I would like to use the AD4008 16-bit ADC from Analog Devices to sample a load cell. I also need a linear regulator and driver amplifier and instrumentation amplifier.

I am curious to know if these two parts have a low enough noise spec to be suitable for a 16-bit ADC application.

For the linear regulator I have chosen the MIC5205 from Microchip. It has 260nV/√Hz output noise.

For the instrumentation amplifier, I have chosen the INA821, which has 7nV/√Hz noise.

Should these parts both be suitable for a 16-bit ADC (assuming they are implemented correctly), or should I be looking for lower noise parts?

  • \$\begingroup\$ It depends broadly on two factors: (a) the required signal bandwidth and (b) the gain of the InAmp. For the linear regulator, a schematic would be needed and an analysis of the low frequency noise (not the spectral noise). Ditto the InAmp for low frequency noise. To a large extent the voltage reference will also be of interest when analysing the noise you can accept. It initially boils down to you providing a schematic. \$\endgroup\$
    – Andy aka
    Oct 2, 2020 at 11:06
  • \$\begingroup\$ @Andyaka - But there's nothing in those components themselves which rules them out? \$\endgroup\$ Oct 2, 2020 at 11:11
  • \$\begingroup\$ There's nothing that rules them in either. \$\endgroup\$
    – Andy aka
    Oct 2, 2020 at 11:13
  • \$\begingroup\$ Let's see your noise budget calculations. \$\endgroup\$
    – user16324
    Oct 2, 2020 at 11:55
  • 1
    \$\begingroup\$ That linear regulator has an extremely low drop out voltage, but noise rejection drops off very quickly with frequency. Do you need to be able to run with a 50-100mV voltage drop on the regulator? If not, I might look at parts with better PSRR within your frequency band of interest. \$\endgroup\$ Oct 2, 2020 at 14:51

1 Answer 1


The maximum non_permanent_warping of a strain gauge is 1%. And the maximum Vout_differential of a strain_gauge, safely, is 1% of VREF. (From what I recall reading). Thus you should plan on using Gain=100 (40dB) in the INA.

Power Supply Rejection will be a big deal.

Establish an error or noise budget. Assume 10 microvolt Peak error. Allocate 1/3 of that for the Power Supply intrusion. (or you pick your favorite numbers).

Have 1/3 of the budget for Power rail trash, ripple, hum, rectification transient ringing.

The INA821 has 80dB PSRR at Gain of 100, at 1KHz, RTI, for the -15 volt supply. Thus you need a series RC filter (Local Battery) on the Minus Rail.

The RTI referred to input means the GAIN_OF_100 (40dB) will boost the -rail trash. Thus equivalent PSRR (referred to output, into the ADC) is only 40dB.

For 3uV total integrated trash, you need total integrated 1milliVolt on the Rail (assuming 80dB applies at all frequencies, which is wrong; 10KHz PSRR is only 60dB. Thus at 10,000Hz VDD trash, the equivalent PSRR is only 20dB. That means the total integrated VDD trash (DC to 10,000Hz, assumed flat contribution; the lower decades get swamped out by the highest decade) must be 100 microVolts.

And total integrated, over 10,000Hz, tells us the budget per rtHz is 1MicroVolt. Thus your chosen LDO, at 0.26uV/rtHz, shuld be fine.

But you need to examine the 60Hz ripple, and any harmonics thereof.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.