Voltage divider on the output of a voltage reference. Can I ?

[Jad.z]

Hi all...

So I'm putting together a 50ksps data acquisition system. For that, I'll be using ADS1271, 24Bit DAC from TI.

This DAC takes 2.5v as its reference. The thing is, I have about 60 or so ADR293E, 5v reference, left over from a power supply project I worked on a while back.
So, question is, is there any good reason why I shouldn't just put a voltage divider at the output of the 5v reference so I could get the needed 2.5v ?



Presuming of course that the resistors are perfectly matched and have the same temperature coefficient.

Thanks 

[olsenn]

This should be okay. Make sure to use large value resistors as those reference IC's aren't typically meant to source much current

[EEVblog]

That is commonly done. But usually as good practice you'd buffer the resistor divider output as shown. But you might be able to get away without a buffer if the resistance is low enough for the ADC (they usually expect a low impedance source).
You'd have to check the specs.
That ref can drive 5mA, so two 560 ohms resistors on the divider is possible.

Dave.

[amspire]

I did some test recently to see how resistors of the same batch tracked each other.  Using 50ppm resistors, I found that the batches I tested tracked in a range up to 5ppm. The average match was probably about 3ppm. What this means is that unless you use special low temp coefficient, you could easily increase your temp coefficient to double or perhaps a bit more.

These were through hole type axial resistors. The situation with SMD may be a bit worse as they may end up with different mechanical stress once they are soldered into the board.

If you do not know the resistors are from the same batch, then you can end up with up to 100ppm difference and that is 30 times worse then the reference IC.

The resistor temp coefficient ratings are not like accuracy ratings where the measured value are near the middle of the tolerance range with only a few near the edges of the tolerance range.  Of the two 50ppm metal film resistor batches I tested, one had a coefficient averaging -45ppm and the other one averaged -49ppm. So a resistor rated at 50ppm/C temp coefficient can easily actually have a 50ppm temp coefficient. Don't assume that they are typically much better then the rating.

10K  metal film resistors with a volt or more across them do start drifting significantly due to heating from the current and this will make them even more reliant on matching temp coefficients. I would stick to over 10K. 100K divider resistors would be much better and with a very low input current opamp like the OPA350 to buffer the reference, the input current would only cause an error of less then 1uV.

If you need the 3ppm stability, I think you better buy new reference ICs.

Richard.

[free_electron]

to make an accurate divider you don't really care about the absolute value. it's the relative value that matters.
Panasonic and susumu have special 4 pin smd resistor in an 0805 like package specially made for this purpose. the resistors are laser trimmed so they match each other. i believe they go doewn to 0.1% (matching , not absolute )  or below

they also have 4 resistor arrays. the center two are matched and the outer two are matched. the outer pair is tuned to be a multiple of the inner pair. this is done to easily make instrumentation amplifiers. you get them with 'weighted' values of 1:1 1:2 1:5 1:10 and others.