Strange Case of a Vanishing Spectrum, The

Signal spectrum vanishes after about 10 min of runtime. Please, help me find the mysteriously vanishing spectrum!

Hi paleicse,

Could you post a psdata file rather than an image, so that we can try and replicate and analyze what is going on.

Thanks,

Gerry

Nope, no can do! Here is the reason why:


What I can do, though, is post another screenshot in case someone on your end will attempt to gainsay the fact and claim that I am hiding the channels:

After your original post a colleague set up a test on his PC running as best a match to your settings as he could, and it ran for a few hours without showing any issue. There may be a setting within the software that he wasn't able to replicate, which is why a psdata file would have been helpful, or it may be related to the test setup you have.

Can you try the test again, and this time, when the trace disappears, close the software down and then locate the trace.xml log file and post it here, or send it through to us at support@picotech.com

To find the trace file check Section 3 of the PicoScope 6 FAQ

Not sure why you can't post the psdata file here, the limit is bigger than that, a task for someone else to look into.

Thank you, Martin. See the attached trace.xtml file below. Website still won't allow me to place anything larger than 512 KiB.


Additional information:
Problem manifests itself after 5 – 15 min of runtime, definitely within a few hours, also on another computer I have tried.

Here I posted the link to the psdata – click the link below to download:

https://mega.nz/#!jB0WTISK!yNpaqqfKP448SVLgQ_866XMcLLEUstDlODb8xvlgUxg

Hi, after testing with your .psdata file we believe that the reason for the average spectrum disappearing is due to the buffer on the device overflowing when the recorded data cannot be transferred quickly enough to the computer. The high sample rate and the relatively small internal buffer of the 2204A cause this to happen quickly following an interruption in the data transfer.
This overflow then causes the device to mark the missing data which cannot be operated on by the FFT returning an error which causes the spectrum plot to have unplotable values which are carried forward to all subsequent average spectrum plots.
The backup and subsequent overflow could be due to other applications running on the computer, other devices plugged into the USB ports or through other processes on the PC such as power-saving or screen locking.

Reducing the load on the computer could solve the problem as might changing power-saving settings to try and prevent the data transfer interruptions.

OK, OK... I see... In other words, there is absolutely nothing wrong with either the scope or the PicoScope software: everything works as designed. Is this statement correct?

The scope is behaving exactly as we would expect it to.

The issue is that with a sample interval of 1 usec and two channels enabled, the FIFO on the device can be filled in 4 msecs. This FIFO is used to provide some protection when the O/S is busy doing something else, or when another device has control of the USB bus. In most cases this is fine, but on occasions, an overrun occurs and a LOST data indicator is added to the data stream.

I will report this to the development team to see if there is something that can be done when calculating the averaging in spectrum mode and some of the data is marked as lost.

Martyn:

Thank you for your help!

At the very least I would expect PicoScope to issue a warming telling me what the problem is, and, as an added bonus, what I need to do to make the spectrum stay: do I need to change the sampling frequency, if yes, how? Do I need to change the size of the FIFO buffer? Do I need to unplug the USB devices?

Also, please note that the computer I have is a very new and powerful one:
http://www.dell.com/en-us/shop/productd ... 10-desktop

It runs on Win 10 (64 bit), apart from the scope only one other device is plugged in which act as a signal generator:
http://www.ti.com/tool/tmdx5505ezdsp

Slowing the sampling speed down would reduce the chances of this happening, as would turning off unnecessary applications and services on the PC. However it would never be possible to totally remove the risk, particularly with the small memory devices, so I will wait for some input from the development team on handling the situation in a more user friendly manner.