Appa 503 Flickering on low range

[Bakaman]

Hi,
On the Appa with a low voltage offset (circa 20mV, all ranges) that I've questioned on another thread, I'm experiencing a low voltage flicker on the display which lasts for some time.
On the lowest voltage range I get something like a +/-1% voltage flicker (between 20-30mV delta).  It is obvious on the display, because it is jumping between say 2.73 to 2.75V once or twice per second, it isn't very good.  But it does seemingly settle down.
Images are a screenshot with time stamp showing the useful software logging of the data against time.  The unstable/flickering and stable sections are pretty obvious.
After being off for 20 mins, I get flickering for 4 mins, before settling.  The second image shows a warm meter where the flickering lasted for 15 seconds.  This might be normal 'warm-up' behaviour for lower voltages/higher precision.  It is annoying that my Fluke or BM867 don't show this sort of characteristic.  The meter has an alleged DC accuracy of 0.03% +20d on this range.  Maybe flicker is an 'acceptable' part of the 20d, and warmup is what I should be learning to live with in this era of instant gratification.
I can live with the flicker as long as I know about it, and as long as it doesn't persist, it is in the low-ish end of the displayed reading range.  If it became constant, then I wouldn't be so happy.
Does anyone have an opinion on this, is it likely to get any worse?  The meter and the software logging is a handy low-ish cost part of my toolset and for an experimental engineer it will come in useful but I don't like the idea of a fault likely to get worse over time.
There is a short video of the flickering and effects of the DC offset her: https://photos.app.goo.gl/H8rEpexEcvgqUd8d8
Thanks in advance.

[Bakaman]

Well, by means of an update, I've been doing some more testing.  The flickering can last for a looooong time after a cold start with the meter, but gradually gets shorter as the meter's off time reduces.

BUT, I found that I can stop the flickering immediately by pressing the Range button to force it out of its Auto AC/DC startup mode (when it has 'TEST' showing in the display).  So, in my case, with a 1.2V battery, I get 1.234 digits by the meter's default display on power up.  and it is the lower two digits that flicker.  RANGE takes it out of Autorange and into manual, and for this battery, the manual range is 1.234, e.g. no different from what it was before pressing the RANGE button, except that the flicker stops, and the word TEST (and AUTOrange) disappear from the display.

This seems to point to me to be a software flicker rather than a hardware noise issue, the meter's startup filtering might be doing something to small signals that is not quite right, but force it to change range with more digits and the problem stops.

An interesting problem, and an example of how software manipulation of values can sometimes give you unexpected results.
And another reminder to me, from my earlier T&M days (40+ years ago, cough) that I shouldn't be surprised when values in the lower part of a range could have significant errors in them (e.g. 0.03% +20digits, is a large error for a small signal.  Drop it down to the mV range, and you get real again.

How tempting it is to search for every increasing digits /resolution in the belief that it gives you higher accuracy, and potentially makes you lazy when choosing the best range for your particular measurement.

[shakalnokturn]

Does it use the ES51966A ?

Not that it will necessarily provide any solution to the problem but it would be nice to know what's going on...
First thing that comes to mind is to compare ADC supply ripple in auto/manual ranging.
If there are measurable  differences in DMM current consumption between auto/manual ranging could point towards poor ground routing.

I'd also be curious of the communication between CPU and ADC in both cases, maybe in autorange the CPU rewrites the range and/or zero calibration register(s) after each conversion without leaving enough settle time before the next sample.

[Bakaman]

Hi, yes it does use that chip (though I haven't taken mine apart, it's 100k count big brother the 505 does).
How would you propose to measure the ADC current, by monitoring the overall battery current draw during operation?  I'll be able to check that out in a few days, I have an assignment that I really must conclude first.
I was intending to do some longer term mains supply voltage datalogging over a few days out of interest (well why wouldn't you?  I can't be the only sad, curious engineer that reads the EEVBlog can I?), and slipped some Energizer NiMH cells in there for the purpose.  This caused the meter to show one bar less on the battery display (unsurprisingly), but I did note that the fluctuations were worse with the lower battery supply voltage.  What's more to the point (with shorted leads) the startup fluctuations were worse with the NiMH, and they didn't disappear with the range change.  On that sort of disappointing news, I went to bed.

[shakalnokturn]

Yes I'd measure overall supply current auto vs. manual ranging. If there is a difference you won't know if the ADC or CPU or both are causing it, but it can give you a clue to think over.

If things are worse on NiMH couldn't it be as simple as poor power supply? (Defective regulator or leaky filter caps?) Definitely scope ADC supply rails if you can.

I don't go to bed until that sort of news is sorted-out or at minimum understood. It makes me pretty inefficient the next day though...