How do the electrical characteristics of an ADC degrade over lifetime?

Question

How do the electrical characteristics of an ADC typically degrade over lifetime?

And is there a way to do a somewhat realistic worst-case estimate for that before doing aging tests over a big batch?

Example data, probably for begin of life, taken from the PIC16(L)F1855/75 data sheet is:

There are also application notes available what can be compensated away and what is remaining, for the case one does not already know.

However, I never see any information about degradation due to aging.

Also no information how much the characteristics worst-case change over the operational temperature range.

Lifetime of an ADC may be several decades, how should a manufacturer get information about degrading of the ADC in only a few years? — Uwe, Jul 20 '22 at 16:47
Consult thru buyer with OEM for reliability and design validation test reports. There may be Vref and offset vs T data. Also noise is a factor for DNL and linearity — Tony Stewart EE75, Jul 20 '22 at 17:00
@Uwe Via accelerated aging tests, as usually done for assembled PCBAs at least for safety relevant applications. — Torsten Knodt, Jul 20 '22 at 18:34
If it’s crucial that the ADC perform to a certain spec over time try one of my tricks: connect a stable and fully qualified reference to one of the ADC inputs and use that to measure the ADC performance. You may need several to get an idea of linearity and the like. — Bryan, Jul 20 '22 at 18:35
@Bryan This is clear, the question was how the results can be expected to degrade over time ... assuming they do in a magnitude which might need to be considered. — Torsten Knodt, Jul 20 '22 at 18:37
Accelerated aging tests have been done for reliability issues, but are they applicable for ADC degrading over lifetime? Acceleration is done at higher temperature, but is ADC performance at this temperature the same as at the specified temperature range? — Uwe, Jul 20 '22 at 19:25
@Uwe For sure it needs to cool down again before the measurement. At least this is what we are doing in my company for aging tests on system level. — Torsten Knodt, Jul 20 '22 at 20:03
For this chip the ADC accuracy is specified for 25 °C only, I looked for an 10 bit ADC chip from analog devices the AD7810 and found a temperature range from -40 to +105 °C. — Uwe, Jul 21 '22 at 01:30
Indeed, as @Uwe says, external ADCs often run circles around integrated ones. Even when you have these peripherals provided, it sometimes pays to use external ones (e.g.: faster; better accuracy or noise; more bits; more flexible supply/reference; onboard gain/buffer stages; wider temp range; etc.). — Tim Williams, Jul 21 '22 at 02:27
Maybe the OP should explain *why* that is important for him. Also does the ADC exist in "sole isolation", or are there a lot of other circuits around? — U. Windl, Jul 21 '22 at 15:05
@U.Windl Right now it is even in an MCU and around are some RC filtering and scaling circuits. For the case it is imported, also there is some FET control. The intention is to get a feeling whether the ADCs can be expected to be sufficiently accurate also over lifetime. — Torsten Knodt, Jul 21 '22 at 15:17
Voltage references usually have specs for aging and I think you’ll find that limits time between calibrations more than the ADC itself. — Navin, Jul 30 '22 at 20:35
@Navin Thanks for the information regarding the voltage source. I once learned that even, but forgot over the years after university. Unfortunately for the reference voltage source they also only give one value and again only for room temperature. And it seems to be the same also for competitors. This is why I wondered whether there might be something like a rule of thumb I just do not know of. — Torsten Knodt, Jul 31 '22 at 10:27
@TorstenKnodt If you're taking about the ADC's internal voltage reference, it doesn't have any specs because it's crap. You usually have to spend $1 more for a separate voltage reference chip that has specs for ageing. (For really good specs you'll have to spend $100 on the reference and ovenize it) — Navin, Aug 01 '22 at 05:23

score 8 · Accepted Answer · answered Jul 20 '22 at 16:49

I'm not aware of any ADCs that degrade, nor what mechanisms would lead to such (while meeting abs. max. ratings). (But then, ADCs are usually super-secret sauce, and I'm hardly expert in VLSI to begin with.) This is a question for the manufacturer, not random people from the internet.

This may be relevant: http://ww1.microchip.com/downloads/en/DeviceDoc/00000169J.pdf

Offhand, the only indications I know of, regarding reliability, are made explicit: to wit, I've been using the AVR64DA64 lately, and errata show a lower than expected Flash lifetime (100k --> 10k erase cycles), and "The offset of the DAC output buffer can drift over lifetime if the device is powered with the DAC output buffer disabled". No such notes about the ADC.

If you need guaranteed device lifetime, expect to pay for it. Perhaps you can obtain some MIL/aerospace rated parts with more extensive testing, proven operation, and a paper trail to prove it? I'm sure an FAE can help figure something out.

score 6 · Answer 2 · answered Jul 20 '22 at 17:09

It'll either cost real money to get parts where this is known, or you can use some intrinsically accurate self-calibration process. A ramp from an integrator is an excellent way to check ADCs for static linearity errors. As far as scaling goes - try for ratiometric operation if at all possible, then you only deal with resistor drift, if any.

The best way to "deal" with such problems is to design them out.

How do the electrical characteristics of an ADC degrade over lifetime?

2 Answers2