Ok, thanks to all for your comments. For now I'll keep the sample rate 
at 1 ms and see how that behaves.If I'm not satisfied I'll remove the 
jitter from the output update cycle and see if that improves anything.

Cheers,
Bas
---

Darren Conway wrote:
> Hello
>
> All standard control theory books assume constant sample rates with no 
> jitter.   This makes the maths solvable.  This consistent treatment in 
> the books may lead to the assumption that jitter is bad.  Jitter is 
> noise and noise is never good but is it significant?  Probably not so 
> long as jitter is less than a sample period.
>
> All the theory says that delay destabilizes control systems (with some 
> exceptions).  So the first aim is to achieve a high sample rate.
>
> If I was writing a software driver, I would be aiming to maximise the 
> sample rate first.  I would be aiming to output the results of 
> calculations as soon as they are available and not necessarily at the 
> same time. I wouldn't worry about jitter.
>
> Jitter might be important for feed-forward systems where errors 
> (noise) tend to get amplified with the potential to produce an 
> unstable system. So long as the feed-forward calculation is running at 
> the same sample speed as the PID loop, there shouldn't be too much of 
> a problem.
>
> In short, just because the theory assumes zero jitter, doesn't 
> necessarily mean that this is a practical requirement.  I suggest you 
> focus on keeping the sample rate high and don't worry about jitter.
>
> Regards
>
> Darren Conway
>
>
> Jon Elson wrote:
>> Bas Laarhoven wrote:
>>   
>>> Ah, thanks John, you made it very clear! So in my situation the input 
>>> sampling moment is fixed (at the start of the thread) and only the 
>>> update of the output varies with the variations in processing time. You 
>>> state this will probably have no noticeable effect, but as this 
>>> variation is a significant part of the PID cycle time, wouldn't  it be 
>>> better to reduce the variation (noise) and work with a fixed although 
>>> increased delay? Another option is to lower the loop update frequency as 
>>> that reduces the noise in more than one way. Do you know of a rule of 
>>> thumb for the cycle time in relation to the system response time ?
>>>     
>> The "pros" usually have a setup so that a hardware clock or 
>> interrupt timer both samples the encoder position and updates 
>> the DAC velocity values at the same instant.  The velocity 
>> output is thus delayed until the NEXT servo cycle, but the 
>> timing has very much reduced jitter.  You might be able to 
>> obtain similar result by holding your velocity update back in 
>> the driver, and writing it to the device registers just after 
>> reading the encoder data.  Then, the computation would follow 
>> these actions, and the result would be held until the next servo 
>> cycle.
>>
>> Jon
>>
>>     


-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Emc-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/emc-developers

Reply via email to