dimented24x7

iTrader: (2)

Its for emmisions as a whole. The closed loop calibration keeps the engine targeted at a desired AFR at P/T and running relatively clean and at an AFR thats safe for the cat, even in not so ideal conditions, like an engine fault of some sort. It also provides compensation for other emmisions equipemnt, like the charcoal can and EGR, which can cause the AFRs to vary.

Its sort of a moot point if theres no emmisions equipment, though.

11sORbust

.....thought people would reject that idea when I suggested it. LOL

Well, closed loop does serve on other function. To aid diagnostics. That is not important though, esp when using a wideband and/or "knowing" the engine.
no cat = open loop
The End

rooster433

quess we shoudl write FAST and tell them how pointless closed loop O2 feedback is.

JPrevost

Why? It's only pointless if you are learned and have the patience to tune open loop. Good results can be done with a narrowband o2 sensor in closed loop but the engine would run better if you tuned it in open loop so that you could define what AFR your engine wants/needs and not just stick it at 14.7

rooster433

I guess I just wasn't clear. If haven't read anyone of my post.. at least read this one. I'm getting around the just sticking it at 14.7

Sure you can make a great tune in open loop. The problem is you can leave the car on a awesome tune.. let it sit in the garage for 2 months and when you pull it out again it will be off... Not to mention its hard to cover every possible load/RPM senerio in any tuning session. There are things you just can't forsee in a open loop tune.. Like how a aftercooler picks up 120 degree heat off the road in the summer..

Thats where this adjusting thresholds comes into play..

Here has been my process..

1. Give the car a really good Closed loop, 128/128 locked tune with the wideband.
2. Create a desired AFR table
3. Use both on board sensors (NB and WB) to create a lookup table. I also made a few histograms to visualize what was happening with both types of sensors.
4. Adjust thresholds to a "best guess" with the lookup table I made specifically for my car.
5. Unlock BLM's/INT's.. Create a histogram of where the AFR went vs. Where you want it vs. BLM/INT avg
6. Finish tweaking the thresholds


Its just that simple. Now the car is keeping the AFR I built in step 1. with the ability to go - 15% (108 blm) to +25% (160 BLM). Of course its not desireable to see them go that far. The idea is you take the car out and the fuel trims are going up and down say 114-136 keeping the car inline all the time.




Now if the weather changes, I'm in a wierd loading area that I didn't cover when I tuned the car, the aftercooler is just complety heat soaked. WHATEVER the senerio the car is going to pull the AFR I prescribed earlier with the ability to learn +/- 15% with block learns and INT's.


For those worried about it driving the BLM's up or down during PE just add a reset bite for <128. (all mask i've seen have it for >128 already)

JPrevost

kevm14

How about this for people who want to create an open loop tune that deviates from stoich, but do not have a WB nor have the ability/desire to fool with the NB treshold voltages:

Tune VE for BLMs that are the percentage off from 128 that you want to deviate from stoich. So if you wanted 13.0, you'd tune BLMs to 113 in the area you wanted that AFR. Then you could either lock the BLMs at 128 in CL or just run regular open loop using the VE you made. Would this work at all?

If the commanded CL AFR is, say, 13.0, would that affect the BLMs? I know it's been stated, but I want pure english. If the initial BPW is changed by the 13.0, then the ECM will have to correct the AFR to stoich by leaning the mixture out and ending up with a BLM under 128 (like, say, 113), right? Would this also work for creating an open loop or locked 128 CL tune?

Perhaps I am way off but I wanted to get this down before I forgot.

69 Ghost

iTrader: (1)

I was also thinking of tuning for OL but then run CL for cruise purposes only. The idea is to force it to lock the BLM's with small throttle changes. Not sure if this is possible by using the delta TPS.

JP86SS

iTrader: (1)

So why not have thresholds looked up based on the BLM cell ?
Or some other qualifier. (maybe cycle time restriction?)
Just a thought.

rooster433

It already works this way. Depending on mask BLM's are learned per fuel trim cell (which you can define also). Thresholds can be defined by gm/sec or map and RPM depending on mask.

As long as the BLM's are not maxed out its hitting the threshold programmed.. So in other words 114 BLM <> 13.0


To say you can create a open loop tune that runs perfect year round is pretty funny. But hey, i'm not learned..

JPrevost

I don't think it's funny, maybe you hit the head on the nail with telling me you arn't learned.
Nobody, I repeat, NOBODY, can make a tune one day under certain conditions and expect to run perfect under different conditions but then who said a tune was done in 1 day? I didn't, nobody did, infact tuning never stops and can always get better. From your statements it sounds like you just keep tuning the VE for the BLMs and haven't paid much attention to the compensation tables. If that's the case, then it's no wonder you are running closed loop. If you, like RBob said, tune the compensations when you get the opertunity to do so then there should be virtually no need to touch the VE under different conditions. Instead, adjust the compensations until your AFRs are back as commanded! Done, end of story and yes, open loop tunes are the best for every engine.

11sORbust

I did that two years ago.

There are correction factors for barometric pressure and ambient temps (esp on map systems). CTS and AIT correction is also used. So when tuning in open loop, you just adjust the CORRECTION FACTORS. That way your are not always recalibrating for different conditions. Once the correction factors are set, AFR will not vary like you think.

A/F ratio will vary when using a feedback sensor though, especially when using it to command anything other than 14.7 . First it constantly toggles rich/lean, extremely inefficient. Then it's readings will vary based on exhaust gas temperature, fuel contamination and age. Something like an exhaust leak would throw the whole target a/f ratio in the trash. Not to mention, zero to one volt is **** for resolution(vs 0-5v).

It's not a bad idea though. You seem to have the NB controlling A/F "ok". The bottom line is you need to run test under different conditions before concluding this is a "great thing". Even go as far to create an exhaust leak and monitor the effects on a/f ratio. How many test have you ran? What was the EGT readings? What was the test conditions? Coolant temp?

kevm14

Seems you don't like switching O2s much. When you say toggling rich/lean is inefficient, are you referring to the 2nd comment I quoted?

How do you suppose maintaining constant AFR is a lower average pulse width? If maintaining a constant 14.7 averages a lower PW than toggling between 14.4 and 15.0, then math is wrong somewhere...in other words, if the AFR spends more time on the rich side of stoich than the lean side as it toggles, then it's not really averaging stoich. I think GM specifically set the O2 thresholds to average stoich with their O2 sensors.

Z69

Bingo.....

But a lot of us have engines that run better at <14.7.
Trying to do that with a switching sensor is less efficient.

kevm14

But what do you mean by less efficient? I could understand harder, less accurate, less precise, less desirable...I guess I'm struggling with the language here. That and pedantry is my middle name.

11sORbust

Narrow band O2's are designed to constantly switch, or toggle. The further one goes from 14.7AFR, the less amount of switching is going on.

rooster433

in voltage, not afr.

69 Ghost

iTrader: (1)

Note that the current tables in $8d for open loop are AFR Vs. Map and AFR Vs. Temp. The newer LT1's use a single 'lookup' table that is not a combination of those but a AFR Vs. Temp Vs. Map for better resolution. This allows individual control at each location instead of a combination of them. I spent some time looking at these tables and think the overall variables are conservative on both ends. Not to hijack the thread but for a open loop tune and better control it would it not make sense to incorporate this into the S_AUJP hac? Currently my AFR's for running temp vary from 14.6 to 12.4.

blue86iroc

iTrader: (2)

I never thought about applying O2 threshold changes to affect the entire closed loop cycle (my main interest always lied with adjusting them under idle conditions). You obviously can't expect a NB sensor to be accurate at other calibrated AFRs, but I was curious how the PID logic would affect the switching states of the controller. I made a graphic to help illustrate my question.

Since NBO2 response curves are sigmoidal (S-shaped), they're clearly more accurate closer to the point of the curve's inflection. This is the 14.7:1, stoichiometric, cat-friendly AFR. From what I've gathered by reading RBob's '747 fueling paper and just a general understanding of PID loops, when the ECM intentionally overshoots (or undershoots) the median voltage of 450 mV, it determines the amount of error and applies a correction to bring it back in line and cross over to the next state, be it rich or lean. Because of the gain used and the inherent nature of the integral method, the car is always going to oscillate between rich and lean states to ensure it's roughly maintaining a 14.7:1 AFR. It doesn't try to achieve 450 mV, but rather stay away from it. In the $6E mask, the oxygen sensor needs to be operating either above 699 mV or below 199 mV to even be considered functional. If it stays within those bounds for more than ten seconds, the ECM switches to open loop. It's interesting to note this range is right around the linear region of the response curve.

For the sake of PID gain and the error term, the switching point should be located at the point of inflection of the curve (as it is). On my graph, this means that whether the sensor is reading rich (above 450 mV) or lean (below 450 mV), the error term will always be roughly the same absolute distance from the setpoint (450 mV). In other words, err_0 is equal to err_0.

So let's say you have a WB, figure out approximately what NB voltage equates to the AFR you want to run, and change the O2 thresholds. For the sake of example, assume it's 790 mV. This "adjusted stoichiometric" AFR is now what the ECM will try to achieve. However, you've moved out of the linear region and into a logarithmic one. The error terms err_R (too rich) and err_L (too lean) are at approximately the same AFR offset from the adjusted stoichiometric AFR, but they may or may not have the same magnitude for a given AFR away from the adjusted stoichiometric AFR. What I see happening is the ECM not being able to hone in on the setpoint as easily because one error voltage could be any number of actual AFRs. It seems that this could cause the ECM to either over- or under-compensate and throw the cycle into wild oscillation or grealy slow the performance of the loop.

One could probably remedy this situation by having individual gains when the error is either rich or lean, but I don't think GM has implemented it since the sensors were always designed to be run at their default switching point. Even if it were possible, it wouldn't be perfect.

I'm not claiming that changing the switching point won't work (rooster433 proves that it does), but it's not an ideal scenario. Most of the posts have discussed the NB's lack of resolution at anything but 14.7:1, although it seems like the switching behavior might also become dysfunctional, negating the point of even moving the thresholds.

RBob

iTrader: (1)

I've stated in other posts and even in the PID paper that the O2 switch points can be manipulated within reason. GM also does this. From what I can tell of their tunes it is dependent upon what CAT is installed (2 or 3 way), and the allowed emissions vs. drivability and mileage as secondary.

Here is an example of a '92 f-body with a 3-way CAT:

The same values are used off-idle from 0 to 40 gms/sec aiflow. That is a tight straddle of 450 mV. After that they slowly increase to 492 mV and 530 mV at 128 gms/sec.

As for using the NB switch points at WOT, I'm with you blue86iroc. Just not a good idea.

You asked about rich/lean magnitude past the switch point. There is a correction value in the calibration for the rich side. It is set up to reduce the magnitude of error. Here it is for the $8D code.

It is about a 12% reduction is the rich side O2 value. This reduces the O2 error term when rich.

RBob.

69 Ghost

iTrader: (1)

It would seem that the real issue here that users will apply the O2 constants to change the idle characteristics. Off idle is less important or not important. I think the Stoich constant is kind of the same way why do you want to change it and what do you expect from it? I am at a point where I am thinking the opposite of GM -remember emissions and economy for CL. I think that you can tune in open loop using a WBO2 to tune to get the motor running like you want just like a carb setup. This should entail mostly just using the standard tables -VE, AE, PE, etc. Once you get the OL tune in line then slowly make the car run closed loop starting with a very narrow range of parameters -for cruising or steady state conditions. This would be of course based on temp and then probably delta TPS, etc. Grumpy always states give er what she wants. Point is that with OL once you get the motor where you want it go CL and note the differences like the VE or MAP is way out of wack then start messing with the O2 constants and Stoich, etc to note the effects. I have gone the opposite direction -Idle problems go here. Don't know yet but I think once the OL tune is down you should only see minor changes for CL if your OL tune is on. If it is not you have a baseline as to where you want to be on the tune with the car.

blue86iroc

iTrader: (2)

RBob, it's interesting that the '92 code you posted shows GM really trying to maintain the 450 mV median. I've always encountered a slight rich tendency for these values, presumably for reliability issues. The $8D code looks to be a bit higher in the 600 mV range. Do you think this explains why there's a separate error correction for the rich side of the sensor? The $6E code I use doesn't have separate O2 values for idle (unfortunately), but at idle airflows it's very similar to $8D; a fair bit higher than 450 mV.

69, I agree that most people will use these to adjust idle characteristics if they are forced to run CL. I don't have a WB right now so I'm limited as well, but it seems that OL with a wide-band is really the way to go. Assuming the NBO2 was working properly, I don't see closed loop being much different than open loop IF you've got it tuned for 14.7... I believe that's what Fast355 has done (up the thread a bit). That would apply only to part-throttle conditions, of course, since WOT uses its own AFR.

RBob

iTrader: (1)

I believe the difference in the target O2 values is whether a 2-way or a 3-way cat is in use. A 3-way cat requires a tight stoich mixture.

In the same '92 f-body code that I shows the idle O2 values, here is the rich O2 correction term:

Here is the same term in ANLU, an '89 Caprice running $61 code:

What is interesting here is that when in idle both terms are used. In this manner the idle rich error magnitude is greatly reduced.

I always thought that this correction was because the O2 sensor output was a little lopsided. With a tendency of a slower and greater response to a rich AFR.

RBob.

blue86iroc

iTrader: (2)

Interesting stuff. Do you think the reason for a slower, rich, AFR response is due to the inherent design of the sensor (i.e. less "resolution" in a richer state; lopsided curve) or just the nature of the S-curve if the stoichiometric AFR voltage is pushed a bit rich? Maybe both in some cases?

JP86SS

iTrader: (1)

I was thinking that the reason may be to limit the amount of correction when removing fuel. The reaction of the control may cause the swing lean to be too great so a fudge factor was put in the code to give some control over the amount of swing in the cycle. Possibly for an inherent overcorrection when going from Rich to Lean.

JP86SS

iTrader: (1)

"Narrow band O2's are designed to constantly switch, or toggle. The further one goes from 14.7AFR, the less amount of switching is going on."

This statement still bugs me in that the switching function is a function of the threshold windows that are calculated, not by the sensor itself. The term "switching" only relates to the very narrow linear range that it operates at. Anywhere within the linear range, the operation is the same. Once out of that range the nonlinearity makes control difficult because of the change in slope needs different control parameters at each point along the curve.
This would be true if the sensors reaction within the linear range is consistent. If the reaction time of the sensor changes at different points (going R-->L, L-->R)
Which may or may not be the case.
Does anyone have any data to prove the reaction times of NBs in and out of the linear range?

tpep

I have also noticed some unusual switchpoints in my stock calibration. Here are the relevant O2 variables for a '91 LT5 calibration (BFXB). Note that the Rich/Lean voltages are all in the 300's.


This application has dual catalytic converters, and heated O2 sensors, located just aft of the exhaust manifolds. Even with these thresholds in the code, it does seem to toggle around 14.7 on the wideband, so the values work for this application.

Todd

dimented24x7

iTrader: (2)

To further add to that, its the steep response and non-linearity that forms the basis of the control stratagy. The sensors response around 14.7:1 is very steep, once you start to stray away from stoich, the sensors rapidly heads toward .9V or 0 volts. This made it easy for early computers with low-res A/D converters to see the change in AFRs and allowed the closed loop to maintain a reasonable ammount of control over the AFRs without having complex sensors, hardware, and software routines. The switching primarily allows for fast response and also allows for easy diagnostics. No switching = dead sensor.

antman89iroc

RBob, Still feel this way? I know this is an old thread and positions sometimes change over time. Maybe you've learned something new.

The reason I ask (and you've commented about some of my posts) is that I am running a non-cat no EGR setup and can't get my arf where I want it. Sometimes I think I get close then I end up chasing my tail. I've been running closed loop until now and am seriously considering your advice..

BTW this is for my 165 $6e maf car.

18436572

I know that the question was directed towards Rbob but...... I feel this way 100%. I switched to open loop tuning about 9 years aggo because I run leaded fuel at the track sometimes. This is by far the best way for me. I run $8D. I will never go back even if I quit running leaded fuel!

RBob

iTrader: (1)

The biggest advantage to open loop fueling is the lack of the required O2 cross-counts. With large injectors it is difficult to get just the right amount of proportional gains to prevent surging and/or a strange feel to the engine.

The forced oscillation of the AFR can be felt once one starts to switch between open & closed loop. With open loop being so much smoother. I'm sure that this is also due to being able to run slightly richer AFR's.

antman89iroc, in closed loop the ECM targets the AFR via the O2 sensor feedback. There are some tables in the calibration that are used as the target.

RBob.

Street Lethal

iTrader: (16)

Like everything else, resolution is the factor in terms of how efficient the ECM can run the engine. In Closed Loop, the '7730 has what, maybe a 6% O2 correction capability? There are ECM's out there that have close to, and perhaps even over, 20% O2 correction capabilities, and Closed Loop with those ECM's are much more smoother than an essentially stock OBD1 ECM in Open Loop...

18436572

Depends on the application and the $$ you are willing to invest. The 7730 can be open loop tuned to perfection especially for a car that gets beat on at the track alot. My ride is proof.

Street Lethal

iTrader: (16)

Too many variables, to be honest. Remember that the ECM switches to Open Loop anyway under Power Enrichment, and the biggest complaint with Closed Loop is the various timing constraints during part throttle when certain conditions aren't met, but that can be adjusted for. This is why people will always ask why their cars seem faster when just started up, they have no idea that the ECM is pulling timing based on other factors. I myself prefer to tune in Open Loop, but then run then engine in Closed Loop after a few changes are made...

Carburetors can be tuned for perfection without O2 correction, too...

18436572

I don't agree.

A carb can never be as precise as EFI, even in open loop. EFI will also make more power than carb on a given set up.

antman89iroc

OK Street Lethal I am trying to make sure I understand what you are saying. CL is better than OL if you have the o2 correction available and tuned properly?

I am running the 165 MAF oem computer with $6e mask. Basically a stock 89 IROC system. I am/was attempting to tune in CL but targeting lower than 14.7 afr's for idle, cruise, light accl and heavy accl (not in PE). Do you believe that with THIS computer/system I am better to run in CL with o2's tweeked or just bite the bullet and go OL? Again, this is a mostly street, sometimes track driven car with no cat, egr, AIR etc. Still have AC cruise power brakes etc. Not a daily driver but driven year round weather permitting lol.

antman89iroc

RBob, I noticed that when I forced OL my BLM's were at 118 across the board, which is where I believe they were on my last CL tune. Should I force BLMs to 128 for OL operation? And one more thing, should I clear (ie remove power) my computer for a "fresh start" when starting this OL tune stratigy?

RBob

iTrader: (1)

There are several ways to handle this. The easiest is to set the min/max BLM values to 128. This way they never move from that value and if they do from a different BIN, they get reset back to that value.

In $6E ($8D is very similar), there are only two of the 16 BLM cells that get saved. The other 14 get reset back to 128 at key-on. The two saved BLM cells have a key-on init min/max values.

Note that in ARAP the min/max BLM values are 108 & 160. This provides a minus 16% correction and a plus 25% correction to the fueling.

RBob.

Dominic Sorresso

Antman,

My LT-5 is ported and runs larger intake cams, headers and no cats. However, I am using C/L for tuning. From what I can tell, not too many other tuners have done this before on an LT-5 w cams. Using the information from Rbob and Trax re: O2 sensor window and Prop gains, I can tell you my LT-5 idles smoother and drives better than other LT-5s with similar setups. I've been able to "fix" someone else's tune and alleviate things like idle/PT surging and idle richness.
Lots of great info from RBob's white paper on PID controls for the $42 def which translates well to the SD LT-5 definitions. So count me in as a proponent of
C/L particularly with cammed motors. Just my $.02 but I think Rbob, Trax, and others have contributed a great deal to the understanding of using C/L for higher performance motors. As an FYI, the current state of tune for my motor puts it at
435rwhp @ 6800 still gets 23-24mpg at 75mph cruise.