I've noticed in my datalogs that my AFR at WOT might vary +/- 5% in a given pull, excluding a few outliers. I feel like this is higher than it should be: at a target lambda of 0.85, that's 0.81 to 0.89. Assuming stoich of 14.73, that would be a target of 12.5 and ranging from 11.9 to 13.1.

I've known for awhile that my RPM signal is a bit noisy in my datalogs. However, I've recently connected the dots that this might be related to my AFR issue. I've observed multiple instances of datalog RPM remaining the same or even decreasing from one point to the next in a WOT pull, with injector pulsewidth following this trend. For example, I see two adjacent cells in a WOT pull where MPH went from 68 to 70, but RPM actually decreased from 5700 to 5650. In another, MPH goes from 66 to 67, but both cells report 5250 RPM and the same injector PW.

Are these AFR fluctuations normal? I've had to replace my tach filter before - did I possibly get a bad unit?

 

Well whats your ve table look like? Is it smooth? Are the pe mode enrichments also smooth?

 

Yes to both. To clarify, it's not just that AFR is jumping around. It's going rich and lean relative to what it should be; I watch for this by taking the ratio of WB AFR to target AFR, and that ratio might vary from 0.95 to 1.05 (+/- 5%).

 

Well as rpm goes through peak torque then up to peak hp and past, cyl pressure changes and fuel requirements change since air mass changes. as you know. So it could be afr is changing thru some points of rich lean because you may need more or less fuel in areas thru the rpm curve. Hard to say but most widebands i have used and seen do not vary that much at wot. 11.9 to 13.1 is a wide variance.

could be log signal. Possibly position of sensor in the exhaust. Is in it collector where it might only see 2 cylinders more so than entire average? Further downstream of collectors to allow more time to mix streams and average out is ideal. Theres potential that the sensor may read pockets of gas from some cylinders which can be rich or leaner due to air distribution issues

its not uncommon for a cylinder to sway rich and lean throughout a pull since induction dynamics arent always even, and headers arent always even and may “tune” to different rpm ranges. That requires 8 o2’s and sequential efi

 

Yeah, of course. I've done a ton of logging and tuning though, and I think I have my VE curve reasonably dialed-in. I've probably gone through 20 revisions of the WOT VE curve since installing the WB, with recent datalogs showing average WB readings right around the target.

I just re-ran my MAT/CTS inverse term script focusing on recent WOT runs, and I see that table seems to need some tweaking. That should get me more consistent, but I'm still concerned about the error in my RPM readings - fueling is never going to be accurate if the ECM is wrong about engine speed.

Interesting thought on the wideband location. It's in the Y pipe right after the flange that bolts to the collector. Off the top of my head, maybe 4" after the primaries merge. Do you think it would be better to move it to the base of the Y pipe right before the cat where it's getting all 8 cylinders? I thought you wanted O2s as close to the head as possible to keep exhaust gases hot, but widebands are heated so maybe that's not an issue?

 

I like it in the single exhaust so all 8 are read so if any one cyl goes out of wack you’ll see a change. But I understand this isnt always feasible or convenient at times.

if the collector flange isnt leaking then it should be ok. There some potential for turbulence around the sensor and getting mixed pockets of gas but you are right the rpm trace is still somewhat important. I too have seen somewhat erratic rpm traces in tunerpro on some of my older ecms but never really investigated it. Figured it was just signal issues and processing speeds of the tech. But my widebands didn’t fluctuate at wot

 

The collector flange is a tiny bit split (think el cheapo exhaust shop overexpanded the tube when they made the Y pipe), but it doesn't seem to be leaking. Of course, I can't really lean my head out the side of the car with it moving down the road, so it could leak under load but not at idle. I might have it replaced just for peace of mind, and I can get a bung added at the base of the Y when I do that.

Good to know that it's not just my ECM glitching with the RPM readings. I'm curious if anyone has found a way around this. It would sure make all my datalog analysis scripts work better. Maybe once I get ALDL mode 3 working (higher-frequency report of fewer values) I'll have some more insight into what's going on.

 

This erratic RPM can be an issue. I've had distributors in a lathe connected to an ECM on the bench, and the RPM values from the ECM were dead on solid.

The problem is that once the distributor is in an engine, there is just too many other motions involved. Crank shaft through a chain & sprockets to the cam, then cam twist, plus the oil pump hanging off the bottom of the distributor. It is a real mess.

Making sure the distributor is in good shape is helpful. That is the magnet on the star wheel is solid, no cracks or pieces missing. No extraneous bits of metal or magnet stuck onto it. Shaft bushings with no play. It can be helpful to shim the distributor shaft end play to be about 14 thou (.014") cold.

The ECM calculates the engine RPM by measuring the time between distributor reference pulse (DRPs). Some simple math and out comes the RPM. This hardware counter runs at 64 KHz, so has a resolution of 15.26 micro-seconds.

RBob.

 

Timing chain play, and spark scatter from the oil pump gear pulsations can mess with the signal. I went with the Melling Shark Tooth pump for that very reason.

You should ensure the mechanical condition of the components is capable of generating a solid signal before you condemn the electronics.

GD

 

Interesting. The timing chain is a relatively new double roller, so that's probably not an issue. However, my cam gear is a bit chewed up - Lunati said the distributor gear I had was compatible, but it definitely wasn't. I replaced it with a new gear, making sure to shim the gear for endplay, and I shimmed the distributor itself to get it relatively centered on the cam gear. When I get a chance, I'll pop off my distributor cap to make sure all looks good under there.

Any reason why the DRPs wouldn't correspond to the reported RPM? Looking at a log, I have:
RPM | Time between DRPs | Vehicle speed | 60/(4*10^-6 * DRP)
5375 | 2792.58 | 68 | 5371
5950 | 2624.72 | 70 | 5715
5675 | 2594.20 | 71 | 5782

The RPM calculated by DRP is still a little jumpy, but at least it proceeds in the right direction (increasing from point to point), whereas the RPM changes non-monotonically (increase then decrease).

 

The ECM RPM is in increments of 25 RPM. This is from 0 RPM through 6375 RPM , with an 8-bit value representing RPM.

There are two reasons for the mismatch in the ECM versus DRP RPMs.

The ALDL stream is picking up the RPM and DRP period at two different times. A spark plug could have fired between them which has the ECM using a different DRP period for the RPM calculation. The way the Time Between DRPs is being represented means that there is math being done on it within the data logger.

So there can also be rounding errors. Using the actual DRP count to RPM is easy: RPM = 983040 / DRP count

RBob.