Your SDR receiver ADC performance depends on number of bits and a good understanding of signals versus noise.
Think of your analog-to-digital converter (ADC) as a little guy named Bob sitting inside the signal chain. He has a camera and a binary ruler. On a regular schedule , Bob takes a picture of the incoming signal. The picture shows a dot representing the amplitude of the signal at that exact moment. Next, Bob pulls out his binary ruler and measures the distance of the dot from the bottom of the picture. Finally, he announces his measurement and prepares to take the next picture.
The faster Bob takes his pictures, the more signal frequencies he can report on. This timing is called sampling rate. The more detail available in Bob’s binary ruler, the more details his measurement reports. This detail is called bit depth.
Faster sampling rate and higher bit depth resolve ambiguities. If the sampling rate is too low, Bob can’t differentiate between different sinusoidal frequencies. Sampling rate gives you bandwidth. Too few bits on the ruler, and rounding errors swamp the measurements. Bits give you dynamic range.
Bit depth defines the range of numbers the ADC provides. 8 bits gives 0 to 255, 12 bits gives 0 to 4095. A bigger range provides better signal voltage resolution. This resolution is around 10 μV on my Perseus, versus 400 μV on the RTL-SDR. So, when I tune in the same signal on both these receivers, the dongle generates a lot more quantization error, or noise. This explains why its dynamic range is more than 30 dB less. It illustrates the different SDR receiver ADC performance between 12 and 8 bits.
SDR Receiver ADC Performance – Dynamic Range
In theory, a perfect ADC provides about 6 dB of dynamic range per bit. So, an ideal 14 bit ADC provides DR = (6.02 times 14) + 1.76 = 86 dB. This is the theoretical SNR for the ADC in the ICOM 7300 and Perseus receivers.
But, in the real world, we need to account for additional noise and distortion products introduced inside the ADC while sampling. This accounting is done with SINAD, or the ratio of signal to all other spectral products including noise, distortion and harmonics in the conversion process. Taking these into account tells a different story, The ADC in the ICOM 7300 has a SINAD of 77 dB. This produces an Effective Number of Bits ENOB = (77 – 1.76)/6.02 = 12.5. So, the ADC in the ICOM has the same performance as an ideal 12.5 bit ADC.
SDR Receiver Performance – Correlation and Dithering
There is a huge statistical difference between signals and noise. Signals are sinusoids. These are periodic – they are correlated over time. On the other hand, noise is random. Noise is not correlated over time.
Most of the time, quantization noise is random and spread uniformly over the sampling bandwidth. But, if the signal and the sampling clock become harmonically related, quantization noise becomes correlated. This is bad. Correlated noise is a spurious signal, and these decrease dynamic range.
When we are sampling, we want to make sure that quantization error is not correlated. This is why your radio adds dithering. Dithering is adding a small amount of noise into the signal before it enters the ADC. Random noise randomizes the output error, reduces the amplitude of harmonics and spreads them out. The result is improvement in Spurious Free Dynamic Range, sometimes a the expense of a slight increase in SNR.
Keep these ideas of correlation and noise in mind. They will come in handy when we discuss processing gain and wideband filtering shortly.