On 20/01/2016 15:47, Michael Black wrote: > So I don't see how the scaling helps SNR or quantization since that > should be unaffected by linear rescaling.... It hinders not helps. Remember that the sample stream before it gets to WSJT-X is 16-bit signed integer and dividing those samples by a constant reduces the resolution and increases to contribution of the quantization error. For example a sample of, for example, roughly half scale might be +16383, that represents a voltage 16383 +/- 0.5 steps up from zero (modulo ADC linearity and offset errors). If that were the peak signal and we halved it by means of a digital volume control it would then have the value +8191 which has introduced a truncation error than cannot be repaired. It has also doubled the quantization error relative to the sample because the +/- 0.5 LSB error suffers the same integer truncation error.
> And since the decode is done in floats there shouldn't be any > quantization produced by dividing or multiplying. True but an intermediate digital volume control in a 16-bit signed integer sample stream does. As I said: "the decoder gets whatever arrives in the audio stream, no gain adjustment is made" I am not saying that the WSJT-X gain slider has any impact on decoding, I am saying that having to use is is indicative of a problem upstream. 73 Bill G4WJS. ------------------------------------------------------------------------------ Site24x7 APM Insight: Get Deep Visibility into Application Performance APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month Monitor end-to-end web transactions and take corrective actions now Troubleshoot faster and improve end-user experience. Signup Now! http://pubads.g.doubleclick.net/gampad/clk?id=267308311&iu=/4140 _______________________________________________ wsjt-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/wsjt-devel
