Feedback for version 1.0a here

[Mike H]

Hi Steve, here are the next three plots. The reference level is included in the filename and I have included a snapshot of the settings alongside the main plot Cheers

-20dBm.png (102.41 KiB) Viewed 193314 times

-30dBm.png (105.85 KiB) Viewed 193314 times

-40dBm.png (104.91 KiB) Viewed 193314 times

x2, Mike.

[Mike H]

-80dBm.png (108.6 KiB) Viewed 193315 times

-60dBm.png (105.26 KiB) Viewed 193315 times

-50dBm.png (105.92 KiB) Viewed 193315 times

[SteveAndrew]

I had attached the plots (or so `I was told...) I will try reattaching them in two submissions this time. It is not so much the noise floor that increases, but the shape of the curves as well, especially around marker 'C'. Many thanks, Mike.

Hi Mike

Regarding the noise floor, the plots are as I would expect them to be for each reference level setting. As the reference level is decreased from 0dBm to -100dBm, the noise floor will drop, and more of the signal will be visible. I have no answer for the anomalous signal that appears at 103.85MHz at reference level -30dBm. Apart from that particular plot, everything looks fine to me. Am I missing something ?

Steve

[Mike H]

Hi Steve,

Yes, that particular peak at 103.85MHz @ -30dBm reference was the aspect that set me off thinking that something was wrong and some more complicated interaction was going on. In a way, I feel comforted that the peak at 103.85MHz is not expected behaviour (by you at any rate), and I will try repeating that measurement.

I think my initial confusion came from the brevity of the release notes where the reference level is described by "Ref – dBm Sets the reference level at the top of the display screen." I maybe took this too literally to essentially be a screen centring function (a bit like the y-adjust in a scope), but I now think that this is what the offset function does in the display scaling panel. I am now interpreting the reference level as a means of changing the upper limit on the screen display, but it also changes the full-scale value for the ADC conversions. So increasing the reference value from -80dBm gradually towards -20dBm increases ADC quantisation (in Volts) so smaller spectral components start becoming part of the noise floor and ADC impairments & quantisation noise form a larger percentage of any spectral tone. If this is correct then I think the fog is lifting.... I think I may have been spoilt with some of the Agilent PSA equipment costing $40k upwards in a lab I once used.

Thanks again for your considered reply and again congratulations on a great piece of software.

Mike.

[Mike H]

Hi Steve,

I found some more info on the peak at 103.85MHz... It is repeatable, but... If I start the analysis with averaging using the 'Start' button on the top panel then the peak is not present. However, if I cycle through the reference values while in averaging mode then after 10-20 changes in the ref value, the anomalous peak will become present. It is not aways visible, and -30dBm seems to show it particularly well. It is not always in the same place and seems to shift frequency. Stopping and restarting the analysis using the button on the top panel seems to fix the problem.

I thought you'd be interested...

Mike.

-30dBm(3).png (101.32 KiB) Viewed 193259 times

-30dBm(2).png (101.31 KiB) Viewed 193259 times

-30dBm(1).png (100.41 KiB) Viewed 193259 times

[SteveAndrew]

Hi Steve,

Yes, that particular peak at 103.85MHz @ -30dBm reference was the aspect that set me off thinking that something was wrong and some more complicated interaction was going on. In a way, I feel comforted that the peak at 103.85MHz is not expected behavior (by you at any rate), and I will try repeating that measurement.

I think my initial confusion came from the brevity of the release notes where the reference level is described by "Ref – dBm Sets the reference level at the top of the display screen." I maybe took this too literally to essentially be a screen centring function (a bit like the y-adjust in a scope), but I now think that this is what the offset function does in the display scaling panel. I am now interpreting the reference level as a means of changing the upper limit on the screen display, but it also changes the full-scale value for the ADC conversions. So increasing the reference value from -80dBm gradually towards -20dBm increases ADC quantisation (in Volts) so smaller spectral components start becoming part of the noise floor and ADC impairments & quantisation noise form a larger percentage of any spectral tone. If this is correct then I think the fog is lifting.... I think I may have been spoilt with some of the Agilent PSA equipment costing $40k upwards in a lab I once used.

Thanks again for your considered reply and again congratulations on a great piece of software.

Mike.

Hi Mike

The ADC quantisation is not changed by the Ref-dBm control, at all times, the full ADC range is used. Changing the reference level changes the overall receiver gain via lookup tables of LNA and IF gain/attenuation settings. A lower reference level of say -100dBm will result in a lower noise floor than a reference level of -10dBm. The tables have been optimised by SDRplay to provide the best overall performance for any given radio, reference level and frequency.

I found some more info on the peak at 103.85MHz... It is repeatable, but... If I start the analysis with averaging using the 'Start' button on the top panel then the peak is not present. However, if I cycle through the reference values while in averaging mode then after 10-20 changes in the ref value, the anomalous peak will become present. It is not aways visible, and -30dBm seems to show it particularly well. It is not always in the same place and seems to shift frequency. Stopping and restarting the analysis using the button on the top panel seems to fix the problem.

Thanks for the information Mike. That seems like a very odd one. I've added it to the list of strange things to investigate.

Due to time constraints, the manual was very interim. A full manual will be ready for the beta release.

Steve

[Mike H]

Hi Steve,

Many thanks for the detailed answer. Cheers, Mike.

[ja8scd]

Hi Steve,

1) The "Ref-dBm" control sets the reference level at the top of the screen. In addition, it sets the LNA and IF gains, according to a pre-defined look-up table. At higher reference levels, the total gain will be low, the lower the gain, the higher the noise floor will be. At lower reference levels, the gain will be higher, resulting in a lower noise floor. This is normal.

2) I have the filter buttons on the debug list already. They are working here, but I'll run through the code again.

Thanks for your reply.

1) Now I understand the function of "Ref - dBm". I contacted Tech Support, but as of this writing, they don't know the reason(s) why the RSPduo shows different spur level patterns, either. Anyhow, hope you and Tech Support will kindly work together to solve this issue for the next beta version.

2) Noted, thanks.

By the way, I have measured an ADS-B (1090MHz) BPF filter with both RSP2 SAS 1.0A and VNWA (vector network analyzer). Please see the attached. The results with SAS 1.0A almost coincide with the ones with VNWA. (I used a transparent display function of VNWA to overlap the traces.) If a through calibration function would be provided for the future versions, it should be perfect!

Many thanks to you for the development of such a nice spectrum analyzer software!

73,


Yas - JA8SCD Tokyo

[SteveAndrew]

1) Now I understand the function of "Ref - dBm". I contacted Tech Support, but as of this writing, they don't know the reason(s) why the RSPduo shows different spur level patterns, either. Anyhow, hope you and Tech Support will kindly work together to solve this issue for the next beta version.

Hi Yas

The RSPduo spurs are on the list now.

By the way, I have measured an ADS-B (1090MHz) BPF filter with both RSP2 SAS 1.0A and VNWA (vector network analyzer). Please see the attached. The results with SAS 1.0A almost coincide with the ones with VNWA. (I used a transparent display function of VNWA to overlap the traces.) If a through calibration function would be provided for the future versions, it should be perfect!

Those screen shots make me feel very happy - Thank you

Can you explain what you mean when you say a 'through calibration function' ?

Steve

[ja8scd]

Hi Steve,

I mean such a function by which the white trace becomes flat (0dB - reference line) so the yellow trace will be displayed in relative to the white one. This is because noise generators will not give out uniform output from 1kHz to 2GHz, and also we don't need absolute dBm readings when it comes to measuring filters.

73,


Yas - JA8SCD Tokyo