Down-ranging always occurs at a fixed / digital count of about < 5000 counts.
Up-ranging occurs at random counts. In DCV mode, that's about > 53125 counts, in Ohm modes it's about > 53300 counts.
In the first place, I tripped about other contributions here, which reported >56000 counts, or so.
Up-ranging is therefore initiated by an over range condition, not by a fixed upper digital A/D value, like in other DMMs (e.g. 1 200 000 ... counts for all HPAK bench DMMs).
Obviously that over-range detection inside the BM869 is realized analogously by a window comparator (OpAmp) which probably monitors directly the input signal over an additional input resistor.
I think what we’re seeing actually is digital switching points for the Up-ranging after all.
What gives the impression of somewhat diffuse switching points that doesn’t seem to occur at the same value each time - is caused by noise and/or low (lower) resolution from the ADC.
What I meant earlier by “dynamic filtering” should rather be called “two stage filtering”. 121GW uses this too - the first stage is the HY3131 Sinc2 filter which is set to give 40 SPS output rate and is used for the Bargraph. The same value is then filtered by the STM32 MCU but with a simple Sinc1 stage and this gives the display output rate of 5 SPS. And 121GW uses this 5 SPS rate + some delays for the auto range which makes it rather slow.
I believe BM869 is configured in a similar manner - but here the first stage filter gives 80 SPS for the Bargraph but this value is then also used for the auto range and will fluctuate just like the Bargraph does when subjected to interference. Then for the second stage filter BM869 likely uses a Sinc4 (or higher) as it has quite noticeably better noise rejection than i.e. 121GW.
It’s also possible the auto range has its own filter settings but one that’s still not nearly as effective as for the final value shown on the display - all to make the auto range faster. But doing it this way the display will never show the exact value that actually caused the range to change.
If BM869 is in fact configured like this it should also be fixable in firmware. It quite strange though that current FW allow auto range to act this fast and hunt back and forth between two ranges without some kind of damping routine i.e. stopping the hunt if it changed back and forth several times within a couple of seconds.