AnsweredAssumed Answered

localhost GPIB Latency?

Question asked by lcalderin on Aug 21, 2012
Latest reply on Aug 24, 2012 by algoss

Like • Show 0 Likes0
Comment • 3

Hi,

Looking online, I've heard of GPIB latency being on the order of 30us, but I assume this is from a PC to an oscilloscope. If I were to run a series of SCPI commands locally on the oscilloscope, through a C program, does anyone know what the latency might be around? I am using a 54855a scope.

Thanks,
lcalderin

No one else has this question

Outcomes

3 Replies

algoss Aug 21, 2012 6:45 PM
Mark Correct
Correct Answer
The short answer is, "This is not specifiied, and will be highly variable".

Where did you get the 30 usec number? Are you talking about GPIB, or SCPI? What do you mean by latency? Does it include transmission time, decode time, turnaround time, etc.?

In the scopes, all commands go through the same internal pipeline. This includes USB (slave), LAN (local or remote), and GPIB. This means that commands from all sources take about the same amount of time to be decoded and recognized by the scope. Actually getting data from the scope will take differing amounts of time, just due to the differing maximum speed of the various transport mechanisms.

The only way you'll be able to figure this out will be by writing some code. If you use a socket connection between your program and the scope, you should be able to run it from anywhere (local or remote), using any transport (USB, LAN, GPIB), and compare the results.

Al
Like • Show 0 Likes0
Actions
- lcalderin @ algoss on Aug 23, 2012 3:27 PM
  Mark Correct
  Correct Answer
  Thanks, Al, I had gotten the 30us number from a quick google, and hadn't thought about it in detail.
  
  I wish to perform some simple analysis (finding amplitude of pulses, timing between pulses) on a group of pulses, and each group of pulses is separated by 50us, so I need to be ready to capture and analyze the next group 50us after the last one came in. I plan to save ~100 measurements locally, and then transfer that data to a computer. At first, I executed a series of SCPI commands through GPIB, but I found the latency was too high, so I am looking for another way to do this.
  
  I've seen code examples where the SCPI commands are sent locally, from the oscilloscope, and assumed the latency would be lower, but have not yet tested this. Do you know of another way to solve my problem, in case the local execution will not?
  
  lcalderin
  
  Edited by: lcalderin on Aug 23, 2012 4:26 PM
  Like • Show 0 Likes0
  Actions
  - algoss @ lcalderin on Aug 24, 2012 5:15 AM
    Mark Correct
    Correct Answer
    Use segmented mode to capture all of the pulses, then process all of the captured pulses in 'batch mode'.
    
    Segmented mode breaks up scope memory into a bunch of smaller sections, from 2 to thousands, then puts one trigger event in each segment. The re-arm time is around 10 usec. By setting the depth and delay correctly, you will be able to capture just the pulses, and skip the time between them. Each segment is time correlated, so that you can measure the time between the pulses, as well as the width, pretty easily.
    
    You don't say what language you are using, but if it's MATLAB, there's a segmented memory example in the Mathworks File Exchange, written by a colleague. I think it's also referenced in the 'Scripts' section of the Scope Forums. Even if you don't use MATLAB, it provides the info you would need to write in Python, C, or some other language.
    
    Al
    Like • Show 0 Likes0
    Actions

localhost GPIB Latency?

Outcomes

Related Content

Recommended Content