Building a 10/7,5/5/2,5V reference power supply - but how do I check it?

[Ysjoelfir]

Hello everyone,

recently I found a AD584JH in a big pile of salvaged stuff from past days. Since I don't have any voltage reference I thought about building a simple switchable voltage reference with this little thing while sticking to the circuit shown in the attached analog devices datasheet, page 6, and the switching chart on the first lines of page 4. Since I would fiddle around with it a bit, I have to check if everything is alright. But - How can I check a self made precision voltage reference, when I do not own any calibrated measuring equipment?
My thought was actually to calibrate my instruments using this thing... hmm.. damn 
What would you do? any advice?

thanks!

[eas]

Well, lets start with what instruments you have?

The AD584JW has an initial accuracy of +- 30 mv on the 10v output, 20, 15 & 7.5 on the 7.5, 5 and 2.5v outputs, respectively. On top of that, tempco is 30ppm/C (ie 300uV on the 10v) and stability in the first 1000 hours of operation is 25ppm.

The simplest thing is to find someone with higher precision and more recently calibrated equipment who will record the actual readings off your finished circuit

If you look over the datasheet in more detail, you'll see that there is also information that would allow you to trim the output to a more accurate value, which will, again require access to better equipment while doing the adjustment.

[Ysjoelfir]

I have many Instruments, but none of them is somehow near to being calibrated. 

- TEL DM1000-CA
- 2x Voltcraft M-3850D
- Voltcraft VC655
- Metravo 464 N digital

Also some Oscilloskopes:

- Hameg HM 512
- Hameg HM 412
- LeCroy 9400A (defective, needs repair)
- Tektronix D10

And some others.

Sadly I don't know anyone around with a calibrated instrument. Even the university I am studying at doesn't have any calibrated instruments.

[babysitter]

I'm at Aschaffenburg and I have some access... and I know somebody even closer. PM me.

[eas]

Before you worry about checking the accuracy of the AD584, I'd suggest working through the uncertainty calculations I started above for both the uncalibrated reference and your meters. It should help inform your path forward.

[Ysjoelfir]

Hi eas,
which calculations do you mean exactly? As you can imagine looking at my mentioned gear above, I never had to work at that "high precision" range. I would love to learn how to get startet with something like this on a professional way - you know, we electronic engineering students nowadays don't learn such stuff anymore As Dave mentioned in EEVblab #10, it is more about learning how to stick a shield into an arduino, what is a pretty sad developement in my opinion.

[eas]

I don't know what they teach electronics engineer students today vs the past, since I wasn't trained as an electronics engineer. My education was in biology, which included a bunch of chemistry, and that is where we dealt at greatest depth with the uncertainty of measurements. Really though, the core question is one of philosophy, specifically, epistemology: How do we know what we (think) we know?

The answer to this question is frequently taken for granted, but it shouldn't be. If you don't take it for granted, it should be clear that it applies whatever instruments you have, and that "high precision" is a vague, if not misleading term. Measurement uncertainties apply whether your meter 3.5 digits, 6.5 digits, or 8.5 digits -- the least significant digit(s) will always be in question.

The other day, before I suggested you finish calculating the uncertainties, I first spent 10 minutes digging up documentation on your instruments so I knew the limits of their precision and accuracy, and how that compared to that of an untrimmed, uncalibrated AD584JH. If you want to follow through with your initial impulse to go beyond the limits of your education while making something you can use to calibrate your meters, you'll need to do the same: look at the specifications of your equipment and understand their implications. I'm sure you can do the first part on your own, and I think you can probably make a good attempt at the second part.

If that's all too abstract, start by getting a fresh, new 9v battery. Measure it with each of your meters on multiple ranges (on my meters, I'd look at it on the 10v and 100v ranges). Do the measurements in different ranges on the same meter agree, what about among your meters?

[Ysjoelfir]

Hi eas and thanks for your explanation. I will go through your advice and write down what I think is the result of my thoughts about this.

[retrolefty]

It's Volts-nuts all the way down. 

[eas]

It's Volts-nuts all the way down. 

There is still hope for him, but its a choice he has to make himself. Best to do it now. Even if he gets sucked in, the step to 5.5 digits is still affordable, and if he'll do Ok if he gets sucked into the next order of magnitude, provided he only gets one 6.5 digit meter.

[Ysjoelfir]

At first, sorry for the long delay, I was to busy and didn't get to look up the datasheets until now.

So... I guess I got what you are up to.

If I check the datasheet for my multimeters I can get the total precision I can get out of those.
For example my "main" Multimeter, the Metex M-3850D (branded Voltcraft but those are manufactured by Metex as I found out) it says for the volts range I am looking at:

Range...........Precision..........Resolution
400mV......+-0,3% +1dgt.......100µV
4V...................".................1mV
40V.................".................10mV
400V..............."..................100mV
1000V......+-0,5%+1dgt........1V

As far as it concerns the "reference-problem", we only look at the 4V and 40V ranges.

The datasheet says, at 4V there is a precision of +- 0,3% + 1dgt. So if I measure the assumed bang on 2.5000000000~ volts from the reference, there could be a reading somewhere between 2.500 +/- 0,076V -> 2.492 to 2.508.
At 5 volts it would have changed ranges to 40V, so the resolution is only 10mV, which means that it could show something between 4.85 to 5.15V. (That sounds like total crap o.o)

Put to a chart this would be like this:

For M-3850
VRef.....Min...........Max
2.5V....2.492V......2.508V
5V.......4.84V.......5.16V
7.5V....7,26V.......7.76V
10V.....9,69V.......10,31V

For VC655 (precision always +- 0,5%+1dgt, resolution as the other one...)
VRef.....Min...........Max
2.5V....2.374V......2.626V
5V.......4.74V.......5.26V
7.5V....7,12V.......7.88V
10V.....9,49V.......10,51V


So lets compare this to the AD584JH in standard configuration without any trimming or compensation at a fixed temperature.
Precision of AD584JH:
VRef........Tollerance........Min.............Max
2,5V...... +-7,5mV.......2,4925V......2,5075V
5V........ +-15mV.........4,985V.......5,015V
7.5V..... +-20mV..........7,48V........7,52V
10V...... +-30mV.........9,97V.........10,03V

If we correlate this now to the "better" of the two multimeters, the MD3850, we see, that the allowed tollerance of the AD584JH is much smaller than the deficit in precision of the multimeter. Which means, I should be able to build this thing up without thinking about any problems at all, as long as the measured reference voltages are within the precision range of the multimeters.
I should have considered the temperature drift of both the AD584JH and the multimeters, too, but as far as I don't have any device to measure more precise, this uncompensated, non temperature controlled device should do the work just fine!

Did I get it right? 

And by the way, does anyone have a datasheet for the Gossen-Metravo 464 N? I would love to see the specs for that one, but couldn't find anything.

[eas]

I think you are on the right track with your conceptual model, however, I think you have some calculation errors, and you might reach different conclusions from the correct numbers.

If you do end up building an untrimmed reference, I'd suggest designing it so you can easily add trim later.

I see that the Metex may have the ability to log data to a computer. That could be an interesting opportunity to explore how the combined system of reference and meter drifts over the space of hours or days.

[Ysjoelfir]

ah... I guess i found the mistake - I forgot a zero for the percentage and calculated most numbers with 3%, not 0,3%... So let's redo this:

For M-3850
VRef.....Min...........Max
2.5V....2.492V......2.508V
5V.......4,96V.......5.02V
7.5V....7,48V.......7,53V
10V.....9,97V.......10,03V

For VC655
VRef.....Min...........Max
2.5V....2.488V......2.513V
5V.......4.98V.......5.03V
7.5V....7,47V.......7.54V
10V.....9,96V.......10,05V


Precision of AD584JH:
VRef........Tollerance........Min.............Max
2,5V...... +-7,5mV.......2,4925V......2,5075V
5V........ +-15mV.........4,985V.......5,015V
7.5V..... +-20mV..........7,48V........7,52V
10V...... +-30mV.........9,97V.........10,03V

If I compare this now I come to the conclusion that the precision of the voltage reference matches nearly perfectly the tollerances of my multimeters, what now sounds actualy quite decent. If i think about the worst case possible, for example the reference being always at the lower precision limit and the meters at the upper limit, I would have a maximum inaccuracy of 19,5mV, 18mV, 6mV and 7mV for 2,5 to 10 volts reference voltage. That sounds quite decent for my deeds.

I actually tried that thing with the three named multimeters and on everyone it was bang one with little to non difference throughout the multimeters, even with the minimum needed configuration (eg just the IC, nothing else...). Guess my worries were unneccessary?


I will build it with the possibility to trim the values, yes. and I guess I would think about some sort of  temperature controll heating-thingy. just for the fun of building.

and yes, the meters have a RS232 connection. I never tried that function though.

[babysitter]

I can do some measurements at your reference, with a few digits more and with some idea about being quite close to the SI volt. You know I am not that far away.  Also, I recommend Kuhne Electronic QH40A crytal heater to stabilize to ~40°C, price about 12€. Consider going to the Weinheimer UKW-Tagung next weekend there is a big electronics flea market (google ist dein Freund). Kuhne and me are there.

[Ysjoelfir]

Yep, I know We talked about that per private message and I will definitely take that oportunity in the future