[Skimmertalk] SNR Explained

[Pete Smith]

Alex responded quickly to my query about Jim's SNR data.  Here, by 
permission, is his response:

"Jim does not explain how he manipulated his signal to make CW Skimmer 
decode the callsign and show the SNR in the spots posted to Telnet. If he 
used an unmodulated signal and read the SNR on the status bar, then his 
report has nothing to do with the SNR in the spots, since these two values 
are completely different.

The SNR on the status bar is an instant value measured in the current 
bandwidth of the DSP filter and then normalized. All signals present in the 
filter bandwidth contribute to the estimate. It was added for the debugging 
purposes and is not supposed to be used by the operators.  I use the 
"debug" SNR on the status bar to calibrate the AGC, this is the only 
purpose of that indicator.

The SNR in the spots is computed in the 50 Hz bandwidth, using only the 
signal during the dots and dashes but not during the spaces; a Kalman 
filter is used to accurately estimate the mean signal strength over the QSB 
peaks and nulls. This SNR is normalized to the 500 Hz key-on SNR.

The algorithms used for both SNR estimates are linear up to the 7-th digit 
and do not have any knees."

[Pete again] It sounds to me as if a different measurement technique is 
needed, using a calibrated CW signal source sending text ("CQs" with 
callsigns) that produces spots.  The benefit is that such data will, in 
fact, be much more accurate than the instantaneous numbers displayed in the 
status bar, because of the QSB compensation described above.

The "AGC" Alex mentions above is solely on the output side of the sound 
card that produces headphone or speaker audio - it has nothing to do with 
the SNR reported in spots.

The only really specific information on what's inside the SDR-IQ box is 
contained in the interface specification at 
http://www.moetronix.com/files/sdriqinterfacespec100.pdf . I see nothing to 
suggest that any AGC is being used there.  The linearity of the active RF 
stages and the A/D converter in the SDR-IQ could affect the accuracy of the 
ultimate SNR measurements.  Maybe Pieter can enlighten us there, but it 
seems as if linearity of <1 dB would be adequate for amateur requirements - 
we're not really trying for an antenna range level of precision here.

73, Pete N4ZR
the World Contest Station Database is back, and improved...
www.conteststations.com