Slight misfire after throttle is pressed

[racingmini_mtl]

This is just to let everyone know that I have been able to reproduce this on the bench (thanks to James) and that I'm tracking the problem.

Jean

[gslender]

While you are looking at the code, I found this odd section that could be a bug (not sure if related, but does involve seq inj).

Code snippet from 3.1.2 beta at lines 2413 in ms2_extra_main.c (and the same section exists in the 3.1.1 release)

Issue is that when tmp_pw3 and tmp_pw4 are individually set for EAE calcs, that the section below this ( ** Set pulse widths 3 and 4 and Add VE trim) essentially puts them back to just whatever tmp_pw1/2 are and makes the code section above irrelevant for inj 3/4.

So either the section for EAE calcs is inefficient and could reduce code size (and a code jump) or the section below is changing the values of 3/4 inj when it shouldn't....

Code: Select all

 
            run_EAE_calcs();
            if ((flash10.staged & 0x7) && !(seq_inj_ctrl & SEQ_STD_INJ)) {
                unsigned long ttmppw1, ttmppw2;
                ttmppw1 = tmp_pw1;
                ttmppw2 = tmp_pw2;
                tmp_pw1 = tmp_pw3;
                tmp_pw2 = tmp_pw4;
                run_EAE_calcs();
                tmp_pw3 = tmp_pw1;
                tmp_pw4 = tmp_pw2;
                tmp_pw1 = ttmppw1;
                tmp_pw2 = ttmppw2;
            }

            /**************************************************************************
             ** Set pulse widths 3 and 4 and Add VE trim
             **************************************************************************/
            if (flash8.seq_inj & 0x03) {
                if (no_inj > 2) {
                    tmp_pw3 = tmp_pw1;
                    tmp_pw4 = tmp_pw2;
                }
                if (pg8_ptr->seq_inj & 0x10)
                add_vetrim();
            }

I'll let the developers decide which is relevant etc.

G

[racingmini_mtl]

I don't want to go in details at this time but it is correct. When staging is enabled the number of injectors is not computed the same way so the code won't undo what it just did.

Jean

[racingmini_mtl]

I think I have resolved the issue. I have attached both the MS2 and Microsquirt s19 files so if you had the issue, give it a try. Hopefully, you won't see it anymore. By the way, this is based in the 3.1.2 version so you'll need that ini (attached)

Jean

[gslender]

I don't want to go in details at this time but it is correct. When staging is enabled the number of injectors is not computed the same way so the code won't undo what it just did.

Jean

Cool - a fact I didn't know is if flash8.seq_inj & 0x03 would be true when staging is enabled (but I guess it isn't), otherwise it would undo it.

Great to hear you've found the problem - are you able to share the code or segment with the bug/fix?

G

[racingmini_mtl]

The code will be in the next release assuming it is completely fixed.

Jean

[jagdude]

The code will be in the next release assuming it is completely fixed.

Jean

Jean,
Does the code you corrected affect only the sequential setups that the Miata installations use??

Jim

[Zaphod]

Thank you very much Jean (and James) for this... I am going to try ASAP.

I call back on how it works.

[Reverant]

Any possibility of a diff -u patch being posted? I want to try this ASAP but I'm running modified code (and don't want to lose the extra features even if temporarily).

Dimitris

[richyvrlimited]

Great job Jean.

My car is off the road so I cant test to verify, but great job nonetheless

[Zaphod]

Hi Jean and James,

I think I am not able to express how much we have to thank you...

I just come back from the first drive with the bugfixed code and I am stumped - you gave me back the joy of riding my Miata!
There are no more misfires, the car generally seems to run even smoother... Great, great, great!

Again, THANK YOU!

[richyvrlimited]

Good news Zaphod

Great work Jean & James. Thanks from all of us in the MX5/miata community!

[Reverant]

Excellent news! Waiting for the source!!!

[BenGTT]

Hello Jean,

What was exactly the issue? Is it only affecting Mx5/miata users or is it a "general" bug and it is worth to upgrade ?

Ben

[jsmcortina]

What was exactly the issue? Is it only affecting Mx5/miata users or is it a "general" bug and it is worth to upgrade ?

The situation that I observed and conveyed to Jean was that the injection start event was right at the change point between two teeth. Under certain circumstances a recalculation at just the wrong time could cause a missed or duplicated injection event. As the Miata wheel is so inaccurate with very few teeth this is quite feasible. On higher tooth count wheels it is still possible, but the incidence would be very low if at all.

Jean can no doubt explain in more detail, but I ran into these same kind of problems about a year ago on MS3 and fought the code for a while to resolve them.

James

[BenGTT]

Thank you James, this is a very good explanation and interesting

Ben

[racingmini_mtl]

Good to see that my bench test results are confirmed.

In addition to what James mentioned, there was also a problem with scheduling events just after the current tooth. In some cases, that allowed the timer to roll around which either pushed the injection pulse much later or increased the pulse width. The solution was to force injection for those cases and schedule the events only for those slightly further in time. This could also have affected spark events so this was also updated. However, since the sequential injection events are scheduled later in the interrupt, they were more likely to be affected.

There will be a beta release soon so the code will be available. And those using sequential injection with any wheel type might want to upgrade due to what I mention above.

Jean

[Reverant]

Is the bug fix in ms2_extra_inj.c only?

[gslender]

Just diff the entire msextra folder ..... .!?

[Reverant]

Exactly what I am trying to avoid.