Different readings for an AD586 voltage reference

[giosif]

Hi,

I am playing around with an AD586 5V voltage reference (the M version) and I am seeing something I find very strange: different multimeters give considerably different readings for the reference output
* Gossen 26M - 4.924V
* Greenlee DM-810A - 5.000V
* Fluke 85III - 4.93V
* Fluke 87V - 4.936V
* Fluke 8842A - 4.9996V
So, given the specs of the voltage reference, it would seem that only the Greenlee and the Fluke 8842A are providing proper readings.
However, if I set my power supply to 5V and check that with all my meters, they agree on the actual value, so it's not that the meters are out of spec or something like that.
The next thing I thought of was input impedance on the meters but, with the exception of the Fluke 8842A, all others have roughly the same 10 Mohm impedance value.
Also, changing the imput impedance on the Fluke 87V to ~1 Gohm does not change the readings from the voltage reference output.

Could someone please help me understand why I see these differences?

[retrolefty]

considerably different readings

 And someone might say they are insignificantly different. Which of those readings are outside the published tolerance of each meter?

[c4757p]

considerably different readings

 And someone might say they are insignificantly different. Which of those readings are outside the published tolerance of each meter?

However, if I set my power supply to 5V and check that with all my meters, they agree on the actual value, so it's not that the meters are out of spec or something like that.

No retro, something else going on here. Can you be more specific about the exact test setup?

[giosif]

No retro, something else going on here. Can you be more specific about the exact test setup?

Yeah, realized later I should've mentioned the setup, which is the most basic one: power supply + AD586 IC.
Nothing else (i.e. I'm reading the output voltage straight from the IC).

[michaeliv]

Yeah, realized later I should've mentioned the setup, which is the most basic one: power supply + AD586 IC.
Nothing else (i.e. I'm reading the output voltage straight from the IC).

What power supply are you using for both cases + what is the input voltage in both cases ?
Are you consistently getting the same readings + are the readings stable ?
Do you have multiple AD586 that are doing the same things ?
Try loading the reference with a 50k load and see if you get the same results.

[Kalvin]

What happens if you hook the Fluke 8842A to the voltage reference output and then measure the output voltage one by one using the other multimeters. Does other meters affect Fluke 8842A reading too? Can you hook multiple multimeters to the voltage reference output at the same time and how the readings differ then?

[Wirehead]

Do you have the additional noise reduction cap connected? Add a 1uF film cap from pin 8 to GND. It's probably noise that your DMM's are picking up.

[giosif]

Apologies for the delay in replying!
Happy New Year, BTW! 

What power supply are you using for both cases + what is the input voltage in both cases ?

I'm using the same bench (regulated) power supply for all setups.
The input voltage is the same 15V for all setups.

Are you consistently getting the same readings + are the readings stable ?

Readings are stable per power cycle and, apparently, the same across power cycles.
Although, today, I am getting a different reading compared to before (although, still consistent during and across subsequent power cycles): 4.975V with the Gossen meter.

Do you have multiple AD586 that are doing the same things ?

I have another AD586, but I have doubts the current one is faulty, since the behaviour is different across different multimeters, not with a given meter.
And I'm trying to find out what difference between these meters can explain the difference in readings.

Try loading the reference with a 50k load and see if you get the same results.

I might try this, but what's the objective with this (since some difference in meters influences the readings)?

What happens if you hook the Fluke 8842A to the voltage reference output and then measure the output voltage one by one using the other multimeters. Does other meters affect Fluke 8842A reading too? Can you hook multiple multimeters to the voltage reference output at the same time and how the readings differ then?

I tried this and the interesting thing is that connecting any of the other meters at the same time as the Fluke 8842A did not affect the readings of the latter.
It's quite the opposite, actually: the Fluke 8842A seems to impact the readings of the other meters (the other meters showed 5.000V).
Any ideas why?

Do you have the additional noise reduction cap connected? Add a 1uF film cap from pin 8 to GND. It's probably noise that your DMM's are picking up.

No noise reduction cap.
I am aware of the function of pin 8, but I'm not convinced it's noise we're dealing with here (might be wrong, though).
The reason I'm not convinced: I've added an LT1001 op-amp connected as a buffer to the output of the AD586 and guess what reading I'm getting from the output of the op-amp (with all my meters): 5.000V.
If there was noise on the output of AD586, I would expect that to have been propagated by the op-amp, right?
Also, the fact that using a buffer on the output seems to make the issue (i.e. difference in readings) go away, I am inclined to think the problem is with load introduced by the meters...
But, again, all my meters in the list above have an input impedance of ~10Mohm...

[Kleinstein]

I see two possibile explainations:
It could be the reference circuit is reacting the HF signals (e.g. AM radio) coming from the meters or just the cable.
It could also be just the capacitive loading that could make the output unstable - some OPs get upset with just 100pF of capacitance. Many meters together with the capbles might provide about that much.

[Andreas]

Hello,

Most probably the bench power supply is a switchmode supply which generates RF spikes which are rectified within the reference.
Try a battery supply (e.g. 2 9V Blocks).
Try proper decoupling capacitors at input + Output of the reference.
(100nF + 10uF electrolytic each time).

With best regards

Andreas

[giosif]

Hello,

Most probably the bench power supply is a switchmode supply which generates RF spikes which are rectified within the reference.
Try a battery supply (e.g. 2 9V Blocks).
Try proper decoupling capacitors at input + Output of the reference.
(100nF + 10uF electrolytic each time).

With best regards

Andreas

I stand corrected: it is indeed noise introduced by the power supply.
Using either 2 x 9V batteries or a decoupling capacitor on the output made the issue go away.
Thank you all for the answers!

[Andreas]

Hello,

thanks for the feedback.
The AD586(LQ) is one of my favourate references.

with best regards

Andreas

[retiredcaps]

I stand corrected: it is indeed noise introduced by the power supply.
Using either 2 x 9V batteries or a decoupling capacitor on the output made the issue go away.

Glad you got it figured out.  With that lineup in post #1, there was pretty much no way they would be out relative to each other by that much.