# Minimum voltage reference for ADC

I am looking to measure very low voltage signal (0-0.5V range) and was wondering if it is possible without amplification. I am a bit confused about what the minimum references of ADC can be. Looking at datasheets, for all ADCs I looked at so far, the range of Vref seems to be from 1.5V. Are references as low as 0.5V not possible? Also, are negative reference voltages possible? Like if I use a bipolar ADC instead which has a negative to positive range for ADC references, can I use a reference of 0.5V then?

• You say no amplification, but in any case you should buffer your signal to normalize the impedance feeding the ADC, right? [normalize, and make it low!] Commented Jun 23, 2021 at 9:28
• Are references as low as 0.5V not possible? Yes, if the ADC was designed for that. are negative reference voltages possible? Yes, if the ADC was designed for that. can I use a reference of 0.5V then? Only if the ADC is designed for that: see the datasheet. Since it is "trivial" to amplify a signal and/or convert it to a positive voltage, few designs need a "low reference voltage ADC" so it might be that they're simply not made. What prevents you from using an opamp based circuit to amplify the voltage just like everyone else does? Commented Jun 23, 2021 at 9:32
• You can also look for an ADC with a built in PGA (programmable gain amplifier). Some even work with single ended signals. Commented Jun 23, 2021 at 9:34
• You also need to consider that most ADC needs to be driven if you have a low impedance input. Commented Jun 23, 2021 at 10:09

Read the datasheet. No, distributors won't catalog this information, and yes you have to look at a lot of datasheets. We do too. That's just how it is.

Most continue to work at low voltages, with corresponding degradation of performance (INL/DNL being due to fixed internal offset voltages / charge injection, thus being relatively more significant at low VREF). Put another way, below the minimum recommended value, you aren't gaining any SNR or ENOB, the reading is equivalent to one taken at VREF(min) and scaled up numerically.

Some ADCs have preamplification available, like, ADS1220 I think is such a case. This has a differential input, so can still be used single-ended by tying one to the common/ground reference (but do mind input voltage range limits). Even some MCUs have features like this, like the XMEGA family of AVRs provide selectable gain, without impairing INL/DNL much (as far as I know).

It is difficult to go under 1.5V of reference due to process limits… for example a typical schottky diode would cause a 0.3V loss; noise is an issue too. Using micropower techniques there are maybe ways to do that but usually you lose some other good quality (like speed or precision); it's always a tradeoff. In fact even the MSP430 (which is a micropower MCU) has the minimum to 1.4V for external Vref (internal is still 1.5V, by the way).

Depending on your bandwidth requirements they make really cheap fixed gain amps: current shunt amplifiers have pretty decent specifications and often cost less than a similar general purpose amp.

Another way is to simply use a lot of effective bits and throw away the upper range of your ADCs (usually not doable with the junky internal ADCs of most MCU)