High Stability Current Source

[EPAIII]

I wish to build (or buy) a current source suitable for the calibration of analog meters in the range of +/-1% accuracy. Therefore it does not need to be excessively accurate; 0.01% would be great and 0.1% would be acceptable. I am aware that I may and probably will need to use an outside calibration facility for the initial adjustment of this device and also for periodic re-certification. Or perhaps not if I can obtain known valid Voltage and resistance sources.

And let's assume that I can enclose it in a temperature controlled container that is capable of keeping it at +/-0.1 degree F or better. The actual temperature would probably be a bit above ambient for my shop which does have both heat and cooling so the temperature there remains near 70 F year round. I would probably aim for about 75 to 80 degrees F (24 to 27 degrees C) to keep power consumption reasonable.

What I need is some suggestions as to how to proceed. An internet search yields many results, but that is part of my problem as 99+% of them are aimed at powering LEDs where a fairly large variation in the initial and over time current values would be acceptable. Almost any Voltage regulator can be turned into a current regulator with a single resistor added from the output to the common terminals. But I have to wonder about the initial accuracy and longer term stability of the common regulator chips. Any suggestions as to which one(s) may be better for this?

Or, what other devices or circuit ideas should I consider?

Or, are there any commercial products that may fit my needs at a reasonable price, which in my case would be a few hundred US$. I am willing to consider used instruments.

[Cerebus]

Go and find manuals for older bench constant current supplies, ones old enough to have full schematics in and study how they did it. Keithley and HP/Harrison/Yokagawa are good brands to start looking for manuals for, they all produced comprehensive and comprehensible documentation; they also all made precision constant current sources. I have learned a lot about precision electronic instrument design from doing just that.

A scan of KO4BB will probably tun up lots of material: http://www.ko4bb.com/getsimple/index.php?id=manuals

[pqass]

I'm no expert but if I was to DIY,
I'd use a good voltage reference and turn that into a current source; not a jellybean voltage regulator.

For example, I just happened to have purchased a few AD4530s last year (it's a 3.0V vref). Its initial accuracy is ±0.02% (600uV) with a TC of 2 ppm/°C. With quality opamp and passives, you should be able to make a current source that meets your 0.1% accuracy requirement.   You may not need to oven-ize it if the chosen components have enough initial accuracy and TC range at room temp ± say 10°C.

What current ranges do you need/want? 

Google comes back with lots for "precision current source". 
My first hit is from TI; app note SNOAA46 "Precision Current Sources and Sinks Using Voltage References"

ADDITIONALLY:
Do you need to dial-in a specific current? or, are just a few discrete currents required?  In the latter case, it could be a simple project box with one vref, with an opamp, transistor, and shunt per banana jack for each discrete source and a common. In the former case, it would mean a DAC, MCU, UI, etc.

Can you get by with just a current sink? or, do you also need to provide the "compliance" voltage (ie battery) too? 

[bdunham7]

The current supply basically just needs to be stable and then you control it somehow.  Is this a once-in-a-while deal or a continuous need?  When I need to do something that is outside of the range of my calibrator, I simply use a good linear CC/CV PSU and an accurate (enough) DMM in series.  Set a reasonable compliance voltage on the CV adjustment (4 volts is typically plenty) and then dial in the current you want.  Most decent linear PSUs will have short term stability of the 0.1% you need, if not your DMM will tell you.

[Kleinstein]

A normal lab supply, suitable (possibly high power) resistor and a DMM with sufiicient accuracy would normally do. Most lab supplies are good for 0.1 % stability, at least in voltage mode. A series resistor would convert that to a current. Directly in CC mode the stability of the usual lab supplies may not be so great, especially not at the lower ranges.

It also depends on the current range in question:  some nA to 100 mA are relatively easy.  > 10 A it can become more of a challange due to heat. The pA range does not have many analog meters. 

[David Hess]

The common operational amplifier and transistor current source or sink will easily deliver accuracy without adjustment commensurate with the accuracy of the current sense resistor and voltage reference.

What level of current were you interested in?

[Dr. Frank]

Hello,
when I needed several precision current sources for thermometry, @ university over 30 years ago, it quickly turned out to be a problem.
Those on the market e.g. from Time Electronics, Keithley, or others were much too expensive and not that great in terms of stability and uncertainty either.
Also, it was not necessary to have several digits of resolution on these instruments, only some fixed currents of each decade.

So I built several cheap (switchable 10µA, 100µA, 1mA) and stuffed the whole physics institute.

Here's a more versatile one, which I redesigned some years ago:

https://www.eevblog.com/forum/metrology/recycling-of-precision-current-source-noise-reduction-for-low-burden-shunts/msg1432410/#msg1432410

Frank

[SiliconWizard]

If only a few fixed values of current are alright, this IC is pretty cool: REF200.
It's not the most accurate or stable source/sink you can build, of course, but it will beat the pants of any *simple* opamp-based one.
 ±0.5%, ±25 ppm/°C, 2.5 V to 40 V

[bdunham7]

Another thought for a cheap and easy current reference in the 0.5uA to 10mA range is that most decent bench DMMs use a fixed precision current source in their resistance ranges.  You need to measure the current with an accurate meter because on closed-case-calibration instrument the currents will be very precise and stable, but not exactly the nominal value. 

[David Hess]

If only a few fixed values of current are alright, this IC is pretty cool: REF200.
It's not the most accurate or stable source/sink you can build, of course, but it will beat the pants of any *simple* opamp-based one.
 ±0.5%, ±25 ppm/°C, 2.5 V to 40 V

I do not see how it could beat a "simple" operational amplifier based one.  Errors will be limited to the reference and shunt resistance so no-trim accuracy can be 0.1% or better, and the operational amplifier will not suffer from self heating when the compliance voltage changes.  The later is solved in the REF200 by adding a cascode transistor.

[EPAIII]

I looked at the spec. sheet for the REF200 and saw that 0.25% is the typical accuracy, but 0.1% is the guaranteed value.

I may have to settle for that, but would like to get it a little better if possible without breaking the bank. It's just a piggy-bank, not Ft. Knox.

But thanks for the discussion and I am going to try to look at some schematics as Cerebus suggested. I am sure this is a doable thing.

[magic]

Save for exotics like REF200, you would normally use a voltage reference and a resistor.
If you can get these two in satisfactory precision (the tolerances add up, of course) then you are set.

Note that even cheaper grades of reference ICs can be adjusted given a trusted reference, but their thermal drift is (almost) inescapable.

BTW, a day has passed and still no information about output current requirements

[Kleinstein]

The ref200 is lmited to a small current range ( 100 or 200 µA). It this fits it can be a simple solution, but AFAIK there is no easy way to trim the current.

The more logical way is a combination of voltage reference and resistor and than use the resistor value to trim to the exact value wanted. To check  analog meters one may want a fixed values for the end and maybe ideally also a few values like  1/4, 1/2 and 3/4 in between.

If really wanted, there is a way around the temperature effect, but it adds extra effort and costs.

[Cerebus]

By the way, the standard first book on this subject is "CURRENT SOURCES & VOLTAGE REFERENCES" by Linden T. Harrison, pub. Newnes.

[EPAIII]

"...output current requirements"?

I guess I am guilty of ignoring or at least delaying an answer on that. This is probably because I am in the process of what could be called initial thinking on this and I do not have any firm numbers in mind. But I do know that actual requirements are needed for good suggestions.

As I said, this would be for checking and/or calibrating ANALOG meters, mostly VOMs. These instruments typically have ranges that go from perhaps 50 or 100 uAmps up to 10 or 20 Amps full scale. So the question then becomes, just what is needed for that task. The lowest value, which is more or less the naked sensitivity of the meter movement itself or with just one or two shunt and/or series resistors used for calibration of that movement, is probably a definite yes. So I would want one or several outputs in the uAmp range. 10, 20, 50, 100, etc. would be possible values but perhaps only one would be needed for any particular VOM.

Then there are the mA ranges and, with most meters, finally a 10 A range. It would be nice to have one output available for each of these, probably full scale which would almost always be a "1" value as opposed to "2" or "5" or other values. So, 1, 10, 100 mA and 1 and 10 A. A "25" value may also be useful as many meter scales do use numbers based on that (25 mA, 250 mA, etc.).

If I build something, I would probably use multiple copies of one or more simple circuits, each producing a different current. The selection of the current would then be accomplished with the selection of an output terminal post (banana or miniature banana posts) with one post for each available current. This avoids any problems due to actual switch contacts and gold plated terminal posts and banana plugs can avoid much there.

With such a wide range of currents, uAmps to tens of Amps, I may need to use two or more different circuits.

I am not at the present time thinking about the clamp-on meters that go up to 250, 500, 1K Amps and even more. There must be some reasonable limit on this project.

Whatever I do, I will probably use a box and construction technique that will allow for future changes and/or additions. Probably even space for multiple power supplies as 10 Amps is going to be a lot different than 1 mA. I will probably proceed in stages unless everything is so dirt cheap that I can just incorporate dozens of outputs in the first version; after successfully bread-boarding it anyway.

This is my current thinking. But I am open to any and all suggestions. I am not decided on any of the details and that includes the exact number or current values of the output(s).

BTW, someone said that the REF200 was not easily adjustable to other values. I only spent about a half hour looking at the spec sheet, but it does seem to be. Numbers like 50 and 200 mA were mentioned there multiple times and there apparently was some reference to other numbers being possible. I need to read it again, with more time available to mull it over and do some calculations. But somehow I do not think it will be able to provide a 10 A current or even a 1 A one. So, again, different circuits may be, probably will be needed for different current ranges.

And BTW, happy new year to all. I am hearing the fireworks starting as I type.

Save for exotics like REF200, you would normally use a voltage reference and a resistor.
If you can get these two in satisfactory precision (the tolerances add up, of course) then you are set.

Note that even cheaper grades of reference ICs can be adjusted given a trusted reference, but their thermal drift is (almost) inescapable.

BTW, a day has passed and still no information about output current requirements

[SiliconWizard]

The ref200 is lmited to a small current range ( 100 or 200 µA). It this fits it can be a simple solution, but AFAIK there is no easy way to trim the current.

Yep. You can arrange a small number of values of current with it, but it's pretty limited (it has 2 100 uA floating current sources + an accurate current mirror, so a few combinations are possible.) Of course, you can also use it as a current reference for a more adjustable current source, but then the benefit of using it as far as stability and accuracy goes would start to be questionable. But, I think it's still a nice part to have as a current reference.

The more logical way is a combination of voltage reference and resistor and than use the resistor value to trim to the exact value wanted. To check  analog meters one may want a fixed values for the end and maybe ideally also a few values like  1/4, 1/2 and 3/4 in between.

If really wanted, there is a way around the temperature effect, but it adds extra effort and costs.

The whole point is that, if you want something that doesn't have to be trimmed, is accurate, stable and doesn't drift too much (especially with temperature), things start becoming a bit more involved indeed.

[EPAIII]

Yes, "more involved indeed". I guess that is why I am asking the question here. There seems to be a number of pretty savvy folks on this BB.

The ref200 is lmited to a small current range ( 100 or 200 µA). It this fits it can be a simple solution, but AFAIK there is no easy way to trim the current.

Yep. You can arrange a small number of values of current with it, but it's pretty limited (it has 2 100 uA floating current sources + an accurate current mirror, so a few combinations are possible.) Of course, you can also use it as a current reference for a more adjustable current source, but then the benefit of using it as far as stability and accuracy goes would start to be questionable. But, I think it's still a nice part to have as a current reference.

The more logical way is a combination of voltage reference and resistor and than use the resistor value to trim to the exact value wanted. To check  analog meters one may want a fixed values for the end and maybe ideally also a few values like  1/4, 1/2 and 3/4 in between.

If really wanted, there is a way around the temperature effect, but it adds extra effort and costs.

The whole point is that, if you want something that doesn't have to be trimmed, is accurate, stable and doesn't drift too much (especially with temperature), things start becoming a bit more involved indeed.

[Kleinstein]

Some trimming would likely be needed and not so difficult. It still needs some care to keep the trim range small and maybe do the coarse part with selected resistors. Resistors in the 0.1 % range (especially for higher power) get expensive and rare. Measuring the current to 0.1% with a modern DMM is usually not so difficult.

It would most likely be at least 2 ranges: the small currents up to some 1 or 10 mA  can be with battery power and use a bit higher voltage at the critical resistor.  The range of highest currents (1 A to 10 A) would be quite some power and thus likely mains supply and to keep the power low also a much smaller voltage at the critical resistor. The heating of the resistor is definitely a point to whatch, already at 10 mA.

For a current source the contact resistance is not ciritical and a combined ciruit could provide multiple currents.
There is no big problem switching the shunt in a constant current sink circuit, at least for a few ranges. The switching may also be with using different terminals. So one could have 1 regulating circuit with terminals for lets say 10 A, 1A, 100 mA and maybe 10 mA and than a 2 nd circuit with terminals for maybe 10 mA,  1 mA, 100 µA and maybe 10 µA. An additional switch could chose something like 100%,50%, 25% to also get some settings in between.
The 0.1 % accuracy level is still not too tricky as 0.1% of 100 mV as a reasonable shunt voltage for the higher currents is still 100 µV and thus well within the accuracy of precision OPs.

Due to the low needed voltage (1 V would be sufficient for the high currents) one could even use battery power for the higher currents. The regulating circuit would than need a separate supply.

[SiliconWizard]

I'm sorry, but the OP needs to tell us, or at least reflect on, the following point:
"I wish to build (or buy) a current source suitable for the calibration of analog meter"

Basically, they want to build a calibrator. I the calibrator itself needs to be calibrated, then (at least as an end-user) how useful would that be really? And, if they own what it takes to calibrate the calibrator, why would they need to build one?

So, I suggested a simple part, not too expensive, that doesn't need any calibration. (But sure, with limited values of current and, not quite as accurate as they were apparently looking for.)

But I really don't know otherwise what the point would be. So yeah, you can think about buying a commercial one, such as this: http://www.extech.com/products/412300A (which should fit your requirements without being too expensive.)
(and have it calibrated every couple years in a specialized lab.)

[David Hess]

If I build something, I would probably use multiple copies of one or more simple circuits, each producing a different current. The selection of the current would then be accomplished with the selection of an output terminal post (banana or miniature banana posts) with one post for each available current. This avoids any problems due to actual switch contacts and gold plated terminal posts and banana plugs can avoid much there.

At low currents, the same circuit with the non-inverting current configuration can cover 1, 2, and 5 with 1, 10, and 100 multipliers.  With some care, the number of trims can be minimized by for instance making the 1, 2, and 5 stages use multiples of the same resistor in series and parallel combinations.

This also makes calibration much simpler because each range, or multiple ranges, may only need one calibration.

With such a wide range of currents, uAmps to tens of Amps, I may need to use two or more different circuits.

At high currents a lower value shunt resistance is required so a separate circuit.  Anything else adds more complexity than having two separate circuits.

[Terry Bites]

 If you want a traceable calibration unit, get a source meter. SMU

[SiliconWizard]

Well, if the OP owns a good precision multimeter, and wants to be able to calibrate other multimeters, I admit that building their own current source could make sense: that's the part not usually available (at least directly) in multimeters and would require buying specific equipment. But in that case, they don't even need to build anything that stable or accurate, as long as it's stable enough for the time it takes to calibrate what they want calibrated with it. They'd just need to trim the source using their reference multimeter just before using it. And in that case, I agree with what some others have said: you can just build a simple source around an opamp and a transistor.

Now if you want long-term stability and no need to calibrate the thing at least for one or two years in a row, that's another territory. But unless you want to build a commercial solution, or do calibrate a large number of multimeters as part of your daily activity, the above should do the trick.

[EPAIII]

I can see some food for thought here. Use a high accuracy digital meter to set an adjustable source and then just use it. For a double check, I could put it on the digital meter after checking the analog one. Hummm!

It occurs to me that I could even have the two meters, digital and analog, in series so the exact same current flows in both of them. Little or no chance for error there. And that could work for both AC and DC. No need for precision resistors even. Just one or more stable Voltage sources and some good quality pots or switchable resistors. Hummm!

It is the end result that is important, not the exact method I wind up using. And it may just give me the justification to buy a nice, new digital meter. Win, win, WIN!

OK, now someone shoot holes in that. It can't be perfect, can it? I am already drooling over the idea of a new meter.

BTW, I thought I had fully admitted above that I was in the preliminary stages of this project while my hands are occupied with actually building other projects. I am in the process of thinking it out and this discussion is helping a lot. My sincerest thanks to all.

[Kleinstein]

For DC things may be about that easy.  For the higher currents (e.g. 10 A or 1 A) the extra drop from the DMM may be to high to have to have 2 meters in series from a low supply (e.g. 1.2 V NiMH cell). A constant current regulating circuit may still be easier than a high power adjustable resistor. So even with the meter in series to measure, the constant current circuit may still be easier than just voltage source and an adjustabel resistor.

For AC things get a bit more complicated, as there is not such easy AC source and the mains wavefrom may not be good enough. So AC could be quite a bit more tricke. There is also no such simple circuit to stabilize an AC current.

[daniellemill]

I had a similar problem too! I personally use a high precision digital meter to set the regulated source! And this is enough for me!