TD_D mod journey. From stock to the 'bastardmobile'

[Td_d]

YES!

With a little bit of lateral thinking, some well postioned and acquired serial connectors, and a whole lot of reading - I now have wideband logging, stand-alone to the SD card, in CAN mode! And I didn't even need to splice open an OBD2 extension cord (which is on its way from the States soon) - which means I can use the cable as it is intended, to move the Tactrix away from the footwell, and reduce strain on the ECU harness.

The resolution is amazing... 8 to 9 times the data compared to K-line, gives you a very detailed picture of what exactly is happening in WOT (the datapoints are about 30-50 RPMS apart).

I'm also happy to report that the AFRs are nice and clean - dropping from the 12's in the upper 2000's, and down to target by 3200 as the turbo is spooling up - and then holding target quite neatly at 11.5 all through, as the meth / water mix is spraying.

I'm stoked!

 

[HolyCrapItsFast]

Awesome!!!

 

[Td_d]

It was relatively straighforward - once I'd read the 3 million web pages to get to the correct approach :tard:

At least something good came out of the the first AFR sensor (the LC1) not working - I was left with the rs232 to 2.5mm jack cable. Then I had a thought... :lol: the OP2 has a 2.5mm jack for RS232 inputs, but had not to my knowledge been implemented yet (which is why I was thinking of going the OBD2 cable splicing route).

So, out of curiousity, off I went and procured a DB9 male to male RS232 adaptor, and used that cable to connect to the new cable that the MTX-L has (so basically the wideband signal went all the way through to the 2.5mm jack - checked with a multimeter, was getting varying milivolt readings, so I knew it was functioning). Then I read up on what the logcfg.txt parameters for the LC-1 inputs were, which are:

type = inno
paramname = mylc1.afr ; you can call this parameter whatever you want
paramid = 0x0101
scalingrpn = x,14.7,*

Tried it, still couldn't get a reading. Then I found an obscure posting on the OpenECU forums about someone who had managed to do this. Basically:

- you need to upgrade to the very latest ECUFlash beta version - it's not even up on the ECUflash site - but it is available on the EvoScan site - http://www.evoscan.com/ecuflash/ecuflash_1433252_win_beta.exe
- Connect the OP2, and let it upgrade the firmware and drivers as well
- Reflash your ECU with the latest ROM you are using, so that the mode23 patch can be reimplemented with the new firmware
- Fire her up!

It was such an awesome feeling to see it having logged the wideband - I now will have constant data to be able to properly analyse the tune at any point - cruise or WOT.

 

[Td_d]

So, in the ongoing obsession in getting my tune absolutely perfect, I'm continuing to systematically examine each area that the previous tuner 'tuned' and critically analysing it for flaws (which, if you've been reading, I have found many...).

With fueling now pretty much perfect (till the new injectors come through at least :tard, I keen to really tear apart the timing and AVCS tables. I've obviously been tweaking these all along to deal with errant knock, but I'd like to review the tables more systematically, in other words testing his assumptions.

So - expect a barrage of questions

First thing that strikes me is this- in the stock ROM, and from others I've seen (including Holy's) the total timing at very low loads (i.e. the first 2 to 3 columns) is in the region of 47* to 49*. Mine is sitting at 40* . AVCS tables, at least above 3600RPMs are also identical to stock. Why on earth would he do this? Surely lower timing would result in higher EGTs - which I get when you're trying to spool up at 3000 Rpms and higher load - but at these low loads? Doesn't make sense to me.

 

[Td_d]

So now that I have a permanent source of data including wideband, I have an endless source of data to analyse. Been using Airboy's timing interpolation worksheet to try and make sure the timing tables are smooth and even, and to check for areas of knock.

I've notice a weird series of hesitations that only happens under very heavy load between 4000 and 5000 RPM, in 4th gear - it's a like a series of small hesitations (that the engine pushes through) as it ramps up the engine speed. So I though - either fuel, or timing. Interpolated the AFRs - tracking very neatly against target. Then I check against timing - and I'm getting a very weird occurence (which would also explain the phenomenon). My mapped timing moves down neatly through the range - but my logged timing suddenly jumps 3 degrees up (from 16 to 19.5*) above the target. What I just noticed is that this happens when load moves from 2.5 to 2.64. The only other factors that I know are supposed to impact on this are IAT and ECT - which are not being triggered.

Any ideas?

I'll post a shot now of what I mean.

 

[Td_d]

The dip at the end is a small knock, accounted for.

 

[HolyCrapItsFast]

Yes... it is adding dynamic advance. If you take a look at your knock correction advance table you might see that corresponding increase in timing at the observed load and rpm. If it is not that the only other thing it could be is learned ignition timing. It wouldn't be a compensation because a compensation applies itself only after total timing is calculated and will effect timing globally.

So now that I have a permanent source of data including wideband, I have an endless source of data to analyse. Been using Airboy's timing interpolation worksheet to try and make sure the timing tables are smooth and even, and to check for areas of knock.

I've notice a weird series of hesitations that only happens under very heavy load between 4000 and 5000 RPM, in 4th gear - it's a like a series of small hesitations (that the engine pushes through) as it ramps up the engine speed. So I though - either fuel, or timing. Interpolated the AFRs - tracking very neatly against target. Then I check against timing - and I'm getting a very weird occurence (which would also explain the phenomenon). My mapped timing moves down neatly through the range - but my logged timing suddenly jumps 3 degrees up (from 16 to 19.5*) above the target. What I just noticed is that this happens when load moves from 2.5 to 2.64. The only other factors that I know are supposed to impact on this are IAT and ECT - which are not being triggered.

Any ideas?

I'll post a shot now of what I mean.

 

[Td_d]

Yeah - I don't think it's dynamic advance - as the red line is already base and dynamic advance added (and looking at the logs it's fully advanced - tracks the dynamic advance table where the jump happens). It looks like there might be other, undefined variables at play - been doing some research, it looks like there are beta defined 'per gear' compensations that might be playing with the numbers.

 

[Td_d]

I think I may have stumbled onto an important missing link in sorting out the classic 'stumble' in the 08' + Sti's. Will only be able to confirm tomorrow with some more logging, but initial logging and driving seem to confirm theory.

Trawling through the Romraider forums, I discovered an experimental / beta set of table definitions for Ecuflash that clicked a light bulb in my head - the tables are for the 'fuel-pump duty' - and essentially define what 'low' duty and 'medium' duty is on the fuel pump. Effectively, that table sets low at 33% and medium at 66% (an as yet undefined table for high is obviously 100%, and and what load low and high is defined is as yet undefined as well). This means that at low to medium RPMs, the fuel pump is not working at full capacity. Now this may work fine for stock fuel pumps and setups (or not, frankly), however in the situation where you have hacked the MAF scale to fit larger injectors and an aftermarket fuel pump, can result in you starting to go WOT with pump duty not at 100% (given that perceived load is lower than what the 'real' load is for the purposes of making sure you stay under the 08' limit of 4 g/rev).

Having set the pump duty cycle at 66% low, and 100% medium resulted in an instant smoothing out of driveability in the 2000-3400 RPM range. Initial logs also demonstrate an interesting trend - whilst I will need to readjust the load compensation tables, across the entire 1400-3800 range, the trims at all manifold pressure points are now all much closer to 0% ( mostly in the 1% to 2% range - which I've never seen). This could very well explain why the load compensation tables alone were not fully dealing with the stumble (and trims were very erratic) - as if the pump was not working at full capacity, insufficient fuel would be delivered.

I would like to drive for the next day or so to confirm the theory, but I think I could be on to something here.

 

[Td_d]

Thanks Fuji - I've been seriously geaking out on the Romraider forums, its a treasure trove of information if you have the patience, and also the temerity to try out experimental definitions...

 

[HolyCrapItsFast]

Awesome info!!!

You aint kidding. All kinds of bells and light bulbs are going off in my head as we speak. :idea:This relates to so much more. Especially as it relates to E85.

 

[Td_d]

Holy shit... I really do think I've stumbled on the 'stumble' holy grail here - I'll still only believe it after a couple of days, and some good solid long logs, but see anything different?

 

[Td_d]

If you're curious, here's what the table looks like:

 

[Td_d]

Yeah, that's what I thought when I saw it - went back to look at the number of data points - enough to feel comfortable that it's valid...

 

[HolyCrapItsFast]

Oh my god this is huge. Great work TD. I'm jealous that I didn't think of it. :bang:

The nice thing is ATR has all these tables to :thumbsup:

 

[HolyCrapItsFast]

Okay... Here is an interesting fact.

When you effect a change to these tables and the pump duty cycle, you also effect a significant change in fuel pressure. I'm currently in the process of experimenting with this right now but I would imagine I would have to re-calibrate all my MAF calibration and load comps. I would also imagine that my WOT MAF scaling need not be touched since that region of the map sees 100% duty cycle any way...

For my car 100% seems extreme for the mid range so I am making it 75%. For the low range I am making it 45%. Any higher than that and I go rich because of fuel pressure but it also could be that I have not scaled the MAF for the new pressure it is seeing at the injector.

What do you think of all of that!

 

[Td_d]

Yeah, that sounds right - my lower end MAF scale was 7-8% rich as soon as I implemented these tables - and yes, WOT figures remained unchanged, as they were likely at 100% already.

Keen to see what your results are - I really hope that this does what I hope I'm currently seeing in the logs...

 

[HolyCrapItsFast]

Well if either of you have insight, I'm trying very hard to wrap my head around what I am observing. It doesn't make any sense at all.

I expected there to be a change in fuel trims as a result of the higher fuel pressure and I did, but the change was not significant at all. The thing I can't wrap my head around is the fact that the trims went positive at the settings I am at now which is 45-75-100. I was expecting the trims to go negative in response to the higher fuel pressure and remove fuel as a result.

Here's the other thing I can wrap my head around. With the pump set to 66.7-100-100, fuel trims were expectantly negative.

Anyway I zeroed out my load comps and rescaled the MAF and I can say it is amazingly smooth. I honestly thought it couldn't get any better than it was but I was proven wrong. The engine is so smooth at an idle and even with 1000cc injectors I can rest a cup of coffee on top of the manifold and not have it spill. I can't wait to see what defining my load comps will do.

 

[Td_d]

Yeah - that does seem a little counterintuitive... it seems the main effect is in keeping the fuel pressure at a higher level more even, resulting in greater consistency in the fuel ratio. Well, that's my uneducated guess :tard:

Well - it's definitely comfirmed - a days worth of driving, 3 fairly long logs, at least 50 data points in each cell, here's the comp graph (and I didn't even start from scratch with zeroed load comp tables). Un-fucking believable... I bet if I did start with zeroed comps, redid the MAF and then load comps, I could probably get it very close to zero trims...

 

[Td_d]

Fuji - I was thinking of the issue that would arise with very large injectors - I think you'll probably be ok - from what I'm seeing the biggest impact on driveability is from raising the 'middle' load, where all the issues with resonance etc. happen - you could probably get away with stock figures for low. Otherwise I would imagine your pulsewidth at idle would be very small, same problem one would encounter with just ramping up fuel pump pressure.