Members Login
Username 
 
Password 
    Remember Me  
Post Info TOPIC: Basis of autotuning feature for map tuning ...


Guru

Status: Offline
Posts: 2338
Date:
Basis of autotuning feature for map tuning ...


Having tuned close to 20 busas with ECUeditor I have started to think of having an autotune feature in ECUeditor - just to simplify the process and make it less prone to human errors like tuning for an incorrect rpm range.

Basically the autotune is easy: Just log AFR at any particulary Load/RPM or TPS/RPM position and then make changes to the maps automatically.

There is anyhow a couple of factors to be included..., this is my initial thinking:
1) The exhaust gas travels at different speeds at lower and higher RPM:s/Loads so adjustable compensation for WBO distance must be made
2) The datalogging that VB provides using timers at 100ms rate may not be fast enough. RPM acceleration may be so fast that this logging rate is unreliable. MTS chain is 10ms if I remember correctly so closer to 20ms could be enough.
3) The required change ratio between MAP points and AFR points is unknown, and most likely be changed e.g. because of additional compensations
4) As the acceleration pump feature of the ECU is likely to enriched or lean the ecu when turning the throttle, then only steady state TPS positions should be logged.
5)  For any particular TPS position, e.g. 25%, some tolerance may needed to be built so that only AFR points for 24-26% should be included

Any comments, ideas and suggestions are welcome.



__________________

When asking a question, you can also consider posting it to facebook:

http://www.facebook.com/ecueditorcom



Guru

Status: Offline
Posts: 741
Date:

that sounds like a very good idea Petrik. i'm currently logging RPM, TPS, AFR, INJECTOR DUTY CYCLE, BOOST AND FUEL PRESSURE. with my DL-32 all my tuning is real world no dyno....so this feature will help guys like myself..smile 

__________________


Guru

Status: Offline
Posts: 2338
Date:

The shifter function is key for us right now - if we can make that working it opens a completely new world for various ideas to be implemented already within ecu.

Locally the chaps over here are also very keen on logging intake manifold pressure particularly for TPS<10%. Almost everyone is reporting tremendous driveability of bikes that have the IAP map area tuned.

This is certainly an area to be looked into in future. Anyhow we are constrained by available time and coding resources to implemente everything once. This is a non profit initiative and a hobby that is driven by persons who have been active in coding some 10-20 years ago before this project. So we are slow in implementing many things smile.gif

-- Edited by PetriK at 10:14, 2008-08-10

__________________

When asking a question, you can also consider posting it to facebook:

http://www.facebook.com/ecueditorcom

YAG


Veteran Member

Status: Offline
Posts: 50
Date:

Are you still planning on implementing this?

__________________


Guru

Status: Offline
Posts: 1344
Date:

petrik,any ideas about autotune feature?

__________________

09 busa.????? now what....still got what it takes.......!

I got what you need...!
www.poweredbyford.com

www.marc@poweredbyford.com

 



Guru

Status: Offline
Posts: 963
Date:

While the basic idea of having software monitor engine performance and adjust the maps for optimum performance seems simple there are several major obstacles .

First before anyone can write any software someone will need to describe, in detail, the required function. In other words what is optimal performance that you are auto tuning towards? Believe me I'm no tuning expert but from what I understand just auto tuning for 14.7:1 AFR will not give you the best performance. Optimum AFR varies with load and rpm doesn't it?

So before someone could write an auto tune program someone would have to come up with a target AFR map. Would that map vary from bike to bike? Are we not just back where we started; having to customize a unique target AFR map so the software can auto customize the maps?

As a software engineer I can tell you if a person can do something you can write software to do it but first I would need to know exactly what IT is. Someone would have to describe step by step how they tune a bike manually and without using a dyno if that is how you want the software to work.

Assuming we come up with software that can quantify generic optimal performance then comes the next obstacle, the ECU design. The ECU has very limited real time tuning capabilities and you can only flash the CPU a limited number of times. Auto tuning is a lot easier on a piggy back unit where the maps are held in RAM memory and can be modified while the bike is running. This allows for the software to change map inputs in real time to engine outputs and search for the optimum.

I don't see any way to auto tune without first providing the ability to real time tune to the current ECU software. Real time tune would itself be a major improvement to tuning and because it would be required before auto tuning real time tuning should be our first priority

-- Edited by RidgeRacer at 14:36, 2008-10-29

__________________


Guru

Status: Offline
Posts: 1344
Date:

rr,real world tuneing meaning able to flash while running,like dynojet and ect?



__________________

09 busa.????? now what....still got what it takes.......!

I got what you need...!
www.poweredbyford.com

www.marc@poweredbyford.com

 



Guru

Status: Offline
Posts: 963
Date:

Yes REAL TIME tuning is being able to modify the ECU while the bike is running on the dyno.

You can do this with the Yoshbox but in a very limited way.

What we would actually have to do in the case of the ECU is come up with a hybrid. Use temporary ram maps while the bike is running to tune then turn the bike off and transfer the ram maps to the flash maps.

__________________


Guru

Status: Offline
Posts: 1344
Date:

rr,is it that the yosh box stores a copied map to be modifyed then transfered?

__________________

09 busa.????? now what....still got what it takes.......!

I got what you need...!
www.poweredbyford.com

www.marc@poweredbyford.com

 

YAG


Veteran Member

Status: Offline
Posts: 50
Date:

Tuning would be done exactly the same way that I do it with the Innovate datalogger, you bracket the tuning in by logging, comparing with a self made A/F map, and calculating the changes. You have to do the whole process several times.

Optimum A/F is going to be different for different applications. As long as you can set up your own target A/F map, autotune with this method would be great.

__________________


Guru

Status: Offline
Posts: 963
Date:

No, the yoshbox works completely different.

In addition to the 256k of flash memory inside the ECU's CPU there is an additional 512 bytes of serial eeprom memory on the ECU circuit board. While the CPU flash can only be written 100 times the serial eeprom can be written to 10,000 times.

The yosh values are stored in the serial eeprom. When the ECU first starts up it copies the yosh values into ram and the fuel injector calculations then use those ram values to add or subtract fuel.

The Yoshbox does not change any maps inside the CPU it has its own kind of little (4 cylinders by 3 values) map in the serial eeprom.

-- Edited by RidgeRacer at 20:32, 2008-10-29

__________________
YAG


Veteran Member

Status: Offline
Posts: 50
Date:

Sorry I didn't mean the yosh box, I meant modifying the firmware maps based on logged data, and uploading that firmware to the ECU.

__________________


Veteran Member

Status: Offline
Posts: 27
Date:

It sounds similar to the Bazzaz "Self Tune" option.  The user is able to go in and set the Target A/F ratio to whatever number you want and turn on the log option.  Until you turn the log option off, the Bazzaz box stores the average A/F ratio for each box and then makes a suggestion of what changes it would like to make to achieve those numbers.  It will continue to log until you stop the log feature so it doesn't restart once you shut the unit off.  If you want to make the changes you just save the changes to your current map. 

The way I used my Bazzaz box with the ZAFM kit was to load a zero map and go out and ride.  It would make suggested changes, and I would go into ECU editor and make changes in those general areas and go back out for another ride.  My goal was to get each box as close to Zero as possible so once tuned I could remove the Bazzaz box and run only off the ECU.  It would be nice to be able to have those adjustments made real time or a way to have the ECU output a calculated fuel map to achieve whatever the A/F ratio is desired which will be different for each bike.  It sounds like it will most likely be a feature designed to work with a dyno station with software running on a PC but who knows, you guys have built some cool stuff in already.

Based on the amount of Innovate items I sell, I find it hard to believe there are many out there that self tune that don't have one version on their bike!  it would be cool to build the interface to work with that plugin. 



__________________


Member

Status: Offline
Posts: 12
Date:

The AFR Tuner does real self mapping in real time. http://www.afrtuner.com

__________________


Guru

Status: Offline
Posts: 2338
Date:

Yes - there is usually an optimum afr which is at around 13.2 for maximum power, depending on ignition etc. At lower rpms for economy and emissions we may want to tune leaner, but targt is most often a straight line above 4000rpms. Definately not any richer for all motors. For turbos and nitrous we usually target to 12.5 - but those are tune according to EGTs to anyway.

The biggest benefit of autotune is simple - you get a straigh line to start working on for ignition and other stuff. That line is easy to be moved up and down - but often it takes a few runs to get it there. Particularly busas seem to have a dip at around 7k and to run very rich below 4k - so getting those areas right are absolutely a key for a smooth running engine. Particularly the vacuum map is fairly tricky to tune without an eddy current dyno, so therefore considering adding the autotune map particularly to make vacuum (IAP) map tuning - but there is so much to do and so little time. ECUeditor has no known (with some practical improvements) for doing this because of the serial protocol making it possible to track the exact map position.

The way how it works is simple and biggest unknown relates to the time the exhaust has travelled for the RPM. At low RPMs the gas velocity and volume is lower so it takes longer time. E.g. Dynostar software autotuning the Rapidbike piggybacks simply adds a multiplier there based on RPM to estimate the velocity. A simple concept and seems to work pretty well.

I have been using or seeing used a couple of autotune packages now and being particularly familiar with KMS and Megasquirt related packages it looks really easy to do. We dont need any interaction between ecu and dyno station computer for anything else than mapping. As long as the ecu gives the exact map position (as it does with the serial protocol already on ECUeditor tuning screen) then all we need to do is write the AFR for each cell within ECUeditor.

The buell etc. in box autotunes are a bit different, if the ecu program based on a target afr then there is usually a max/min injector pulsewidth that are allowed for the internal map and usually only for cruising conditions. Even the busa ecu can determine cruising condition dont see building anything like this inside the ecu. Many cars also have the target afr map, but not for WOT (wide open throttle). All WOT maps I have seen so far have been so far based on calculation that has no feedback from the oxygen sensor. I can understand this approach as oxygen sensors get faulty and seems to need constant free air calibration to work correctly.

Sorry if I may sound too harsh in the above text. Opinions may, will and should differ in this matter - this is just my personal view based on practical experience on ECU tuning cars from early to mid 2000. Bikes I have ECU tuned or being involved tuning this year in a dyno include in addition to Busas also a Triumph, R1 with race ecu and Buell with autotune which was not used - and of course a few with a KMS fuel computer installed drag bikes. Last years (2007) there was also one R1 with Megasquirt, but that was put in a mini cooper smile.gif. The mini was a very interesting project as the engine originally came from a sidecar.

Mottys have not tried out. What I have read it seems to work and would like to understand the algorithms used a bit better. I like concepts where ideas are put in real life test - it proves the ideas to work.


__________________

When asking a question, you can also consider posting it to facebook:

http://www.facebook.com/ecueditorcom



Guru

Status: Offline
Posts: 963
Date:

I think we are talking apples and oranges.

To me autotune means tune while you drive. A kind of plug and play, no tuning knowledge required tuning.

It sounds like what PetriK is describing is dyno tuning automation; a software tool to speed up and aid manual tunning.

I think such a tool to give a quick automatically generated baseline maps based on AFR, should not be too hard to figure out.

__________________
YAG


Veteran Member

Status: Offline
Posts: 50
Date:

It would definitely make it a lot easier.

__________________


Guru

Status: Offline
Posts: 2338
Date:

Oh - sorry - yes its apples and oranges.

EU K2-K7 Busa as stock has a narrow band oxygen sensor keeping the AFR at around 14.7 when cruising, so called closed loop feature. Therefore I did not even consider that autotuning would in this case mean internal to ecu or as more common referred closed loop map. We know already today how to change the size of this closed loop map and also by using the oxygen sensor table to change the target afr. So in that context busa has it - for the limited area which is described as Oxygen sensor window in advanced setting of ECUeditor.

In practise we can make the closed loop map larger in size as well as change the afr - as sometimes changing from closed loop map to open loop map means a sligh hesitation from the engine its very handy that we can do this.



__________________

When asking a question, you can also consider posting it to facebook:

http://www.facebook.com/ecueditorcom



Guru

Status: Offline
Posts: 953
Date:

I really liked the way megalog viewer worked for megasquirt stuff. You would go take data, then setup an A/F table (engine rpm vs load type of thing) with your a/f values you were looking for, it would go back through the data using filters you would apply and tune for you, keep doing that and it would get closer and closer to where you wanted to be.

Greg

__________________


Guru

Status: Offline
Posts: 2338
Date:

Yes - very easy to do if e.g. innovate LC-1 is in use through the innovate api. Other issue is the speed of engine data interface at higher rpm - may not give enough points to work on. Anyhow as autotune is really (at least for me) for idle to low acceleration area you can get there fairly easily. The filters is important - in innovate datalogs I ended up writing fairly complex filters to be able to tune based on dataloggin information. Its so difficult to see from the innovate datalog if the bike was going uphill or downhill (i.e. what was the load). Therefore I started to use a laptop and ECUeditor screen on top of the tank to get the exact map position known.

Anyhow I miss one important piece of information at this stage: How much AFR is changed by changing the map value by 1 unit - anyone having spent some time into logging that ? (3pts = 0.2AFR units ?)

Btw - really like the innovate afr sensor, its very fast. I often get to see dyno charts from major names which have a seeminly straight afr curve after autotuning feature. When those are analyzed with an innovate sensor all the smoothing goes away and faults in the curve comes so visible.





-- Edited by PetriK at 08:43, 2008-11-01

__________________

When asking a question, you can also consider posting it to facebook:

http://www.facebook.com/ecueditorcom

YAG


Veteran Member

Status: Offline
Posts: 50
Date:

I agree, the LC-1 is very fast... makes it easy to spot small trouble spots in tuning, and the price is VERY good compared to everything else out there.

__________________


Guru

Status: Offline
Posts: 1247
Date:

PetriK wrote:

.

...Anyhow I miss one important piece of information at this stage: How much AFR is changed by changing the map value by 1 unit - anyone having spent some time into logging that ? (3pts = 0.2AFR units ?)...







-- Edited by PetriK at 08:43, 2008-11-01






Not sure on the AFR sensitivity relating ot map tuning. Megasquirt has a variable which allows the O2 sensor sensitivitiy adjustment, allowing it to be agressive and make major changes or only to tweak within a small amount.

If the "autotuning" is looked into further, perhaps having an adjustable "Weighted value assigned to the current AFR vs AFR target would allow this to be fine tuned.

__________________


Guru

Status: Offline
Posts: 2338
Date:

I have a vague recall of 0.2AFR being about 1-2 map points at top RPMS from those various dyno runs - just have never paid attention to it at lower RPMs. Anyway I remember that we changed the resolution earlier this year. Initially the map values were divided by 24, now those are divided by 48. Thats because with the previous resolution it was more of 0.3-0.5AFR points per map point change. But never really tested this and now its too cold outside.

Anyway because of the fuel map being rather VE map than milliseconds it will never be this straight forward to do this conversion. By this I mean that the map value is multiplied by ambient temperature, ambient air pressure and some other factors so map values can not be directly milliseconds without a fixed temperature and pressure (=air mass / volume).

But for user perspective it would be very good to have the map values somehow related to AFR values.

If I remember correctly the Innovate wide band oxygen sensor control circuitry is patented and partciularly faster than others.

Would not have looked into this earlier but the DJ fuel curves looked so straight before put into the DS dyno which showed immediately some bumps. There seems to be a lot of smoothing going on in DJ and even the scale seems to be much wider than oxy sensors are capable reading of.

So this brings the biggest issue of autotuning into the discussion. From what device would the wide band oxygen sensor to tuning software for automapping be available from ?





__________________

When asking a question, you can also consider posting it to facebook:

http://www.facebook.com/ecueditorcom



Guru

Status: Offline
Posts: 1247
Date:

Not sure what you are asking PertiK?

Are you asking which O2 sensor or asking if someone knows of a different method to link AFR values to map tuning?

If you are asking which wideband O2, then I would suggest the LC-1.


I guess if an onboard interface is used, one could link the current engine values, along with the map cell tracking to an external O2 datastream, perhaps with a modified innovate logworks plugin to "suggest" map value changes?

Maybe a new map file could be created and saved on a PC as a sort of temporary file while tuning the running the bike through its paces and recording the AFR for each area rather then trying to do a real time burn. Then the bin could be written after each test. While that is not real-time and would still need to be iterative, it would at least remove the interpretation of actual AFR to target AFR for each map cell.

__________________


Guru

Status: Offline
Posts: 1247
Date:

Perhaps something fairly simple like this that then links the actual AFR vs target AFR to a "delta map cell"




__________________


Guru

Status: Offline
Posts: 2338
Date:

Yes - LC1 is one possibility. My experience with both LC1+DL32 and LM1 says that those are very good. Also having written already the link between Innovate SDK and Hayabusa ECUeditor I now that to generate a map like that is very easy. It is already possible to log a dyno run with LC1 and see the real AFR for each point with the existing version.

The point rather being here that instead of writing an all compassing software by myself would rather start building plugins for various hardware options and share some of the workload - and that may also invite different hardware options being plugin. Based on experience the dyno stations require different kind of software and particularly service level than us hobbyists.



__________________

When asking a question, you can also consider posting it to facebook:

http://www.facebook.com/ecueditorcom

YAG


Veteran Member

Status: Offline
Posts: 50
Date:

Any new news on this?

__________________


Newbie

Status: Offline
Posts: 1
Date:

I have 10+ years of tuning experience, and the first issue with your idea is that you want to base it off an inferior wideband. I've done back to back comparisons with AEM, Innovate,MOTEC, FJO,PLX, and AFM1000. There are huge descrepancies between all these widebands, and not all in the same way. For instance, the innovate widebands have way to much filtering and poor sensor control for acurate AFR tuning. The AEMs AFR scale is completely off anything richer than about 12.5AFR I've seen them be off 1.5 points. Some engines are less sensative to AFR and in those cases it works ok, but other engines are not, and you'll never get a nice smooth powerband with a wideband that doesn't have enough resolution. If your going to make a true professional autotune you need to use either the MOTEC PLM, or AFM1000. Those 2 widebands have excelent sensor controll, very acurate scaling, and excelent noise filtering that doesn't interfear with actuall readings. You also need to make sure you choose a sensor that has a greter sampling capability like the older LSU 4.0 or the L1H1 L2H2. The newer 4.2 and up sensors have much less sampling ability in exchange for longevity which is a bad trade off for a professional tuner.

Secondly your gonna want to do part throttle areas steady state and higher load areas as acceleration pulls with proper loading on the vehicle. With the steady state areas sensor timing won't be an issue since you will want to hold the cell long enough for everything to stabalise anyways. Acceleration pulls, some AFR timing offset would probably be good, but honestly on a bike the distance from exhaust port to end of exhaust is so short you should be able to get it nailed with no time compensation. This is even more so if the engine is properly loaded by vehicle weight and drag coeficient ( Such as a mustang dyno would provide)..

If you insists on doing AFR timing delays, you could easily setup your software to gather data to figure out the delay.  Basically hold the bike at one RPM make a change, and have to software calculate the time between when the change is made and when it sees a noticable difference in AFR.. Do this at multiple RPMS through the rev range, and you will probably have a nice curve when your done. The issue with this, is it will change with different header configurations (more/less EG velocity), and AFR sample placment, IE a bung 6 inches from the ports, VS a tailpipe sniffer with a vaccum pump on it. So it would have to be recalibrated every time you use it.  Almost no point, seems to time efficient to just use your head.

So to do it right you really need a good wideband, and I really wouldn't be concerned about timing delays. I'd have 2 modes, a steady state mode, and a acceleration mode.

Steady state mode would log AFRs and RPM, detect when they are steady (within certain bounds of change) then make appropriate change based on target AFR table, then re-check/readjust. Once fueling is complete at that breakpoint, Lock cell and continue on.

For acceleration, I'd do a loaded acceleration pull, maintaining a steady load point (Hold TPS still 50,60,70,80,90,100), record data though run, avg changes in adjacent cells, apply changes, then pull again. This will probably get you pretty darn close to target on 2-3 pulls.

Just some ideas.. Hope this helps.

__________________
"The Coolest Shit Happens Outside The Box!!"


Newbie

Status: Offline
Posts: 3
Date:

Hi,

I'd like to buy mentioned LC-1 of Innovate.. will I be able to make some charts of RPMs/AFRs in LogWorks? or only AFR value I've got from it? logically it seems that only AFR values are be read, but maybe there is some way to read additional value of RPM from ECU and combination of these values?

Regards.

M.

__________________


Newbie

Status: Offline
Posts: 3
Date:

I answer mysefl - NO :) I also need SSI-4 to data log RPM/TPS in conjunction with A/F..

Regeards, M.

__________________


Veteran Member

Status: Offline
Posts: 66
Date:

Hi,
Try Zeitronics AFC. It you could try to download their program to run it and deside if its good for you. it has many inputs at the datalogger and it's very strong for bike use


__________________
A Man Chooses His Weapons According To Their Abilities
Page 1 of 1  sorted by
 
Quick Reply

Please log in to post quick replies.

Tweet this page Post to Digg Post to Del.icio.us


Create your own FREE Forum
Report Abuse
Powered by ActiveBoard