OddUnit

OE Development ECUs...

Page 2 of 2 Previous  1, 2

Go down

Re: OE Development ECUs...

Post by mrbeige on Sun Feb 28, 2010 3:18 pm

So, been giving this some more thought. Currently the strategies that I have are fairly sound, the only two things I don't have in there is wideband EGO and knock sensing. What I plan to do is use some adaptive parameters and use the feed back from the Wideband EGO sensor to adapt the VE map. Furthermore I can use knock sense to adapt the ignition map too.

_________________
Stu
"Take one old performance car, modify it so that it is as good, if not better, than a new performance car"

£300 Corrado 1.8 16v
The good lady wife's Golf VR
avatar
mrbeige
.:Admin:.

Male Number of posts : 1384
Age : 41
Location : Near Ely, Cambs.......but another planet according to my wife....

View user profile http://www.oddunit.com

Back to top Go down

Re: OE Development ECUs...

Post by mrbeige on Mon Mar 15, 2010 10:15 am

Loom spec in progress for a HIL (Hardware In the Loop) test bench. ECU end of the loom is done, but need the rest of the spec done and I need to work out the I/O board layout of the HIL.

_________________
Stu
"Take one old performance car, modify it so that it is as good, if not better, than a new performance car"

£300 Corrado 1.8 16v
The good lady wife's Golf VR
avatar
mrbeige
.:Admin:.

Male Number of posts : 1384
Age : 41
Location : Near Ely, Cambs.......but another planet according to my wife....

View user profile http://www.oddunit.com

Back to top Go down

Re: OE Development ECUs...

Post by mrbeige on Tue Mar 23, 2010 1:55 pm

So, loom spec is done. And so is the loom for the test bench. Some tweaking is required as I have started to adapt the model I have from one newer (more expensive) ECU to the G400 ECU I have got (free!).

So I've been through and reworked the pin allocations for the inputs and outputs. Pretty sure I've got a fairly stable pin-out now for running the 16v normally aspirated.

Currently I'm working on adaptive mapping. I think the best approach is to fix the target AFR map and use that and compare the output of that map to the current AFR reported by the UEGO. Only when operating in a 'steady state' can it then 'adapt' the VE table at the appropriate load site.

That's where I am at the mo.

_________________
Stu
"Take one old performance car, modify it so that it is as good, if not better, than a new performance car"

£300 Corrado 1.8 16v
The good lady wife's Golf VR
avatar
mrbeige
.:Admin:.

Male Number of posts : 1384
Age : 41
Location : Near Ely, Cambs.......but another planet according to my wife....

View user profile http://www.oddunit.com

Back to top Go down

Re: OE Development ECUs...

Post by mrbeige on Sat Mar 27, 2010 4:00 am

Right, well I thought I was making some progress. Currently the strategies I have have provide a closed loop fuelling model. This operates by reading from a narrow band lambda whether it is reach or lean. If rich, it switches rich, followed by a ramp, then pause, all whilst adjusting/amending a 'multiplier' until it switches lean, then ramps, pauses and switches rich etc. etc. What this is doing is learning where Lambda 1 is (Stoic) by adjusting the 'multiplier'.

So, do I turn off this closed loop fuelling during the VE map adaption phase, or do I say, wait for several AFR 'switches', then use the multiplier that is produces to feed into a map to create the gain for the VE adaption? I was thinking the former, as I could use the measure wideband lambda reading an compare to the target AFR to determine the size of the VE adaption at a particular load site. For information, as it stands now, VE adaption will only take place when the engine is in steady state RPM and steady state Load for a calibrateable period of time

_________________
Stu
"Take one old performance car, modify it so that it is as good, if not better, than a new performance car"

£300 Corrado 1.8 16v
The good lady wife's Golf VR
avatar
mrbeige
.:Admin:.

Male Number of posts : 1384
Age : 41
Location : Near Ely, Cambs.......but another planet according to my wife....

View user profile http://www.oddunit.com

Back to top Go down

Re: OE Development ECUs...

Post by dirtytorque on Sun Mar 28, 2010 12:51 pm

If this was my undertaking... Smile

I'd try and start it open loop first and then add subtletly after.It's how I always approach my software projects.
I admire you for taking this on.
Allow me to play devils advocate please.
How are you going to carry out VE adaptation with a narrow band sensor just flicking back between rich or lean?
Is this a strategy that you employ at work?
What will you use for your start point?
avatar
dirtytorque
.:Charged:.
.:Charged:.

Number of posts : 1101

View user profile

Back to top Go down

Re: OE Development ECUs...

Post by mrbeige on Mon Mar 29, 2010 1:55 am

dirtytorque wrote:If this was my undertaking... Smile

I'd try and start it open loop first and then add subtletly after.It's how I always approach my software projects.
Well, I can try it one way and then if it doesn't work well, I can change it, I suppose that is the beauty of the OpenECU system. I like the idea of starting simple, then increasing complexity though. I have some other ideas, but they can wait (EGT, boost control, dual maps etc etc)
dirtytorque wrote:I admire you for taking this on.
Seems like an excellent way to learn Smile
dirtytorque wrote:Allow me to play devils advocate please.
How are you going to carry out VE adaptation with a narrow band sensor just flicking back between rich or lean?
Is this a strategy that you employ at work?
What will you use for your start point?
I probably wasn't clear. The original strategies employed by Pi are for naturally aspirated engines and only use a narrow band, and do not have adaption in the same sense as what I'm trying to achieve. I'm essentially trying to achieve 'self-calibration' by adapting the VE map based on AFR target error.

As far as starting point, well that's more a case of 'finger in the air'. Probably start with a some 'guestimate' values.

_________________
Stu
"Take one old performance car, modify it so that it is as good, if not better, than a new performance car"

£300 Corrado 1.8 16v
The good lady wife's Golf VR
avatar
mrbeige
.:Admin:.

Male Number of posts : 1384
Age : 41
Location : Near Ely, Cambs.......but another planet according to my wife....

View user profile http://www.oddunit.com

Back to top Go down

Re: OE Development ECUs...

Post by dirtytorque on Wed Mar 31, 2010 2:32 am

mrbeige wrote:
I probably wasn't clear. The original strategies employed by Pi are for naturally aspirated engines and only use a narrow band, and do not have adaption in the same sense as what I'm trying to achieve. I'm essentially trying to achieve 'self-calibration' by adapting the VE map based on AFR target error.

As far as starting point, well that's more a case of 'finger in the air'. Probably start with a some 'guestimate' values.

Yes but the problem you have is a lack of a good reference to calibrate the error %.
All you'll have is an error,but you won't know the magnitude of it.
At some point you will need access to equipment to help you get close to ball park figures.
avatar
dirtytorque
.:Charged:.
.:Charged:.

Number of posts : 1101

View user profile

Back to top Go down

Re: OE Development ECUs...

Post by mrbeige on Wed Mar 31, 2010 3:00 am

Ah, I see where you are coming from, but again, I think I may have not been clear (need to concentrate on my written communication Laughing )

I am employing a wideband lambda, and that will give me the AFR error based on my target AFR map. I'll use that to determine the increment to the VE map.

What I want to do is check the AFR error, then depending on how far away it is from the target AFR, map that to an increment size and then make that incremental change to the VE map. I wouldn't intend on making large changes in one go, but rather smaller steps at a time. depending on the error will determine the size of the step though.

In fairness though, it would actually be possible to use a narrow band to achieve the same thing, it's just that it would take longer. The original strategy generates a 'multiplier' value that is based on learning stoic, so it is possible to then use the magnitude of the multiplier to adapt the map accordingly. However, it only learns stoic, and not actually AFR, as a wideband can do, so would only work for target AFR map's load sites that were mostly, if not all 14.7!

_________________
Stu
"Take one old performance car, modify it so that it is as good, if not better, than a new performance car"

£300 Corrado 1.8 16v
The good lady wife's Golf VR
avatar
mrbeige
.:Admin:.

Male Number of posts : 1384
Age : 41
Location : Near Ely, Cambs.......but another planet according to my wife....

View user profile http://www.oddunit.com

Back to top Go down

Re: OE Development ECUs...

Post by dirtytorque on Wed Mar 31, 2010 5:16 am

mrbeige wrote:Ah, I see where you are coming from, but again, I think I may have not been clear (need to concentrate on my written communication Laughing )

I am employing a wideband lambda, and that will give me the AFR error based on my target AFR map. I'll use that to determine the increment to the VE map.

What I want to do is check the AFR error, then depending on how far away it is from the target AFR, map that to an increment size and then make that incremental change to the VE map. I wouldn't intend on making large changes in one go, but rather smaller steps at a time. depending on the error will determine the size of the step though.

In fairness though, it would actually be possible to use a narrow band to achieve the same thing, it's just that it would take longer. The original strategy generates a 'multiplier' value that is based on learning stoic, so it is possible to then use the magnitude of the multiplier to adapt the map accordingly. However, it only learns stoic, and not actually AFR, as a wideband can do, so would only work for target AFR map's load sites that were mostly, if not all 14.7!

Using a Nb would be limiting I think Just couldn't see where you were getting your reference from for afr,but your using a wideband so that explains that. Cool
avatar
dirtytorque
.:Charged:.
.:Charged:.

Number of posts : 1101

View user profile

Back to top Go down

Re: OE Development ECUs...

Post by mrbeige on Wed Mar 31, 2010 5:25 am

I agree. For standard closed loop, narrow band is fine, but for what I'm trying to achieve wideband is defo the way forward.

I know that Megasquirt has a self-calibrating feature, but what does it adapt? Is it the VE map?

_________________
Stu
"Take one old performance car, modify it so that it is as good, if not better, than a new performance car"

£300 Corrado 1.8 16v
The good lady wife's Golf VR
avatar
mrbeige
.:Admin:.

Male Number of posts : 1384
Age : 41
Location : Near Ely, Cambs.......but another planet according to my wife....

View user profile http://www.oddunit.com

Back to top Go down

Re: OE Development ECUs...

Post by dirtytorque on Wed Mar 31, 2010 6:35 am

Yes,you have a the autotune feature to help get attain your target afrs.This directly changes the VE table.
Once you have the VE tune in I don't think the closed loop feature adjust the vE table anymore.
I think it just moderates another multiplier to correct any error read back from the lambda.
avatar
dirtytorque
.:Charged:.
.:Charged:.

Number of posts : 1101

View user profile

Back to top Go down

Re: OE Development ECUs...

Post by mrbeige on Wed Mar 31, 2010 7:37 am

Ah well, then, I wasn't going to have any closed loop fuelling apart from the adaption, but I think that would be daft. I will calibrate that back in then Laughing

The adaption that I do have has a secondary adaptive map. I think it would be good to know how long you have been adapting at any given load site, so the secondary map has the same axes, but the adaptive 'increment' is the fast task rate (10ms) and will only increment when 'adaptive' is active. This will allow me to know which load sites I am neglecting, and thus drive accordingly to given RPM and load

_________________
Stu
"Take one old performance car, modify it so that it is as good, if not better, than a new performance car"

£300 Corrado 1.8 16v
The good lady wife's Golf VR
avatar
mrbeige
.:Admin:.

Male Number of posts : 1384
Age : 41
Location : Near Ely, Cambs.......but another planet according to my wife....

View user profile http://www.oddunit.com

Back to top Go down

Re: OE Development ECUs...

Post by Sponsored content


Sponsored content


Back to top Go down

Page 2 of 2 Previous  1, 2

Back to top


 
Permissions in this forum:
You cannot reply to topics in this forum