# Why would smaller LSB reduce SNR performance in an ADC?

As a follow up to my previous question there is another argument I have difficulty to understand. First here the related information from the text:

Above what I understand is there are two ways to reduce LSB hence to obtain better resolution. One is to reduce the reference voltage. The second way is to use a higher resolution ADC.

But the text says for both ways when the LSB is reduced SNR performance degrades. As far as I know the quantization noise decreases with better resolution as the formula is (6.02n + 1.76) dB where n is the the number of bits available for discretization.

But the text is talking about loss in SNR performance due to lowering the LSB value. Loss in SNR performance I guess means SNR gets worse.

How does making the resolution better degrade the SNR performance? I dont get why would SNR of the input signal decrease/degrade by doing that. Can this be explained in a clear fashion?

• Its YOUR job to lower the noise. There is no magic about that. Commented Apr 25, 2019 at 16:43

## 2 Answers

As you reduce Vref the quantization size reduces but the input signal has to be reduced (the max input is probably equal to Vref) so the effect of analog noise in the input circuitry is increased and so SNR is reduced. Reducing Vref is not the same as increasing the number of bits where you could keep the input signal at the same level.

• I see what you mean. But the text also says using a higher resolution ADC.(I guess for the same input) also degrades SNR.What do you think about that? Commented Apr 25, 2019 at 14:25
• @atmnt Reading that statement, it sounds it's dancing around the idea of a noise floor. If the noise floor is well above your LSB size, then you won't get your ideal SNR from a 12-bit ADC.
– W5VO
Commented Apr 25, 2019 at 14:45
• @W5VO I dont understand:( can you illustrate what you mean maybe as an answer. Commented Apr 26, 2019 at 8:31

There are multiple sources of noise. Quantization noise is one source and it is reduced by decreasing the size of the LSB. Other sources of noise may remain the same (think of noise from an input amplifier) so you may see a gain by reducing Vref until the point where you can no longer accommodate the signal peaks and have to start attenuating the signal. After that point there is no advantage to be had, and usually disadvantage (because of those other sources of noise).