I am on the road until Saturday and cannot do any more testing until then. I have taken the opportunity on the planes to read up on the definition of noise floor. The methodology in the ARRL handbook requires measuring the RMS value of the noise at the speaker terminals of the receiver as a reference. Then a signal generator is tuned to the receive frequency and the signal attenuated until the measured audio power is exactly 3 dB above the reference level. This is the "noise floor" and the MDS by the ARRL definition.
I suspect the HDSDR baseline noise floor may turn out to be the same but I want to confirm that when I return.
Note this measurement is highly dependent upon the filter bandwidth selected, as is the HDSDR baseline noise. I have tried to stay with a filter bandwidth of 2.5 KHz because that is the reference used by Joe Taylor for his SNR measurements.
I further suspect, in the WSPR example for instance, a measured SNR of "-20 dB" may well be measured on a signal that is visible above the HDSDR baseline because it would take significant power in a single frequency spike to double the noise power in 2.5 KHz..... but again, all I can do is speculate until I get back on the ground.
Finally, I have observed that I cannot attenuate the signal generator completely at -130 dB and I have always put this down to signal generator leakage. Now, with Victor's input, I am wondering if this is indeed leakage or if it is LSB quantization error.
So many things to learn..... and so little time. This is really starting to impinge on my night life! :-)
Warren Allgyer - W8TOD
--- In email@example.com, "Alan" wrote:
> ----- Original Message -----
> From: "victor"
> Subject: [softrock40] Re: USB sound card dongle with stereo INPUT
> > Warren, the 96 dB is indeed the dynamic range in which an ideal ADC will convert a signal with error smaller than 1/2 LSB. Any
> > larger signal will be positively clipped.
> I tried to duplicate your tests using your method. The generator I used did not have a sufficiently high output but with HDSDR I
> seemed to get a range of at least 109dB.
> Delta 44 with 16 bit MME driver.
> With the levels I used the noise floor remained steady.
> But using Rocky, SDR# and SpecLab I could not get results anywhere near this. I think more tests will be needed. HDSDR seemed to be
> responding linearly but I should see the same with other software.
> > I certainly felt an improvement of less spurs and better clean spectrum when you receive a dense spectrum with many strong signals
> > after changing the audio card to 24 bit.
> Sorry, I am reluctant to accept subjective assessments. In the absence of proper tests I'd at least like to see simultaneous
> spectra. One of 16 bit and one of 24 bit taken at the same time. Preferably on identical cards but even two cards in the same
> computer can have different levels of spurious pickup.
> What SDR were you using for this check?
> 73 Alan G4ZFQ