Experimental Physics and
Industrial Control System

Mark Rivers via Tech-talk <[email protected]> · Fri, 15 Feb 2019 15:38:14 +0000

That shows that the device is responding within about 10 ms.  However, it is then another 1 second before the next write.  That is probably because the higher level software is waiting for more input and then finally timing out after 1 second.  Perhaps it expects \r\n rather than just \n as the terminator?

devGPIB is obsolete and no longer supported.  You could reproduce this communication using a single ai record and StreamDevice with a single protocol.  You could also reproduce it with just a single asyn record.  Try one of those and see what performance you get.

Mark

-----Original Message-----
From: Sobhani, Bayan <[email protected]> 
Sent: Friday, February 15, 2019 8:46 AM
To: Mark Rivers <[email protected]>
Cc: [email protected]
Subject: RE: How fast can EPICS communicate with a device?

The IOC I am using is an old keithley IOC I found online which seems to be using devGPIB:

https://epics.anl.gov/download/modules/keithley2400-1.1.tar.gz

The messages when I turn on asyn trace are:

2019/02/15 09:45:00.087 192.168.56.7:5025 write 7
:READ?\n
2019/02/15 09:45:00.097 192.168.56.7:5025 read 13
4.203473E-02\n
2019/02/15 09:45:01.099 192.168.56.7:5025 write 7
:READ?\n
2019/02/15 09:45:01.106 192.168.56.7:5025 read 14
-8.474424E-02\n
2019/02/15 09:45:02.112 192.168.56.7:5025 write 7
:READ?\n
2019/02/15 09:45:02.121 192.168.56.7:5025 read 14
-2.418922E-02\n
2019/02/15 09:45:03.126 192.168.56.7:5025 write 7
:READ?\n
2019/02/15 09:45:03.135 192.168.56.7:5025 read 13
7.028082E-02\n
2019/02/15 09:45:04.140 192.168.56.7:5025 write 7
:READ?\n
2019/02/15 09:45:04.149 192.168.56.7:5025 read 13
6.516452E-02\n
2019/02/15 09:45:05.160 192.168.56.7:5025 write 7
:READ?\n
2019/02/15 09:45:05.169 192.168.56.7:5025 read 13
6.202125E-02\n
2019/02/15 09:45:06.172 192.168.56.7:5025 write 7
:READ?\n
2019/02/15 09:45:06.181 192.168.56.7:5025 read 13
1.992785E-02\n
2019/02/15 09:45:07.187 192.168.56.7:5025 write 7
:READ?\n
2019/02/15 09:45:07.197 192.168.56.7:5025 read 13
9.599727E-03\n
2019/02/15 09:45:08.202 192.168.56.7:5025 write 7
:READ?\n
2019/02/15 09:45:08.211 192.168.56.7:5025 read 14
-5.647328E-02\n

-----Original Message-----
From: Mark Rivers <[email protected]> 
Sent: Friday, February 15, 2019 9:25 AM
To: Mark Rivers <[email protected]>
Cc: Sobhani, Bayan <[email protected]>; [email protected]
Subject: Re: How fast can EPICS communicate with a device?

Sorry I missed that you added .01 second in the dbd file, so ignore that question.

I suspect you are getting a timeout if you only get 1 Hz. Turn on asynTrace to see what is happening on the communication link.

What device support are you using? StreamDevice, asyn device support, etc.?

Mark

Sent from my iPhone

> On Feb 15, 2019, at 8:20 AM, Mark Rivers via Tech-talk <[email protected]> wrote:
> 
> There is no limit in the speed of communicating with EPICS, only the speed of the link and the device latency.
> 
> How did you get SCAN=.01 second? Did you redefine the EPICS scan rates?
> 
> If you want to go as fast as possible you cannot use periodic scanning. You can use a set of FNLKs that for a loop or other mechanism to scan faster.
> 
> Are you using an asyn serial or IP port driver? If so you can turn in asynTrace to get time stamped messages to measure performance.
> 
> Mark
> 
> 
> Sent from my iPhone
> 
> On Feb 15, 2019, at 8:05 AM, Sobhani, Bayan via Tech-talk <[email protected]<mailto:[email protected]>> wrote:
> 
> What exactly are the limitations on how fast EPICS can communicate with a device?
> 
> I have a device I can communicate with around 200Hz if I send it commands with a python program that uses the telnetlib module. I was wondering if it would be possible for the device to communicate with EPICS close to this rate.
> 
> I tried modify EPICS database definition files to make records process faster. To test this, I made a PV that does no device communication, but counts up when it was processed. When I added a line in the dbd file to add a scan option of .01 second, and set the SCAN field to .01, this record increases by 100 every second when I run the IOC, which is expected behavior.
> 
> However, when I tried having this record send commands to the device, it was much slower, around once per second. In general, when I process PVs that communicate with a device, it seems to take a while to get a response, whereas when I send commands directly through telnet I get the result back almost instantaneously.  Is there a limit to how fast EPICS can communicate with a device, apart from the SCAN field options?
> 
> Alex

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System