First touch - vertical zoom and spectrum

Having problems ? let us know the details here
Post Reply
oxd
Active User
Active User
Posts: 10
Joined: Fri Aug 17, 2018 8:16 pm

First touch - vertical zoom and spectrum

Post by oxd » Fri Aug 17, 2018 9:05 pm

Hello,
I've measured dc voltage after 3.3V regulator with PicoScope 2207B. I wanted to zoom just the area around 3.3V out to fullscreen, but didn't find out how. It was possible just horizontally. The zoom does not work vertically. Why?

So I switched to AC mode. I wanted to see the frequencies presented in the signal so I added the spectrum view and noticed that there was a 47 kHz frequency with -31dBu and the sampling rate 30 kS, but when I changed the sampling rate to 40 kS, it disappeared and there were none above -72 dBu! Is is working correctly? Nothing else changed just the kS. How can I be sure that there are particular frequencies if the spectrum changes so dramatically with the sampling rate?
Attachments
dcvoltage.png
50KS.png
40KS.png
30KS.png
20KS.png

Gerry
PICO STAFF
PICO STAFF
Posts: 424
Joined: Mon Aug 11, 2014 11:14 am

Re: First touch - vertical zoom and spectrum

Post by Gerry » Mon Aug 20, 2018 11:47 am

Hi oxd,

To answer your first question, you can zoom into that exact rectangle using the 'Windowed Zoom' tool (the one that looks lie a magnifying glass with a square in it, next to the hand icon) by selecting the tool and then clicking on a corner of the rectangle and dragging the tool to the opposite corner.

For the Signal Bandwidth that you chose, your Time Domain plot appears to be predominantly noise (it's difficult to be sure without being able to zoom in to the data file to look for anything that looks less random, or modify the FFT). So, a 47Khz signal peak in the Spectrum Plot seen at one sample rate but not another could be due to the spectral density that happens to coincide with the width of the FFT bin, and/or the windowing function. The peak, when it is displayed, could be discrete (very narrow band) signals hidden in the noise with a high enough amplitude and close enough spread, for a frequency bin, that has been created at the 150kS/s sample rate, and is the ideal size and location in the spectrum to fully encapsulate the signal spread. So, you end up with an artificially elevated signal level with one peak, as you have seen. When you change the sample rate, you also change the bin location and size, and could instead, for instance, have parts of 2 larger bins straddling the high density region of signal peaks, which will even out the signal level over those 2 bins and therefore significantly reduce the peak. Or it could be the windowing function which smears the peak for one position of the window but not for another.


So, to answer your question, you can test the Window smearing by changing the Windowing function to a Hann window in the FFT Options, which may or may not give better consistency. Alternatively, the only way to be sure if what you are seeing at 47kHz is a signal spread (in Spectrum Mode), would be to reduce the noise floor in the FFT plot to reveal what is buried within it. You can do this by keeping the same bandwidth and increasing the number of bins (in the 'FFT Options' window) which lowers the noise floor due to process gain (for more information on process gain see here: post86081.html#p86081).

Note that what you would be doing here is analogous to increasing the signal resolution and reducing the noise floor in the Time Domain, which you can do in Scope Mode by using resolution Enhancement, in the 'Channel Options' window. This would reveal any hidden non random signal just below the noise floor (but is not as effective as Process Gain, because the signal would have to be within 6dB of the top of the noise and less than 1/256th of the maximum bandwidth for the Timebase/Sample Rate you are using (for more on Resolution Enhancement see here: https://www.picotech.com/library/oscill ... nhancement).

Alternatively you could post your psdata file and we could look for you. :-)

Regards,

Gerry
Gerry
Technical Specialist

oxd
Active User
Active User
Posts: 10
Joined: Fri Aug 17, 2018 8:16 pm

Re: First touch - vertical zoom and spectrum

Post by oxd » Sun Sep 02, 2018 4:48 pm

1. How could I overlook that :-) Thanks!

2. What do you mean by FFT bin? Is it the sampling buffer?
How is the FFT calculated actually? Is it just from the displayed sampling buffer or from all the buffers?
If I want to just see any frequencies in an unknown signal, what is the best way to display them to be sure they are really present in a signal and they are not artificially created by PicoScope?

3. I had the similar experience with the noise and the sampling rate. At 300 kS it showed me a low noise, at 400 kS it showed me a bigger noise and at 500 kS it returned back to the low noise. Why did the noise change at the 400 kS rate? I would expect even bigger noise at 500 kS since it could have taken a bigger spectrum. See the pictures.
strange_sampling_01.png
strange_sampling_02.png
strange_sampling_03.png

Gerry
PICO STAFF
PICO STAFF
Posts: 424
Joined: Mon Aug 11, 2014 11:14 am

Re: First touch - vertical zoom and spectrum

Post by Gerry » Thu Sep 06, 2018 2:32 pm

Hi oxd,

1/ Easily done, as I did with your last post.

First of all, you can disregard the comments I made in my last post regarding the difference between true signals and ones created by the Picocope, I overlooked the important detail in my previous explanation about the 49kHz signal peak. I didn't pay attention to how high the peak was, because we were talking about zooming and I assumed you had zoomed in. The explanation that I gave can account for much smaller peaks, but the one you have is more likely to be a result of the signal and/or and the sampling process. So, where you are probing the 3.3V regulator or where you are taking the measurement you are probably also picking up a harmonics/interference that it is too high in frequency for the sample rate that you're using, so what is happening is the sampling process is creating Alias components, because when you select a higher sample rate the 49kHz signal is gone, and the higher the sample rate that you use the less signal peaks you see.

So, to answer your questions:

2/ How you can be sure that the signals are true signals, is (a) when dealing with interference you can make sure that you're performing the measurement far enough away from any potential sources of interference, (b) when dealing with unwanted signal components, you can probe the part of the circuit that has the least unwanted components, and (c) when dealing with Alias components, you can use a high enough sample rate (greater than twice the highest signal frequency that expect to be capturing).

To answer your other questions in (2) and then follow on from my previous answer:
An FFT is calculated from discrete values of the signal in time, so it uses the samples that have been captured in the buffer. A bin is the small band of frequencies into which the signal bandwidth is split based upon the number of points that are to be used for the conversion, so it's width is the frequency resolution of the Spectrum plot (smallest difference in frequency that can be measured between any two adjacent points. The number of bins you select (in 'Spectrum Options->Spectrum Bins') determine how many samples from the buffer will be used and how precise the Spectrum plot will be (when the Spectrum Plot is the Primary View, the number of samples will be just under twice the number of bins). If you have more than one buffer then each one will have a spectrum plot that you can scroll through using the 'Buffer Overview'.

The more bins you select for your Spectrum Plot, the more of the true signals you will see (as explained here: post86081.html?hilit=process%20gain#p86081). So, regarding the effect of windowing creating small signals at specific bin widths, the way to be sure this isn't happening is to use small enough bin widths so that the height of the signals of interest are large enough to be completely distinct from any noise.

3/ Regarding what you are seeing by changing the sample rate, I wouldn't expect to see that either, so could you take a photograph of your test setup and post in an answer.

Regards,

Gerry
Gerry
Technical Specialist

Post Reply