It depends on your ADC sampling rate. Generally it doesn't matter that much as long as the analogue voltage signal is constantly moving through a range of many ADC counts (between taking each ADC sample). Then just keep one bit of the ADC sample, the LSB.
My experiments with a PIC input ADC module showed some bit bias, especially where the input noise was weak in drive ability (high source impedance). I scrapped the voltage noise source idea early on and made something better.
If you are going to use an ADC make sure the noise signal has been buffered by an opamp etc first, so the signal has good drive ability.
We really need to know what you need to achieve, so we can suggest a good way to do it. One fo the most important things is the datarate; how many bits of randomness do you need per second. That will be a big decider in choosing the best technique.
Also we need to know just how good the entropy needs to be. Is this for an electronic dice or do you want to beat the NSA?
My experiments with a PIC input ADC module showed some bit bias, especially where the input noise was weak in drive ability (high source impedance). I scrapped the voltage noise source idea early on and made something better.
If you are going to use an ADC make sure the noise signal has been buffered by an opamp etc first, so the signal has good drive ability.
We really need to know what you need to achieve, so we can suggest a good way to do it. One fo the most important things is the datarate; how many bits of randomness do you need per second. That will be a big decider in choosing the best technique.
Also we need to know just how good the entropy needs to be. Is this for an electronic dice or do you want to beat the NSA?