While I agree with your conclusion (and the moral of the story), Alex (KR1ST), I don't agree with some of your specific comments on the audio side of the world.
The radio's passband curve is something I hadn't considered and you've raised a good point. Very cool, thanks for that. However, regarding your sound card comments: > >A 5khz sinewave sampled at 10 kHz, as mentioned in the example >above, will be reconstructed (either in memory to feed >applications or to, say, a DAC) as a square wave. At 48 or 96 >kHz sampling rate, the resulting representation of the >sinewave will look a whole lot more like the original sine >wave than the one sampled at 10 kHz. > Obviously -- The Nyquist frequency is the MINIMUM sampling frequency. I agree that sampling a pure 5KHz sign wave (with no added noise) at 96KHz would be a good thing, cuz you'd get a nice smooth curve. But sampling a complex signal that contains information above 5KHz at 96KHz is going to get you a lot "extraneous information" (noise and artifacts) that you're just going to have to "deal with" later. >> It seems to me that the key criteria for choosing a sound card for >> digital use would be: >> >> - Flat frequency response from (some low frequency such as) 100HZ to >> 5KHZ > >Every card has this property. > I guess it depends on one's definition of "flat", yes? I don't consider +/- 1db "flat" >> - Dynamic range in the area of 100db (and hence, a very low noise >> floor) > >The noise floor of what? > The sound card as used in the overall sound system. Some sound cards have an abysmal noise floor, with self-generated and induced noise (from the computer power supplied, graphics card, cable routing, etc) that I'd guess (without measuring) is on the order of -30dbFS. Again, while I admit I haven't measured this (wouldn't be hard), just taking the output from a cheap or poorly installed sound card and plugging it into a good audio system demonstrates this to be the case: Hiiiiisssssss... Just horrific. >Dynamic range is defined by the sample size, the number of >bits per sample. Weeeellll.... the MAXIMUM dynamic range of which the card is capable is defined by the sample size. The overall dynamic range of the sound card had to be within this value, but is limited by other factors (such as internal noise). >> Additionally, it seems like it would probably be a good idea if the >> sound card had filtering to remove frequencies below 100Hz and above >> 5KHz. > >There's no need for the sound card to do so. This is already >done in the radio. > Assuming the audio signal path in the radio is well filtered, doesn't generate any noise above 5KHz, and the cables between the radio and computer are well shielded, I agree. > >A flat response is not as necessary for a narrow bandwidth >mode like PSK31 or CW, > Except when you want to look at a 5K swath of bandwidth and select a signal on your waterfall to which to listen, right? >modes. People should be much more worried about a flat >passband curve of the IF filters than the frequency response >of the sound card. The latter is usually extremely flat, while >the former is usually not. > Ah... I thank you for this, because this is something I wasn't aware of. Though I *have* seen people refer to this issue in passing, I didn't really understand what they were referring until seeing your graphs. Very cool, and much appreciated. Good input, thanks, de Peter K1PGV
