SNR meter vs S-meter
This discussion was created from comments split from: Web address shortcuts for less used features.
It looks like you're new here. If you want to get involved, click one of these buttons!
Comments
S-meter reading in SNR? I don´t know if that´s "mission impossible" but I would like to see the signal strength in SNR dBs because different Kiwis have different noise levels. SNR would make reading easy to compare between different receivers. Any thought about this...?
So I think the problem is establishing the narrowband noise floor for a particular frequency at a particular time. You have to subtract a noise floor value from the S-meter value to get an SNR. Remember that the S-meter is calculated purely on the data from the audio channel.
One way is to take the S-meter value during a period of no modulation by the signal of interest. But how do you reliably determine that? And what about when there are a lot of static crashes etc making a quiet period difficult to determine?
Is there any other SDR that actually does this?
SDR-Console App has an SNR meter option.
"Is there any other SDR that actually does this?"
By default SDR# only has and SNR indication, although there is a signal strength meter add-on.
Personally I like the option of both, as a strength indication is useful when testing transmit antennas and the SNR is useful when testing receive antennas.
Maybe you could measure the difference between the maximum (peak) and minimum signal level within the receive passband, but the user would have to make sure that the passband was wider than the actual signal, so that it also included a portion of the noise floor ?
Another method may be to tune to a quiet bit of spectrum near to the wanted signal, measure the noise floor, store the value and then tune to the wanted signal. The stored frequencies could then be automatically recalled on a timed basis and the delta averaged.
This is a bit like how the 'priority' function used to work on scanning receivers. Every few seconds the scanner would quickly tune to the priority channel and check for activity, before returning to the tuned frequency or range being scanned. The priority channel check would only take a fraction of a second and you hardly noticed a break in the audio on the main channel. With modern DSP you may even be able to 'mask' such a transition.
Regards,
Martin
IMO, this is an important problem. wsprdaemon calculates noise using two different methods, one looks at the (expected) quiet period at the end of the 110 second WSPR frame and the other looks at a value calculated over the entire frame. Not surprisingly these two methods give different results but within those results additional information about noise type/cause can be discerned. This turns out to be rather useful as some types of noise, e.g. T-storm crashes, become obvious.
But in general the questions of 'what is noise' and 'what is signal' seem to enter in. I find it useful to examine both narrow and wider spectra for both level and shape, and each of these vs. time. For example, by watching the two wsprdaemon noise measurement algorithms over a diurnal cycle it can become much more obvious when a station is limited by propagated noise - the holy grail of an HF system. But even this is more complex than a single-valued litmus test since propagated noise comes in with azimuth, elevation and even polarization characteristics.
There are some notes about noise sources, measurement and coupling mechanisms that are available on the wsprdaemon web site http://wsprdaemon.org/technical.html
I know we all love simple answers but I think that all of these facets of SNR can be important in order to understand what is going on, a station's limitations and how performance might be improved.
There is some description on how the narrow band SNR is derived in Simon's SDR console app at https://www.sdr-radio.com/s-meter
This is done in the frequency domain for the IQ data within the bandwidth of the selected mode and the process is similar to the estimation of wide band SNR calculated hourly in KiwiSDR.
Well, interesting discussion, even for a layperson like me. The question raised upon me because I frequently get comments that a station is coming for example S9+something in dBs. Well that does not tell me anything. Just that maybe it is good. If the noise level is S8 or the noise level is S1 there´s hell of a lot difference in those two signals coming like S9+10dB. But yes, I do use also SDRConsole and SDR# and they both have SNR readings of a signal. That´s why I thought it would be easier. It does not have to be absolute truth with the SNR but if it is the same in all KiwiSDRs, they would be comparable. Just an idea...but if not possible to achieve easily, we can forget it and concentrate on things with more importance. I love my Kiwis. :)
One of the great benefits of an SDR like the Kiwi which has a controlled and known gain path from the input connection to the detector is that it is, in theory, possible to make absolute signal and noise measurements. Furthermore if the transfer function from the real part of the radiation resistance of the antenna to that connector is known it is possible to make absolute field strength measurements. This is perhaps more highly to be desired than just knowing more about the noise.
It is only the real part of the radiation resistance of an antenna that delivers power. For electrically small antennas this can be known pretty well and for a broadband antenna, at least, is vastly different from what a VNA measures at what are normally considered to be the antenna terminals. This is a vital issue for field measurements since the gain and aperture of a small antenna is close to constant all the way up to a full have wave dipole (see my notes mentioned before). Small antennas intercept the same power as larger ones but they are much harder or impossible to match to well.
If we all knew this calibration factor for our antenna systems we could meaningfully compare not only noise but propagated signal, among ourselves and against ITU noise levels and more. But knowing this accurately is not a trivial thing and to my knowledge not many Kiwi systems can provide this so relative measurements such as SNR across frequency and time may be the best that we can reasonable hope for.
my waveform professional buddies advise that the SDR-Console SNR method is pretty sound. See comment from Benson above.
From Simon's SDR-console discussion of how the SNR meter works:
Noise floor: Purists may not be happy with the logic used to determine the noise floor - they rarely are, but it works for me.
So, no magic then. It's FFT based and works just like the current waterfall SNR measurement technique. On the Kiwi we get audio FFT for free because the excellent audio passband filter is FFT convolution based. So in principle this shouldn't be too difficult to implement (famous last words..)
Hi John,
This is good news, and hopefully relatively easy to implement.
If you do get it to work, could you please at some stage also consider adding a second plot line to the S-Meter extension to graph the SNR in addition to the signal level ?
This would be very useful when making comparisons between antennas.
Thanks,
Martin
Jon,
Clint has added an experimental SNR readout to the KFS WebSDR, upgraded version, which is now in test. Send your email address to me at w6drz@arrl.net and I'll pass along the access link.
Craig
W6DRZ