Probably a one of these.
Tim
-- Seven Transistor Labs Electrical Engineering Consultation Website: http://seventransistorlabs.com
Not so good at DC though...
The OP could try a Hamon divider.
High-end multimeters are 24+ bits.
For AC I guess one of those ultra-low-distortion sine source ideas? And FFT.
-- John Devereux
I own several Gertsch AC ratio boxes, but I wouldn't know how to use them to test a 24-bit ADC. They are great for LVDT and synchro work.
I can't recall any lectures about AC ratio boxes. I doubt any of my profs ever heard of them. How did Sloman get to be such an ass?
It's still what was used by the only guy I know who had tackled the problem seriously. Standards labs use their ratio transformers at quite low frequencies. There are lots of ways that you could use them at DC, but I don't know what was actually done.
It's still resistors. The windings in properly constructed ratio transformer present rather more stable and predictable ratios.
So?
That does seem to be the idea.
-- Bill Sloman, Sydney
20Ad%20Free.pdf
patience.
ous values.
lip the string, and remeasure. Something like that.
pretty good.
Neither do I, but I could probably find out. When I last heard, the ADC cal ibrator guy was still working at Cambridge Instruments (or whatever it is c alled now), and appreciative of the nice things I'd said about him to some of my Dutch contacts.
s ever heard of them. How did Sloman get to be such an ass?
How would John Larkin know what his professors had heard about? And the fac t that John Larkin can't recall any lectures that mentioned ratio transform ers could reflect that fact that he'd skipped those particular lectures, or hadn't managed to get his head around anything transformer-based at that s tage in his career.
And how does my knowing stuff that John Larkin doesn't make me an ass? It's a rather self-damaging insult, if you think about it.
-- Bill Sloman, Sydney
Get? Was there change somewhere in his life time?
Jamie
This more practical:q
Great but you have not shown us how to do it near DC (like, sub-hertz, the only way you are going to get 24 bits out of a D-S).
A Fluke 720A KVD is stable and ratiometric to 0.1ppm, few hundred dollars off ebay, ~$1k if you can't wait. Needs a high impedance chopamp type buffer here. I think someone already posted the LT app notes but AN86 has a lot of lore about precision DC.
.
It would help the OP verify the performance directly, at least at DC. The short-term stable adjustable voltage reference is left as an exercise for the student.
-- John Devereux
Jamie fails to recognise that he really is an ass, and prone to make serious errors of judgement.
John Larkin probably has enough sense to recognise that I'm not an ass, but he does lash out when his vanity is injured, which happens quite a lot, to the extent that it is fun to provoke him.
-- Bill Sloman, Sydney
ormer present rather more stable and predictable ratios.
e only way you are going to get 24 bits out of a D-S).
No. You'll have to find that out for yourself. The only thing I can tell yo u is that it has been done.
I posted
which was about testing 18-bit ADCs. There are better versions of the Wien bridge oscillator around. As I said earlier the National Semiconductor/Texa s Instruments LME49710 integrated circuit operational amplifier introduces quite a lot less distortion than the Linear Technology parts that Jim Will iams was constrained to use.
and their bank manager. You can buy a Josephson-junction based system which is hard to beat - and hard to pay for.
The liquid helium supply and refrigerator are extra.
-- Bill Sloman, Sydney
No it's not that bad, only needs to be short-term stable and does not need to be an absolute standard like a JJ. Just stable long enough for a meter to measure, and be compared to the ADC output.
A precision PWM source should also work. There was a nice thread about ppm-accurate precision PWM, evolving from a "SMD TC" thread a couple of years ago.
Don't know how to link to it properly with google groups but could try this:
(If that turns out to embed my amazon credit card login please buy me a nice book).
-- John Devereux
At Datron, building a calibrator for their 7-digit bench-top A/D converters.
I'd forgotten the name of the company, but John Larkin just posted a link to an old thread, which dates back to the time before I'd forgotten.
-- Bill Sloman, Sydney
rote:
0Ad%20Free.pdfatience.
us values.
ip the string, and remeasure. Something like that.
ef) so no external artifacts are needed to do calibration. It's like balanc ing weights.
retty good.
use ratio transformers. If John had paid more attention to his lecturers, or what he might have read here more recently, he'd have known about them.
I've now remembered that I did once work out a technique for getting very p recise DC outputs from a transformer, back around 1975.
It was a three-winding transformer, one winding to carry lots of current, one over which to measure the DC output, whose output you used to control t he - rising - current in the high current winding to get the desired output voltage, and a third winding which generated the same DC voltage in a diff erent context - in my case on the other side of a galvanic isolation barrie r.
You had to monitor the current going through the high-current winding, and when it got high enough to make the winding warm or bring the core close to saturation, you disconnected your outputs, and drove the coil hard with th e maximum practical voltage of the opposite sign, and sustained that voltag e until the current current through the high-current coil had not only drop ped back to zero but actually reversed and got a high in the opposite direc tion as it had been when you started the switch-over.
The nice thing about this approach was that you didn't have to worry about the rapidly changing voltage drop in the driven coil - the feedback system dealt with that.
Once you'd got there, you resumed generating the desired DC voltage across the sense winding, and re-connected your outputs, and kept going.
With a ratio transformer, you'd use the feedback to develop a voltage equa l to your reference voltage across all the identical segments in your bundl ed winding, and take the outputs to be digitised from the different taps al l the way down the bundle. The driven winding would be machine wound onto t he toroid and your bundled ratio windings would be hand-wound on top of tha t.
You might be able to get away with the idea I'd had back then, but have nev er been able to test, which was to make the "bundled" winding out of round- to-flat ribbon cable, with IDC connectors on either end .
The most labour intensive part of making a traditional ratio transformer is matching each wire sticking out of one end of the bundle to the correspond ing wire sticking out of the other end.
The winding density of ribbon cable sucks, but with this approach none of these wires is carrying any significant current, so you couldn't care less.
You do want as many turns as you can fit in to allow you to sustain a given voltage for a long as possible before you have to got to sample and hold o r start over, but a big ferromagnetic toroid would get you off to a good st art.
I used a relatively small ferrite core in my device, and could sustain a co nstant DC voltage form 2V to 10V for a few msec - it was to be part of 1-5m A industrial control system.
For calibration work, you'd probably used a big iron- or mu-metal-cored tor oid, and could sustain the voltage for quite a bit longer - long enough to use reed relays in your sample and hold circuits, if you needed them.
I don't know if anybody has used this approach, but it ought to work - we g ot remarkably good results out of my circuit. I tested it with a 6-digit DV M and couldn't see any errors at all, despite the fact that I was using che ap op amps to do the work. The amplifier that drove my high current coil wa s a 748, and I got the recovery voltage by pulling the compensation pin dow n to the negative rail.
Very crude, but very cheap. I'd left Cambridge Instruments before the idea got turned into a real product, and the friend of mine that did that did ma ke some changes - mainly better sample-and-holds, IIRR. He now lives in Per th, Australia, and we still exchange e-mails every year or so.
-- Bill Sloman, Sydney
Can't you just switch the resister (1m, 10m, 1ohm, etc) for constant max voltage sense error, say with relays, automatically if necessary. It seems that you probably only need 1% actual accuracy? Looking for 1uv is daft, when not required.
Figure out what you are actually trying to do, and solve *that* problem. Getting people to solve "how do you test a 24 bit ADC" is the wrong problem.
Kevin Aylward B.Sc.
There is a limit, but only in terms of F(accuracy, speed, power) = 0
I do some hand waving to come up with:
Po > (f^2/error).h/2.pi
Kevin Aylward B.Sc.
spoilsport :)
-- John Devereux
If it's a conventional resistive shunt, the shunt itself will dominate the error budget, so much that the ADC doesn't matter. The shunt will self-heat, and a good manganin shunt will have a tc in the 10s of PPM/degC. Thermoelectrics and 60 Hz pickup could well be worse than that.
On the low end, 1 mA across 1 mohm is 1 microvolt, hard to measure with any accuracy.
Common-mode voltages will make this even harder.
DCCTs are great, fully isolated and super accurate, but expensive. If
100 amps made 10 volts, 1 mA would be 100 uV, a lot more reasonable.-- John Larkin Highland Technology, Inc jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
24bit converter? Does this require expensive, calibrated, super precision e quipment, or is there a simpler (cheaper) way to do this?
was wondering if there was an easier way, maybe this is just not an easy t hing.
100A. Taking a reading across a 1milliohm shunt.
any I'm getting.
gain stages to get my range.
the error budget, so much that the ADC doesn't matter. The shunt will self
-heat, and a good manganin shunt will have a tc in the 10s of PPM/degC. Th ermoelectrics and 60 Hz pickup could well be worse than that.
ny accuracy.
100 amps made 10 volts, 1 mA would be 100 uV, a lot more reasonable.DCCT's are indeed the cat's pyjamas. The less ambitious can use Hall effect sensors.
The problems of wanting to measure microvolts can be deal with more simply by using a pair of make-before-break change-over contactors to reroute the current through the shunt in the opposite direction from time to time.
It may be easier to use four contractors in a bridge - two are closed to ro ute the current through the shunt from right-to-left, then all four are clo sed - briefly, to get the make before break action - then the first two are opened so the second two route the current through the shunt in opposite d irections.
DC offsets and thermocouple effects cancel out, and microvolts become measu rable (though not all that easily).
-- Bill Sloman, Sydney
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.