The tacho TPS is a very good idea when you want it but it's awful when you don't. I'd appreciate some way to choose normal tacho or not while in Autotune. Preferrably defaulting to normal, but that's me.

Maybe what we need is 2 autotunes, 1 diagnostic (as it is now) and 1 normal use (ie normal tacho)?

Or maybe it should be a normal tacho unless the DIP switch is set AND we're in Autotune? That is possibly a really quick fix in code.

I'm think of adding some Optimiser configurations and this would be one of them.

Personally I don't see how you could accurately use autotune without it. Its easy to nail RPM and then use speedo to track it but throttle is too hard to nail without the tacho feedback.

_Cliff_ wrote on 05/13/09 at 08:50:16:

Personally I don't see how you could accurately use autotune without it. Its easy to nail RPM and then use speedo to track it but throttle is too hard to nail without the tacho feedback.

It's probably vital on a dyno and I use it on the road for quickly fixing the most used areas in the map.

On the other hand, I have made several thousands of kilometers in autotune without caring the least for breakpoints, with very good results. The advantage is that it nails a lot of the more uncommon map cells, sooner or later. Maybe my way of using autotune is not so common but I have good results from it.

I should probably revise the tuned cells as often as possible, setting their targets to open loop but I have yet to do that. I had an idea of an "auto tune once" mode (as soon as a cell is tuned, the corresponding target cell is put to open loop) because it would need much less attention between rides. I love that idea, Luhbo doesn't  :)

Yes, I don't. Because I found that cells are changing dependent on their neighbour cells. After some time, when bigger parts, bigger areas are "flatened" the correction factors in fact  go down to max. 2 or 3 percent. In the beginning not.

Hubert

Luhbo wrote on 05/14/09 at 04:33:11:

Yes, I don't. Because I found that cells are changing dependent on their neighbour cells. After some time, when bigger parts, bigger areas are "flatened" the correction factors in fact  go down to max. 2 or 3 percent. In the beginning not.

Are you damping the LC-1 to 1/3 second or something? My experience is that autotune only alters the map after a stable reading almost spot-on a target. I see what you mean but I don't see it ending up significantly wrong. When large parts of the map has been put off autotune, you could reset it and do another pass. I'm pretty sure mine would come out pretty accurate after that.

Anyway you could still do a limited pizza dough rollout first at the most important areas, then manually set that good block of cells in open loop (or "no auto tune" per below), and only then you enter this mode.

Also, let's say the software won't put a cell off autotune until the correction was much lower than limit, like 2 or 3%.

Maybe there could be another 'bit' in the O2 flag of each cell, setting it as "no auto tune" but leaving the target, and closed loop, intact. This is a separate idea that could have advantages even if you do the rest manually. I'm just thinking loud here, I'm not sure if it would do good or bad. If nothing else it would stop autotune from endlessly alter good parts of the map with 1-2%.

Cliff has probably tried a lot of algorithms and methods over the years, maybe he already tried some of this and ditched it...

Raz I must admit I do the same as you just ride and let autotune get on with it, I appear to get better results than by trying to watch the screen and hold speed or throttle (too many potholes round here !!!).
I only refer to the screen if I find a problem then I try to remember the location so I can fix it later :)