Test and Measurement Forum

Waveform averaging - the PicoScope way

Forum for discussing PicoScope version 6 (non-automotive version)

Re: Waveform averaging - the PicoScope way

Postby lab!fyi » Fri Aug 11, 2017 12:36 pm

Noticed averaging keeps same "sample interval" as current sample rate dictates. But does not keep phase with original signal. When starting capture it assumes some random position within total "play" approx same as sample interval. Average trace is in magenta, and max play observed is marked with cursors. Sometimes, randomly, it starts in correct phase and keeps it.

PS6__bug__averaging__phase_shift__NOTOK.jpg
Average trace phase shifted


PS6__bug__averaging_phase_shift__OK.jpg
Average trace ok, lucky strike


Perhaps averaging should also have [option of] higher res horizontally, than current sampling rate, to work sort of like ETS works.

PS. Please increase allowed attachment size - 256KB not enough for screenshot PNGs and even JPGs, if do not compress to garbage!
lab!fyi
Active User
Active User
 
Posts: 21
Joined: Mon Dec 12, 2016 6:58 pm

Re: Waveform averaging - the PicoScope way

Postby lab!fyi » Fri Aug 11, 2017 1:13 pm

When observing some other similar product noticed that sample points on normal trace are aligned with trigger pos. When enable averaging it also keeps same sample interval as regular trace, and also keeps sample points aligned with trigger pos. Getting more smooth trace is governed by "oversampling" control (2X, 4X, etc).
With this system averaged trace does not shift in phase.
Probably Picos "non aligned" triggering system is more accurate but need to account for it in secondary processing.
lab!fyi
Active User
Active User
 
Posts: 21
Joined: Mon Dec 12, 2016 6:58 pm

Re: Waveform averaging - the PicoScope way

Postby Martyn » Fri Sep 01, 2017 8:13 am

Thank you for the comments which will be passed on to the development team for further discussion.

There are many data paths from the analog front end of the scope through to the display of data on the screen so, although a suggestion may seem easy to implement when looking at it in isolation, it may not be as straightforward in context of the full system.
Martyn
Technical Support Manager
Martyn
Site Admin
Site Admin
 
Posts: 3036
Joined: Fri Jun 10, 2011 8:15 am
Location: St. Neots

Re: Waveform averaging - the PicoScope way

Postby Sensorman » Tue Feb 20, 2018 12:41 pm

Having recently switched from using ADC212-100 (running under XP) to a 5444B and PicoScope 6, I found this thread very helpful - for a moment, I thought averaging was not possible!

Just want to check a couple of points. Having created, say, "average(A)" as a Math Channel, and displaying only this channel while working, then:

When saving in native .psdata format, a current displayed average trace will NOT be available (stored) if you select "current waveform only". It seems that you need multiple traces in the buffer in the saved file, otherwise the "average(A)" trace will be identical to the single normal trace. This implies that a displayed average from a saved file is created from the average of the buffer traces (whereas a live "average(A)" trace may include data not held in the buffer).

But if you "save as" another format, such as .csv, then the current displayed average trace IS exported when "current waveform only" is selected. Actually, I haven't checked whether this is exactly the displayed trace, or an average of all traces in the current buffer, but the difference would be minor in my case.

A slight peculiarity I have observed is that, using the Math Channel Wizard, unchecking the Range tickbox "Override automatic range selection" does not seem to be persistent. I found that auto range was quite often overridden when arranging the display, and I had to go back into the Wizard quite frequently to uncheck this box again.

Finally, and this may be an obvious comment, note that the averaging facility enormously increases vertical resolution, while maintaining original sampling rate. Say you select 8 bit resolution to get fastest possible SR (1 ns, in my case). I found resolution was enhanced by 6 bits in the "average(A)" trace, giving apparent 14 bit result. And for 12 bit original resolution, averaging appeared to enhance this to >25 bit.
Sensorman
Newbie
 
Posts: 0
Joined: Tue Feb 20, 2018 8:31 am

Re: Waveform averaging - the PicoScope way

Postby Martyn » Fri Feb 23, 2018 10:10 am

Yes you are correct, average works from start of run when live, not just the buffer memory, but obviously can only work with the data buffers from the saved file. We have a request with the development team to just average the buffer memory as an option, but this work has not been scheduled yet.

The range tickbox issue is already in the system for investigation by the development team.
Martyn
Technical Support Manager
Martyn
Site Admin
Site Admin
 
Posts: 3036
Joined: Fri Jun 10, 2011 8:15 am
Location: St. Neots

Re: Waveform averaging - the PicoScope way

Postby Benedito » Mon Feb 26, 2018 2:15 pm

In reality the Math Channel Averager are a powerfull COMB FILTER
Benedito
Newbie
 
Posts: 0
Joined: Sun Feb 25, 2018 12:08 pm

Re: Waveform averaging - the PicoScope way

Postby eilert » Wed Mar 14, 2018 7:30 am

Benedito wrote:In reality the Math Channel Averager are a powerfull COMB FILTER

Hi Nedi,
When I read your message first I wondered what made you think of a comb filter.

Normally these are used to separate equally spaced channels, so how should they do the desired averaging.

But investigating how comb filters are implemented and how the Picoscope averaging actually works made things clear, and you are right.

Picoscope averaging adds up buffered samples of the signal. So this can be seen as a time delay. The triggering makes sure that the starting point of the repetitive signal is the same for all samples. Therefore the resulting comb fillter will have its pass bands at the harmonics of the repetitive signal and erase the frequencies in between, thus clearing the signal from distortions.

This explanation requires good understanding of signal and filter theory.

One could also explain the signal improvement simply by looking at the properties of a noise signal.

While the noise amplitude randomly changes at equally spaced points, summing up a number of such samples eqals to almost zero, while the underlying repetitive signal amplitude will be the same at these times and therefore multiplied by the number of sample sets.(Needs to be divided afterwards to get the correct amplitude.)

So the signal will be restored while the noise is averaged away. However both ways require proper triggering.

Kind regards
Eilert
eilert
Newbie
 
Posts: 0
Joined: Wed Mar 14, 2018 7:25 am

Previous

Return to PicoScope 6 for Windows

Who is online

Users browsing this forum: No registered users and 1 guest