When exporting 15-sec capture into csv file using GUI → right click → Export all data, I expect to get 15M lines text file. Actually I get 14,961,377 lines, but last line time stamp is 14.999999. For some reason there is a time gap in csv file: lines between time 0.961376 and time 1.0 are missing in the file.
Can you zoom into this same region in the Waveform widget? Does the data look good?
It is possible that this duration corresponds to a sample drop on your host machine. Another way to confirm is to inspect the Joulescope UI log file and search for sample_id skip.
Just reproduced this again.
Here is zoom, data look all right:
Here is dump.csv file excerpt containing time gap:
Line numbers slightly different this time, but gap is here. End of file looks like this:
Hi @inlevis - I was able to duplicate what you have observed:
And the data looks just fine in the Waveform widget:
I tried again, and got a slightly different gap:
Interestingly, if I save to JLS and then load the JLS, it looks fine BUT the duration is short. Here is the end of the JLS from the second export:
That looks like 14.9681 to me…, which is the same as the CSV gap.
I tried adding dual markers set to nearly the full extent, but not the full extent. The export works:
By default, the range tools operate on 1 second data blocks. It appears that the first sample block may be truncated. The range_tool implementation looks good on a quick inspection. However, it’s not surprising that any issue would be in the view or stream_buffer implementation that actually retrieves these samples. Due to how the JS220 works, all 15 seconds may not be available as the different channels update in separate data blocks. My suspicion is that the existing buffering implementation does not take this into account and just returns the samples it can skipping the initial “missing” samples. I captured more than a full buffer and paused. Here is what I see at the beginning:
This confirms my suspicion that it is the initial data that is missing, but the CSV exporter timestamp just blindly increments from 0. One way to “fix” this is to ignore the CSV timestamp.
We are currently in the process of restructuring the entire way that sample data is stored in the Joulescope UI. This change will affect all range tools, which includes the export feature. (Un)fortunately, we will be discarding the very code that is likely the cause of this issue. We hope to have an alpha release in about two weeks.
As a workaround in the meantime, I recommend either ignoring the CSV timestamp column or exporting using dual markers with at least 1/2 second buffer on each end.
Thanks @mliberty for quick investigation.
Good to know that V and I data have no gap actually, thats what matters.
Ignoring csv time stamps is simple and easy solution for me.
1 Like