Let me preface this thread with the setup.

Setup: '90 miata, bp4w +kraken +GTX2860 + id1000 injectors on a brain built trubokitty ms3x. Fully sequential fuel and spark. Walbro 190 fuel pump on stock FPR. IAT sensor is mounted on cold side of IC so it doesn't "heat soak". Using incorporate AFR for VE, MS 1.4 firmware (yeah, pretty old).

Issue: Basically put there's a discrepancy in fueling between when the car first reaches operating temp (I.E. when it first goes from 179 - 180*) vs when it's thoroughly heatsoaked. And when I say "heatsoak" I mean underhood temps, not the sensors. My IAT sensor is mounted on the cold-side of the I/C and never really heat soaks in that regard. And when I say discrepancy, I mean like 10-15% fuel discrepancy across the board.

The big issue here is that the car is able to maintain AFR target until goes out of WUE at 180 (thermostat), at which point it goes lean until thoroughly heatsoaked, then it's on target again. We tuned it while very heatsoaked/flogged, as to simulate closer to track conditions. AFRs are super consistent and dialed when hot, but there's this "post warmup but pre heatsoaked" area that we just can't get to obey. EGO of course is a band-aid here, but something seems fucky.

Here's an example from a recent drive, only one example but you can trust me (or at least @redursidae ) that this is consistent.

JUST had hit 180* CLT maybe a minute or two before, so "warmed up" but not "heatsoaked". 2683RPM, 75kpa, 27%tps, 92* MAT. Way lean off target.




Another example, after letting the car "fully warm up", I.E. a bit more driving. Maybe parked it to get coffee and got back in it. Here's 186* CLT, 2631RPM, 75kpa, 26%tps, 96* MAT. Right on target, err well, close enough.



Looking through some older logs this has always been an issue on my car for some reason, though the issue sees worse on sequential fuel, especially since we were able to lean it out a lot in cruise.

I don't have an eye on fuel pressure, which is a concern of mine, maybe something is fluctuating there. But it's so damn consistent and temperature related it seems like maybe there's something else we're missing. Just looking for opinions/thoughts/etc.

I do have a coolant re-route, with CLT sensor in the stock place, but a remote mounted thermostat. It's about 1/2 way down the long coolant re-route pipe from the back of the head. I've speculated that this could be part of the issue, but not super sure how since the CLT is in the same place and on the correct side of the thermostat.

Attached are some more logs and .msq. The commutetowork log is a cold start and commute, the other log is fully heatsoaked where the issue isn't present. Anyways just looking for ideas!

 

A few more details to help diagnosis:

- AFR gauge and tuner studio agree within 0.1 AFR.
- Checked and fixed grounds in the engine bay.
- Coils are grounded separately from sensors.
- No vacuum leaks around the manifold, throttle body, turbo or wastegate.
- Found and fixed issues with the throttle body, but made no difference (it was sticking).
- VE3 is within 1-2% when the engine is heatsoaked. I made VE1 the same for now as we haven't ran 91 through it.

I'm hoping I've missed something, because I don't know what else to look for. The way it changes from a minute to another is very odd to me, and backwards to what I would expect.

 

Are you both users Fireindc and redursidae?

My understanding is that the condition of different fueling requirements between coolant reaching temp 180F and underhood heatsoaked exists. When underhood heatsoaked, AFR's get lean. I tuned fuel for when car is heatsoaked, so my AFR's are typically rich before that happens. A recent post pointed towards fuel temperature and fuel pressure, suggesting a vacuum referenced fuel pressure regulator solves this issue.

I think you are describing something else. I haven't looked at the attached tune and logs, but based on the screenshots I wonder if you are focused on transient AFR's during application of throttle. I see when throttle is applied, there is a momentary lean spike above target (not heatsoaked) versus momentary lean spike to target (when heatsoaked). It my understanding that short lean/rich spikes in transient AFR's are no big deal, but if you wanted to go down that rabbit hole its AE related. I think this represents the condition that hotter engine bay leads to more fuel atomization, less adhere-to-wall. In your case maybe increase Cold AE % multiplier, and look at adhere-to-wall, suck-from-wall, and CLT correction.

 

I’ve seen this on link tunes too. Maybe not 10%, closer to 5. Seems to be closely related to fuel temp, tracked with an ethanol sensor. Above ~90-100 degrees, I find it’ll need more. Maybe batch fire was firing injectors more often and keeping them cool?

 

First, thanks for the replies guys. Really just want this car to run right, and this ones pretty annoying.

I'm the owner of the car in question, and @redursidae is my friend and tuner of the car - hence his chiming in.

I agree about transitions in general not being a good example, though that screenshot was pretty light throttle just rolling through the VE table without any AE or rapid transitions. Generally I'd agree with what you said, most cars need more fuel when heatsoaked, not less. This car when thoroughly hot and thrashed needs less fuel than when it first hits operating temp, which is what's so perplexing.

The problem with any setting that you tweak when "cold" is that it's not cold according to the sensors. It's up to temp so enrichments are off and AE is using the operating temp settings.

We thought of that as well, but it's not that unfortunately. Here's a log snippet that shows less than 1* fuel temp variation, I'm cruising full lean after first getting up to temp, then after a couple of gears it goes to normal. The only correlation Ricardo pointed out is there is that the thermostat seemed to open at 185F, and then temps dropped to 179, which is an effect I have noticed with my 180* thermostat in-line about 16" down the coolant re-route line off the back of the head. It does swings like that for a while before it stabilizes. I've considered a 190* tstat but not fully convinced it will fix the issue. It's easy enough to try so I probably will.

Wayy off target, just after coming up to temp:



Roughly 60 seconds later in the same log:



Of course we're going to tune around this the best we can come up with, but something does seem odd here.

 

EGO isn't a bandaid, it is how OEMs accomplish tight fuel control. FW 1.5 and above allow you to distinguish between adding fuel via ego (good) and pulling fuel via ego (bad).

Watch pw, the screen shots above actually have good visibility here, same pw different AFR.

Most likely scenarios are delivering less fuel than intended or change in burn. I don't open logs anymore, but fuel pressure and battery voltage are key parameters to monitor.

Edit: third scenario - change in AFR measurement accuracy

Honestly, unless you have variability in voltage or pressure, my recommended solution is to update to 1.5+ (careful if on mslabs), give max EGO add authority while limiting EGO pull authority, and then tune EGO well.

 

I'm watching this with interest because I think I'm experiencing a similar thing going on.

I haven't tracked my car this year yet - I've been tracking down and "killing" a bunch of problems that needed dealing with before subjecting the car to that kind of abuse. BUT with the high ambient temps in the area lately I've been experiencing a situation where the AFR numbers go more lean as the temps (IAT and/or CLT) rise. I have my CLT "mystery" mostly solved, so I think what I'm seeing is IAT-dependent. All of this led me to try to tune this out using the MAT Air Density Table (which may, or may not be the right place to go - but this is the only place that I know of where fueling is temperature dependent).

But here is the problem; when I tune the downward-sloping density correction to make the hot IAT condition have reasonable AFR, the colder condition is too rich. But when I tune the density correction to make the cold IAT condition "right", the hot AFRs are too lean. The tipping point seem to be around 80-85*. Below that, it wants low-ish correction numbers. Above that, it wants high-ish correction numbers. I'm not talking about massive changes, it's the difference between 100 (low) and 104 (high), but it makes all the difference.

I don't know if this has any correlation with what @Fireindc is experiencing (and excuse me for butting in), but I thought I'd put this into the MT-universe before starting a different thread.

 

Just to clarify, the OP's condition is AFR get rich as underhood heatsoaked.
The more commonly discussed condition is AFR get lean as underhood heatsoaked.

Can't successfully tune it out until identify exactly what is being affected by heat and throwing a sensor on it since its not CLT.

 

First, thanks for everyone's input. Second, a big breakthrough has happened and I'll report back but things are looking promising. The CLT sensor was somehow miscalibrated. I know at some point I had calibrated it, but there's noone to blame about this but myself. IDK what calibration it was using but @redursidae pointed out a 10* difference in CLT vis IAT in one of his logs when the car was cold and not yet started. Seemed minor enough but he pointed out that 10* is exactly what we're chasing here.

I assumed a grounding offset or issue, but first step was to re-calibrate. I got into my garage yesterday after work, used my IR gun and tested CLT, block, and the adapter where the CLT sensor lives. All 3 showed a range of 74-78*. Then I turned the key and looked at the readings, the IAT said 73* but CLT was saying a whopping 88*. Nothing is that warm, the high of the day was 80* and by the time I got home it was maybe 73* so there's no way that was right. So I went into the calibration menu and told it s4/s5 rx7 sensor (which is what I've always set these to use, even on my ms1) and boom, 74* CLT. That means we had a 14* difference in actual CLT vs sensor CLT when cold.

Today I drove to work and things are looking way better. It seems the high end of the temp sensor was close enough, because the 180-185* temps seemed accurate when flogged and hot with my 180 thermo. However, with the sensor re-calibrated I can already see more clearly the car getting up to temp, coming back down, up to temp, etc. as the (admittedly problematic) remote t-stat opens and closes.

At this time I think I have an issue that was mostly the sensor, but also I do believe the remote T stat pulling the car in and out of warmup is a bigger problem than I had thought. On this mornings drive I saw it hit 182* pretty quick, drop CLT to 165* when the t-stat opened, rise up to 185*, drop to 172, rise to 185, etc. By the time I got to work I was sitting closer to a 172* CLT and not actually warmed up, where the old sensor calibration definitely though this was up to temp.

I will be updating this thread with more data once we get things ironed out, but I do think the slow-acting thermostat is problematic even with the sensor working correctly. And also my cooling stack overcools and I should be on a 195* t-stat.

Thanks again folks. And again I've got no clue what calibration it was using, it could have been non-linear or completely different for all I know. It was just close enough to not throw any red flags until closer scrutinized, and I'd say this thread motivated both of us to keep digging, so indeed thanks again.

 

As much as I thought/hoped this was resolved, it was not. We're now on MS 1.6.1, with a SM coolant reroute and proper/confirmed sensor calibration (no more weird warmup behavior and CLT yo-yo's). We still see this happen, and it's almost as if something "switches" and all the sudden fueling is correct:

Right here is a good example. Car is fully warmed up before even starting the drive in this log, this log is about ~7 minutes into a drive on a warm 80+* day. Right here you can see EGO go from a mostly lean/adding fuel trend, to a on target/pulling fuel trend.




This zoomed out view of the same log gives you a better idea of what I mean. Sorry about the screenshot, I moved ego correction plot to its own frame for clarity.



Summary of things I've confirmed:

• Grounds are all in order. Sensor grounds are grounded at the ECU. The stock NA "grounding loop" issue where sensors and high voltage are grounded at the same spot was "fixed" by cutting the black/green ECU wire.
• All sensors are in working order, tested, calibrated, and confirmed good.
• All logs show no unusual noise in any of the readings.
• Injector data double, triple checked, and even updated to the new MS1.6 way of handling PW's.
• Firmware on ECU up to date.
• Cooling system remote thermostat funkiness resolved.

Things on my list to confirm:

• Fuel pressure (sensor on order)
• WBo2 gauge, will try moving the ground on existing setup, but also now that I'm on the newer firmware we're going to try the spartan CAN controller.

Fuel pressure is my top theory, though it doesn't explain why this issue only shows in cruise. In boost, even when this issue is happening, AFR's come to target with minimal EGO intervention. This makes it feel like a small pulsewidth issue. One thought was heatsoaked injectors, but my rabbit hole into that idea revelaed that head-soaked injectors do the opposite, and "slow down" so you end up with more fuel, not less.

I think we're running out of ideas, but going to keep looking into this, kinda feeling determined to get to the bottom of it. There's gonna be something we're missing here.

 

Not sure what you mean by heat soaking slowing down and increasing fuel, all of the heatsoak injector situations I have dealt with are an increase in dead time, effectively reducing the fuel delivered for a given pw, resulting in a lean condition.

edit: battery voltage?

 

He typed it out backwards. You're correct and we agree about heatsoaked injectors having longer deadtimes which is why this issue doesn't make sense. The fuel table was tuned on a heatsoaked engine on a warm day. I would expect it to then be richer during this first start/run of the day with the WUE, but it has this seemingly "switch" behavior where it goes from lean to normal per Nate's screenshot.

Battery voltage is a good point, but it is stable. On a cold start it is ~14.2V, then 14.0V by 180F, and sometimes dips to 13.8V on a very heatsoaked idle. Unless the voltage at the fuel pump is changing compared to what the ECU is reporting, which is possible and I hope the fuel pressure sensor will shed light on that.

 

Thanks for keeping me honest. What I meant was we tuned for the heatsoaked injectors in their "slowed down" state, and if anything I'd expect it to be rich when not heatsoaked. We're seeing the opposite.

And voltage is perfect. No fluctuation when this issue happens.

 

Probably appropriate to confirm that injector pw is in fact greater when in "lean" state than when in "rich" state at otherwise equivalent operating conditions.

Also check deadtime with the same criteria

 

We forgot to respond and provide closure to this. The issue is resolved. TLDR: it was the wideband controller.

Ted, we verified that the PW and deadtimes were nearly identical and so were the conditions. It was a good suggestion that lead us down to the solution. I had been suspecting the wideband controller, and your comment gave me reassurance that something was messing with the pulsewidth MS3 was commanding. We replaced the old AEM EGO LSU 4.2 (!!) the car had for my old Spartan 3 w/ LSU 4.9. My guess is that the LSU 4.2s heater was sinking enough current into the circuit to mess with the pulsewidth, or something else along those lines. After switching the controller this discrepancy disappeared and the tune behaves as one would expect, richer during the first heat cycle.

 

Yea the car is perfect now and works as expected. Thanks to @redursidae for pressuring me to change out the WBo2 and controller, I had serious doubts that would do the trick. But sure enough, here we are. Thanks for concluding this thread for me Ricardo!

FWIW, I also installed a fuel pressure sensor at the same time and fuel pressure is consistent and working exactly as expected.

 

Nice

 

My car is absolutely the same for years, but it's also on a shitty Innovate MTX-L Plus, which likes to burn sensors every other year, so maybe I should change to an X-series or something.

Thank you for all your work!