1

Consider the ADR421. From the data sheet the initial accuracy is shown to be + or - 3mV of the nominal (2.5V). I understand that if you power on one of these chips with conditions at 25 degrees C, Vin = 5-15V, you will expect output in that initial accuracy range.. But does this mean that the output voltage is expected to change from + or - 3mV if kept at those same conditions?

I know that there are things like long term stability and thermal hysterisis but i'm asking strictly about the initial accuracy.

enter image description here

SamGibson
  • 17,870
  • 5
  • 40
  • 59
BobaJFET
  • 764
  • 4
  • 16

2 Answers2

2

You would expect it to stay at the initial voltage, with a slow change (seconds) as the die heats and a slower change (minutes, hours) as it drifts, give or take the noise.

Typically "long term" drift is much higher initially and tails off, not linear, so 1000 hours may not be indicative of where in time the bulk of the change takes place.

It can also be affected by soldering as Andy mentions, and by mechanical stress (related or not).

These are references that are based on a whiff of electrons on a floating gate so presence of certain types of radiation may also have a profound irreversible effect.

Spehro Pefhany
  • 397,265
  • 22
  • 337
  • 893
  • 1
    If you were to turn off the device, then turn it back on with identical conditions, would we expect the same initial voltage as before? Or could it be different + or - 3mV? – BobaJFET Feb 13 '23 at 16:54
  • @BobaJFET I believe it depends upon things like how long between times when you turn it on. Normally, if accuracy is truly important then plan some burn-in time and calibration steps because, as Spehro mentions, larger changes will take place in the first few hundred hours before settling down into the long term stability specification -- typically gathered by manufacturers after at least 1000 hours -- maybe more. "Fire and forget" with "a hope and a prayer" may not be such good practice. Even given expensive high precision devices with initial accuracy specs that look nice on paper. – periblepsis Feb 13 '23 at 18:46
  • @periblepsis So it sounds like a "set it and forget it" solution doesn't work here. I'm wondering if adjusting the trim with a microcontroller and high resolution DAC would be an appropriate solution to correct the drift. – BobaJFET Feb 13 '23 at 18:54
  • @BobaJFET Maybe not the same, but close. Certainly not +/-3mV. A real situation might be much worse than a laboratory situation with no stress and very little thermal change, but still more like tens of uV than mV. That's an educated guess. I would not be gobsmacked to see a 0.1mV change in a real situation though, especially the first time power was cycled. – Spehro Pefhany Feb 13 '23 at 18:55
  • @BobaJFET well, the problem with autocalibration is that you need access to a much better reference than the one you are calibrating. If you have one built-in, then you probably would just use that one. It's rare you need as complex a solution as a DAC unless your goal is to produce a reference voltage output and even then I don't see why it would have to be particularly high resolution. If it has to adjust +/-3mV then 10 bits is +/-3uV. If it's used as a measurement reference, then you'd just multiply by a calibration correction factor in firmware (or via a trimpot if there's no MCU). – Spehro Pefhany Feb 13 '23 at 19:00
  • @SpehroPefhany In my case I need to adjust the trim on this reference voltage with very small adjustments... About 100 micro-volts per step. So you're saying that the DAC or digital POT would also need an equally precise reference voltage? – BobaJFET Feb 13 '23 at 19:30
  • Not to adjust it, no. Just to know how to adjust it, assuming you're trying to create a certain voltage. Ten bits is probably adequate, and there's no reason not to use the ADR420 itself as the reference for the trim. – Spehro Pefhany Feb 13 '23 at 20:30
0

Note that the initial accuracy of the A version is 0.12% or 1200 ppm. Major parameters that can cause the output to change (drift: 50 ppm, hysteresis: 40 ppm, temp drift:10 ppm/deg, line and load regulation: 35 and 70 ppm) are only a fraction of that value. Since all of the quoted parameters are maximum, and in general may partially cancel each other, you can expect the part will meet its initial accuracy specification for a long time.

Barry
  • 15,967
  • 1
  • 26
  • 28
  • So then the initial accuracy only "guarantees" that the output will stay within that range at those conditions? So if I were to trim the output to "exactly" 2.500V, would it still be reasonable to expect drift within that accuracy range, + or - 3mV at the same conditions? – BobaJFET Feb 13 '23 at 17:06