MicroSquirt.com read only forum (Jan.15,2018)

Thanks Al. I suspected something like that with an offset of 8* and a base time of 8* in my ignition table.

Joffery, I was able to get a more steady idle using the default only if the predictor gain was down near 15 or so. I am not sure which is better, Last Interval with a high predictor gain or the defaul with a low predictor gain.

bleoh wrote:Thanks Al. I suspected something like that with an offset of 8* and a base time of 8* in my ignition table.

Joffery, I was able to get a more steady idle using the default only if the predictor gain was down near 15 or so. I am not sure which is better, Last Interval with a high predictor gain or the defaul with a low predictor gain.

I would try to get the default or at least first deriv prediction to work - there is no problem setting the predictor gain to 0 - this is a separate correction that seemed to do some good on a simulator but I'm not sure it is needed on a running car.

One other thing I should mention: if you haven't already, everyone should update to vsn 2.35 because there were some corrections made that could have an impact on these problems.

Sorry if this is a dumb question, but vsn 2.35 of what?? I see under the MS2 downloads there is still the 2.34 stuff. Under embedded code I see there is v3.00 and under MegaTune the 2.25 version is still the official release.

edit: I found a link to it in the manual. I guess the download section of the forum is a step behind. I am guessing the v3.00 embedded code is for MS1.

I will have to check on getting al the code versions synchronized, but I did check that v2.34 was the last one that had a change pertinent to the problems on this thread . So if you have 2.34 (MS 2 code version) then you won't be helped by going to 2.35, but help is definitely possible by upgrading if you have a code version lower than 2.34.

I have code version 2.3 so I'm gonna put in 2.35

About the predictor algorythms. Whats wrong with using "last interval" ? Is there somewhere that tells me all about the different algorhthems and what the predictor gain does?

If you are accelerating at several 1000 rpm/ sec and you use last interval your timing may lag behind what it should be. The algorithms are ways to predict in advance when the tach input will arrive. I would use at least first derivative prediction. The predictor gain is for a Kalman filter which is a separate predictor algorithm that, when a new tach input comes in, looks at the error from the last prediction and adds in a -%gain* error to the normal preditcion. The %gain is a user input. If it is set too high it may cause overshoot and oscillation. I would keep it below 20-30%.

So if I got it right.

The rpm drops are probably from missing input signals.
Or a wrong offset.


But you can 'solve' them with switching between the predictor algoritms.

But this isn't the real solution, the VR sensor must be adjusted, maybe the distance between sensor and teeth is not right.

Or too much clearance on the axe of the distributor

Or the isolation of the signal wire isn't that good.

Nick

Ye said: "The rpm drops are probably from missing input signals.
Or a wrong offset."

No, if I understand correctly there are no missing input signals. Using embedded code version 2.3 and predictor algorithm "1st high rpm, 2nd low" caused some of the output pulses to be dropped or moved far enough away from where they should be that they are effectively not there. More predictor gain allows it to move the pulses further which is a factor. Now that I have code 2.35 loaded it will run on "1'st high, 2nd low". "1st derivative" still idles better though, so I'm using that for now.
I didn't change my offset.

Okay, but I have that problem to, and I already have 2.35 loaded.

But I'm going to try a few things next friday, I will post my solution here.

Thanks for all the advice on this forum.

Nick

When I changed my algoritm to last interval or 1ste derivative, the rpm was very steady, with the 1st high 2nd low, I can't get it right.

But my problem is solved, thanks to you guys.

nick