Not sure if this the correct forum / category - move to the proper.
I have a fundamental question being a new user to SDR in general.
I have noticed that the noise floor or noise in general increases with the higher sampling rate. This makes sense to me since its increasing the spectrum bandwidth. However, I am not seeing an perceivable difference in signal improvement nor any improvement with any of the DSP functions. and in fact S/N degrades as you increase the sampling rate.
Can someone explain to me the benefits? OR what I should expect to see increasing the sampling rate?
I am on latest build of Thetis and Protocol 2 with a 7K MKII.
Thanks
Mike
WA2COP
Fundamental Question - Sampling Rate
Re: Fundamental Question - Sampling Rate
Mike,
The good news is that changing sampling rates is not increasing noise levels. It is merely changing how noise power is displayed, and you have control over this.
The fundamental thing (hah, a pun!) you need to understand is that RF energy is measured per unit bandwidth. This is most easily observed for monotonic signals (those with constant power vs. frequency) like noise. If you measure noise power in a 2.93Hz bandwith, and measure the same noise power in a 5.86Hz bandwidth, in the latter case you have twice as much energy and the reading will be 3dB higher. This is why engineers often report noise as, for example -174dBm/Hz. With that number, it is possible to calculate what the total noise power is in any bandwidth since you know what it is normalized to 1Hz. See this topic for more math
: viewtopic.php?f=6&t=2463
From a more practical perspective, you have a lot of control over how noise power levels are displayed.
First, when you increase (or decrease) your sampling rate did you also change your FFT bin width (Setup > Display > RX1) to remain the same (because it will change automatically, unfortunately)? If not, then the displayed average noise level will appear to go up or down depending on the bin width.
Second, consider also your zoom level. If a single pixel corresponds to two FFT bins, then it shows a noise power twice that (3dB higher) than if it that pixel corresponds to a single FFT bin.
Finally, this all gets a little harder to understand with non-noise signals. If a CW signal fits in a bin, then it fits in two bins, or three bins, and the measurement of that signal is the same either way. It's only when a signal is wider than a single FFT bin, or filter, or processing channel when you start having to pay attention to math, and of course noise is infinitely wide.
I encourage you to read the link posted above.
73,
Scott
The good news is that changing sampling rates is not increasing noise levels. It is merely changing how noise power is displayed, and you have control over this.
The fundamental thing (hah, a pun!) you need to understand is that RF energy is measured per unit bandwidth. This is most easily observed for monotonic signals (those with constant power vs. frequency) like noise. If you measure noise power in a 2.93Hz bandwith, and measure the same noise power in a 5.86Hz bandwidth, in the latter case you have twice as much energy and the reading will be 3dB higher. This is why engineers often report noise as, for example -174dBm/Hz. With that number, it is possible to calculate what the total noise power is in any bandwidth since you know what it is normalized to 1Hz. See this topic for more math
: viewtopic.php?f=6&t=2463From a more practical perspective, you have a lot of control over how noise power levels are displayed.
First, when you increase (or decrease) your sampling rate did you also change your FFT bin width (Setup > Display > RX1) to remain the same (because it will change automatically, unfortunately)? If not, then the displayed average noise level will appear to go up or down depending on the bin width.
Second, consider also your zoom level. If a single pixel corresponds to two FFT bins, then it shows a noise power twice that (3dB higher) than if it that pixel corresponds to a single FFT bin.
Finally, this all gets a little harder to understand with non-noise signals. If a CW signal fits in a bin, then it fits in two bins, or three bins, and the measurement of that signal is the same either way. It's only when a signal is wider than a single FFT bin, or filter, or processing channel when you start having to pay attention to math, and of course noise is infinitely wide.
I encourage you to read the link posted above.
73,
Scott
Re: Fundamental Question - Sampling Rate
Thank you sir for the detailed explanation and fantastic read!
Am I mistaken by your statement (" It is merely changing how noise power is displayed,") but is the sampling rate only affecting the visual aspects of reception? Are there no other benefits?
Am I mistaken by your statement (" It is merely changing how noise power is displayed,") but is the sampling rate only affecting the visual aspects of reception? Are there no other benefits?
Re: Fundamental Question - Sampling Rate
Sorry, I totally forgot that part of your question! 
The benefit of higher sample rates is that more bandwidth can be monitored simultaneously. For example, at a 768KHz sample rate, the resulting 768KHz of bandwidth available allows one to monitor the entirety of every band except for 10 and 6 M. This is super handy for seeing who's who in the zoo. Is CW active? Phone? FT8? What do the levels look like, etc? Combined with CTUN display mode and a good mouse or MIDI controller you can work an entire band in a heartbeat and maintain full situational awareness of everything that is going on. In contrast, on radios with spectral displays that only cover 48KHz or less it is like looking at the world through a soda straw.
Remember, on our radios sample rate equals bandwidth because the sampling is done in quadrature, i.e. each sample produces a separate I and Q sample, hence Nyquist limits are not exceeded.
Also, Thetis with Protocol 1 firmware only support a maximum sample rate of 192KHz. Protocol 2 firmware is required to use faster sampling rates.
The benefit of higher sample rates is that more bandwidth can be monitored simultaneously. For example, at a 768KHz sample rate, the resulting 768KHz of bandwidth available allows one to monitor the entirety of every band except for 10 and 6 M. This is super handy for seeing who's who in the zoo. Is CW active? Phone? FT8? What do the levels look like, etc? Combined with CTUN display mode and a good mouse or MIDI controller you can work an entire band in a heartbeat and maintain full situational awareness of everything that is going on. In contrast, on radios with spectral displays that only cover 48KHz or less it is like looking at the world through a soda straw.
Remember, on our radios sample rate equals bandwidth because the sampling is done in quadrature, i.e. each sample produces a separate I and Q sample, hence Nyquist limits are not exceeded.
Also, Thetis with Protocol 1 firmware only support a maximum sample rate of 192KHz. Protocol 2 firmware is required to use faster sampling rates.
Re: Fundamental Question - Sampling Rate
Thank you again. Makes perfect sense!
Much appreciated!
Much appreciated!