Evolution Stu

iTrader: (2)

Kind of on topic... BMW 635D, this is my own car that I have been using to learn more about EDC16 C35 so it gets a lot of development and endless miles on and off the dyno datalogging boring stuff like inj stroke, SOI and fuel economy. (Power is the easy bit).

Standard car.


Last week.


An extra 10mpg purely from adjustment to wastegate and SOI.
Sadly, most of the remapping companies out there just decalibrate the NM-IQ table so the fuel computer reports better as its reported the amount of fuel it "thinks" is going in per stroke. Not the amount really going in.

I had better go to lunch before I bore you all to tears and ruin a good topic. LOL
Sorry guys... I do like to waffle at times.

john banks

Nice work, 80 tables is still a hell of a lot to juggle, 3 or 4 tables interacting at once can blow my mind. The tables I put into RAM to allow realtime tuning with the engine running on the R35 GTR for Cobb could map any car from just about 500 to 1000 BHP without a further reflash beyond the initial one to load the modified code:

Wastegate
AFR
Ignition timing
Injector scaling
Injector latency
MAF scaling
Intake valve timing

Boost limit in ROM, but that is about all you needed to reasonably alter on the vast majority, sometimes injector timing to avoid smoke on some bigger injectors.

I've avoided German ECUs because of complexity!

Evolution Stu

iTrader: (2)

Nice work mate
Excuse my ignorance as I wasnt aware you were also a calibrator. (Besides being admin, im not exactly a regular on here )

The biggest headache with the newer Bosch, Siemens & similar systems is the production versions cant be emulated at all so its literally hour upon hour of cal, flash, test, cal, flash, test, its mind numbing.

I would much rather be calibrating OE systems like this all day every day like I used to... (This is the Old Weber Marelli L8 on a Sierra Cosworth)


But sadly, those days are almost all but gone for me nowadays my mapping is far more "educated guesswork" by trial and error than genuine "recalibration".

And that trial and error comes after spending 40hrs a week for 3 weeks deciphering the tables and trying to figure out what part does what in a system you havent seen before.
(Unless we are lucky enough to have the A2L of which i have about 20gb but even then, the A2L is only accurate with the bin file it came with and doesnt usually relate to the car your working on without some educated extrapolation and even more assumption.)

Matteeboy

We were doing mostly 90 from Cornwall to Folkestone. Much faster in Europe (France and Germany) saw 30ish; better rush our 335d for similar speeds.

john banks

That is a lot closer then Matteeboy.

Stu, I like your video, no messing around, quick trim of each zone and move on. I'm not working in engine management at the moment as the R35 work was completed as much as I could on my 2009 by 2011, day job is a GP, but the work for Cobb bought the GTR. The R35 is probably the easiest car to tune with so few tables to actually alter and pretty much direct control. I do admit to ripping out several boost control tables that did ridiculous things like model the size of the wastegate to just finding where in the code the wastegate duty final value is used and then hijacking the last table in the chain to just write to it directly. Table is MAP vs RPM, worked smoothly and easily, works on road or track and in lots of different conditions. The fuel table is just a load of lambda targets and the factory almost widebands just do it and you log the trims to workout where to polish the MAF curves with an intake, or just use speed density. It is almost painting by numbers but I made it a simple set of tables deliberately with the well understood code in the background helping the user and altering stuff that didn't, like boost limits that didn't go high enough, load limits etc. I've still not been tempted into tuning or developing Bosch stuff, glad to see you are doing it properly. I can't imagine having to flash every change though, how long does it take or are people now managing to emulate on the new stuff? Doing this with modified code on the GTR was work I was very proud of, the competition have caught up since and surpassed in many areas but still aren't doing this, it works nice and slick like your emulator. I was like a nutcase awake at night working out how to code it though. By the time it was done my own car was thoroughly tuned using 8 minute flashes for each change, we got that down to 20 seconds just flashing changes, but nothing like doing it under load and seeing/hearing/feeling the response.

Carnut

Surely the majority of the gain in mpg has a lot to do with the journey, one was 15miles, and with a average speed of 70mph I would of thought it would of involved higher speeds compared to the one at 46 miles with a similar average speed also taking into account the journey times.

On second thoughts, have you zeroed the trip computer once you have got to 70.

Evolution Stu

iTrader: (2)

Thanks, its the only way to do it IMO.
If its a mile out then I do 2000rpm and 4500rpm columns using the dynos RPM lock feature so I can move the sites vertically via throttle load then interpolate the lot to get me something like and go from there. My cells cooling is pretty damn awesome with 40'000cfm of outside ambient pressure air going through a closed one way cell even if I do say so myself but extended periods of loading is still going to cause thermal damage to something and I am very **** about controlling thermals. (Im just in the process of integrating some Infra red tyre temp monitors into the cell that will allow me to monitor tyre temps with warnings on the dyno software actually. )


Wow, talk about chalk and cheese. I like it... bet you dont get to talk about mapping very often at work then? LOL


Sounds familiar. I was deciphering an EDC 17 today and am currently up to 38 boost tables and am pretty sure another 4 odd looking ones are boost related too. Its absolute madness, yet absolutely necessary in an odd kind of way. I just love the control modern systems have over every aspect of the engine. Maps for everything, throttle, hills, altitudes, fuel temperatures, EGT's, sport modes and then the whole lot again for
regeneration, EGR active and not active, the list is almost endless but the result is, Joe Bloggs can pretty much do as he pleases and nothing will blow up.


I still favour speed density for ease myself, cant beat an old fashioned "Just add the fuel i asked for at x psi boost and forget about the rest of the world" kind of system. LOL The problem arises when Fred Bloggs goes away, gets bored of his 500bhp and decides one rainy Saturday to dremel his cylinder head and leans her out... followed by blaming the mapper for it, along with his recent puncture. LOL

If you ever do mate, id be interested in what you get up to and there would certainly be a market for it because for road tuning, Bosch EDC is where its at for the next ten years or so id say as we tune about 40 diesels to every petrol now and the vast majority are on Bosch EDC15,16 and 17 systems. I miss petrol...

It can take months... depends what inside knowledge you can beg, steal or borrow and just how different it is to previous ones you have worked on. When the BMW 330D first came out we hired one for 6 weeks and learnt all we could about its EDC15 and In reality, we just about learned how to get the power output up and advance the SOI a little and that's it. The rest remains a mystery until you have thousands of hours with it as its just so damn complex, but of course, there is no real need to understand most of what it does.

No - there is still no way to emulate new stuff. (That I know of)

The EDC16 system is actually quite an innovative system that models the engine and outputs its tables in torque. Basically you have an Injection Quantity to NM table (well, a few actually to compensate for various fuel weights and pressures) that essentially gives the ECU a quantity of fuel to inject at a given pressure if the driver requests "X" Nm of torque.

Then you have the "drivers wish" maps, that look at throttle input and convert that to torque that the driver wishes to have. The rest of the few hundred maps are mainly shown in torque (Nm) and are known as torque limiters.

This makes it a little simpler to understand than the older EDC15 that was Injection Quantity based. For example...

Operating example:
Lets assume the standard OE map allows a max of 650Nm of torque at 100% throttle.... So we modify the drivers wish table to ask for 800nm at 100% throttle input, we see that the gearbox input limiter table is already happy with 800, the steering angle torque limiter is also happy with 800 and so on through many hundreds of tables, but the current exhaust backpressure limiter while normally fine with 800nm is now limiting to 700nm because the DPF is partially clogged and showing 1024mbar back pressure... that amount of fuel, enough for 700Nm, being the lowest of all the tables is what will be delivered. If you never find that table, you will never tune this ECU past 700Nm.

There is also an "Engine Drag" table that models the engines lost torque at any given RPM and oil temperature so the system knows how much to add on for parasitic losses, IE: Driver wants 800 but the engine oils quite cold and losing -125nm instead of the usual -75nm so it delivers enough for an internal torque generation of 925nm to compensate for the colder oils.

There are many other models in the ECU to model loss through Aircon, Power steering, etc etc so at any time the ecu knows precisely how much power is required to give the driver his exact wish but it will only allow the lowest output from its various tables.

Our headache, is finding that lowest variable amongst tens of thousands of tables and switch points. Thousands of which are single value. IE: SVBL = Single Value Boost Limiter. Without the A2L I doubt anyone would ever find that. Its always near a main boost table but just searching for a figure in millibar will return thousands of results and of course, you don't know what that millibar figure is.

(for example, Its Dec 02950 on a BMW 330D... so if you find that number in the bin file, is it meaning 295Nm? is it 29.50 mg of fuel? or is it temperature related in kelvin? 16Bit Dec *0.1 -273.14 = Deg C so it could be 22deg C as well, maybe an air temp?... Groan.)

Usually, without some inside help, you never get a foot under the table, and it can cost big money to get that info.

Sounds awesome mate and so you should be bloody proud of it.
Certainly no mean feat and sounds a great system indeed.

Evolution Stu

iTrader: (2)

I always do my tests like this:

Attain speed... apply cruise control, reset trip counter.
Its about the only real world way to measure gains. I only have 70mph ones and below of course as I dont get to Germany very often...

I get the basic SOI maps right on the dyno but a lot of cars (mine included) wont allow cruise control on a dyno due to differing wheel speeds. Fords however dont care so i run them with a script on the dyno which is nice as I dont even have to sit in it. LOL

Saxo Boy

Cliff Notes:

Tuning seems tough.

john banks

Stu, the scenario you describe could be solved by disassembling the code. IDA Pro is what I used for the Evo and GTR. The processor on both was the same which helps as experience was transferable. On both these I loaded up the flash ROM image which contained code and maps. If they are separate files you can load both once you work out the offsets. Reading the chip documentation you can usually find the vector table and an entry point for the code and also for timer and other interrupts. Then IDA (with the correct module for the processor) will go through the code in a moment or so and find all the branches and cross references. Then with a known flash image you can mark the tables you know and find the code that cross references them. With the GTR we had a few tables incorrectly defined by the IT guys who had worked out how to flash via OBD. An equivalent to an A2L would have been immense and made me unnecessary, but the boost couldn't be controlled properly and tuners were just changing numbers without proper understanding and there was no way I was trusting my VR38 to that which is when I got on board. Without an A2L, we still got a lot done by just spending months reverse engineering code line by line, the flip side is the logic is less OTT in terms of torque models. There are of course complications if the compiler used does weird things with memory addressing. Then on a series of ECUs you recognise the code that accesses certain tables and can search by a series of instructions or you recognise the tree diagram of the subroutine and then find the variable. Steep learning curve but beats WinOLS and pattern recognition. Are you doing this already? Usually the guys that are on an ECU can do code modifications to add new features. Forgive me telling you how to suck eggs if you are already doing this. If not, getting a reverse engineer/disassembler on board could help. From what I have seen the guys that tune don't usually disassemble, and the guys that disassemble don't tune, and because there is lack of communication between the two camps, there is lack of results and imagination. I carved a niche doing both, but I couldn't sustain it and the day job, and working for myself already didn't want to do it for someone else, yet to start up something myself would have needed a few years of development and staff/premises before ROI, so I bought more of the medical practice instead. GTR also came along at the right time as a really inspiring and affordable tuner car, so live mapping with engine running, massive load and boost, multiple switchable maps, launch control, speed density were interesting to work on and see the results. I've not found a really interesting alternative car to get involved in developing and nothing coming out floats my boat so these days my spare time is DIY with a big project for about another year.

Evolution Stu

iTrader: (2)

Afternoon John,
Yeah, IDA Pro is exactly what we did with some older less complex ECU's in years gone by. Nowadays its preferable to us to work it out with an A2L from a similar architecture and engine type/manufacturer and of course we have all the expected software to give us a leg up the ladder when we come across something we havent already worked out. (Swiftech/ Winols / Race Pro)

Thats not to say I wouldnt be interested in a fully disasembled EDC16 code but they change so often nowadays it wouldnt arguably be of much use to me unless I specialised in a car that used it and did a lot with just that type of system.

john banks

I could only do one car at at time, would go mad trying to juggle in this depth. With good A2L in English (I don't speak German and auto translate didn't work for me well with the file sizes) though and your present methods it sounds like you're making really good progress.

Evolution Stu

iTrader: (2)

Thats the problem of course, I rarely get time to just spend good solid time looking at only one system as we cover pretty much everything out there so we are always swopping and changing focus.

AOn the topic of translation, the problem is, technical German is not at all easily translated.. Ive had native germans look at them and not be much help at all so Google had no chance... However, winols has a nice A2L German to English Translation plugin on later versions that helps a great deal with technical German translation if you ever decide to take a look and if you drop me a PM id be happy to give you an A2L to get you underway... ive got them for most German Vehicles now.

Over the 20 years ive been doing this professionally Ive learnt enough German suffixes to enable me to identify the most common tables I need in an A2L such as "Mi1" "Engprt", "CoEng" "qPrtLim" but Winols and Swiftech made life a LOT easier in recent years.

john banks

Thanks.

DT-SPD

iTrader: (7)

It's interesting to read how you reverse engineer the Bosch systems.

I've been working with Bosch EDC16 and now an advanced version of EDC17 for the last 10+yrs at OEM level, firstly for Bosch and now I'm a Calibration Engineer working with JLR.

Even with the fully emulated ECU's, A2L files and a 20k page software guide that Bosch give us it's still incredibly complicated. To start from just the hex/bin file and work back is impressive stuff

john banks

Do you think the ECUs are getting unnecessarily complicated to the detriment of reliability, bugs, good calibration? Certainly I find it fascinating the obsession with consistency of torque. I figure that the torque between gears and inclines is variable enough that I don't notice a driveability improvement on an ECU that does things like raise boost on a cold engine, or raises boost on a hot day. If the torque can be higher in some conditions and is within the rated cylinder pressure, turbine speed and torque limits of transmission etc, I want it. As a driver I want my engine to be the best it can be, not dumbed down. If the engine has a natural torque peak, I want to enjoy it, not have it smoothed out. Is there any role for a more organic, enthusiast type mapping on German ECUs by OEMs perhaps not for the shopping models ?

DT-SPD

iTrader: (7)

Yeah I do think they are getting too complicated. Was about 8k variables on EDC16 and now over 40k in the latest EU6 level software in EDC17.

ALi-B

iTrader: (1)

Its certainly causing issues with cars in later life.

When items such as variable valve timing don't perform as they should, or inlet ports clog up with crud the engine doesn't have the same power curve it once had and the ECU can have a tough time compensating. Same goes with sensor aging, there is aging algorithms for oxygen sensors (especially on diesels), petrols use post cat sensors to trim pre-cat sensors (fine when they both work ok, not so good if one goes out of kilter and the ECU's programming can't figure it out). The same goes for autoboxes as they age their shifting ability deteriorates (i.e the classic ZF 1-2 shift thump and DSGs with worn clutches that won't move, then suddenly lurch away).

The end result is a car that just doesn't pull smoothly, holding back, surging, blocking up DPFs prematurely, slip or thump between gear-changes....but with no fault codes to say what's wrong. The ECU's programming has to have a tolerance for variance between engines and sensors/actuators - if it were too rigid it'd throw codes all the time.

Classic example is BMW diesel thermostats...DPF keeps clogging up. Why? One of the thermostats is stuck open - ECU "should" know how quick the engine should take to warm up, its possible in this day and age of electronics to predict this, but nobody thought it'd be needed. Hence lots of BMW throwing up DPF faults - no temp guage. Not that a temp guage is any use anymore - as they are all "dumbed down" to stop customers thinking their engine is too hot/cold when it goes 5 degrees above spec. So the manufacturer makes the gauge do nothing and stay in the middle between 70 and 100+degrees. It only moves if its outside these extremes - about as useless as a oil pressure light that's switched at 1psi.

I've caught my Golf doing weird stuff when the oil requires changing (or the wrong spec is used), all I can assume it VVT related. But why, I'll never know (VVT was all replaced when the chain were done anyway) - I just change the oil and its fine again

Evolution Stu

iTrader: (2)

I can't tell you how much I'd love to spend a day calibrating with you on the Bosch software, it would be so damn useful to see how it works with the genuine interface and would help me no end to understand the system so much better than I do now. Having to make presumptions about the way a system will work makes for very slow work. Even simple things like smoke limiting tables, figuring out which one of the 8 I have identified is actually active is a matter of flashing each one to 26:1 and running it on the Dyno to see if it did anything to the fuelling in the condition your looking for.

It's a slow, long winded and frustrating process. I Was on the Dyno yesterday from 7am until 11pm last night doing just that on a new car we are working on. Mind numbing stuff and so frustrating when some bugger out there has a secret instruction manual... LOL

Littleted

I have to agree, Wifes zafira tourer is a 60 or 70 L tank, 1.4 turbo petrol, my old Scooby was 50L she gets around 400 to a tank, I got if I was careful 280, consid power diff and fun factor Scooby all the way.

Also my old E350 had a massive tank 70 80 L did around 40+ on motorway 265 BHP twin Turbo V6 , my new A45 does 30 to 37 at 70MPH on motorway and has a 56L tank. ill be honest and say unless your cruising the motorways then its horse for courses.

That and DPF EGR issues I think diesel is starting to be caught up.....

john banks

Hopefully this will sort it, diesel being traded this week My first V8, just a bit excited.

alanbell

iTrader: (6)

alanbell

iTrader: (6)

it will be good for your feedback with this , as i may get the same soon , ish.

john banks

I haven't even driven one, but it is really just a change of engine, addition of sport rear diff you already benefit from but my TDI doesn't, uprated suspension/brakes, 21" wheels (would have preferred standard 20s and might swap them if the ride on my 19" winters is good but it still handles well), more standard spec and trim/options. The seat massagers will be interesting whilst doing 3.5s 0-60 with noise cancellation whilst plotted on Google earth. Power on most dynos seems to be about 570BHP stock over the quoted 520, with remap to 650BHP on SUL. Apparently it understeers over 0.9g but I'm working the GTR to pull much more than 1g held rather than peak, but I haven't experienced unintended understeer on my A8 on the road, it feels quite neutral even without the fancy rear diff, and the criticism of Audi steering feels seems to translate to other marques with electric steering. The car I am getting is the last of the 2013s before the 2014 facelift, still hydraulic power steering, they changed most or all of the range to electric, not sure on the S8.

Given how good the A8 is, how bad can the S8 be with the power of a remapped GTR?

jameswrx

iTrader: (4)

Jees, they don't hang around do they! Check the video from 4:15 onwards, ties in nicely with the R8 thread...

john banks

That one seems to be about the most negative review on the handling. I think the handling needs to be seen in context of the luxury of the car and road rather than track use. I think I will be happy with the compromise, if delighted will sell the GTR, if not will keep both.

jameswrx

iTrader: (4)

I didn't watch the handling part, just saw the title and skipped to the race. But does it really matter what a journo says about handling when thrashing it round a track? I'm sure it's more than capable in the real world, it's a luxury big cruiser at the end of the day.

Saxo Boy

I think it's borderline ridiculous to expect an S8 to be 'agile' given the size of the thing and the expectations with regards to build quality and creature comforts.

john banks

It is quite ridiculous to say, but now I have driven it, it does feel agile, although 235kg heavier than the GTR. On 21" wheels it goes firm enough and the ride softens enough in the appropriate setting where it feels like the firmest setting on the A8. It is not at all shy of direction changes and the steering is fine, more direct than the A8 which I was happy with. I think it has more traction and cornering stability/force than the GTR in the wet, it is more confidence inspiring as you aren't fighting the rear end, but because of that it might not feel as challenging to drive in a few years as the GTR still does. It weighs similar to the RS6/7 despite dripping with kit as there is a lot more aluminium, and the engine is similar but with 0.1bar less boost, for now. I will reserve judgement as to whether to sell the GTR, not sure yet.

In terms of perceived firmness from 1 to 10:
GTR on run flats 10
GTR on Michelin 9
Mk V Golf GTI on 18" 7
Evo IX on Michelin 17" 7
E46 M3 on 19" 7
Ford Puma 6
B5 RS4 6
C5 RS 6 6
C6 RS6 on 20" 6
E92 335i on 19" 6
E39 5 series SE on 16" 5
D4 S8 on 21" 5
Peugeot 406 V6 on 15" 4
D4 A8 on 19" 3
Rover 75 3

I would say all that I have tried with adjustable suspension vary +-1.

john banks

In terms of agility:
Evo 10
GTR 9
Puma 8
Fiesta Zetec 7
Golf GTI 6
S8 6
B5 RS4 5
C5 RS6 4
406 3