EPICS Controls Argonne National Laboratory

Experimental Physics and
Industrial Control System

1994  1995  1996  1997  1998  1999  2000  2001  2002  2003  2004  2005  2006  2007  2008  2009  2010  2011  2012  2013  <20142015  2016  2017  2018  2019  2020  2021  2022  2023  2024  Index 1994  1995  1996  1997  1998  1999  2000  2001  2002  2003  2004  2005  2006  2007  2008  2009  2010  2011  2012  2013  <20142015  2016  2017  2018  2019  2020  2021  2022  2023  2024 
<== Date ==> <== Thread ==>

Subject: Re: ePID deadband on output
From: Bruce Hill <[email protected]>
To: "[email protected]" <[email protected]>
Cc: Core-Talk <[email protected]>, "[email protected]" <[email protected]>
Date: Mon, 25 Aug 2014 16:32:30 -0700
There's no need to use the same time constant for both, and
you're right that they shouldn't be coupled.  The feedback loop
should be independent.   However, as Matt pointed out, using
a simple dead band doesn't allow the PID to settle on the target.
- Bruce

On 08/25/2014 03:38 PM, [email protected] wrote:
But the "I" in the PID is already an IIR filter.

Putting additional filter inside a feedback loop is tricky.  Too short a
time constant make the filter ineffective, too long a time constant makes
the loop unstable.

The motor itself could be another integrator. Having 2 integrators and
addition phase lag from an IIR filter in a feedback loop is not a good
idea.

Dead band with hysteresis is much simpler.

Ken

We have a similar PID controlling a stepper motor.

Since we don't have a amplifier for the stepper motor I actually
think we probably could allow the motor deadband to handle this,
but the "owner" of the stepper motor is more comfortable with
us shutting it off while we're close to the target position.

Instead of using time within the deadband to turn ours on and
off, we use an IIR filter similar to the smoothing algorithm in
the AI record with an OFF threshold which can be quite small
so we don't shut off till we're centered, and an ON threshold
which turns the PID back on when it drifts too far.

Time still enters into this calculation via the time constant of
the IIR filter factor, but you can ensure your PID gets closer
to the target position.

Regards,
- Bruce

On 08/25/2014 10:24 AM, Pearson, Matthew R. wrote:
Hi Folks,

I initially tried setting MDEL on the ai record that provides the
pressure readback (which is the input to the ePID). I found that the
ePID only controlled the output (the stepper motor) so that the pressure
sat at the low or high limit of the MDEL, depending on which side it was
approached from. So I needed some time factor to take the settling time
into account (to reach the center of the deadband).

The motor record has deadbands for the monitors (MDEL and ADEL), but not
on the input (except for the NTM logic to deal with backlash
compensation).

On Aug 25, 2014, at 11:14 AM, "Dalesio, Leo" <[email protected]> wrote:

Time can be captured in the integral term of the ePID. If that term
continues to push the change up, it will exceed the position.
Is that the argument for putting it in the ePID? We don't integrate the
error if it is inside the deadband? Is this the only term that does not
care inside a given deadband?
The output would be completely off (FBON=Off) if we have been within the
deadband for N seconds. It would only turn on again (and reset the
timer) if it drifts outside the deadband.

The idea is that the controlled variable may sit within the deadband for
minutes or even hours without any further control needed (which in my
case means switching a stepper motor amplifier on and off).

Cheers,
Matt



This looks like a good discussion for either the code-a-thon or a side
meeting at the EPICS meeting.
________________________________________
From: Mark Rivers [[email protected]]
Sent: Monday, August 25, 2014 11:03 AM
To: Dalesio, Leo; 'Benjamin Franksen'; [email protected];
'Core-Talk'
Subject: RE: ePID deadband on output

That's an interesting point.

The motor record already has a deadband.  Matt, did you try just using
the deadband on the motor record?

I don't think it has the concept of the time inside the deadband
though.

Mark


-----Original Message-----
From: Dalesio, Leo [mailto:[email protected]]
Sent: Monday, August 25, 2014 9:51 AM
To: Mark Rivers; 'Benjamin Franksen'; [email protected];
'Core-Talk'
Subject: RE: ePID deadband on output

Does the deadband belong on the ePID or the motor record?
It seems like more of an output issue - not an algorithm issue.
Bob
________________________________________
From: [email protected] [[email protected]] on
behalf of Mark Rivers [[email protected]]
Sent: Monday, August 25, 2014 10:39 AM
To: 'Benjamin Franksen'; [email protected]; 'Core-Talk'
Subject: RE: ePID deadband on output

Sorry for the delay in replying, this seems good to me as well.

Mark


-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Benjamin Franksen
Sent: Monday, August 25, 2014 5:14 AM
To: [email protected]; 'Core-Talk'
Subject: Re: ePID deadband on output

[cross-posting to core-talk]

On Thursday 14 August 2014 18:13:15 Pearson, Matthew R. wrote:
I'm using the ePID record to control a needle valve filling a cryostat
with helium. The needle valve is controlled by a stepper motor. The
ePID is giving good results, but I found the need for a deadband on
the PID output. This is to avoid continually driving the motor to and
fro by just a few steps. I implemented it via set of database
records, and included a timer to tell me how long the helium pressure
has been within a deadband, and then I disable the PID output (by
setting FBON).

However, I'm wondering if's a common enough use-case to be implemented
in the ePID record. It would mean adding some additional fields,
something like:

DBS - the deadband size
DBTS - settling time in the deadband (to avoid sitting on the edge of
the deadband) DBT - time in seconds since we entered the deadband
window
DBIN - binary flag to indicate we are in the deadband

If ABS(CVAL-VAL)<=DBS, and DBT>=DBTS seconds, then disable FBON.

If ABS(CVAL-VAL)>DBS then reset DBTS and enable FBON.

And to maintain current behavior, keep DBS=0 (the default).

The database version is working fine, but it's quite a few records and
I suspect it would be much simpler in the soft device support.
Sounds like a pretty good idea to me.

Cheers
Ben
--
"Make it so they have to reboot after every typo." ― Scott Adams


________________________________

Helmholtz-Zentrum Berlin für Materialien und Energie GmbH

Mitglied der Hermann von Helmholtz-Gemeinschaft Deutscher
Forschungszentren e.V.

Aufsichtsrat: Vorsitzender Prof. Dr. Dr. h.c. mult. Joachim Treusch,
stv. Vorsitzende Dr. Beatrix Vierkorn-Rudolph
Geschäftsführung: Prof. Dr. Anke Rita Kaysser-Pyzalla, Thomas
Frederking

Sitz Berlin, AG Charlottenburg, 89 HRB 5583

Postadresse:
Hahn-Meitner-Platz 1
D-14109 Berlin

http://www.helmholtz-berlin.de



--
Bruce Hill
Member Technical Staff
SLAC National Accelerator Lab
2575 Sand Hill Road M/S 10
Menlo Park, CA  94025



"The beaver told the rabbit as they stared at Hoover Dam. 'No, I didn't
build it myself. But it's based on an idea of mine.'"
Charles Townes


--
Bruce Hill
Member Technical Staff
SLAC National Accelerator Lab
2575 Sand Hill Road M/S 10
Menlo Park, CA  94025


Replies:
RE: ePID deadband on output Dalesio, Leo
References:
ePID deadband on output Pearson, Matthew R.
Re: ePID deadband on output Benjamin Franksen
RE: ePID deadband on output Mark Rivers
RE: ePID deadband on output Dalesio, Leo
RE: ePID deadband on output Mark Rivers
RE: ePID deadband on output Dalesio, Leo
Re: ePID deadband on output Pearson, Matthew R.
Re: ePID deadband on output Bruce Hill
Re: ePID deadband on output fong

Navigate by Date:
Prev: Re: ePID deadband on output fong
Next: RE: ePID deadband on output Dalesio, Leo
Index: 1994  1995  1996  1997  1998  1999  2000  2001  2002  2003  2004  2005  2006  2007  2008  2009  2010  2011  2012  2013  <20142015  2016  2017  2018  2019  2020  2021  2022  2023  2024 
Navigate by Thread:
Prev: Re: ePID deadband on output fong
Next: RE: ePID deadband on output Dalesio, Leo
Index: 1994  1995  1996  1997  1998  1999  2000  2001  2002  2003  2004  2005  2006  2007  2008  2009  2010  2011  2012  2013  <20142015  2016  2017  2018  2019  2020  2021  2022  2023  2024 
ANJ, 17 Dec 2015 Valid HTML 4.01! · Home · News · About · Base · Modules · Extensions · Distributions · Download ·
· Search · EPICS V4 · IRMIS · Talk · Bugs · Documents · Links · Licensing ·