Another short episode from the Fun with TEA dept:
I've connected this reference board (indeed this is five references in parallel to reduce noise and average the nominal 5V) to the four HP34401As:
This is the result over a few hours with less ambient temperature change than yesterdays experiment:
One vertical division equals 1ppm (5uV out of 5V). So they're all agree within 17ppm of their reading, while the average is about 20ppm from the nominal 5V. Pretty good for non-adjusted 5V reference chips.
DMM #3 (blue) has more noise than the others. It's also got an older firmware than the others, don't know if this correlates.
DMM #1 (green) shows some jumps of 1 ... 2ppm at some points, so does DMM #2 (yellow) but less amplitude.
DMM #4 (red) just looks fine.
All curves drift roughly the same direction and amount, so I'd attribute this to the 5V references, not the DMMs.
In yesterdays experiment, the green curve showed these jumps, and the blue one was more noisy - same as here, so it's the DMMs property.
Vertical division equals 1ppm here, too.
BTW
The logged data has more digits (seven) than the HP34401A shows (6 1/2) and is specified for (5 1/2). They default to show 5 1/2 digits after power on, you can set them to show 6 1/2 digits. IMO these plots show the 6 1/2 digits are indeed useful with these instruments.