Precision Current Source using MAXIM MAX6126

[mTron47]

Hi,
I need a precision current source to calibrate some DMMs. Having read MAXIM's datasheet for their precision voltage reference MAX6126 (https://datasheets.maximintegrated.com/en/ds/MAX6126.pdf I decided to follow their suggestion in Fig. 2 on page 17. To avoid the base current error I used a small MOSFET instead of the BJT. Much to my surprise the current was always 30 microamps too high. These 30 uA "disappeared" into pin 6 OUTS. I expected this pin to be a high-impedance input, but no. Interestingly enough MAXIM says nothing about input impedance or input current on that pin which they call "voltage reference sense output". Strange... Can anyone of you comment on my findings?
Cheers
Peter

[dmendesf]

Seems reasonable. The device has a "not quite high" input impedance at the OUTS pin because it doesn't matter when used as a voltage reference. You can solve that by buffering it with a voltage followers (unitary buffer). I would chose a chopper amplifier.

[tggzzz]

Are you sure you can substitute a FET for a BJT?

30uA in 100uA or 100A?

What makes you think the 30uA went into the sense input?

Could the 30uA offset simply be the consequence of the 0.06% voltage accuracy spec?

[mTron47]

Hi tggzzz

Thanks for your post!

Are you sure you can substitute a FET for a BJT?

Yes, I am pretty sure I can substitute a FET, because the currents to produce are 200uA, 2mA and 20mA. I use an IRLML0030PBF which is good for 1.6A and its turn-on voltage + output voltage (5V) + Dropout voltage are well below the input voltage of the MAX6126.

What makes you think the 30uA went into the sense input?

My microampere meter  I cut open the link between the MOSFET source and pin 6 of the MAX and true enough, there were the 30.6 uA.

What kind of astonishes me is that MAXIM suggests an application example and forgets to mention that current leak. Not to talk about its reproducibility, precision, long-term stability etc. 30 uA offset in a 200uA current source does not really fulfill the claim of "ultra-high precision" as the datasheet's title implies. (Just my 2 cents)

Peter

[mTron47]

Hi dmendesf

Thank you for your post, which kind of confirms what I suspected. I will attempt to solve the problem by inserting a OPA189 as a "zero offset" voltage follower between the MOSFET source and the OUTS pin of the MAX6126 - just as you suggested.

Nevertheless the fact that a sense input (sorry, MAXIN says "sense output") is not very high impedance astonishes me. But well, problem solvable...

Peter

[MasterT]

Confusing:

OUTF, OUTS, NR.......-0.3V to the lesser of (V IN + 0.3V) or +6V

page 2. 
Does it mean, that IC can't drive OUTF above 6V, in case OUTS is grounded and IN is high enough 12V or so?
It would imply BJT only buffer, MOSFET with 6 - 5 = 1V G-S operational voltage hard to find.

[magic]

Yeah, taken literally that would mean that the output might be capable of killing itself

What happens to the OUTS current if you short G-S, i.e. configure the reference for normal 5V operation fully within specs? 30µA seems plenty lousy TBH...

[mawyatt]

Hi dmendesf

Thank you for your post, which kind of confirms what I suspected. I will attempt to solve the problem by inserting a OPA189 as a "zero offset" voltage follower between the MOSFET source and the OUTS pin of the MAX6126 - just as you suggested.

Nevertheless the fact that a sense input (sorry, MAXIN says "sense output") is not very high impedance astonishes me. But well, problem solvable...

Peter

Hi Peter,

Been reading this with some interest, but never used the MAX6126. Agree the ~30ua omission is poor documentation!!

If you are going to the trouble of adding a precision op-amp, why not use the op-amp as the core of a standard type current source with a bipolar or MOS device and leave the  MAX6126 as a standard voltage reference?

Best,

Mike

[mawyatt]

Peter,

Here's a DVM and Oscilloscope calibrator I've been working on. The current source is in the upper left schematic corner.

Please note R5 should be 10K, not 1K.

Best,

Mike

[Kleinstein]

When building a precision current source, one should watch the power dissipation of the precision resistors. Especially the small form factor ones should be used with only low power. Otherwise they would heat up and this changes the resistance. So the voltage over the reference resistor may have to be smaller - more like 1 V or even less for the higher currents.

[mawyatt]

Yes good point.

If your referring to the schematic posted, that's why there are five 1206 resistors to make a 100 ohm (100ma), and a 1206 for 1k (10ma), the 10K (1ma), 100K (0.1ma) and 1Meg (10ua) are 0805 types.

A lower reference voltage becomes a tradeoff with the errors due to IR drop in the board, various jumper & contacts and also the offset error of the op-amp. With a 10 volt reference these effects are reduced and why the decision to go with 10 volts from either ADR01 or scaled LM399.

Also tried to stay with parts that I had or weren't expensive, since this isn't intended for serious use. The PCB hasn't been fabricated, since another version is in the works that includes an oscilloscope calibrator but does away with the high-precision voltage divider in the upper right and removes the LTC1043 to keep the PCB size down.


Best,

[Kleinstein]

10 mA at 10 V is still 100 mW. This would not burn an 1206 resistor, but would still give quite some temperature rise and not good for precision use.
One has to remember that the power rating for SMD resistor is with a significant board area (e.g. 1-4 square inch). So 5 x 1206 close together would not have 5 times the power rating - more like 1.5 times.

Reasonable modern OPs have a noise and drift level well below the LM399. So one could use a voltage considerably lower than 7 V to produce the current sources. I would consider something like 1/2 or 1/4 of the 7 V as a suitable level. Ideally one would have the jumpers in a way that there resistance is not part of the reference resistance.
Even with some 2 V the 100 mA would be 200 mW and for high precision this would call for resistors with a rated power of > 5 W. This would be more like a TO220 case resistor with heat sink.

[mawyatt]

Good points, thanks!!

Think I'll abandon the 100ma scale and just go with 10ma with in the PCB design underway.

Best,

[mTron47]

Hi - and sorry for my long absence! 

@mawyatt: Thank you for your suggestions! As a reply to your post here is the schematic of my DMM-calibrator which has been completed in the meantime and calibrated against a 7 1/2 digits Keithley 2010. It works satisfactorily for my purpose: calibration-check of 3 3/4 digits hand-held DMMs.

[ Specified attachment is not available ]

@Kleinstein: You certainly have a point mentioning the error source of power dissipation in precision resistors. In the schematic above you see that the most critical resistors are R3 and R9 in the 20 mA path, each dissipating some 50 mW in a 1206 size. Be it what it is, the calibration exercise showed an error of 0.0167% during a measuring period of some 20 seconds. This is a factor of 9 worse than the precision of the 200uA output, but for me is acceptable.

[David Hess]

What kind of astonishes me is that MAXIM suggests an application example and forgets to mention that current leak. Not to talk about its reproducibility, precision, long-term stability etc. 30 uA offset in a 200uA current source does not really fulfill the claim of "ultra-high precision" as the datasheet's title implies. (Just my 2 cents)

It is the sort of thing which I expect of companies which do not even provide a simplified schematic and function diagram.  Maxim is particularly bad about it but Texas Instruments often does it as well.  I deliberately avoid their parts where possible.

[mTron47]

@David Hess
 I deliberately avoid their parts where possible.

My opinion, too. Maxim often provide really nifty, clever circuits - but they are so unique that there is no second-source. For me that's a clear no-go for series production. The situation is different if you want to solve a specific problem, one-off style, and if you are prepared for some "surprises" as in the case discussed here.