Debate: Closed Loop vs Open Loop - Narrowband vs Wideband

[TheMechanic]

Don't get me wrong G, it seems as though you are breaking my 30+ year "undeniable truths" I have in that if a properly functioning O2/HO2 gives me a voltage in a data stream it may be reporting a different value/voltage than what say my DSO is reporting hooked up directly to the sensor. Or that the sensor is just inaccurate at given conditions from its' rich lean switching point generically held as 14.6 AFR with E0 fuel. If so it is because of a bunch of modifiers in a PCM/ECM that is giving a calculation rather than a displayed reading.
To add putting in a WB being truly accurate during any and all conditions baring reversion or exhaust leaks of course.

[GHuggins]

Don't get me wrong G, it seems as though you are breaking my 30+ year "undeniable truths" I have in that if a properly functioning O2/HO2 gives me a voltage in a data stream it may be reporting a different value/voltage than what say my DSO is reporting hooked up directly to the sensor. Or that the sensor is just inaccurate at given conditions from its' rich lean switching point generically held as 14.6 AFR with E0 fuel. If so it is because of a bunch of modifiers in a PCM/ECM that is giving a calculation rather than a displayed reading.
To add putting in a WB being truly accurate during any and all conditions baring reversion or exhaust leaks of course.

You're not wrong with this. Widebands have their own offsets and brand to brand and how they're wired changes this.

I prefer for all of my customers to use the AEM CAN setups for this reason. They are the most controlled in my opinion. I also did direct testing on them vs some others when they came out and the AEM was the closest to my own $4500 Lambda Pro for reading transients at that time than any other I had ever seen. Again this was when AEM got the software from the Doc on here and a few of us were testing them for him. I think I wound up having one with his original edited software and then got 2 more from aem when they came out with them. I ran one in each bank on the dyno at that time. Gen 3's throw their own variables into this due to not being able to use CAN and having to use analog. Jslic at that time discovered AEM puts their own offset into their software that makes it display or read .2 leaner than it actually is too so yet another offset to add into things. All I can say is work to get them closer. They don't have to be perfect. Then go from there. Makes things a lot easier and cleaner in the cal as a whole.

[GHuggins]

So, you're not ignoring the wideband in areas where it disagrees with the narrowband fuel trims, and you're not compensating the wideband with math.

When you make an adjustment to the narrowbands does the wideband report change as well? Do you know you're good when the two converge or something? With that there's no point of reference, unless what you're saying is the wideband is what's constant and you make narrowband fuel trims match it.

Seeing a log like in post #2 what would be your thoughts on what's going on?

Hard to say on this one just for the simple fact that the O2's are already prioritizing to the rich side and not switching uniformly. This can be a wideband that needs to be cleaned or just an offset issue. You should only need to use a math to correct the wideband if an offset exist and even then I would want to gather data at a higher rpm to verify this as the O2 and wideband errors alike start coming more in line with one another at higher rpms so an offset issue would be more evident here.

As long as you know for a fact that the wideband is reading correctly, which is the whole reason I like to stick to one brand and kind, then once the engine is changed in any way shape or form the wideband then becomes the constant and all else needs to be corrected to it. Why use it for wot tuning?

This is a log from a LT header car where the MAF was really relocated from this morning if it helps you any. Again the AEM CAN is being used and this is a 4th gen 2SS camaro. There is one area off idle on this one where the two greatly differ, but everywhere outside of that they are identical O2 to wideband. The off idle is a transient issue at lower maps that I still need to sort. I just got injection timing to clean up the MAF curve and fix a lean issue that it was having around 120 to 140 kpa. Injection was too retarded in this case in that specific rpm range.

AEM O2 error.jpg

[GHuggins]

Jeez I'm starting to feel like King here making all of these long post.

Anyhow, here's another thing to digest or think about and it both argues in your favor and against. Lets look at the Ford Biasing tables for wideband controls. Why do Roush or Whipple change these if they are only adding a supercharger and cai tube? I know both companies use Ford engineers from time to time, so why do they change them if the OE settings should be held to the be the most accurate?

Another point. Gen V's... Are you aware that the torque model can control or influence the defined lambda of 1? Yes I had to see it myself to believe it. But the torque model being wrong will influence what the O2's switch around. AFR stays the same, but the trims and the EQ error both change and sometimes greatly. So, do I believe the O2's in this case skewing the airflow tables while ignoring the EQ error or do I instead correct the calibration? I prefer to correct the torque model and bring things back closer but again, that's me.

[TheMechanic]

I'm getting the feeling that the only true, good reading is going to be a 5 gas. It just has that delay plus the slow reaction time. I guess for a steady state test for NB's and WB's. 30 or so years doing Smog tests here in SoCal and I let that crap go about 15 years ago. Just no money in it.
Glad I got the AEM CAN version.

[edcmat-l1]

I've used O2 proportional and airflow vs mode in the past to try and tame low speed surge. Especially back in the day with all the LG long tubes that put the O2s 2 feet from where they were. I've used that method with some degree of success on larger injectors to try and limit the oscillation of the trims that you can feel at lower speeds. That being said I've never even considered that for trying to dial in a wideband. Not meaning to argue but I find it hard to believe I'm going to pull a WB to stoich using this method in say a LS7 with a stage 3 or 4 cam. Or anything similar. Was working on a GM crate 6.2 today and at idle the wideband (AEM) is nearly pegged and the narrowbands are oscillating nicely.

All that being said, aftermarket widebands are notoriously inaccurate in general. Greg has touched on it. The discrepancies between brands, etc. The only real way to calibrate one is using a 5 gas or some other lab grade method. That's the main reason I almost always use 2 when doing WOT. And I own and have owned a 5 gas for over 25 years. I've done quite a bit of tuning and testing with it. Using 2 widebands continuously I always see a difference in them no matter what. It's always at least a couple tenths. I always split the difference if it's acritical, forced induction, higher HP, etc.

This discussion goes beyond the accuracy of sensors though. I like how y'all are trying to keep it "tech" but it's more than that. It's about philosophies of how to do things. To me, for all of my career building and/or repairing cars it's been about doing things at an OE level. I've always been anti-cat delete. I almost always prefer milder cams for driveability and emissions. I prefer to "tread lightly". Least intrusive. Even big stroker higher power builds. If it's a street car, it's going to be fairly mild as far as camshaft selection.

I am strongly opposed to people who encourage others to delete all their smog shit, turn off half their codes and toon their vehicle with a mallet because they think they have a better way. I will continue to call those people out with fervor. They're not good for our industry. I don't want them influencing newbees and up and comers with the copy n paste and the circles n arrows.

Sorry for the rant but I feel strongly about this.

[GHuggins]

I've only ever used a 5 gas analyzer a few times as I don't own one. Stupid right. I own probably $6000 worth of widebands and don't have a 5 gas Oh well. This is why I keep saying to tune things more into one another instead of just relying on one, however most times I will lean towards the wideband more so just because so much changed from the motors original calibration standpoint.

I agree with the rest. I want my customers to be able to go out and fire their car up in the negatives in the winter and then beat the crap out of it in the summer and it keep going and stay the same for those extreme variables.

[SiriusC1024]

Anyhow, here's another thing to digest or think about and it both argues in your favor and against. Lets look at the Ford Biasing tables for wideband controls. Why do Roush or Whipple change these if they are only adding a supercharger and cai tube? I know both companies use Ford engineers from time to time, so why do they change them if the OE settings should be held to the be the most accurate?

It's in favor wouldn't you say? That the control side has to be modified means it matters. They're not running a passive wideband OL.

Are Transport Delay settings along with PI tables being changed? That makes sense because the Load axis in Ford tunes is airflow load. Sensor bias is no different than adjusting switchpoints on GM for richer or leaner trims. I doubt FAOSC has been disabled.

Yeah I'll get loud about stuff sometimes, but I figured out a long time ago that I learn more by not investing my ego in being correct. The early points from kingtal0n have been dispelled with technical explanations. I don't have an opinion on reality. If I'm wrong I'm wrong and I get to learn something.

Thanks for taking the time to show us your method, though I still don't understand it entirely. I think it enriches the learning experience for everyone.

[SiriusC1024]

I'm getting the feeling that the only true, good reading is going to be a 5 gas. It just has that delay plus the slow reaction time. I guess for a steady state test for NB's and WB's. 30 or so years doing Smog tests here in SoCal and I let that crap go about 15 years ago. Just no money in it.
Glad I got the AEM CAN version.

You may or may not know, but I'll explain. CAN Low isn't a ground wire. CANBUS signal is held at 2.5V to show 0. For 1, CAN Hi goes to 5V and CAN Lo goes to 0V simultaneously. About the 0V, right? CANBUS is offset-tolerant by design. That means 0.026 at the ECU and 0.103V at the wideband controller doesn't affect the integrity of the data. CAN Lo (-) doesn't bond the various devices and make them equi-potential.

CANBUS reference is a floating ground that is capacitively coupled to ground. This filters out noise. For DC a capacitor is an open circuit, but for AC it's a reactive element that acts like a resistor with impedance of 1/jwC. At the end of each branch of a CANBUS circuit is a 120ohm terminating resistor between CAN Hi and Lo.

Imagine if CAN Lo was given ground. This would signal a 1 all the time on Lo, but Hi would keep doing its thing. A fault would be indicated because no communication on any CAN device on that circuit would cease. Hi and Lo have to mirror each other.

Another way to look at it is what if the ground wire from battery to the wideband controller got cut? If CAN Lo was really a ground, then the wideband sensor would be grounded through the tiny CAN wires and burn them along with potentially damaging the chips in every node between there and ground.

[GHuggins]

It's in favor wouldn't you say? That the control side has to be modified means it matters. They're not running a passive wideband OL.

I still think it shows both in favor and against.

In favor because it means that even the wideband needs correcting for airflow changes.

Against because it means the OE settings are no longer accurate and the wideband/factory O2 needs correcting for airflow changes

[SiriusC1024]

Oh I don't dispute tuning the closed loop parameters for a different setup. Post #18 here has a link where I was working on that with someone. Thing I don't get is if O2's are oscillating correctly you're saying there still might be a problem?

[GHuggins]

Ahh, I see. Then yes. Depends where they are oscillating. All goes back to the same reasoning for tuning the closed loop parameters.

[veeefour]

Wut in tarnation is going on here bois? :P

All I know the earth is not flat...but Closed Loop fueling with narrow band was always making me smile.

It was funny 20 years ago when I started working converting my motorcycle to EFI on Megaqsuirt...good times...and now 20 years later
when I have to hook up an external wideband on brand spanking new 2022 Dodge or GM makes me more cringe than smile...

Merry Christmas!


P.S...this is how the VE(not using NN) looks like on 6.4 twin turbo Hemi we just did....trims under 5%...runs like stock.

image.png

P.S/2
Wideband always reads leaner at low load.

[SiriusC1024]

Haha. I used to know a guy who believed in flat earth AND hollow earth. Like how's that even supposed to work? Then there's that flat earther guy who built and piloted a steam powered rocket to prove NASA was lying with their photographs showing curvature. Steam powered rocket. Probably tuned it OL with a wideband, but we'll never know because the chute failed to deploy...

[SiriusC1024]

HPT Core ECU same thing. Closed Loop with the widebands.

core ecu.png

[TheMechanic]

You may or may not know, but I'll explain. CAN Low isn't a ground wire. CANBUS signal is held at 2.5V to show 0. For 1, CAN Hi goes to 5V and CAN Lo goes to 0V simultaneously. About the 0V, right? CANBUS is offset-tolerant by design. That means 0.026 at the ECU and 0.103V at the wideband controller doesn't affect the integrity of the data. CAN Lo (-) doesn't bond the various devices and make them equi-potential.

CANBUS reference is a floating ground that is capacitively coupled to ground. This filters out noise. For DC a capacitor is an open circuit, but for AC it's a reactive element that acts like a resistor with impedance of 1/jwC. At the end of each branch of a CANBUS circuit is a 120ohm terminating resistor between CAN Hi and Lo.

Imagine if CAN Lo was given ground. This would signal a 1 all the time on Lo, but Hi would keep doing its thing. A fault would be indicated because no communication on any CAN device on that circuit would cease. Hi and Lo have to mirror each other.

Another way to look at it is what if the ground wire from battery to the wideband controller got cut? If CAN Lo was really a ground, then the wideband sensor would be grounded through the tiny CAN wires and burn them along with potentially damaging the chips in every node between there and ground.

Been disconnected from the internet for Christmas and family time. I always just looked at CAN (as I first learned it in GM school in Burbank in 2005 or so) was nothing more than a communication platform. Not an analog signal but a digital communication. 0 and 1 stuff. Much like the difference between a record and a CD. Much less likely for interference to screw with its intended outcome. Now ground coupling I see as we are using an analog voltage signal that is of a very low and easily messed up signal through induction of wires running parallel to a signal.

[SiriusC1024]

Been disconnected from the internet for Christmas and family time. I always just looked at CAN (as I first learned it in GM school in Burbank in 2005 or so) was nothing more than a communication platform. Not an analog signal but a digital communication. 0 and 1 stuff. Much like the difference between a record and a CD. Much less likely for interference to screw with its intended outcome. Now ground coupling I see as we are using an analog voltage signal that is of a very low and easily messed up signal through induction of wires running parallel to a signal.

That's a reasonable thought. Thing is between classic serial and CAN wideband this doesn't apply. Look at the schematics. The only place where signal interference can occur is between the controller and the wideband. That circuit is identical between the two being that LSU4.9 is used. The difference is the data output. Both serial and CANBUS are digital. There is no analog offset to worry about.

aem wb type.png

What's interesting is if you look at the first link at Post #1 here, OP is using a digital output wideband output. It's an AEM 03-0300, and it's not matching up well with the STFT's. The spread is nearly 10% down low.

The second link has a raw wideband voltage configuration measured through EGR. It's an actual analog signal being measure, yet there is no offset correction needed.

Sensor type and controller logic play as much of a role as installation location. The point is that wideband needs to be verified against narrowband fuel trims. Having CANBUS output doesn't preclude a user from needing to do this. Analog to digital conversion doesn't filter error from the wideband.

[horsepowerguru427]

That's a reasonable thought. Thing is between classic serial and CAN wideband this doesn't apply. Look at the schematics. The only place where signal interference can occur is between the controller and the wideband. That circuit is identical between the two being that LSU4.9 is used. The difference is the data output. Both serial and CANBUS are digital. There is no analog offset to worry about.

aem wb type.png

What's interesting is if you look at the first link at Post #1 here, OP is using a digital output wideband output. It's an AEM 03-0300, and it's not matching up well with the STFT's. The spread is nearly 10% down low.

The second link has a raw wideband voltage configuration measured through EGR. It's an actual analog signal being measure, yet there is no offset correction needed.

Sensor type and controller logic play as much of a role as installation location. The point is that wideband needs to be verified against narrowband fuel trims. Having CANBUS output doesn't preclude a user from needing to do this. Analog to digital conversion doesn't filter error from the wideband.

So, if there is no offset to match up the WB reading with the NB reading then youre suggesting altering NB switch points to match up with the WB? Im a little lost on this thread lol Lots of counter points from several people...

[SiriusC1024]

No. If they match up, and the narrowband O2's are doing their normal full range switching, then nothing to adjust.

[TheMechanic]

Kinda makes me wonder if the reversion from overlap on a big cam effects the pumping action of a WB that is not effected by a regular zirconium/platinum sensor. I guess as I said before a 5 gas should answer the question. Maybe a new set of offsets for the wideband would make it more accurate or maybe the engine itself just likes a different afr. I don't have a 5 gas to test this theory any more. Was going to get a Vetronics many years ago when CARB mandated DSO's and a requirement of license retention