RE: Bidirectional device support

["Elder Matias" <Elder.Matias@lightsource.ca>]

It may very well be an issue of personnel preference and design
philosophy.  On our machine it comes up on a lot of equipment.

The key issue is that you can walk up to a power supply or another
intelligent device switch the device to local mode change the set point
(on its local panel) and then put the device back into remote (EPICS)
control mode.  At this point there is an inconsistency because EPICS is
running with a different set point than the device in the field.  This
is independent of how you handle the readback.

As an example:
	In the control room I set a power supply to 20 Amps.  
		EPICS SP= 20 Amp, PS SP = 20 Amp, feedback = 20 Amp. 

	Go to the local PS panel place the device in local and change to
30 Amps
		EPICS SP = 20 Amp, PS SP = 30 Amp, feedback = 30 Amp.

	Place the device into remote
		EPICS SP = 20 Amp, PS SP = 30 Amp, feedback = 30 Amp.

The end result is the poor operator in the control room is left
scratching his head because he is looking at a screen that says the
device is being commanded to be at 20 Amps, the feedback says it is at
30 Amps.

As for the concern about someone changing the setpoint in the field and
at an OPI at the same time, this is the nature of a distributed control
system.  This problem is no different than two people changing a value
on two different OPIs at the same moment in time.

Elder

To: "Thompson, David H." <thompsondh@ornl.gov>, Andrew Johnson
<anj@aps.anl.gov>, tech-talk <tech-talk@aps.anl.gov> 
Subject: RE: Bidirectional device support 
From: Carl Lionberger <clk@ornl.gov> 
Date: Mon, 12 Apr 2004 10:56:43 -0400 

I think the desire for bidirectional records reflects some personal 
preference; some people think there is some kind of elegance about it.
I 
agree with Andrew that it is not realistic.  I think it can be easily
avoided 
in all cases, as I outline below.

>===== Original Message From "Thompson, David H." <thompsondh@ornl.gov>
=====
>If you are trying to do I/O on the same point you are right.  You can't
>really turn most hardware around like that anyway.  There are two or
>three cases that I think we (at SNS) need to be able to do
>bi-directional I/O; when talking to a smart device and you want to do
>bumpless transfer of control between local and remote,

This sounds equivalent to having a remote/local switch on the device;
the 
switch is implemented in labview.  I thought it was SNS policy to not
have 
EPICS control remote/local switches.  These should be controlled at the
device 
(locally).  EPICS reads them.  Some poor guy locally testing the
hardware 
should not have control taken away from the control room.  Its not safe.

[>]when you have a
>device that can set one state and depends on Epics to set the other
>state,

For this you have an output record that sets it and a readback record
that 
reads it.  I find the least confusing way to do this is to have the
output 
record be momentary, ie, a bo with a HIGH of about 1 so that when you
push the 
button it sets the state (which is read out) and returns to standby.

[>]and when you have more than one PV pointing at the same hardware
>address in the OUT field.

In this case, last gets it. This just says you monitor the val field.
The 
difference from the normal case is that all device supports which write
to the 
same hardware must post monitors when any of them controls it.  Its not
really 
bidirectional control.

Carl

Carl Lionberger
SNS Controls Group
(865) 574-7636