VC8145 mysterious-ish offset

[adx]

This (see photo).

Just bought one, was about to question its calibration, then I look on the web and every unboxing vid or photo shows them sitting there reading -4 to -10 counts fresh from the box. Spec is 5. Mine does it on the 8V range (shorted or not), and 20A range. mA shows 0. mV range is expected to show a few microvolts out, but the 8V range is 100uV counts (no better than your average 3.5 digit meter). Perhaps I was expecting better, because I have the handheld version of this meter, and it always quickly settles to 0 on VDC, even after 14 years. I know the Fortune chips on these DMMs are not rated up with HP and Keithley, but they do have a competent autozero at 80000 counts, and generally seem to function competently and linearly.

There's a REL mode, but I suspect having 400uV of offset when measuring lower voltages is going to bug me, especially as there's no lower range in the standard input impedance and protection VDC mode (without switching over to uV). Fortunately microvolts is just a button press away (the reason I got it basically - the rotary switch of the handheld is one thing I can't automate through the interface, having sniffed the protocol out a decade or so ago), but still.

I wonder what it is.

[nctnico]

It is not just the counts but also the percentage. 100uV at 8V range is 0.00125% and way better than a 3.5 digit meter.

[adx]

Already worked that out, and wondered if I was being picky. 400uV out of 1V is 0.04%, so yeah maybe. But some people are getting 11 counts, and the mV range craps out at 800mV, which means a measurement would be 0.1375% out.

But that's not my point. It is that you can usually trust a meter to be more linear than it is accurate, and if the offset were ignored, it would appear hugely nonlinear at the bottom end (even though the idea of wilfully ignoring a significant offset is absurd - therein lieth the rub, and my point). Also if you had a 3.5 digit meter that read 001.1 on the 200mV range you'd feel shortchanged. If you modded that to read to 9999 but it sat there reading 0011 for zero, people are going to look at you sideways if you promote that on EEVblog as the best thing since sliced bread. A $5 3.5 digit meter will autozero to 1 part in 2000, a VC8145 that fails to autozero to 1 part in 8000 can hardly be called better.

I suspect people expect high-digit meters to be a bit flighty in the lower digit or possibly two, so it gets automatically overlooked in something like this. mA on mine is a steady 0.

I was thinking contact potential in the relays, that could explain it if it is post or part of the divider, but that should drift with time (mine is a fairly steady -3 to -4 counts). Maybe a circuit offset exists and there was no space left in the micro to compensate? I'm not overly complaining, but it does seem unusual.

[adx]

Well looks like the VC8145 simply has a design flaw, or less likely a factory calibration flaw. Many of them show worse than -5 counts, which is out of spec from the factory. At least I've got a 'good' one.

Simplest workaround I can think of is to hack the reading routine to conditionally add 300 or 400uV to the 8V range, otherwise it will skew test jig data and may make some results gathered near the low end on the 8V range fail typical +- x% limits. In fact one of the tests is to self-verify mV (checks say 20000 microvolts reads 19.9 to 20.1mV on the 8V range), that will definitely fail.

I was a bit frustrated I must admit, I was expecting performance on par with the handheld and justified it on that basis (rather than springing for a Keithley or whatever - if I could get one). I haven't even measured anything yet. But the VC8145 is a good package for the price, with the ability to automate input switching between mV, DCV, ACV and temperature that a handheld definitely can't do.

Now is not the time to dig into the fault but one day I might get in there and see if I can work it out - if I do I'll post back here.

[nctnico]

Simplest workaround I can think of is to hack the reading routine to conditionally add 300 or 400uV to the 8V range, otherwise it will skew test jig data and may make some results gathered near the low end on the 8V range fail typical +- x% limits. In fact one of the tests is to self-verify mV (checks say 20000 microvolts reads 19.9 to 20.1mV on the 8V range), that will definitely fail.

With 20mV input you should use the 800mV range. In the 8V range the lowest reading will be 800mV before you switch to the mV ranges. 400uV is 0.05% of 800mV which is within the specifications.

[adx]

My existing tests use the 80mV and 8V ranges. The comparison it does is between those, so what range it could be on is moot. The mV ranges don't load the input the same as V ranges, and I don't want it switching back and forth when there's weakly driven faults. 400uV of my Vici 8145 is within spec on the 8V range but only 100uV from failing spec (+-5 counts). However many of the fresh units I've seen on the web fail this outright, so this is a fault with the product. Obviously not all that concerning to many, and for some uses (microvolts - need per measurement offset measurement anyway) nor does it matter to me. But as a 4.x digit meter, it should be capable of getting a zero below 0.1mV, as it is on its mA range, as is the handheld one, and indeed any 3.5 digit meter I have ever owned.

[Gyro]

Have you checked the display with the input shorted? Your photo shows the inputs open.

If the display drops to zero then the offset is due to input stage current acting across the 10Meg (?) input resistance. I don't know if the Vici 8145 actually has an input offset current spec.

The situation is universal and on high-end DMMs with high impedance (>10G) inputs. It might be that the Vici has enough offset current to show up across 10Meg.

[adx]

I have checked with it shorted (first post). It does it on the 10A range too, so the 0 reading on mA may be a red herring as far as input leakage / offset current goes. The mV ranges are (as far as I remember) very high impedance - without shorting that, it will continually beep and go overrange unless you stay very still, except being much louder the Vici behaved the same way as my handheld.

BTW that's a Sinometer NB80000R (the old version that still shows up at sicom.co.nz if you do an images search). Although sometimes I've suspected it of being a bit out, it has generally held its cal to within 0.1%-ish in 14 years (-0.064% compared to an LT1236 originally calibrated back in the 90s on a Keithley 2000 is the longest continuous comparison I have). I have never adjusted it, just tracked the drift - in part why I would like to get the Vici reading right before I use it. The Sinometer is extreeeeemly slow to settle on some AC ranges, but other than that, is pretty good (does read 0.0000 on VDC). It has precision approaching a decent bench meter, RS232, and good audio measurement abilities. Same chip as the Vici, and if they're both made by Victor it'd be the same manufacturer (not so sure about that now).

[adx]

Looks like I cocked up that mA zero observation, it reads +8 counts (0.0008mA), I blame low contrast on the 7 seg when viewing from the 6 o clock angle while unboxing and tinkering, but really it was more likely me old eyes (or brain). Amps zero is -4 or -5 counts. So it's all over the show. A DMMcheck read 5.0008V in a cold room, so don't take that as gospel, but it's pretty good in that respect (within spec). 1mA was also within spec.

I'll abandon getting in there and working out what it is now. If it's linear and stable I'll be happy enough for what I bought it for.