Hello Members,

There has been good application note MS-2022 regarding signal chain design for ADC. However, I have few confusions after going through that application note. I would appreciate if somebody from the community can clarify those,

1. In step-2 of design, there has been mention of noise density of 416nV/sqHz. But is not mentioned whether it is RTI side or RTO side. If it for RTI then I am bit confused as total noise of 4.16uV divided by 10KHz bandwith comes to be 41.6nV/sqHz.

2. Also gain block is of 10 while there has been mentioned we need to design solution which is 1/10th of source noise so that signal chain noise is much less than noise floor of sensor. So I am bit confused by calculation of step-4 in application note.

3. In step-6 it has been shown that total noise over 10KHz of signal chain is 3.04uV which is not 1/10th of 4.16uV. So it goes against the statement that we need to design at 1/10th of source noise. So again confused whether we need to design such that signal chain noise is lesser than noise floor of source noise or we need to design at 1/10th of that.

I will appreciate if somebody can reply my above points,

Thanks,

jagdish

Hello Jagdish,

I share your confusion regarding MS-2022. I have the impression, that the "1/10 of source noise" and the gain-factor of 10 may have got mixed in the AN and that not all calculations are done consistently (especially the conversion of rms noise to noise density and backwards).

I would calculate as follows:

1) the sensor noise density is 41,6nV/sqHz (i.e. RTI for the amplifier, just as you stated above)

2) in step 4 of the AN a limit for the opamp-noise is calculated, which does not give a signal chain noise which is 1/10 of the sensor noise. The limit is calculated for a signal chain noise which is equal to the sensor noise.

3) finally an opamp with slightly better noise performance is choosen. This gives a total noise density of 30.5nV/sqHz and an integrated rms-noise of 3,05µV. So the numbers in the text of step-6 seem resonable to me. The numbers in the formula (3,04µV/sqHz and 30,4µV) seem strange to me.

I guess that depends on the efforts you want to spend for a low noise measurement. Having a signal chain noise which is just 10% of source noise would be almost perfect, you would hardly see any influence of the signal chain noise on the total noise.

In real systems you will often not be able to build such a "perfect" signal chain or you will not be willing to spend the necessary effort and cost. Designing a signal chain which gives as much noise contribution as the sensor itself (or slightly less, as done in the AN) will oft be a reasonable target.

best regards

Achim