Hi Peter,
A lot of the older technologies are based on the level of precision available at the time of the design. This doesn't mean that the older technologies are no longer useful, but rather newer technologies are often less expensive with higher levels of integration on the chip, and less operating power for hand-held devices.
Delta-Sigma converters often have a built-in PGA and digital filters. This technology allows for less complex anti-aliasing filters, and provide notches for line cycle rejection. With the oversampling/digital filtering combination quantization noise that exists in the higher frequencies is filtered with the low pass digital filter. This allows for resolution in the nanovolt range.
Which device you ultimately use depends on the measurement you want to make. This will depend on passband needed and sensor used. For example, if the sensor is primarily DC output and contains 100uV of noise in a 0-5V range, there is no need to have a 24-bit converter with 298nV resolution. If you want to convert a specific range of frequencies, you may want to choose a flat-passband converter over a sinc filter output.
As far as a bench multi-meter is concerned for 6-1/2 to 8-1/2, you are not going to do this with a $5 part. For 8-1/2 digits you will need a 30-bit or better converter. It is going to take a lot of effort and high precision components. As you've stated the reference is very critical and must have a level of noise lower than what you are attempting to measure. If it were a trivial matter of a good cap and good reference I think Agilent would have more competition and I wouldn't have to pay 10k for their meter.
There are MSP430 microcontrollers with built-in dataconverters for both SAR and Delta-Sigma. The MCP3901 does not have a microcontroller built in, and is more equivalent to the ADS1292. Perhaps you gave me the incorrect part number.
Best regards,
Bob B
