[rusEfi] Miata NA 1990, 1.6L stock #21

[Thommm]

How did you trigger this? I see a big difference in injector dutycycle if you compare before the miss and after, how did you get this scenario?

[AndreyB]

The way you experience this only on acceleration and the way this picture looks like really makes me wonder if an even wider pulse was supposed to be there with maybe above 100% duty cycle or smth like that.

I can reproduce this not often at 2900 RPM and more often at 4500 RPM by just touching ADC pins, probably not the most fair way of testing.
Still working on finding the steps to reproduce with potentiometers which are a more honest way to do it.

[tomiata]

How did you trigger this? I see a big difference in injector duty cycle if you compare before the miss and after, how did you get this scenario?

From driving I could tell there were problems when accelerating from a stop, shifting through gears, reving up to 5000 or 6000 rpm, and also when accelerating from 60mph to 70mph in 5th gear, 3000rpm to 4000ish. I had my son manually capture 10 second traces when I said "now".

I did not try to correlate each trace with a specific section of road. I just grabbed a bunch of samples and looked to see anything unusual. The logic analyzer software I have is fairly primitive. I just found an open source project sigrok that might be better. I'll look into it more.

[AndreyB]

Tom, can I bother you to try 20160903? My Miata is at a shop for transmission work.

I've fixed two fuel miss scenarios pretty sure, I also know that fuel miss is still possible - but the amount of changes if worth testing.
Also there was a small fix which maybe would resolve the spark miss but probably not. Not much progress in understanding spark miss so far.

[AndreyB]

I am afraid we have at least two problems here
1) fuel misses due to defects in scheduling algorithm in case of injection duty cycle crossing magic threshold values: transition from below 50% to above 50% is one of these, transition from below 75% to above 75% etc. With this issue looks like there is an improvement: the older logs had 262 warnings detected over 900 seconds and on the newer log only 3 warnings over 180 seconds. I am aware of a failing use-case still at the transition from above 50% duty to below 50% but should be less annoying since in that we are rich instead of lean. Good news is there is progress and fixing the algorithm is within my control

2) general acceleration fuel calculation. Note the spikes and falls in AFR while sudden acceleration. I am suspecting we are still not aware of the best magic settings to control the three acceleration strategies we have in the code, and bad news is I have no idea how to approach this.

[stefanst]

In order to troubleshoot the fuel accel, it would be helpful if you could post a graph with AFR-AFRtarget (alternatively AFR and AFRtarget), MAPdot, TPSdot and rpm.

Also only tune accel enrichment on a fully warmed up engine.

[stefanst]

Oooops. Didn't realize you posted the logs. Will take a look.

[stefanst]

Looks like you get consistent lean spikes on quick upshifts. Everything else looks pretty spot-on or close to it. Post the tune please!

[tomiata]

Post the tune please!

Here is my tune. I started with an old one from @thommm and have tried to tweak and adjust using TS Tune analyze.
Any suggestions on improvements would be appreciated.

[stefanst]

A few things:
- It appears you are not logging all values. In order to troubleshoot, I think we need to look at 'fuel: tps delta', 'fuel: tpsAccel'. I would also like to see the values for the engine load based accel calculation, but it appears they are not available for logging.

- Your signals are noisy- especially your temperatures (MAT and CLT), but also your TPS signal and your AFR. I would guess this is caused by a poor ground. The noise on these signals correlates very well. A noisy TPS signal will make the thresholding for TPS accel much trickier. Interestingly your MAF signal looks pretty clean. Not sure if that's running off a different ground or something.

- Have you tried a larger adder for load-based accel? Looks like you're running an engine load multiplier of '2'. I'd up that considerable- maybe go for 10 or even 20 and see what that does.

@software developers: Am I missing something, or is it true that laod-based accel calculations can't be logged?

[tomiata]

What is needed to get the other data values logged? Is that an ini change?

My frankenso board is in a case, but is open and not bolted down. I'll try grounding the case and look at the signals more. I have the stock AFM with temp sensor in place for MAT. I guess the MAF is recording AFM signal, but I'm running with MAP and speed density.

I have not done anything with load-based accel settings? I'll try that. Thanks for the tips.

[AndreyB]

What is needed to get the other data values logged? Is that an ini change?

That's on me - I need to go over the values we have an make sure all are going into native logs.

At the moment we have three ways of logging - one is in C code (firmware and SD card logging), another is in java code (console) and the 3rd one in .ini file (TS logging)

Will post an update in an hour.

[AndreyB]

Just commited console version 20160909 (two new fields into console logs) and firmware 20160909 (8 fuel related values into SD card logs), hopefully a bit better consistency between these two. I should have really improved this a while ago

Current description of acceleration logic is at http://rusefi.com/wiki/index.php?title=Manual:Software:Fuel_Control#Acceleration_Enrichment

[tomiata]

I tried closing up the frankenso case and grounding it to see if signals get any cleaner.
And I drove for a longer test run and attached the logs.

CLT and MAT values still look pretty noisy. But looking at vbatt, it looks like its noise or modulation is getting amplified in the values that CLT and MAT show.
I don't have experience looking at car analog signals, but is that normal at all?

log9-9.jpg (420.63 KiB) Viewed 16943 times

Cruising at 60mph is fairly smooth. Accelerating still has some hesitations around 3500 to 4500 rpm.
@russian, let me know if you want to see another test run with the logic analyzer looking at ignitor and injector signals.

I still need to figure out the MAP values showing more than 100kpa. I have an MPX4250 and have it set that way in TS. The log shows MAP going up to 140kpa and I don't have a turbo (yet). Do I need to set up a custom curve for it?

[AndreyB]

And your tune does have X-average enabled for slow sensors like CLT and IAT

<constant digits="3" name="slowAdcAlpha" units="coeff">0.3333301544189453</constant>

Without X-average it would probably be worse.

Let me fix fuel miss scenario #3 I am aware of tomorrow, for fuel miss troubleshooting yes logic analyzer would be needed. To make it easier I'd say there is no need for sudden acceleration testing while I am still fighting fuel misses even with relatively smooth acceleration/deceleration.

I am not the right person to comment on analog noises...

[kb1gtt]

I see the CLT temperature started reasonably noise free, then as it warmed up it got noisy. Also I see your battery voltage is jumping around, largely based on RPM. However I see the CLT noise does not appear to be affiliated with RPM or battery voltage. In the below at 681S, it's min is about 82C and max is about 105C. I see it starts noise free, then gets noisy as it warms up. Could that be a thermostat opening and closing? I'm not sure if this is electrical noise or mechanical noise. I suspect there is a certain amount of electrical noise. Perhaps we could average this over something like 3 seconds to 5 seconds to decrease the noise. Inside the cyl does not change that quickly, so even if this is caused by the thermostat, what rusEFI cares about is the cyl temp, not the water temp. So smoothing it out would be OK. This might be corrected either by replacing the thermostat, or perhaps by adding a software filter.

What is tCharge? I see it looks very similar to the noise on the CLT.

About the MAT, that seem to be affiliated with engine RPM and some other more general shorter term noise. The shorter term noise seem to be about +/- 2C and at a frequency of about 1 second. This appears to be general noise and perhaps it could be filtered via software. Again perhaps a 3 second to 5 second filter would smooth that out. I suspect it would be hard to filter out physically, as it could be sourced from many things like electrical coupling, or short term air flow around the engine block and into the intake. I'm not sure if this is real world noise of erroneous noise. Is there a way to measure if this is simply the real world temperature vs electrical noise?

I'd be more concerned about the 6C rise based on RPM rise. This seems to be erroneous, as I would expect increased air flow at the intake would decrease the temperature not raise the temperature. So I'm suspicious this is electrical coupling. At the end of the day does a 6C rise cause significant differences in fuel delivery? I don't think this Miata is being tuned that close to get every HP out of it, but if you we're pushing the limits this noise would want to be decreased. For now I suspect this noise is not a real issue.

I see at the beginning it's about 22C, then goes up to 27C at the end. Could that be how hot the engine bay is getting?

[AndreyB]

What is tCharge? I see it looks very similar to the noise on the CLT.

http://rusefi.com/wiki/index.php?title=Manual:Software:Fuel_Control#Speed_Density
tCharge = charge_temp(COOLANT_TEMPERATURE, INTAKE_AIR_TEMP, TPS) assumed (air) charge temperature

[AndreyB]

If we want to investigate analog noise we should probably start by car not running log. Attached is mine with the same Exponential Average = 0.3 with 2 minutes of data. VBatt range 12.79-13.01, clt 22.07-22.15C, iat 21.78-21.97C (wow, CLT is pretty close to IAT on a cold engine )

[stefanst]

If you look at the noise on the '90 Miata, you can see the short term order of noise on the MAT and CLT running exactly in parallel. The noise on the TPS is exactly opposite to that. So it's clearly something with the supply. So maybe more low-pass on the 5V side might help here.

@russian: During that log, is anything that is using PWM running? The Vbatt noise looks like it's caused by a PWM somewhere....

[kb1gtt]

If we are looking to resolve noise via software, lets not forget this software filer approach.
http://rusefi.com/forum/viewtopic.php?f=3&t=696&start=102

[AndreyB]

I've made yet another fuel miss fix recently but have not yet had a chance to try it, any chance you can compare the before and current firmware versions?

[tomiata]

Ok, I've been driving this car as my daily on the stock ecu for a couple of weeks.
So I switched out to rusefi to what I previously had from about Sept. 6 firmware, and attached the log from that below, rus279.msl.

After updating to 10/8 firmware I cannot start, and I get an unexpected MAP value error.
Attaching console logs from that...

[AndreyB]

MAP,9.75
vbatt,12.26,IAT,27.62,MAP,9.93

that's unrealistic MAP value especially for car not running

is MAP powered/connected etc?

Just started my car with latest version but it's a sad rain here

[AndreyB]

Can you please post a picture of your MAP sensor and hose? Wonder how you've mounted it and where you do you take vacuum.

[tomiata]

Can you please post a picture of your MAP sensor and hose? Wonder how you've mounted it and where you do you take vacuum.

The set up works with the previous Sept. 6 firmware. See logs posted above.

MAP sensor is installed on the frankenso board and hose routed through firewall to a spare port at the throttle body. I'll add a photo later, out of town right now.

[AndreyB]

I am mostly interested which port on the TB you've used since there are two and it seems like these work a bit differently. Mostly curios, do not imply that there is a problem here.

Oh, right. So same hardware setup works with older firmware and does not work with current firmware. Can you please post .msl file from the console out folder for the failed attempt with latest firmware? Want to compare a working log against a failing log. Regardless, MAP below 10 kPa is not realistic I believe but that could be not related to this issue.

[stefanst]

[...
MAP sensor is installed on the frankenso board and hose routed through firewall to a spare port at the throttle body. I'll add a photo later, out of town right now.

I'm not intimately familiar with the 1.6l intake and TB, but I had some trouble with noisy MAP when using a spare port towards the front of the IM. Sourcing the MAP signal from the rear of the IM (actually spliced into the brake-booster line) seems to work best for me.

[AndreyB]

(actually spliced into the brake-booster line) seems to work best for me.

how does one hack into the brake-booster? Is there a known mcmaster part number for the right kind of a T-junction?

[stefanst]

On the NB it's just a hose routed along the firewall. Cut the hose, put in a T (Any autoparts store has them). It's fairly large ID, at least 6mm, so you may need an adapter to reduce the size to whatever your MAP sensor needs.

[tomiata]

[...
MAP sensor is installed on the frankenso board and hose routed through firewall to a spare port at the throttle body. I'll add a photo later, out of town right now.

I'm not intimately familiar with the 1.6l intake and TB, but I had some trouble with noisy MAP when using a spare port towards the front of the IM. Sourcing the MAP signal from the rear of the IM (actually spliced into the brake-booster line) seems to work best for me.

See step 12 from this megasquirt install guide: http://www.megasquirtpnp.com/docs/mspnp_g2_mm9093.php

I'm doing basically the same thing.

90miata-map.jpg (249.02 KiB) Viewed 16747 times

Also attaching msl file from fail to start config with 10/8 firmware.