The EPID record is supposed to do a bumpless transition when feedback is turned on. It should do the following when feedback is turned from Off to On:
- Reads the current value of the output power supply PV.
- Sets the Integral term of the PID equal to that current value.
- Begins the PID control.
Assuming you set the setpoint equal to the current value of hallprobe when you turn on feedback, then it should not "bump" because the Proportional term will be 0 and the Integral term will be equal to the current power supply setting.
Are you seeing something different from this?
The EPID record does have an ODEL field but it is currently not implemented. I'd be happy to accept a pull request to implement it.
From: email@example.com [mailto:firstname.lastname@example.org]
On Behalf Of Crisp, Daniel
Sent: Wednesday, November 29, 2017 2:48 PM
Subject: EPID bumpless restart
I'm currently using the EPID record to stabilize a dipole fields. It is reading a magnetic field from a hallprobe controller, and driving a power supply current to match the hallprobe reading to
a target value specified by the user.
How can I make the EPID record perform a ‘bumpless restart’?
Currently, if the power supply set point is significantly changed while the EPID feedback is off, the epid record will drive around the supply wildly once turned back on. I have some channels set
up which pause a user’s request to turn on feedback until the ‘<epid_channel>.I’ field can be automatically set with the live value in the supply. And when I monitor the ‘<epid_channel>.I’ value, I see that it in fact does get set, but it then gets overwritten
and ignored. What’s the best way of dealing with this?
How to best implement a deadband? (It appears there was a thread about the ODEL field being present, but not used in the epid support code? Am I wrong? Is this still the case?)
Especially for some magnets that require a less frequent touch, I’d like the epid to only act on the supply when the error reaches a larger percent difference. Specifically, I’d like it to only
react to errors in field greater than, say, .01% of the target field. We had tried to simply decrease the KP value enough that resulting P values are less than what can be acted on by the supply (supply doesn’t recognize a difference in set points differing
by the 3rd significant figure), but this isn’t ideal.