is due to throttle closing, just did a quick log (engine temp 30*C). This was done with stock FKSTT, fuel is E22. The startup reaches 1200 rpm nicely, then the ECU decides this is too much and aggressively rams it down. Now that I think of it I get occasional things like this also on stock maps / fuel, it has something to do with throttle adaptation, and this gets reset each time I flash I believe. So getting the fkstt factors right is one thing, the other one is to be patient enough for the ECU to finds its ways. Possibly there is a way to reduce the throttle reaction instead, have to look into this. And yes, I know that the cranking fuel will also affect the reached start rpms...
Logged
woj
Hero Member
*****

Karma: +41/-3
Offline Offline

Posts: 500


« Reply #219 on: May 18, 2018, 12:48:09 PM »

Found some information of the sort I was looking for:

http://www.ft86club.com/forums/showthread.php?t=94751&page=5

Though totally different technology / cars, there are some numbers that I was expecting to see. I cannot see them mentioning non-linear blending for cranking fuel (but I did not read through the whole thread), yet they only have sensibly looking values down to only 10*C, clearly it's a warm climate down there. But that's what I am looking for at the moment, will not be colder here that 10*C for a good few months. Ball parked my maps based on these and also added non-linear blending scale to play with later on when it gets colder.
Logged
woj
Hero Member
*****

Karma: +41/-3
Offline Offline

Posts: 500


« Reply #220 on: May 25, 2018, 12:59:52 PM »

Fueled up a warm car with some more E85, after stabilising resulted in E67, I was hoping for more, but there was too much of the old fuel (E22) in the tank, I did not want to top it just in case, and I have no clue what the actual content at the gas station is.

Anyhow, no problems so far, but I am yet to try a cold(er) start. My algo to read ethanol content on initialisation does the trick. On the startup log you can see a fixed line first - this is eeprom restored previous ethanol content and stuck because of inactive fuel pump. Then 1s wait after the fuel pump starts working, the read is still stable at roughly the old level for ~5 seconds, then it starts to go up. It's also nice to see the fuel mixing dynamics on this one.

The STFT histogram looks good, that's from a longer drive, STFT oscillation is the same as I had with E22, ltfts are minimal, multiplicative stays within 1.008-1.01, additive -2..0 (unconverted values for additive, I am not sure what the conversion here is).

For the not perfect things - on high load / pulls the ethanol content read drops by 1-2%, sometimes nearing 3% drop. I had the same on E22 now that I checked. I am guessing this is due to the sensor being on the return line and catching some air when  most fuel stays in the rail where it is needed. It's not a problem, effectively it results in ~1% fuel deprivation, on load the car is overfueled by factory anyhow, that's nothing. I am not going to fight it on the software side, but most certainly mounting the sensor on the feed line would remedy this.

Then, I get less retard, but in one spot it was still there - -5.25 on two cylinders. This is less in total than I had before, but to be honest with E67 I would expect to see zero (in one of half pulls it did stay at stiff zero though). A new knock sensor is on the way, will see about that. 
Logged
gt-innovation
Sr. Member
****

Karma: +60/-89
Offline Offline

Posts: 442


« Reply #221 on: May 26, 2018, 12:15:45 PM »

Normally you could use e30 and the retardation could have gone away, I do not know if someone asked before but have you checked for vibrating parts near your engine?

Are you sure the compression on your engine is stock or if anyone has repaired that before? Why don`t you do this the old way use a second knock sensor with a circuit from a hearing aid and good headphones.. Train your ears a bit in some knock sound files you could find on the internet and check.

Once you are sure it is not so bad you could also trim your knock sensors from your calibration mostly on the higher rpm range to allow some more noise to pass through.
« Last Edit: May 26, 2018, 12:18:24 PM by gt-innovation » Logged
woj
Hero Member
*****

Karma: +41/-3
Offline Offline

Posts: 500


« Reply #222 on: May 26, 2018, 01:16:57 PM »

The car has not been power modded by anyone before me, or myself (my small power mods to the maps are now on the shelf, only doing the E85 tests at the moment), the engine has all stock parts. I am a third owner, before me the car was only dealer serviced, and looking at the state of the hose clamps I am 99% sure the head was never taken off the engine. The engine is not young (for the FIRE engine) - closing 140k km. I use factory prescribed plugs. It was the same before I changed the plugs and after, it was the same before I changed all engine mounts and after (at least one of them really badly needed it). I will check for rattling things and around the knock sensor when the new one comes, I do not want to crawl underneath the car more than necessary. I do have a "stethoscope" I made myself, essentially a copy of one that dyno guys had when I did my previous project. So yes, I know what to listen for, but thanks for reminding me, I totally forgot I have this in the depth of my basement. 

There are two things though. First, the roads in Sweden are very rough / made of thick grain asphalt. They do it on purpose here, noise-wise driving fast is unbearable, so folks just slow down (and it still does not kill the noise), and I think the grip in winter is tiny bit better. So I'd have to look into that next time I am elsewhere. The second thing is - another twin-bin from essentially the same engine (well, it has different cams and slightly bigger turbo) has some knock sensor related maps distinctly different (and some tiny bit different), I'd have to check again exactly which these were and how they were different.

Finally, I know other t-jet users around here reported problems with excessive and impossible to get rid of retardation. Perhaps the Italians or Germans messed up something fundamentally with this setup. I now for a fact that the stock program is seriously overfueled for example, even though FIRE engines are fine with, power-wise, and like AFRs down to 11.5, yet I see 10.X  (requested and actual) very often on WOT / high load, and not ~12 which would be natural.
Logged
woj
Hero Member
*****

Karma: +41/-3
Offline Offline

Posts: 500


« Reply #223 on: May 28, 2018, 10:34:45 AM »

Tell me if I get this right, I am tired beyond belief.

After driving almost 100km on E67 I am now convinced that the fuel consumption figures I see on the dash are not correct (expected +30% is not difficult to miss). I looked at the S4 wiki, and surely enough that needs to be taken care of, missed that. Problem is, this ECU does not have KVB or related things. Looked at the definition, in module BGKV found KUMSIRL. This is used twice in the code, once to calculate umsrln, and once to calculate fuel consumption (which is even reported through diagnostics). Obviously this all happens before KRKTE is applied that I modify for E85, so needs to be sorted. Will not touch the first instance of this - 100% sure this will f-up everything in the program. The second one is a simple fix and seems only to affect fuel consumption figure. Couple of lines of assembly will fix this (unless of course the cluster gets this from yet a different place).

My hesitation, but I think I understand - regular Abarth and SS version have different injectors, KRKTE is different by ~15%. KUMSIRL is also different in both, but minimally, one is ~99% of the other (hex 21F4 vs. 2200). There is nothing else (can't be 100% sure) that is changed in between the two bins to account for different fuel consumption. I presume this is right - it is all about relative fuel mass, so as long as the fuel stays the same, nothing else needs to be changed, this is this ECU's way (to use KUMSIRL) to avoid KVB related matters, right?

EDIT: Alternatively, perhaps I should be E85-scaling only KUMSIRL globally and that should be it, and leave KRKTE alone.

EDIT2: My first edit does not seem to be a good idea, KUMSIRL is not about fuel at all. But perhaps try to modify rkg earlier before it reaches KRKTE might work, will look into that.
« Last Edit: May 28, 2018, 11:22:05 AM by woj » Logged
woj
Hero Member
*****

Karma: +41/-3
Offline Offline

Posts: 500


« Reply #224 on: May 28, 2018, 01:38:35 PM »

Still not thinking straight, but I think I sorted it. I now E85-modify rk_w instead of frkte_w, this is effectively the same, apart from the fact that rk_w defines fuel consumption figure, frkte_w -> te does not. The car starts and I drove around the block, all seems fine. But digging deeper in the code made me realise that to do the fuel dosage really right is a bit more tricky, one would have to also deal with the EVAP system. It also goes into fuel consumption calculation and rk_w calculation. I do not feel strong enough to attack this any time soon.
Logged
Pages: 1 ... 13 14 [15] 16 17 ... 26
  Print  
 
Jump to:  

Powered by SMF 1.1.21 | SMF © 2015, Simple Machines Page created in 0.024 seconds with 17 queries. (Pretty URLs adds 0s, 0q)