Please help me understand this, digitizer deadtime question.

[LCHEN5154]

Dear all:

 I have a PCI-5154 high speed digitizer.  I have measured the trigger re-arm time

 and to my surprise I found the re-arm is changing regularly with the captures waveform

 length. please see attached figure.

 In the test I set the trigger as 'immediate' so I can calculate the trigger re-arm time by

 calculating the time gaps between waveforms.

 Each time I captured 10000 waveforms, get the re-arm time and then go for a different

 measurement with a different waveform length.

 Can anyone help me understand this figure? 

 A brief interpretation will help a lot.

 thank you.

[Steven_K.]

Hi LCHEN5154,

The rearm time is going to vary depending on your record length as stated in the NI High-Speed Digitizers Help. The rearm time that you measured is well under the specification of the PCI-5154. According to the NI 5153/5154 Specifications that specification is rated at 1 microsecond.  

[LCHEN5154]

Thanks, Steven.

The help file mentioned there is a dependence but what I want is a brief explanation

of that dependence.  The measured dead time is within 1 us and I am not gonna complain

anything about that.

 What I am asking is if someone can explain this at a fundamental level, for example, this

 dependence could be caused by some data storage optimization or something else.

[LCHEN5154]

Hello, can anyone say something about this?

I repeat the help file in the following, Yes it mentioned "depending on the record length" but what I want is a brief explaination of why or how.

The main benefit of multiple-record acquisitions is that you can acquire numerous triggered waveforms quickly. Multiple-record acquisitions allow hardware rearming of the digitizer before the data is fetched. Therefore, the rearm time, or the time when the digitizer is not ready for a trigger, is extremely small, often from 1 to 100 µs, depending on the record length and the digitizer. This allows you to capture data if the triggers occur 100 µs apart, or if they occur many days apart.

[dansch]

Hello Dr. Chen,

In general, there are some different padding methods and data alignment that takes place in the FPGA and memory streams, and the exact details of what is happening can vary greatly between digitizer models and even firmware versions. It is hard to point to specific values and numbers for whatever cases you measured, but there are some different reasons having to do with padding the data and making sure the data and timing streams are properly aligned. As the number of samples in your record varies, the time the board takes to perform the processing steps and padding of data will vary (generally, a step-like pattern is expected when padding data).

The main take away is that for any case, we spec a value that is an absolute maximum for the board to account for all of the different situations in which the rearm time may vary. I found the graph you provided quite interesting and I would also be very interested in seeing the program/code that you used to acquire that data and get the results that you provided. Additionally, I am interested in knowing about your application and the reasons for testing this information and understanding the details of the process (such as the relevance for your application). I look forward to hearing back from you soon with any additional information you can provide.

Best Regards,

[LCHEN5154]

Hello Daniel:

 Thank you.

 To get the plotted data:  I recorded some data with immediate trigger and then simply calculate the time gap between waveforms to get the  re-arm time.

 The reason for I concerned about this is I am measuring some time dependent events. I need to know the true dead time during the measurement so that I can make dead time correction to the recorded data. Without knowing the exact re-arm time I can not make correct dead time correction.

The correlation between the dead time and the waveform length is very interesting to me too and I am trying to understand this.  From the attached figure, the longer the waveform length the shorter the re-arm time and at a certain point the re-arm jump up for more than 100ns.

I can not understand why, for example, the re-arm time of 300 ns waveform length is short than the 200 ns waveform length but not the other way around which may be explained by setting a maximum re-arm time.  

 I really want understand the zigzag shape in the as shown in the attached figure.

 Thank you.