Band - gap reference for ADC

[dipnirvana]

Band-gap ref for ADC

I have a micro which has band-gap reference of 1.2V . I am powering up the ADC with 3.3V and i want to give reference to my ADC around 3.3V.

Do you think i can use band-gap ref. and use a voltyage follower to make it ~3.3V . Or u may suggest me some other technique ?

Last resort is voltage reference diodes !

Help .

 

[dick_freebird]

Band-gap ref for ADC

You really need to look at how good that reference
has to be. A CMOS microprocessor bandgap may be
just a "free, good enough" one - not to mention the
likely high frequency noise. Every impairment will
find its way into the data.

Maybe you should do a match-up between the "free"
one and a decent, acquisition-grade reference. You
may find some out there that have some sort of
ADC-driver output and are scalable (one with a
pin-strapped, guaranteed-accuracy gain setting may
eliminate the additional errors associated with an op
amp lashup).

What would a simple, filtered 3.3V do for / to you?
Back again to the question, how good does it have to
be?

 

[PaloAlto]

Band-gap ref for ADC

Not sure I understand, is it just a reference or a power supply?

In other words, does that 3.3V node to the ADC has average current consumption?

 

[dipnirvana]

Band-gap ref for ADC

Thanks for replying.

I am talking of voltage reference to ADC . Now i would be using 14bit of ADC. and as per my calculation 0.003% is the requierment of initial tolerance and less than 1ppm is required for the temco as temprature range is -20 to 70degC.

I hope i have done the calculations right!

to add more into it , solution should be cheaper,too.

 

[PaloAlto]

Re: Band-gap ref for ADC

First of all: If your supply is also 3.3V, then it may be complicated to get the desired 3.3 reference for the ADC, maybe you can get 3.1V or so. and if you need 3.3V when your supply is 3.3V-10%, then you have a much bigger problem. If your supply is larger than 3.3V, then forget about all this.

Now, about the error. It is true, for 14bits, you need errors in the order of 100uv or less. That is definitively a challenge for a voltage reference and a buffer (which is probably what you need). But, I think, and I'm not sure here that you may have some sort of calibration to remove offset errors, so you first have to know which contributions are really critical for performance and then take decisions about architecture.

So, first create the right specs, then do the rest.

 

[swolf]

Re: Band-gap ref for ADC

I have the same confusion as dipnirvana, It seems as if the calculation is not wrong.

My present opion is: first, it seems impossible to design such a bandgap to us, even to the big companies such as ADI (deduced from their product datasheets).
secondly, we often do not use the full scale input signals, so it brings no harm when the bandgap ref moves a little.

How do you think so ?

 

[dick_freebird]

Band-gap ref for ADC

Yes, you need a phenomenally good reference to not
have it affect the digitization. But it's my observation
that a lot of people specify much better conversion precision
than the application really needs, and accuracy may be
well less.

For example if you have a 1% advertised sensor tolerance
then you simply do not need all 24 bits of sigma-delta
converter resolution. Not to mention the ambient noise
and such, which goes more to my point of mistrusting
a bandgap built into a microprocessor.