Question about voltage references

[Len]

In Dave’s uSupply, he uses an ISL21070 voltage reference for the DACs & ADCs. Is there a reason to use that type of reference instead of a 2-terminal shunt reference like the LM4040?

[Rerouter]

very very low current requirements

[EEVblog]

In Dave’s uSupply, he uses an ISL21070 voltage reference for the DACs & ADCs. Is there a reason to use that type of reference instead of a 2-terminal shunt reference like the LM4040?

IIRC that choice was simply based on cost.

Dave.

[Len]

OK, thanks for the replies.

[c4757p]

very very low current requirements

The difference in current between the two is going to be completely dwarfed by the power actually being supplied, since it is a power supply. Both ISL21070 and LM4040 are specified at tens of uA.

[Len]

The difference in current between the two is going to be completely dwarfed by the power actually being supplied, since it is a power supply. Both ISL21070 and LM4040 are specified at tens of uA.

I was wondering about the difference in general, not just in a power supply. At first I had the same thought, that a few uA doesn't matter even in a low-power device, but then I thought about it some more. Let me know if my thoughts are wrong.

1. It could be hard to get a shunt voltage reference like the LM4040 down to its minimum current draw because you have to make a close estimate of the load and choose the bias resistor precisely.

2. If the load on Vref is shut off (e.g. a microcontroller goes into sleep mode and shuts off its ADC), the shunt reference will continue to draw the same current, except now it's all going through the shunt. On the other hand, the ISL21070 will only draw its quiescent microamps.

So maybe the difference in power consumption would matter in a small battery-powered device.

[c4757p]

1. This is true for both devices - the regulator-style references will take more current if the supply voltage is higher. ISL21070 datasheet confirms this: it is quite linear - a quick lin. regression on the line gives 800k? plus a constant 6.2uA. To get the minimum current draw you'd have to get the supply voltage down near the reference voltage. In both cases you're playing a game of "how close can I cut it before it becomes nonlinear?"

2. True. I think in this case, for either type of reference, I'd power the reference off an output on the microcontroller so it can be shut off.

You are definitely correct about microamps mattering for low power devices. That's why I hate working with low power devices 

...damn it, that question mark should be an Ohm sign.

[Len]

Thanks again.

You are definitely correct about microamps mattering for low power devices. That's why I hate working with low power devices 

Yeah, in the past I had problems trying to maximize the life of a small robot using Arduino-based bits that weren't designed with sleep modes in mind. Dave may be able to fly to the moon on a dozen uA but I sure can't!