PicoScope 4262 noise floor has significant spurs

Having problems ? let us know the details here
Gerry
PICO STAFF
PICO STAFF
Posts: 1145
Joined: Mon Aug 11, 2014 11:14 am

Re: PicoScope 4262 noise floor has significant spurs

Post by Gerry »

Hi Dan,

The PicoScope 4262 would not have the same sort of front end as the HP analyzer, as the frequency domain processing is not a hardware function (so it is not even a digital spectrum analyzer, i.e. it doesn't capture a spectrum of frequencies and then transform the analogue values of those frequencies into digital values for computation and display), the spectrum is entirely created in software (so the hardware is all time domain only, no log-amps or compensation needed).

Regards,

Gerry
Gerry
Technical Specialist

dnessett
Active User
Active User
Posts: 39
Joined: Tue Oct 16, 2018 10:30 pm

Re: PicoScope 4262 noise floor has significant spurs

Post by dnessett »

Gerry doesn't seem to be responding. Anyone else know the answer to my question?

Thanks,

Dan

Posted later: Oops. Didn't see Gerry's reply as it was on the second page. My apologies.

Dan

Gerry
PICO STAFF
PICO STAFF
Posts: 1145
Joined: Mon Aug 11, 2014 11:14 am

Re: PicoScope 4262 noise floor has significant spurs

Post by Gerry »

Hi Dan,

I had to create a quick answer for you as I was going to be out of the office for a while. So, I've made the below explanation more comprehensive for the benefit of other forum users.

As already mentioned, you have to keep in mind that the Spectrum plot of a PicoScope is not a digital Spectrum Analyzer. It is a mathematical transformation of data captured in the normal way, i.e. the Time domain of an oscilloscope, while a Spectrum Analyzer sweeps a window across the spectrum of a signal in real-time to analyze the content. We can make the analogy that it acts like a digital Spectrum Analyzer because it splits the spectrum up into discrete sections (bins) to convert the information into the frequency domain equivalent, while a Spectrum Analyzer, scanning the spectrum, effectively splits it up into continuous sections to analyze and display the frequency content.

The harmonics of a signal diminish pretty quickly as you increase the number of the harmonic, so the plot of a Spectrum analyzer, or Spectrum mode in an oscilloscope, is most useful when it can bring the higher harmonics into view that would normally be lost off of the bottom of a linear scale, so a logarithmic scale is the preferred way of plotting a frequency spectrum. The Spectrum analyzer therefore uses a log amp for its analysis/plotting, because it is analyzing and plotting the data while it is sweeping across the spectrum. The PicoScope will have already captured the data linearly (using the amplification/attenuation best suited for displaying the chosen input range) and only needs to transform it according to the users requirements, so that it can be displayed in linear or logarithmic form.

Regards,

Gerry
Gerry
Technical Specialist

dnessett
Active User
Active User
Posts: 39
Joined: Tue Oct 16, 2018 10:30 pm

Re: PicoScope 4262 noise floor has significant spurs

Post by dnessett »

Thanks Gerry.

I have had to take some time off as my mother fell and broke her hip. Helping her with rehabilitation has taken a lot of my time. So, I didn't visit this forum for a while, which is why this response is so late.

Cheers,

Dan

Gerry
PICO STAFF
PICO STAFF
Posts: 1145
Joined: Mon Aug 11, 2014 11:14 am

Re: PicoScope 4262 noise floor has significant spurs

Post by Gerry »

Hi Dan,

Sorry to hear that. I hope she's doing O.K.

Coincidentally, my mother did exactly the same thing a number of years ago and had to have a replacement. The most important thing that I learned from was that was that the rehabilitation (physio-therapy) after that is so critical to getting them back to their mobility (even if they don't have much enthusiasm for it), because it can be key to maintaining their quality of life, so I'm sure it was time well spent.

Regards,

Gerry
Gerry
Technical Specialist

dnessett
Active User
Active User
Posts: 39
Joined: Tue Oct 16, 2018 10:30 pm

Re: PicoScope 4262 noise floor has significant spurs

Post by dnessett »

Gerry wrote:
Thu Apr 25, 2019 8:57 am
Coincidentally, my mother did exactly the same thing a number of years ago and had to have a replacement. The most important thing that I learned from was that was that the rehabilitation (physio-therapy) after that is so critical to getting them back to their mobility (even if they don't have much enthusiasm for it), because it can be key to maintaining their quality of life, so I'm sure it was time well spent.
Yes, rehab is the key. I am spending 1 1/2 hours a day at her care facility (she is 94 with dementia) walking with her. She complains a lot, but as you mention, getting her mobility back is critical to maintaining her quality of life.

Regards,

Dan

dnessett
Active User
Active User
Posts: 39
Joined: Tue Oct 16, 2018 10:30 pm

Re: PicoScope 4262 noise floor has significant spurs

Post by dnessett »

Hi Gerry,

I wanted to characterize the noise floor of the PicoScope 4262 in terms of dBm@50 ohms. Since my version of PicoScope 6 runs on Linux and that version doesn't support displaying dBm@ 50 ohms, I saved the data in csv format and, using Octave, converted it to the desired form. Here is the graph:

<>
<>
<>

As you can see the noise floor is between -90dBm and about -115 dBm. This is about 10-20 dBm less than the figure you gave in this message (fixed broken link 4-29-2109). I looked over the various topics in which we discussed noise floors and couldn't figure out a definitive answer to the question whether my calculations make sense. So, I thought I would ask you. I can supply you with the psdata file for the measurement, or you can run the measurement yourself, whatever you prefer.

I also ran corrections for spectral leakage (as discussed in this message). I plotted the corrected spectrum (in red) over the non-corrected version. As you can see the difference is slight, so I don't think corrections are important for my application at least.

<>
<>
<>

Finally, I tried to use the suggestion you made here, but could find no way to select Primary or Secondary view. Is this perhaps a difference between the Windows and Linux versions of the PicoScope 6 software?

Cheers,

Dan
Attachments
PicoScope Noise Floor Windowed Correction Overlay 4-26-2019.jpg
PicoScope Noise Floor 4-26-2019.jpg
Last edited by dnessett on Mon Apr 29, 2019 10:32 pm, edited 1 time in total.

Gerry
PICO STAFF
PICO STAFF
Posts: 1145
Joined: Mon Aug 11, 2014 11:14 am

Re: PicoScope 4262 noise floor has significant spurs

Post by Gerry »

Hi Dan,

Have you corrected for Process gain (I couldn't find any reference to it any of your links)?

If not then go here to find out how to correct your data: topic25101.html?&p=86081&hilit=process+gain#p86081

Regards,

Gerry
Gerry
Technical Specialist

dnessett
Active User
Active User
Posts: 39
Joined: Tue Oct 16, 2018 10:30 pm

Re: PicoScope 4262 noise floor has significant spurs

Post by dnessett »

Gerry wrote:
Mon Apr 29, 2019 3:55 pm
Hi Dan,

Have you corrected for Process gain (I couldn't find any reference to it any of your links)?

If not then go here to find out how to correct your data: topic25101.html?&p=86081&hilit=process+gain#p86081

Regards,

Gerry
Hi Gerry,

I did not correct for processing gain, as I will be comparing the noise floor of the PicoScope with the noise floor of the mixer I am using to make phase noise measurements. I will do this by measuring the phase noise floor (actually an upper bound of its noise floor) of an ultra-low phase noise oscillator (a Morion MV89A) using the mixer and analyzing its spectrum on the PicoScope. I will then note if the result is greater than the PicoScope noise floor. If not, then the PicoScope will limit phase noise measurements, not the mixer.

Nevertheless, you raise a legitimate point. Considering the PicoScope in isolation and comparing it with the figure of -95 dBm you quoted as the noise floor of the 4262 requires taking processing gain into consideration. I used 262,140 samples when capturing the spectrum that I displayed above. This represents 10 Log (262,140 / 2) = 51.175dB of processing gain. Ignoring the first two bins (which have abnormally high dBm values, -38.766 and -49.510, not sure why), the maximum for the spectrum is -90.509 dBm and the minimum is -115.20 dBm.

Correcting for processing gain gives the PicoScope noise floor between -41.56 dBm and -64.025 dBm. This seems to high when compared with the -95 dBm figure. So, I am a bit confused what is happening.

At this point in composing this message, I hit myself on the side of the head because I suddenly realized I multiboot Windows 8 on the machine to which my PicoScope is attached. So, I am able to run the Windows version of PicoScope 6 and do the comparison myself (mea culpa!). I did so and setting the Spectrum output to dBm/50 ohms I got the following image:

<>

Correcting for processing gain, I obtained the following image:

<>

The last plot is closer to the result you gave, but still not quite on the mark. In particular, the minimum and maximum of the processing gain corrected data (ignoring the first 3 bins) is -93.158 dBM and -62.445 dBm. I used the movable ruler to measure the middle of the non-corrected noise plot and it equaled about -139 dBm, which corrected for processing gain is -87.825 dBm.

The bottom line is my calculations for dBm from dBV are obviously faulty and I need to figure out why. On the other hand, I am still confused what the correct dBm/50 ohm values should be, since the values calculated by PicoScope 6 on Windows 8 (when corrected for processing gain) appear to be about 10 dB too high.

Regards,

Dan
Attachments
4262 Noise Floor - Windows 8 - Process Gain Corrected.jpg
4262 Noise Floor - Windows 8.jpg

dnessett
Active User
Active User
Posts: 39
Joined: Tue Oct 16, 2018 10:30 pm

Re: PicoScope 4262 noise floor has significant spurs

Post by dnessett »

Hi Gerry,

I figured out the problem in my conversion program. I was first converting dBV to millivolts by dividing the dBV value by 20, exponentiating 10 by this value, and then multiplying by 1000. To get power I squared the result and divided by 50. This is obviously incorrect. Power is voltage squared divided by the load resistance then multiplying by 1000 to get milliwatts (converting to dBm by taking the log of the result and multiplying by 10). By squaring the millivolt value I introduced a factor of 1000 increase in the final value (i.e., 30 dB). I have corrected the program and the result it returns now agrees with the result in dBm @ 50 ohms produced by the Windows version of PicoScope 6.

The issue of the average noise floor being about 10 dB greater than the figure you quoted still remains.

Cheers,

Dan

Gerry
PICO STAFF
PICO STAFF
Posts: 1145
Joined: Mon Aug 11, 2014 11:14 am

Re: PicoScope 4262 noise floor has significant spurs

Post by Gerry »

Hi Dan,

Just to confirm that the measurements in PicoScope 6 should be correct, as you mentioned running the example of process gain that I gave you the link for, and getting errors in your values, I have recreated the data in the attached file for the latest version of PicoScope 6 and you you can see that the signal to noise ratio measurement and ruler measurement are very similar to the ones in the post, and taking the values from the recreated file: -118dB of noise +42dB of Process gain gives a signal to noise ratio of -76dBc which matches the measured value.
Re-worked example.psdata
(202.67 KiB) Downloaded 153 times

I also terminated the input of the Scope with 50 ohms, to measure the isolated input noise of a channel and got the attached data files showing the Spectrum and Time domain plots, (along with their screen captures). So, working through the Spectrum plot data, we have a noise level of -144dB + process gain of 42dB = -102dBV of noise.
So, 20log(noise voltage/1V) = -90dBV
or, the noise voltage = 10^(-102/20) = 8uV
If you look at the Time domain plot you can see that the RMS measurement of the noise is just over 5uV, which is close enough.
50 ohm terminated input noise on PS4262 - Time Domain Plot.png
50 ohm terminated input noise on PS4262 - Spectrum Plot.png

However, if you are making the measurements w.r.t. 1mW dissipated into a load for dBm, and your load is 50Ω then you need to make the necessary correction to the measurement, in particular, if you want to use a voltage reference then the ref for a 50Ω load will be √(50 * 0.001) = 0.2236V. So, my measurement of the terminated PS4262 input channel in Spectrum Mode should now give a level of:
20Log(5.1*10^(-6)/0.2236) = -92dBm - the process gain of 42dB giving a grand total of -134dBm (which is where the missing 10db that you mentioned are), and is what you will see (or almost) in the attached screenshot of the same measurement, and the data file for it.
dBm measurement of 50 ohm terminated input noise in a PS4262.png

Going back to your previous post where you mentioned that you couldn't select Primary or Secondary view I'm not sure whether you were referring to selecting a second viewer, or deciding which of the second viewers is Primary and which is Secondary, but you just need to go to the 'Views' menu and select 'Add view' and select the type of view you want to add. The Secondary View will then be the view that you are adding to the already existing Primary View. I have run PicoScope 6 on Ubuntu, and done exactly that with the same 50 ohm terminated test to very similar results again, as you can see in the attached Screenshot.
Ubuntu screenshot.png

Regards,

Gerry
Gerry
Technical Specialist

dnessett
Active User
Active User
Posts: 39
Joined: Tue Oct 16, 2018 10:30 pm

Re: PicoScope 4262 noise floor has significant spurs

Post by dnessett »

Hi Gerry,

Gerry wrote:
Wed May 01, 2019 1:10 pm
...
However, if you are making the measurements w.r.t. 1mW dissipated into a load for dBm, and your load is 50Ω then you need to make the necessary correction to the measurement, in particular, if you want to use a voltage reference then the ref for a 50Ω load will be √(50 * 0.001) = 0.2236V. So, my measurement of the terminated PS4262 input channel in Spectrum Mode should now give a level of:
20Log(5.1*10^(-6)/0.2236) = -92dBm - the process gain of 42dB giving a grand total of -134dBm (which is where the missing 10db that you mentioned are), and is what you will see (or almost) in the attached screenshot of the same measurement, and the data file for it.

The mistake I made was using the number of samples rather than the number of bins to compute the processing gain. Specifically:

dnessett wrote:
Tue Apr 30, 2019 12:33 am
Nevertheless, you raise a legitimate point. Considering the PicoScope in isolation and comparing it with the figure of -95 dBm you quoted as the noise floor of the 4262 requires taking processing gain into consideration. I used 262,140 samples when capturing the spectrum that I displayed above. This represents 10 Log (262,140 / 2) = 51.175dB of processing gain. Ignoring the first two bins (which have abnormally high dBm values, -38.766 and -49.510, not sure why), the maximum for the spectrum is -90.509 dBm and the minimum is -115.20 dBm.


...

The last plot is closer to the result you gave, but still not quite on the mark. In particular, the minimum and maximum of the processing gain corrected data (ignoring the first 3 bins) is -93.158 dBM and -62.445 dBm. I used the movable ruler to measure the middle of the non-corrected noise plot and it equaled about -139 dBm, which corrected for processing gain is -87.825 dBm.

Correctly computing the processing gain as 10 Log (131070 / 2) = 48.165 and taking the approximate average of the non-corrected noise plot as -139, the value corrected for processing gain is ~ -91 dBm. Since the estimate of -139 is not precise, this seems close enough to the theoretical value.

Gerry wrote:
Wed May 01, 2019 1:10 pm
Going back to your previous post where you mentioned that you couldn't select Primary or Secondary view I'm not sure whether you were referring to selecting a second viewer, or deciding which of the second viewers is Primary and which is Secondary, but you just need to go to the 'Views' menu and select 'Add view' and select the type of view you want to add. The Secondary View will then be the view that you are adding to the already existing Primary View. I have run PicoScope 6 on Ubuntu, and done exactly that with the same 50 ohm terminated test to very similar results again, as you can see in the attached Screenshot.

Thanks. I didn't explore the "Add view" submenu and therefore missed this feature.

Regards,

Dan

Post Reply