EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: SingedFingers on February 19, 2017, 10:13:02 am
-
I am currently using a 'hope for the best' approach calibrating my power supplies and meters. This isn't working as one of the supplies is drifting off and two of the meters have had extensive repairs. So I need a voltage reference!
I have a friend with a Keysight 6.5 calibrated meter I can check any reference against periodically and get a reading off.
I am currently looking at a an LM399 reference with a precision divider (1% low tempco), buffered by LTC1052. Plan would be to build it, run it for 3 months non stop off a bench supply to age it, get it checked by him yearly and then check my kit every few months.
To be clear this is just to check a 20,000 count meter with 0.02% +/- 1 count spec best case for amateur use. It doesn't have to calibrate Jesus' sandals or anything. Main use case is to ensure I am actually checking everything against a reasonable opinion of what a voltage is and that all my kit has the same opinion.
Are there better options? Am I heading in the right direction.
BTW I do not want to become a voltnut!
-
BTW I do not want to become a voltnut!
Yeah, yeah, they all say that at the beginning. ;D
You probably want to go and browse around the metrology section for a while, there are plenty of current threads on references of various degrees of precision and cost.
Depending on your voltage needs, (10V is a good starting point) I'd probably pick up an AD587 or REF102. They have reasonably good initial accuracies (depending on grade) and are easilly trimmed. Something you could put together on a bit of veroboard. (Edit: both are burried zener based).
The easy bit is done - you have access to a calibrated meter (or sufficient accuracy for your needs) so you won't have a problem with calibrating or periodic checking. One tip.. build it and run it for 200-300 hours before initial calibration, that's where most of the initial drift happens. Ceramic packages are best for long term stability.
-
Thanks for the tips. Much appreciated. I will investigate further.
I'm avoiding the metrology section as I end up with all sorts of expensive things I probably don't need appearing in my Farnell basket :)
-
How much work do you actually do that needs absolute* voltage measurement accuracy better than 1 part in 1000? Give three examples and how often you perform those tasks.
If the answer is less frequently than monthly, your idea indicates impending voltnuttery. :scared: If its more frequently than weekly and you have already rejected the idea of getting a good quality new meter and keeping it calibrated, with a current certificate, you are already suffering from voltnuttery and unless you are highly technically competent, and are willing to accept the voltnuttery lifestyle, should probably consult a psychiatrist! 8)
* Not relative accuracy of measurements taken in same session
-
That's a shame, There is a thread running that you might find interesting in the metrology section. It's about a very simple 10 volt reference. It does use one expensive part, a 1n829 zener, probably 15 dollars or so. Your choice if you dare to look at it or not :)
https://www.eevblog.com/forum/metrology/old-fashioned-zener-10v-reference/ (https://www.eevblog.com/forum/metrology/old-fashioned-zener-10v-reference/)
-
I saw that. The 1n829a is also still available. But then you need a current source and buffer for it and then the next thing you know you've got six keithleys, a voltage calibrator and a bending shelf on the bench :)
I'm basically after a DMMcheck but don't want to pay for it.
I've picked a suitable off the shelf design from the REF102 datasheet (thanks Gyro!) and will use that.
I have got some old buried reference based LM723's in metal cans floating around. This was idea one for ref, using the built in pass transistor and opamp as a thermal regulator ... then sanity resumed.
For ref, cases where this will be used:
1. When my old Thurlby PL drifts by 500mV on the metering suddenly because it's dirty.
2. When I'm palming old kit off via ebay I can say with confidence if it's a lemon or not.
3. When I am taking differential measurements on two meters which is generally a bad idea but difficult with out of phase AC with a single meter.
4. When I get a new meter (see TEA and Fluke 8024B thread) and want a quick confidence check against a known voltage.
5. When I've fixed my PM2524 switches for the 305th time to make sure I haven't made it worse :)
-
One thing you need to realise is that there is a difference between accuracy and stability. (There are other considerations, but they can wait for now). The VoltNuts are typically more concerned about stability, whereas "normal people" are more concerned about accuracy.
VoltNuts are quite happy if their 10V reference is actually 9.51035378V when calibrated against Josephson Junction references - provided it stays 9.51035378V. If their own meter shows it to be 10.0V, they can either change their meter or do the arithmetic. Similar considerations apply to meters, of course.
I suspect you are more interested in absolute accuracy. In that case there are inexpensive voltage reference ics that have an initial accuracy of around 1mV in 5V. They are easily found on Farnell/RS/etc, and there are a range of voltages from 1V to 10V. If you need your PSU to be more "accurate" than that, then why are you using a PSU as a reference in the first place? :)
You are right to note that references will have to be aged over at least a week, and preferably more.
-
Aware of accuracy vs stability. I've picked a 2.5mV initial accuracy REF102. I'm going to leave this untrimmed and get a reading of a cal'ed meter then use that to adjust mine to that. Will age it on a bench supply first :)
The main thing is everything needs to agree across three supplies and four meters. I'm not using the PSU as a reference but I can't really trust anything I have at the moment.
I really am just making an electronic bit of paper with a number on it that I can take to the meters and get them to agree with between that and zero.
Absolute accuracy and agreement at FSD is close enough then.
At the moment I'm making an assumption that my PM2524 is closest to reality which is a bad idea as it's shonky.
-
I have a spare saturated Weston Standard Cell, but obviously I can't sell it on fleabay and can't use a courier :)
PM me if interested.
-
Pass that man a thermometer - preferably Mercury too! :)
I still need to enclose Muirhead K-375-C in a large chunk of Aluminium, with a suitable thermometer too.
(Don't forget measuring impedance too).
Edit: Too, too many 'toos' :palm:
-
The one I'm prepared to release did have a bent thermometer, but it has been broken. Surprisingly the cell seems to have survived, 1.01853V @ 16.5C, according to my HP3468A. (An unofficial cal of that indicated it was reading 9.9985V at 10V, or 0.017% low).
-
The thermal mass of mine is too low to make consistent measurements at the moment. I just can't face sawing through the big Aluminium billet I've got at the moment. Those sensitive bent thermometers are sadly rare these days (I'm still looking!).
-
If a fresh dry cell battery isn't a good enough check for you, you are already a voltnut. If the problem can't be cured with a cheap voltage reference from Ebay you actually have a moderately advanced case.
-
Don't give up on the Metrology section. There was a related thread recently. Check it out: https://www.eevblog.com/forum/metrology/best-out-of-the-box-10v-reference/?all (https://www.eevblog.com/forum/metrology/best-out-of-the-box-10v-reference/?all)
-
Four forays into the metrology section and we automatically issue you an ID card and register you as a voltnutter in the national database.
If you just want to buy a good IC reference, which sounds plenty sufficient to your needs, that last link given should do the trick.
-
Ok built, checked and meters checked against it. Used the REF102 and built a board for it. This is straight from the datasheet with the additional noise reduction capacitor installed.
Capacitors are underneath and are relatively expensive axial tantalums. There is a 1n914 (idiot diode) in series with the supply to stop accidents costing a new ref IC. Uses a star topology on the ground pin of the IC.
Has been checked against a friend's new and calibrated keysight meter (didnt get the model number). It's going to be attached to a bench supply and left for a month to age as well and then will be re-measured.
Thanks for everyone's help. Now to avoid voltnuttery :)
(http://i.imgur.com/P4ttzut.jpg)