Re: SNR meter, bandwidth and gain adjustment for transverters
jdow
See my discussion. I got my wind up and had to work
it out making you guys the victims. But it does explain where you
are off the reservation and where others have gone off the
reservation.
I'll add a little more wind here. In the digital signals world there are two "domains" that make thinking interesting and provide opportunities. They more or less exist in the old analog design days but were not broken out as well as DSPs allow. First we have the time domain. In a collection of digital data the values bear a direct relationship to what you perceive with old analog instruments. As you apply successive filters from the front end input filter to the final output filters you capture smaller portions of the overall analog input. At some point you may meter this information. The meter would jump a lot. So you want to average it in some manner. It all seems to flow with our perceptions. Nyquist and others showed there is a nice relationship between a continuous time set of signals within a defined bandwidth and the data captured with a reasonable number of evenly spaced samples of that continuous set of signals. You can recreate what went in by filtering the output of the digital sampling. The proof depended on the prior work of another fellow. Long ago a gentleman named Fourier invented some math that showed interesting properties with sampled data. If you sample a sine wave of infinite duration and pass it through his math properly you get a singular value out and a whole bunch of zeros. That singular value is the representation in the frequency domain of that infinitely long sine wave. And thus we have invented the frequency domain view of the world. Infinity is longer than I, or anybody else, wants to wait. So this has been trimmed down somewhat. If you take a set of samples over time, apply a digitized version of Fourier's math, his transform, you get a frequency domain view of that short time interval. reversing his transform and filtering to remove harmonics and such restores the input, mostly. Some filtering takes place. Note that the frequency domain picture says nothing about what is happening before or after that set of data. And nothing in the recreated time domain says anything useful about the time outside that original set of samples. SDRs use this effect by taking successive sets of data, transforming to the frequency domain, manipulating the new representation of the data, and then restoring the sets in order. It's difficult to wrap one's head around the concept that what comes out of this is a good and accurate representation of what came in. However, it does work. So this is what I meant, in brief, by the time domain vs the frequency domain in my prior pontification. The frequency domain portion of an SDR breaks a limited chunk of time into a spectrum including amplitude and phase information representative of the limited time set of samples. Manipulating frequency in the time domain is awkward. Manipulating time in the frequency domain is awkward. One data value in time or frequency means nothing in particular in the frequency or time domains respectively. You cannot take one frequency data point and call it representative of signal level during that time interval. You need to use the time domain to get the quasipeak data IARU requires. With regards to your questions about the measurements I suggested with NFM and BFM suggest you are in over our current level of knowledge. FM modulation schemes spread the bandwidth of signals over a wider frequency range than the modulating signal. With a single sine wave modulation the spectrum takes on a (theoretical) infinite number of sidebands on integer multiples of the modulating signal. As the level of modulation increases from zero the various sidebands and the carrier go up and down through nulls in which the carrier or pairs of sidebands, upper and lower for a given multiple, falls to a zero value. In general intuition is validated as high order sidebands have very very low amplitudes until the frequency deviation becomes significant relative to the sideband's difference from the carrier. Because of this effect BFM signals can show "single spectral element" peak selection 20 dB or more below what you see during silence broadcast with carrier only. The SMeter value should not show that kind of change. (And even on AM the SMeter in principle should show more power on very weak signals as you make the bandwidth wider. After all, it is a meter reading signal plus noise in the selected "IF" bandwidth. And hooboy I don't want to get off on digital modulation schemes. Those skills are close to defunct through lack of use over the most recent 20 years.) {^_^} A little more below On 20210705 14:49:08,
oldjackbob@... wrote:
Joanne, Actually the meter will not read the same. Observe the spectrum behavior of a broadcast FM station as it goes from relative silence to full deviation. The meter will drop 20 dB or more even though the power received remains the same. 2) "Take an NFM signal modulated by a sine wave. Run up the modulation from zero until there are three spikes showing within the NFM bandwidth that are all the same level. What is the correct measurement? One of the peaks or the proper sum of all three?" What is this "proper sum of all three" value you mention??? Be specific. Your prized "sum" value remains undefined. This is the problem with reconstructing the "power" reading when you consider power readings of the individual terms of an FFT. Some are incoherent, as with noise, and some are coherent as with the example. With an FM signal the proper sum will remain constant not drop by several dB. 3) "Now continue running up the amplitude of modulation into the BFM realms with the signal spread out over 200 kHz bandwidth. What is the correct SMeter reading?" What is even the point of that question? The Smeter couldn't (and shouldn't) care less about bandwidth. You are wrong here on several levels. I touch on this above. 4) "Does that reading change when the announcer pauses for breath? Should it?" The simple answer is "no", at least not for a transmission with a steady FM carrier, given that by definition the modulation on any FM signal should never exceed the strength of the carrier. Once again  the Smeter should only report the strongest signal measured anywhere in the passband. For an FM (or AM) signal, the strongest signal will always be the carrier. I am SERIOUSLY struggling to not be abusive. But this requires an answer. You are absolutely and comprehensively wrong. I hope my explanations have shown you at least a hint of how wrong you are. You're asking me these questions, but you (or anyone else) can easily see the answers to every one of those questions by simply opening a session of SDRC and click on any BCFM or BCAM station. If the carrier is steady then the Smeter reading will also be steady, regardless of the amount of modulation and regardless of bandwidth. I have an SDR implementation that works the way it should measuring power within the last IF's effective bandwidth. It behaves exactly as I define it, power within the last IF's bandwidth. With FM signals this stays constant as it should. The POWER received for FM modulation does not change with time. It's pretty clear to me that you do not understand what the Fourier Transform or its derivatives provide you when viewing a signal. It is not successive views in time by any stretch of the imagination. You seem to think it is. I hope I expanded your digital horizons a little. And I hope it was worth it. If anyone has gone off a cliff, it certainly isn't me! https://www.youtube.com/watch?v=y50i1bI2uN4 (And, yes, I do fall victim to that from time to time. Everybody does. I figure I know something about radio as I've been involved with them and designing them for over 60 years.) {o.o}

