Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Voltage reference for ADC - best approach?

Status
Not open for further replies.

Smillsey

Member level 5
Joined
Jul 7, 2014
Messages
91
Helped
0
Reputation
0
Reaction score
1
Trophy points
1,288
Activity points
2,121
Hi all

I am interested as to what your approach is when selecting a voltage reference for your ADC’s?

Imagine you have selected a suitable ADC and it’s Vref Range is max 5V.

Should I find a voltage reference with that exact output?

What if I cannot find a reference with a good enough drift characteristic that exactly match the 5V Vref requirements, what if I want to use a Vref with 7V or 10V output due to their drift performance?

1. Is it bad practice to change the output reference voltage of a fixed reference with an external attenuators circuit placed before the input of the ADC?

2. If this is accepted, I have the following concerns;

a. Need to place capacitor close to Input of adc on order to furnish the adc with inrush current during conversion

b. thermal drift of any attenuators circuit could ruin performance of reference

c. Not sure whether to go for an op amp type attenuation circuit for resistor divider + cap.

Any help would be greatly appreciated

Thanks!
S
 

Hi,

Sadly you don't give values...so it's just a theoretical discussion.

But the answers differ very widly according your application.
While with an audio ADC the exact value isn'very important ... here the noise is more important.
But if you want to design a highly accurate, high resolution measurement device...then it becomes difficult.
Low drift, low noise, high accuracy....
10 bits accuracy is easy
12 bits accuracy ... you have to care about some things
14 bits accuracy ... becomes difficult if you don't already have good knowledge and experience in designing precision analog circuits
16 bits accuracy ... is difficult even for experienced designers
18 bits accuracy ... needs very, very good design, and every resistor, Opamp, reference needs to be chosen very precisely...it becomes expensive
20 bits accuracy is about impossible to achieve....

******
Should I find a voltage reference with that exact output?
Some desingners do this. I say: there is no need to.
Choosing a 4.096V reference instead of a 5.000V reference just reduces performance by max 1.7dB...

Additionally.... often you need Opamps for signal conditioning...if your ADC input goes up to 5V, thdn you need an extra power supply fir the Opamp with higher voltage than 5.0V.
Even with good rail-to-rail Opamp I recommend to leave some margin (supply voltage to output voltage) to avoid clipping and to avoid that the Opamp goes into saturation - even for a short time, like an signal overshot.

1) I don't say bad practice. But often the ADC reference inputs need a very low impedance source ... this usually can't be guaranteed with a voltage divider.

2a) in any case I recommend to use a capacitor very close to the ADC_VRef_input

2b) true, drift needs to be calculated. But if you use good resistors from the same production batch, then they should have very matching tc. But ratio is limited to 1:1, 2:1, 3:1...

2c) see answer to 1). Often you can't do without Opamp.

Klaus
 

Hi Klaus

I appreciate the response, this is s great info. I should have given more details;

As an example, I have been looking at the following ADC:

https://www.analog.com/media/en/technical-documentation/data-sheets/AD7357.pdf

The temp co of the reference is 6ppm/deg C (pretty good)
Long term stability over 1000 hours is 100ppm (not good enough for me)
The Vref max is Ved (2.5V nom)

Imagine, for arguments sake I wanted to use an LT1000 due to the long term stability, or another high end reference that does not have a 2.5V reference output.

For the LT1000 I would need to change the output from 7V to 2.5V...

To do that, I am thinking about a resistive divider, using very low temp co resistors and a suitable cap close to the Input pin of the ADC.

Or should I go for an active circuit with very low impedance output?

The Input current draw of the Vref Input on the ADC is 300uA nominal, which is actually quite high so I am concerned about loading on any attenuation circuit.
 

I don't understand how a 7V reference comes into play. For the said AD7357, an external 2.048, e.g. ADR4520 or LTC6655 would be your first choice.
 

Hi FvM

My wife also tells me I often make no sense at all! Which is also true :)

Your point is of course valid and one would not intentionally look for a 7V reference if you can find a “direct match” with the correct voltage reference output if it meets the drift specs of the application.

I suppose my question is more generic, what if you had to use a reference (due to exotic drift requirements) that only a voltage reference with a “non matching” output voltage would meet?

I wondered if people preferred a resistor divider + cap solution in these very rare cases, an op amp solution, or do they find ADC’s with matching Vref specs?

My example was poor, as you quite rightly say the 6655 would be a good solution.

Maybe I am asking a question that doesn’t need to be answered :-(

Thanks so much for your time though!

Have a nice day.

S
 

It's of course possible to scale a reference voltage. I'm doing this quite often when different reference voltages are derived from a single high precision reference. You need however to consider voltage divider t.c. matching and amplifier respectively reference buffer drift.

Look for matched resistor arrays like Vishay Precision Group DSMZ or Linear Technology LT5400xx.
 

Hi FvM

Thanks so much, I had been looking at matched resistor arrays before.

This makes me feel much better as the route I was taking was loosely in line with yours!

That being said, for my application I feel I may be able to use an off the shelf solution as you suggest, with the correct output voltage matching my ADC. I need to run some numbers to see if the 12 month drift will affect my overall system performance between calibration.

In your experience, how does the 1000Hr drift spec relate to 12 month drift? I take t the 1000Hr spec is simply “time”? Or is it “powered on” time?

This is my first serious measurement design
 


Thanks FvM

Fantastic reference, calibration after 300Hrs should do the trick then.

All the best
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top