AnsweredAssumed Answered

Using :DIGitize command to acquire data problem !

Question asked by mulligan252 on May 13, 2014
Latest reply on May 16, 2014 by mulligan252

Like • Show 0 Likes0
Comment • 4

Hi,

Is :DIGitize the only command to use to acquire data remotely from a scope?

We have an experiment where we are trying to acquire waveforms at a trigger frequency up to 10Hz (hardware trigger connected to the scope).
We are loosing some of these waveforms as the trigger frequency increases to 10Hz because the :Digitize command takes so long . :DIGitize is followed by :WAVeform:DATA? in a loop in C code.

If however I don't use :DIGitize and just send :WAVeform:DATA?, there is no loss of data but then the acquisitions happen independent of the trigger. It is of course possible to poll the trigger register remotely in my code to determine when a trigger occurs but then this is more or less a software dependent trigger.

Is there some other command or method to wait for a trigger on the scope side and then transfer the data without using :DIGitize as this seems to take too long clearing the buffers, stopping the acquisition etc.

I would appreciate any help or ideas for this. Many thanks in advance

Edited by: mulligan252 on May 13, 2014 2:06 AM

No one else has this question

Outcomes

4 Replies

algoss May 13, 2014 5:23 AM
Mark Correct
Correct Answer
1) You don't indicate what model or family of scopes you are using.

2) You don't say how much data you are trying to capture and save

3) You don't say anything about what medium you are using to transfer the data (USB, LAN, RS-232, etc)

4) You don't say what programming environment you are using (VBA, Visual C/C++, MATLAB, Python, LabView, etc).

5) You don't say what format you are using for transfer (BYTE, WORD, ASCII, etc)

6) You don't indicate what type of drivers you are invoking (Visa, Sockets, etc)

All of the above will have some effect on your results.

That said...

- DIGITIZE, according to the manual for the Infiniium family, is the fastest way of using the scope. It is an internal macro that combines a RUN with a wait for completion. For other scope families, the answer may be different.

- I doubt that you will be able to trigger and unload data at a 10Hz rate, unless you are using a fast medium (Gbit LAN), with the right protocol (possibly HISLIP), and a very shallow trace depth. I have done some benchmarks in Python and MATLAB, and seen some in C/C++, and there is a certain amount of overhead in getting the captured data pulled out of scope memory and converted into a transferable format. ASCII will be the slowest, BYTE will be the fastest (to transfer, but not necessarily to convert), but least accurate. WORD is the best compromise, but note that you will also need to transfer the Preamble to understand what the bits mean (This is true of BYTE, as well).

- Most of my testing was done using LAN. Some findings: GBit is faster than 100Mbit, Sockets are (usually) faster than Visa, Isolated LAN is faster than Shared LAN, Newer models of scope are faster than older models, Direct connect LAN is faster than going through 1 or more switches, Desktops are (usually) faster than laptops.

More information about what you are doing might be useful.

Al

Disclaimer: For more reliable response, you should consider calling your local Agilent Technical Call Center. The Agilent Forums are monitored on an "as available" basis, and aren't necessarily the fastest way to get technical questions answered.
Like • Show 0 Likes0
Actions
- mulligan252 @ algoss on May 13, 2014 6:12 AM
  Mark Correct
  Correct Answer
  Hi, and thanks for your reply
  
  1) I am using an Infiniium 9000 series scope
  
  2) I am trying to capture 250,000 points of data
  
  3) The medium I am using is shared LAN
  
  4) The programming language that I am using is C and in the Linux environment.
  
  5) Currently using BYTE format but I have tried all three. My program is creating and saving raw BYTE files in real- time and saving them remotely. The conversion and processing of these raw files then is of course done off-line after the acquisition session.
  
  6) I am communicating at a low- level without using any supported libraries and just using sockets.
  
  I have a generator which outputs a signal which changes frequency on every hardware trigger repeatedly. That way it is like time stamping and I can view the processed waveform data after to see the pattern of frequencies to observe if I am loosing any data. The trigger frequency also changes from 1 to 10 Hz. When the trigger frequency is at 10 Hz, I am loosing triggers/captures.
  
  The thing is, if I don't use DIGitize and just keep transferring the waveform data by calling :WAVeform:DATA? from my C code in a loop, I am not loosing any data , but then I am receiving too much because it is not trigger dependent . For example when the trigger is at 1 Hz, I would be receiving the same signal 22 times per second.
  
  I assume the DIGitize command waits for the next trigger before acquiring the new data. Is this true?
  
  If I don't use DIGitize and just poll the trigger register remotely in my code waiting for a trigger and then call :WAVeform:DATA? to transfer the data, I am not loosing anything. So what is really the purpose of DIGitize here if it takes so much time to complete?
  
  DIGitize clears the waveform buffers and starts the acquisition until it is complete, then stops any acquisitions. Why should the acquisition stop? Is there not many waveform buffers that can fill up on every acquisition?
  
  Thanks again for the reply and any help or ideas are much appreciated.
  Like • Show 0 Likes0
  Actions
  - algoss @ mulligan252 on May 15, 2014 8:49 AM
    Mark Correct
    Correct Answer
    1) Make sure your scope has the latest FW. If it's running XP, that's 4.20.something, but you should upgrade to Win7. If it's running Win7, the latest is 4.60.something. The Win7/4.60 SW is faster than the WinXP/4.20 SW.
    
    2), 3) It's doubtful that you will be able to get this throughput. 250K Points is 500K of data ( in WORD, DON'T USE BYTE), which is 4Mbits times 10/second is 40Mbits/second. On a dedicated Gbit connection, it's possible, but marginal. On a shared connection, it might work, if there's only a single switch(and it's a good one), with CAT5e or CAT6 cables, and little competing traffic. On 100Mbit LAN, even on a dedicated connection, there's no way. I have tested throughput on a lot of different network topologies, and seen 25:1 speed differences.
    
    4) The scope captures 8 bit data, and linearizes it to 16 bit, for accuracy. The best results will come from using the WORD format. Using BYTE, the scope truncates it back to 8 bits. In the process, you will end up with missing codes and doubled codes (2 voltages giving the same result).
    
    Make sure the scope is in 'Triggered' mode, and on the channel with the signal that tells you when the frequency changes.
    
    Just doing :WAVeform:DATA? will give you indeterminate results, and may just give you the same data over and over again. You need to synchronize with the DUT in some way.
    
    I have talked to the guy in the lab who is optimizing this stuff, and he says :DIG followed by :WAV:DATA? is the fastest way.
    
    There is only one waveform buffer. The scope captures, stops, then processes, and possibly transfers, the data. If you are in free run mode, only displaying (and not transferring) data, then there are 2 buffers and they ping-pong. They do this to improve screen update rate, and the memory controller (which also does data reduction for the display) can handle it.
    
    If you don't need 'continuous' capture and transfer, you can use Segmented mode. With the default 20M depth, the scope will capture 80 250K segments, all timestamped, before it fills up, and you can unload it.
    
    Al
    Like • Show 0 Likes0
    Actions
    - mulligan252 @ algoss on May 16, 2014 5:45 AM
      Mark Correct
      Correct Answer
      Right now I am not using :DIG but instead waiting for the waveform buffer to be full after a trigger buy using query :ADER? and then getting the data by using :WAV:DATA? .
      
      This way I am successfully able to unload 250 KB of data (using BYTE mode at present as we are not too concerned about accuracy) at 10 Hz with thousands of triggers without any loss. So is this proof that using :DIG is not the fastest way?
      
      We now wish to try and capture more than one channel at the same time and see how it goes. What is the best way/ commands to capture multiple channels on the oscilloscope?
      
      Is there a separate buffer for each channel? How exactly does the scope save more than one channel for example via the GUI when selecting multipurpose on trigger and save all waveforms? I am trying to implement the same in my code.
      
      Thanks
      
      Edited by: mulligan252 on May 16, 2014 4:16 AM
      
      Edited by: mulligan252 on May 16, 2014 6:36 AM
      Like • Show 0 Likes0
      Actions

Using :DIGitize command to acquire data problem !

Outcomes

Related Content

Recommended Content