Auto Tune?

[RalphTrimble]

Bryan,

I've noticed other PID controllers have an "Auto-Tune" function that will learn the proper PID settings for a pit on it's own. This is about the ONLY feature I have seen in another controller that you haven't covered with the HM. I am wondering if you have given any thought to adding a auto-tune function to the HM?

I'm curious about your take on the subject.....

 

[Bryan Mayland]

I have, and it is on the TODO list, but honestly I haven't found an algorithm that works worth a darn.

Rather than take the time to port code from another project or implement a tuning algorithm from documentation, I just manually follow the steps of doing the tuning. The numbers I come up with are always bonkers even after converting them to the HeaterMeter units. Stuff like D should be 500, which would mean that _any_ temperature change would turn the fan either completely on or completely off.

If you can come up with a working process, I'll test it and if it works, implement it.

 

[RalphTrimble]

I have, and it is on the TODO list, but honestly I haven't found an algorithm that works worth a darn.

Rather than take the time to port code from another project or implement a tuning algorithm from documentation, I just manually follow the steps of doing the tuning. The numbers I come up with are always bonkers even after converting them to the HeaterMeter units. Stuff like D should be 500, which would mean that _any_ temperature change would turn the fan either completely on or completely off.

If you can come up with a working process, I'll test it and if it works, implement it.

Do you have any idea if the commercial units that have auto-tune come up with good numbers? I hadn't gone further to check into whether it actually did a good job or not...

I wonder if somehow your manual method of simulating a tuning algorithm is why your numbers come out so wild? Perhaps updates are too slow or two few data points?

 

[Bryan Mayland]

I wonder if somehow your manual method of simulating a tuning algorithm is why your numbers come out so wild? Perhaps updates are too slow or two few data points?

Autotune is a slow process, especially with the grill. All the methods are basically timing how long an oscillation takes, the I input you ran for that cycle, and the high/low point for the cycle. With a cycle taking like 5-15 minutes, there's plenty of time to adjust and you don't even adjust every cycle. It's still something I'd like to put more time into but I just haven't figured out the right way.