On 8/4/22 11:55 PM, F1AMM wrote:
Hello
I follow your discussion with interest but my language is not English - thank you Google. If you could avoid using acronyms such as "ADC". What does it mean ?
Could you develop a little more what follows
sure..
----o----
One other thing..
The "raw measurement" that the NanoVNA makes is essentially a filtered sum of 48 samples. The raw adc samples are probably about 1 part in 10^4 (call it 1E-4 fractional uncertainty) so 48 samples will be sqrt(N) better, or 1.4E-5. (ignoring arithmetic precision).
The way the NanoVNA receivers work is by taking the input signal and mixing it with a LO that is 5 kHz away and doing an analog low pass filter. That audio 5 kHz signal is fed into a standard multichannel audio codec[1]. The audio signal is sampled into 16 bit samples at 48 kHz for 1 millisecond. The Analog to Digital Converter (ADC) in the particular chip used is a bit tricky, but it's easiest to think of it as a straight 16 bit ADC, with a specified noise SNR of 80dB (min) 95 dB (typ), and a THD+N spec of -85 dB (typ) -70dB (max).
80 dB is 1 part in 10,000 for voltage. So I assume that each sample has an uncertainty of 1E-4 full scale. That is, if I put 1 volt into the digitizer, the samples have rms noise of 0.1 mV. (This is approximate, there's lots of caveats on this).
If you average a bunch of samples, the noise voltage is reduced by a factor of sqrt(N), where N is the number of samples. So the estimated noise is 0.0144 mV, not 0.1 mV.
But to turn that into an uncertainty on a measurement, we need to know the level of the signal. We know the noise (after averaging) is 0.0144mV, but we're probably not putting a full scale signal in. I assumed we'd put in 0.1*full scale. So our measurement uncertainty is 0.0144 mV out of 100 mV.
[1] The ADC is a TLV320AIC3204IRHBR multi channel audio chip (Ultra Low Power Stereo Audio Codec, as TI calls it).
That's voltage, not power, relative to full scale.
In reality, the input signal isn't full scale, it's more like 0.1 full scale or 0.05 full scale, so the uncertainty is about 0.1%
The reflection coefficient is calculated as the algebraic combination of two of those measurements, so the uncertainty is roughly doubled (0.2%)
----o----
- thank you
