??? 11/27/04 14:36 Read: times |
#82077 - Answering the question Responding to: ???'s previous message |
The problem with PID systems isn't necessarily the maths of implementation, its the tricks of tuning, handling saturation effects, like integrator windup and noise in numerical differentiation to end up with a stable, fast settling system.
It seems to me we are focussing on the wrong bits here. IF he sees problems with rounding, speed, or program size then worry about Floating point or fixed point. I have used both successfully, but my sampling periods are usually in the seconds area than the msec area. And tuning is STILL an art. One of my ex-employees, and whose help I miss used limit cycle oscillations to tune PID loops, he could find good parameters,but his implementations were poor, and suffered from windup etc etc. I couldn't do the analysis, but could write robust algorithms. Some proper guidance on TUNING would help everyone (especially me ! ) greatly. Bruce, We are using a Peltier cooler on one of our instruments with an analogue loop and a digital loop - we didn't find any improvement from tweaking parameters. What kind of things are you doing with yours ? We have to operate at 7 fixed temperatures, -10.0, 15.0, -20.0, -25.0, -30.0, -35.0 +/- 0.05 C Steve |
Topic | Author | Date |
help about pid | 01/01/70 00:00 | |
Floating point? | 01/01/70 00:00 | |
Keil floating-point performance | 01/01/70 00:00 | |
Search the forum | 01/01/70 00:00 | |
Floating Point Vs Mixed point | 01/01/70 00:00 | |
You mean Fixed Point | 01/01/70 00:00 | |
Mixed not Fixed | 01/01/70 00:00 | |
Warnings about Floating Point | 01/01/70 00:00 | |
Shortcuts | 01/01/70 00:00 | |
Implied aka Fixed point | 01/01/70 00:00 | |
Another Shortcut | 01/01/70 00:00 | |
Answering the question | 01/01/70 00:00 | |
Autotunning? | 01/01/70 00:00 | |
PID Tricks and Tuning | 01/01/70 00:00 | |
Great Stuff ! | 01/01/70 00:00 | |
I followed your idea... | 01/01/70 00:00 | |
Another couple tricks | 01/01/70 00:00 | |
tests | 01/01/70 00:00 | |
limit cycle tuning![]() | 01/01/70 00:00 | |
PID to control a Peltier | 01/01/70 00:00 | |
An idea to try | 01/01/70 00:00 | |
very fast 20ms ??? | 01/01/70 00:00 | |
20.000, not 2000. | 01/01/70 00:00 | |
very fast 20ms ??? | 01/01/70 00:00 | |
in points... | 01/01/70 00:00 |