Hi what is the negative effects of using an audio ADC in sampling voltages in instrumentation use? I know it blocks low frequency like 10Hz but my application uses 100Hz to 10000Hz with input of microVolt. Can it somehow distort the voltage sampling of microvolt? Im using the E1DA Cosmos ADC that uses the following chip.


enter image description here


3 Answers 3


Offset and gain will not be specified with the same level of accuracy (or at all).

If you only do AC measurement, offset is irrelevant. Note the ADC you mention in the datasheet has a digital DC blocking filter (high pass) which can be bypassed via I2C configuration to make the ADC work with DC too. I don't know if the Cosmos allows you to do that, or if the analog frontend passes DC or not.

Gain will probably be specified under "full scale digital corresponds to ... dBV on input" or something like that ; ESS chips have some variation of gain between chips or between channels, which is usually compensated for by the internal volume control. Basically, you have no out of the box accuracy on gain, so you have to calibrate it. That's not really a problem.

Same thing for frequency response, phase, etc, it depends on the frontend circuit in your box.

Boxed ADCs usually have volume pots, which means once the calibration is done, the pot should not be used anymore.

On a "real" measurement instrument you would find rotary switches (or relays, etc) having discontinuous accurate settings, ie you can increase gain by an accurate 10x. With a pot, you can't (you can use fixed attenuators).

You won't have any specs on gain and offset stability vs time, temperature, etc.

This means a signal source of known accurate amplitude, or an accurate AC True RMS voltmeter, are needed for calibration if you want to get "real volts" out of your soundcard data.

Besides that, they work absolutely fine as low noise, low distortion, and especially low cost ADCs. No problem at all.

You can use a soundcard for a wide range of measurements by building your own analog frontends and software. It makes a great low frequency network analyzer, but that's only the beginning.

The main feature in a soundcard is synchronization between DAC and ADC. If you use the DAC to make a stimulus signal, run it through a DUT, then acquire it with the ADC, when ADC and DAC are synchronized you can measure phase lag from the DUT, or average a low amplitude signal over many periods to reduce noise.

Separate DAC and ADC not using the same clock are less useful because the recorded data won't be synchronized with playback data.

  • \$\begingroup\$ You sure about synchronous dac/adc in soundcards? Don't they use an output buffer? However, one can use one if the stereo channels to records the dac and a second stereo channel to record the response. The stereo channels are typically well delay matched \$\endgroup\$
    – tobalt
    Mar 3 at 11:40
  • 1
    \$\begingroup\$ There is a delay between DAC and ADC due to buffering, driver, software, configuration, settings, etc. You need to use direct IO like ASIO otherwise Windows may decide to change your settings on a whim. Then the delay needs to be calibrated out (for example using autocorrelation on a logsweep). But once you do, you get one sample out of the ADC for each sample out of the DAC, which greatly simplifies programming the measurements. \$\endgroup\$
    – bobflux
    Mar 3 at 11:48
  • 2
    \$\begingroup\$ If you use a sigma delta ADC it's going to use noise shaping whether it's officially designed for audio or not... \$\endgroup\$
    – bobflux
    Mar 3 at 12:13
  • 1
    \$\begingroup\$ I've never seen an audio ADC do "shenanigans" but drivers and dsp's and windows... that's another matter. ASIO is a must. \$\endgroup\$
    – bobflux
    Mar 3 at 12:19
  • 2
    \$\begingroup\$ @Jtl Noise shaping is very widely used in high resolution ADCs, both for audio and general instrumentation. It's not a psychoacoustic effect as typically implemented, so it is not limited to audio. Rather it is used to shift quantization noise out of band. \$\endgroup\$ Mar 3 at 14:23

In general, audio is a rather demanding application in terms of INL and noise floor. So high-end audio data converters can be an excellent pick for instrumentation.

However, audio also has some notably irrelevant specs, such as low-frequency noise, offset, offset drift over time or temperature, gain instability over time/temp. Audio parts either don't have any specs along these lines at all, or they tend to have bad specs.

If your application cares about the first class of specs, audio converters are good. If you care about the second class of specs, you'd either have to check parts yourself or just pick parts which do excel in these specs, which are likely not marketed as audio products.


What the marketers think the part is for is irrelevant: what matters is whether the specs meet your requirements. God is in the details. You may, and often should, precede your ADC with some sort of signal conditioning and/or amplification.

The Chandra ACIS instrument uses ADCs marketed for digital audio back in the day. They've been getting the job done in space for almost a quarter of a century.


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.