Powering ADC off a reference

What voltage?

How accurate and stable?

--

John Larkin         Highland Technology, Inc 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

It's probably not a good idea, if you want measurements good to 16 bits.

Whassamatta with a separate reference and linear regulator?

--
www.wescottdesign.com

A precision reference and a linear regulator work in much the same way, but one is optimised for accuracy and stability, the other for current capacity and efficiency (along with price, size, etc.). As long as the current requirements of the ADC are well within the ratings of the reference, I don't see a problem with using it as a supply.

The only risk is that a reference might not cope so well with changes in the load, which could be large if the ADC has a sleep mode between samples - I guess some capacitors would be the answer to that.

Of course it is fine if you want a ratiometric measurement (like a bridge transducer). All depends on whether OP wants absolute or relative measurements.

--

John Devereux

I'm not sure about the "of course" if you want 16 bits of accuracy and the ADC has any memory (which a sigma-delta does) and if the power supply is bouncing around.

It's the OP's choice, though.

--

Tim Wescott 
Wescott Design Services 
http://www.wescottdesign.com

Should be fine, many of the d-s application notes have the reference connected to the positive supply, it seems to be a common practice for bridge transducers. E.g. see the OP's LTC2240 fig 17 and 18 (p23). d-s ADCs are very forgiving in my experience, compared to all other types. The OP likely does not even need a "reference" chip at all just a regulator. Having said that I would want to put a couple of filter stages in to take care of that power supply.

--

John Devereux

Also, the OP wanted to power the ADC from a reference - but that does not necessarily mean that the ADC's voltage reference input will be connected to its power supply input. I imagine he wants to use a shunt reference as a stable power supply, and then would have either an independent voltage reference device or at least some filtering between the power supply and the ADC voltage reference input.

I have also noticed that delta-sigmas can be very stable. I have seen a card with a switch mode supply for 5V, which is used to power a D-S ADC with only a ferrite bead and a capacitor for filtering, and another bead and capacitor for the ADC's reference input - and that ADC is resolving accurately to 20 bits. I would expect that putting a shunt reference between the switch mode supply and the ADC's supply would increase the effective number of bits significantly.

(However, I am not an analogue expert - so if I am wrong, I am happy to be corrected.)

rt

s

ve

the

s

It isn't a good idea to take application notes as examples of good design p ractice. I've got a lot of faith in Analog Devices and their application no tes, but they lead me astray once. I've yet to catch Linear Technology in a similar screw-up, but I'm sure that it could happen

I'd certainly want to isolate the reference input from the varying current drain at the A/D converter positive power input.

--
Bill Sloman, Sydney

Yeah that's kinda what I was thinking, I'm considering just setting up reference with a pass transistor to supply the extra current.

Thanks all for all the input!

Yeah that's kinda what I was thinking, I'm considering just setting up reference with a pass transistor to supply the extra current.

Thanks all for all the input!

If you are going to add a part to supply the current for powering the ADC, why not just use that to power the ADC but still drive the ADC reference input directly from the reference output?

--

Rick

could also be used to power an ADC. I have an LTC2440 (that's the part I have to use) it's a 24bit sigma delta, but I only need about 16bits from it.

ripple.

It sounds really stupid to me. The operation of the ADC will disrupt (corrupt the quality) of the reference so bad that you will lose half the bits of the ADC. Of course it does depend on which power pins, but the principle is valid.

?-(

This guy is getting 20bits out of his, and is running the ADC off the references

Me too. Connecting the Vref input directly to Vcc is dumb. A resistor and a second low-leakage electrolytic capacitor would be a better approach, and a inductor plus enough damping resistance to kill the Q of the LC would be even better.

Only a bare-bones miminalist would skip the two or three extra components. Steve Luce thinks he's getting 20-bit performance, but there are lots of ways of fooling yourself in this kind of work.

--
Bill Sloman, Sydney

I'm not interested enough to read the whole thing, but near the bottom he shows what he describes as "SPECIAL NOTE: A NEW NOISE MEASUREMENT

better than 13 ENOB, not 20.

"Based on the floor noise of +/- 10 and the maximum peak height of +/-

100,000, the signal to noise ratio of this system with a maximum signal present is about 10,000."

His ADC is a very low speed, low power device designed for such apps as weight scales and the like.

The reference he used is a MAX6250 rated for 15 mA of current. So that will power your ADC. But the circuity of the ADC will inject noise on the power rail which will show up on the reference which will affect your measurement. Better to use separate supplies for the reference input and the VCC input. You can use the same part if you want, just use two of them.

--

Rick

could also be used to power an ADC. I have an LTC2440 (that's the part I have to use) it's a 24bit sigma delta, but I only need about 16bits from it.

100mV ripple.

references?

the

the

references

%20ADC.html

You coudn't get me to do that. I am not sure i believe it either.

?-/

Do you believe LTC? See the LTC2400 datasheet.

Essentially all of the example circuits tie VREF to VCC. Even the demo board used to verify performance (Fig42).

datasheet:

AN78: shows many example circuits with Vcc tied to VRef, it is the norm rather than the exception, expecially for ratiometric operation.

--

John Devereux

Application circuits should work but can't even always be relied on to be that good. They don't have to work as well as they might. They certainly can't be relied as examples of "best practice".

--
Bill Sloman, Sydney

I would power the reference input from a reference IC or LDO (depending on whether absolute or ratiometric measurements are needed).

I would probably use an RC from that to the Vcc "power" input, but even that would just be superstition on my part.

The OP just wanted to know if it was OK to power the ADC from a reference at all. A lot of people are saying no don't do that. It is certainly worth trying IMO.

--

John Devereux