Lean burn with negative split timing
05-20-2011, 11:01 AM
#101
I realize this thread is a bit of an intellectual exercise and that there are no "absolute" AFRs - the engine needs what it needs - but the best torque fuel values are pretty well known on the Renesis for all operating loads and RPMs.
Operating the engine at values that deviate from the known values does NOT increase fuel efficiency because maximum torque IS maximum efficiency.
Operating the engine at values that deviate from the known values does NOT increase fuel efficiency because maximum torque IS maximum efficiency.
I'm willing to accept that what you say is true wrt the Renesis. I know you've done the work.
However, for most engines it simply isn't true. Systems like lean burn and EGR improve fuel economy and reduce torque in most engines. Even aside from lean burn, virtually all GM ECUs have a lean cruise mode -- at least when they are sold outside the US.
So, the Renesis is an exception, I suspect it is due to relatively low inherent pumping/throttling losses further reduced by high operating RPM.
It is also worth pointing out that true lean burn engines typically have separate control loops for lean burn. AFRs are slowly raised over several minutes of sustained low-load operation, and the slightest throttle openings trigger an exit and the whole cycle has to repeat. I've read lots of hypemiler types complaining about this behavior.
I don't understand the closed loop PID well enough yet to know if I could replicate this behavior, but it certainly wasn't what I was trying for to start with, so I think my endeavor, even in principal, may have been doomed from its inception.
05-20-2011, 04:50 PM
#102
I'm willing to accept that what you say is true wrt the Renesis. I know you've done the work.
However, for most engines it simply isn't true. Systems like lean burn and EGR improve fuel economy and reduce torque in most engines. Even aside from lean burn, virtually all GM ECUs have a lean cruise mode -- at least when they are sold outside the US.
So, the Renesis is an exception, I suspect it is due to relatively low inherent pumping/throttling losses further reduced by high operating RPM.
It is also worth pointing out that true lean burn engines typically have separate control loops for lean burn. AFRs are slowly raised over several minutes of sustained low-load operation, and the slightest throttle openings trigger an exit and the whole cycle has to repeat. I've read lots of hypemiler types complaining about this behavior.
I don't understand the closed loop PID well enough yet to know if I could replicate this behavior, but it certainly wasn't what I was trying for to start with, so I think my endeavor, even in principal, may have been doomed from its inception.
The problem as I see it is the inability to override the fuel trim regime - although you say you achieved this ? If so how ?
BTW - I do have access to ALL the maps via my protuner software and may be able to find a map that would have the ability to allow higher AFRs . Maybe there is a possiblity to use the fuel trim logic in reverse ? IE have it adjust to a lean condition .
I was surprised to find how quickly the fuel trims do activate when in cruise mode.
Also - have you tried playing with the Ve map in closed loop ranges as MM alluded to earlier ?
Edit : just found this reference . Perhaps MM is onto something but wants to dissuade anyone else from trying
I know recently Jeff Abrams' exploits in hypermiling his own RX8 have proved to be rather informative.
Much different application mind you but I think it shows that the rotary can get good fuel mileage under the right circumstances and I don't mean only when it's not running.
Much different application mind you but I think it shows that the rotary can get good fuel mileage under the right circumstances and I don't mean only when it's not running.
Last edited by Brettus; 05-20-2011 at 05:30 PM.
05-20-2011, 10:44 PM
#103
I do intend to test out lean burn on a long trip, but maybe not for a couple weeks, as I mostly do short trips, and I'm getting some tuning fatigue.
MazdaManiac is a magician, and magicians often depend on misdirection as much as skill.
It isn't quite the same is lying though, and I think he believes what he is saying here.
Here is what I've learned about closed loop.
The algorithm used is a proportional integral derivative controller (PID), or perhaps just a PI controller.
http://en.wikipedia.org/wiki/PID_controller
Closed loop has one main purpose -- get the car to lambda or to oscillate around lambda.
Here is my current thinking on how it functions:
1. Entering the loop, fueling is calculated by lambda x the value in the VE table.
2. A reading is taken from the WBO2 sensor, and the difference between that and lambda is calculated, normally called the error.
3. The value in table A is multiplied by the error.
4. The result is used for a fueling calculation.
5. Table B is just a multiplier to cancel out oscillation when the above correction overshoots the target. So if it was trying to correct from 15 to 14.7, and it got 14.4, it will stop it from goin back to 15 the next time through the loop. It can also be used to cause such oscillation, as catalysts like it better when the fueling switches from rich to lean rather than staying right at lambda constantly.
You can change the fuel ve to alter the initial fuel calculation to make the AFR very lean or rich. Then you can set a very small or very large value in table A to prevent the loop from correcting back to lambda.
Motec uses a similar system for closed loop, so you may find more info on tuning it on their forums, as I'm sure my understanding/explanation is somewhat incomplete.
MazdaManiac is a magician, and magicians often depend on misdirection as much as skill.
It isn't quite the same is lying though, and I think he believes what he is saying here.Here is what I've learned about closed loop.
The algorithm used is a proportional integral derivative controller (PID), or perhaps just a PI controller.
http://en.wikipedia.org/wiki/PID_controller
Closed loop has one main purpose -- get the car to lambda or to oscillate around lambda.
Here is my current thinking on how it functions:
1. Entering the loop, fueling is calculated by lambda x the value in the VE table.
2. A reading is taken from the WBO2 sensor, and the difference between that and lambda is calculated, normally called the error.
3. The value in table A is multiplied by the error.
4. The result is used for a fueling calculation.
5. Table B is just a multiplier to cancel out oscillation when the above correction overshoots the target. So if it was trying to correct from 15 to 14.7, and it got 14.4, it will stop it from goin back to 15 the next time through the loop. It can also be used to cause such oscillation, as catalysts like it better when the fueling switches from rich to lean rather than staying right at lambda constantly.
You can change the fuel ve to alter the initial fuel calculation to make the AFR very lean or rich. Then you can set a very small or very large value in table A to prevent the loop from correcting back to lambda.
Motec uses a similar system for closed loop, so you may find more info on tuning it on their forums, as I'm sure my understanding/explanation is somewhat incomplete.
05-21-2011, 12:09 AM
#104
Table B might also set the rate at which the algorithm seeks to eliminate any steady-state error. I simply zeroed out the whole table because that had little effect, and eliminating a variable made things easier to figure out, but you might have to do that to make it work.
05-21-2011, 12:28 AM
#105
Very helpful:
http://blog.opticontrols.com/archives/260
http://blog.opticontrols.com/archives/260
05-21-2011, 01:27 AM
#106
Even better: http://forum.aempower.com/forum/inde...ic=2183.0;wap2
O2 FB proportional table Closed loop table A
Units: Proportional Gain vs. RPM
Description: Proportional control is a pure gain adjustment acting on the error signal to provide the driving input. The advantage of a proportional-only control is its simplicity. If AFR offsets can be tolerated, the use of a proportional controller may be optimal. However, it will not eliminate the steady-state error that occurs after a set-point change or a sustained AFR disturbance. Note: When tuning the O2 feedback and overshoot occurs, lower this number. If undershooting the AFR target, raise this number.
Typical Use: Used to adjust the speed of the system and reach the AFR target quickly.
O2 FB integral tableClosed loop table B
Units: Integral Gain vs. RPM
Description: Integral control is implemented through the introduction of an integrator. This is used to fine tune the O2 feedback at the AFR target once the proportional has acted in getting close to the target. Note: start tuning the O2 feedback with this option at zero, until the proportional has the feedback close to the target, then step this in slowly until the feedback holds the AFR target.
Typical Use: Integral control is used to provide the required accuracy for the control system.
Units: Proportional Gain vs. RPM
Description: Proportional control is a pure gain adjustment acting on the error signal to provide the driving input. The advantage of a proportional-only control is its simplicity. If AFR offsets can be tolerated, the use of a proportional controller may be optimal. However, it will not eliminate the steady-state error that occurs after a set-point change or a sustained AFR disturbance. Note: When tuning the O2 feedback and overshoot occurs, lower this number. If undershooting the AFR target, raise this number.
Typical Use: Used to adjust the speed of the system and reach the AFR target quickly.
O2 FB integral tableClosed loop table B
Units: Integral Gain vs. RPM
Description: Integral control is implemented through the introduction of an integrator. This is used to fine tune the O2 feedback at the AFR target once the proportional has acted in getting close to the target. Note: start tuning the O2 feedback with this option at zero, until the proportional has the feedback close to the target, then step this in slowly until the feedback holds the AFR target.
Typical Use: Integral control is used to provide the required accuracy for the control system.
Last edited by oltmann; 05-21-2011 at 01:57 AM.
05-21-2011, 04:19 PM
#107
It is difficult to tune closed loop when you have vastly different AFRs at different loads but the same rpm. For example, if I'm running 19:1 at 25 load, and I crack the throttle even a little, load jumps up to the 60s almost instantly.
Say I'm targeting 14:1 at 60 load, well the torque difference between 19:1 and 14:1 is almost instantly sent through the drivetrain and the car bucks. Same effect I reverse when load drops down. You might guess that this is a lean misfire or something, but I've seen the same behavior at lower AFRs, just not as pronounced, but still very annoying.
Even so, the car is drivable, and if I keep a steady foot, I can cruise along just fine at stupid high AFRs. In fact I don't really know how high, because in some cases I can drive while over the limit of the WB.
Also, I can go very lean even on the stock timing table, negative split is not required. I guess this is owed to the good fuel atomization of the Renesis, and the BHR ignition.
The problem? Fuel economy isn't good so far, neither in one short-term test, nor in the injector duty cycles I've logged.
However, I've set my "normal" tune to target .95-.92 in closed loop, and I'm getting excellent power, drivability, fuel economy (based on injector readings) and, best of all, sound. Thanks, Jeff!
There may be some application for lean burn, but if so, it appears to be in a very narrow operating range that I haven't found yet and can't measure and would probably have little effect on overall fuel economy.
While I'm not completely ready to say the game is in the refrigerator, right now lean burn seems to be a tuning technique in the same category as setting the rev limiter to 10krpm. It is cool that the Renesis can do it, but ultimately rather useless.
Say I'm targeting 14:1 at 60 load, well the torque difference between 19:1 and 14:1 is almost instantly sent through the drivetrain and the car bucks. Same effect I reverse when load drops down. You might guess that this is a lean misfire or something, but I've seen the same behavior at lower AFRs, just not as pronounced, but still very annoying.
Even so, the car is drivable, and if I keep a steady foot, I can cruise along just fine at stupid high AFRs. In fact I don't really know how high, because in some cases I can drive while over the limit of the WB.
Also, I can go very lean even on the stock timing table, negative split is not required. I guess this is owed to the good fuel atomization of the Renesis, and the BHR ignition.
The problem? Fuel economy isn't good so far, neither in one short-term test, nor in the injector duty cycles I've logged.
However, I've set my "normal" tune to target .95-.92 in closed loop, and I'm getting excellent power, drivability, fuel economy (based on injector readings) and, best of all, sound. Thanks, Jeff!
There may be some application for lean burn, but if so, it appears to be in a very narrow operating range that I haven't found yet and can't measure and would probably have little effect on overall fuel economy.
While I'm not completely ready to say the game is in the refrigerator, right now lean burn seems to be a tuning technique in the same category as setting the rev limiter to 10krpm. It is cool that the Renesis can do it, but ultimately rather useless.
05-23-2011, 10:52 PM
#108
I've extended my research tentacles to the nihongonet and found some nice charts published by Mazda. I found one where Mazda shows that negative split does improve fuel economy... when the rotor recess is moved towards the trails edge. That design reduced power quite a bit, so they didn't use it.
This is from a 2003 Mazda Tech Report on the Renesis. It is in an SAE paper as well, but this one is free on the web, so I figured it was okay to post it. VERY interesting.
Torque, NOx, lambda from the dual hydrogen/lean burn gasoline Renesis prototype. It has an EGR system added, which seems to be important. Says something about torque control as well, but I can't understand the translation. Talks about emissions and gas prices, but not efficiency that I can see.
Still don't have closed loop tuning down 100%, I can usually get the AFRs I want, but it takes to long to reach them, and I get oscillations some of the time. Working on modeling the system in Matlab so tuning isn't so laborious. It is really nice system when it work, works fast and creates stable fuelling, but I still feel like I'm reading by candle light.
That design reduced power quite a bit, so they didn't use it.This is from a 2003 Mazda Tech Report on the Renesis. It is in an SAE paper as well, but this one is free on the web, so I figured it was okay to post it. VERY interesting.
Torque, NOx, lambda from the dual hydrogen/lean burn gasoline Renesis prototype. It has an EGR system added, which seems to be important. Says something about torque control as well, but I can't understand the translation. Talks about emissions and gas prices, but not efficiency that I can see.
Still don't have closed loop tuning down 100%, I can usually get the AFRs I want, but it takes to long to reach them, and I get oscillations some of the time. Working on modeling the system in Matlab so tuning isn't so laborious. It is really nice system when it work, works fast and creates stable fuelling, but I still feel like I'm reading by candle light.
05-24-2011, 12:26 AM
#110
Of course I've tried this. What do you think I do all day? I put 30 or 40 hours a WEEK into trying stuff out on the OE PCM.
I just don't have the time or patience to sit here and explain stuff that doesn't work.
JUST USE THE REGULAR FUEL MAPS AND THE ENGINE Ve MAPS!!
How much more plain can I be about this?
05-24-2011, 06:50 AM
#112
I was getting mostly ~14mpg. Started getting tunes from MM, and it went to ~16mpg. Latest was 17.6mpg on my latest tank, which was all urban driving and some WOT testing down at my personal drag strip by the abandoned military base.
Not a lot of data, but concordant with what MM has said.
Just to mitigate any confusion regarding the stuff I just posted. Mazda seems very interested in rotaries lean-burn ability, but they don't seem to know what to do with it.
In their papers they like to highlight the ability to run lean, but hedge around the topic of what AFR gives the engines best overall BSFC. For example, maybe 16:1 works best if you gear it to run at 1200rpm at 50mph, but you'd actually get better MPG running at 3000rpm at 14:1.
I was hoping to find some "real" data, but it all seems to be tarted up PR.
05-24-2011, 06:00 PM
#113
Maybe I need to pull out that "cool story bro" pic classic again to remind you how ridiculous you appear to me at times .
Well that is major improvement on how you normally answer questions . See - you can do it if you try !
Well that is major improvement on how you normally answer questions . See - you can do it if you try !
Last edited by Brettus; 05-24-2011 at 06:07 PM.
06-12-2011, 05:06 PM
#116
OK - adjusted the CL maps per Oltmans suggestion and have been running 0.94-0.95 for a few days now . No interference from LTFTs .
My WOT afrs in boost seem richer for some reason (could just be a coincidence )
Will see if my fuel economy improves per MM's advice.
My WOT afrs in boost seem richer for some reason (could just be a coincidence )
Will see if my fuel economy improves per MM's advice.
06-17-2011, 09:41 PM
#118
It seems that, contrary to widely-held belief, the rear O2 sensor is used for feedback to adjust fuelling in most cars.
Under normal circumstances, it has a command authority of only %0.5-2.0, however this may get pushed far higher at time. I think this is where the "phantom" fuel trims that create the lean limit come from.
Additionally, I've read that often the open loop map is actually a set of offsets referencing the closed loop map.
I think that the RX-8 ECU uses the closed loop map "B" and the rear O2 sensor both in open and closed loop to establish lambda, and then fuel targets are calculated from that.
The values stored in the main fuel tables are from 0-255 with 0 being 1 lambda and 255 being fully rich. Much less granular than what the CL tables use.
What probably caused you to go rich is that under open loop, feedback from the WBO2 is turned off, but it still gets feedback from the NBO2. I'm assuming you don't have one, so effectively this is the same as 0v, fully lean. Since you have commanded a rich target, and the NB02 is, effectively, fully lean, it may be giving the max enrichment allowed.
In fact, you may always be getting enrichment on this basis, but you've tuned around it without knowing it. Altering the closed loop tables only made it apparent.
I think this Mazda patent offers a reasonable description of how closed loop works on the RX-8, though it is short on details and suggests some functionality that I don't think is used.
The NBO2 offers faster, more reliable readings, at least within its puny range.
Under normal circumstances, it has a command authority of only %0.5-2.0, however this may get pushed far higher at time. I think this is where the "phantom" fuel trims that create the lean limit come from.
Additionally, I've read that often the open loop map is actually a set of offsets referencing the closed loop map.
I think that the RX-8 ECU uses the closed loop map "B" and the rear O2 sensor both in open and closed loop to establish lambda, and then fuel targets are calculated from that.
The values stored in the main fuel tables are from 0-255 with 0 being 1 lambda and 255 being fully rich. Much less granular than what the CL tables use.
What probably caused you to go rich is that under open loop, feedback from the WBO2 is turned off, but it still gets feedback from the NBO2. I'm assuming you don't have one, so effectively this is the same as 0v, fully lean. Since you have commanded a rich target, and the NB02 is, effectively, fully lean, it may be giving the max enrichment allowed.
In fact, you may always be getting enrichment on this basis, but you've tuned around it without knowing it. Altering the closed loop tables only made it apparent.
I think this Mazda patent offers a reasonable description of how closed loop works on the RX-8, though it is short on details and suggests some functionality that I don't think is used.
The NBO2 offers faster, more reliable readings, at least within its puny range.
