choosing a suitable ADC for the fully differential amplifier

Hello,

I have designed a fully differential instrumentation amplifie with 3.3 V suing the 0.35 µm CMOS technology.
In the design I have fixed the common mode voltage to 1.65 V.

Now I have the chip from the foundry and I need to interface it with the ADC.

The required ADC specifications is to have at least 12 bit of resolution (preferable 14 bits and above), and has to operate with 3.3 V. My analog signal is up to 1 MHz.

Important specification when I search for the ADC is the samping rate, how much I should go safely?

I have done a rough search and I am puzzled with reagrds the VCM=1.65, I couldn't find ADC to accept this reference voltage and the problem is that I fixed this voltage in my amplifier design. May be I did something wrong in my search and I need your kind help to guide me to find the correct ADC based to my constrains

Thank you

Regards

Hi,

Nyquist says you need more than twice of the maximum frequency of interest.

But I´ve done some undersampling applications, where my sampling rate knowingly is below Nyquist.
As so often: It depends on your application.

Klaus

KlausST said:

Hi,

Nyquist says you need more than twice of the maximum frequency of interest.

But I´ve done some undersampling applications, where my sampling rate knowingly is below Nyquist.
As so often: It depends on your application.

Klaus

Click to expand...

Thank you Klaus, you have answered me the part with regards to the sampling rate requirement, I am planning to go at least for 10 MS/s to relax my filter requirement.

Remaining the other issue which are brings me headache, mostly to find ADC that accept 1.65 V as a reference and work with 3.3 supply voltage

Hi,

VCM is comon mode voltage of a differential system.
REFerence is the DC voltage an ADC refers it´s full scale input voltage on. It determines the voltage level for each LSB.

Klaus

Reference voltage is often different from Vcm, depending on the internal reference amplifier circuit. In so far 1.65V reference sounds like an arbitrary and inappropriate requirement. 1.65V Vcm is accepted by most ADC with 3.3V analog front end. Many types from major manufacturers.

Dear Klaus and FvM

Thank you both for your kind help,

Yes Vref represents the full-scale voltage of the ADC. So I need to have an ADC with Vref=3.3 so my analog signal with a common mode voltage of 1.65 will utilize the full scale range.

I have found many ADC with VDD=3.3 but their internal voltage is fixed to 2 V, which means their optimum common mode voltage is 1 V. If tie myself to 1.65 this implies that I am loosing a lot from the full scale digital code and hence effective resolution.

The other problem is my amplifier are optimized during the design to function with VCM=1.65 V, the performance will not be guaranteed if I change it to a different value, for example, 1 V.

I have searched in Mouser and Digikey but couldn't satisfied with my results, please share with me one example of ADC with my requirement so i can study it and move further.

Thank you once again

Junus2012 said:

Yes Vref represents the full-scale voltage of the ADC. So I need to have an ADC with Vref=3.3 so my analog signal with a common mode voltage of 1.65 will utilize the full scale range.

Click to expand...

This is not necessarily true.
And I don´t recommend it. From my experience it´s not a good idea to go to the very limits of the power supply.
Using the power supply as VRef (in case this was your idea) isn´t a good idea anyways.

Junus2012 said:

I have found many ADC with VDD=3.3 but their internal voltage is fixed to 2 V, which means their optimum common mode voltage is 1 V.

Click to expand...

What does "their internal voltage" exactly mean?
Please give a link to an example datasheet, so we can discuss the specification.

Junus2012 said:

If tie myself to 1.65 this implies that I am loosing a lot from the full scale digital code and hence effective resolution.

Click to expand...

We should do the whole math ... just to see what really happens. (I see many designer they try to use most perfectly 100% of the input range, but forget about distortion, offset, noise, nonlinearities...They better use just 90% and get higher quality results)
I mean you talk about 12 bit or 14 bit. So if you are satisfied with 12 bits you still can use a 14 bit ADC while just using 1/4 of it´s analog input range.

***
Requirement.
Please answer: do you need DC performance of your whole measurement system? (Down to 0 Hz)

Klaus

To FVM's point there are a number of processors out there with 20 bits
and their ADCs with input diff stages. Analog devices 24 bit, Cypress
PSOC 20 bit, TI 24 bit......

The PSOC for example has Vref onboard and options for Vref shown -





Regards, Dana.

Hello,

My application requires DC amplification, so currently can not go for AC coupled solutions.

My problem is is better understood from this article from TexasInstrument


As explained by the document, the fully scale voltage of the ADC is mostly different from the supply voltage rails used for the amplifier, so you believe that if for example working with single supply opeartion of 3.3 with ADC has VF=2 V and differential input properties. This implies that the common mode voltage of the ADC is 2/2 = 1V. Hence it is recommended to use the same value as a common mode voltage for the fully differential amplifier as well.

Back to my question, my fully differential amplifier is fixed in chip design to VDD/2=1.65 V, so I don't have the freedom to adjust it.

If I understood the comment correctly from Klaus, it should be no problem, its only a matter of loosing some bits from the total resolution resolution. is that correct?

Hi,

a very good document! Thanks for sharing the link.

It shows the problem when going to the limits.
Fig4. shows the problem at the low end rail (GND). It causes unlinearities and thus distortion and overtones.

The results in fig5 show exactly what I meant:
While Ch A tries to go to the limit and get all of the possible range .... you get a lot of distortion. You can only rely on 9.4 bits (ENOB).
With CH B the ensure some headroom to GND. Maybe they can´t use the full range but now you can rely on 15 bits
ENOB.
In their case the lose about 100mV from +/-2.048V,
so they can use 3.9V from 4V or about 64000LSBs from 65536LSBs
or 15.96 bits from 16 bits. They lose just 0.06 bits, not "several bits".

You lose a bit of range but gain quality! The ADC results are more reliable.
***

What does this mean for you:
* I understand that you need to use an ADC that can accept VCom of 1.65V
(The alternative is to use a fully differential amplifier (maybe with gain of 1 ) that can adjust VCom.)
* don´t try to go to the limits, maybe reduce gain.

Sadly I can´t help you with chosing a suitable ADC.
I guess ADC manufacturers as well as distributors should provide good selection tools.

Klaus

I have been using many 3.3V ADC with Vcm of 1.65 V, e.g. 12 bit/10 MSPS LTC2223. In so far I can't understand the problem.

Dear Friends,

I am very grateful to your help and nice explanation,

With regards to Klaus solution of using an intermediate fully differential amplifier, I have discussed this option with my team leader, he objected on use another amplifier in front of our DUT amplifier, I understtood his concern, he doesn't want to have a trouble of chaining analog amlifiers in that might affect the measurement of our prototype amplifier.

The second solution from Kluas of not using the full gain of our amplifier is looking interesting to accept. but me personally not able to digest it, forgive my poor knowledge please and let me ask this question which is important for this solution

For ADC with Vref = 2V which means his VCM = 1 V, is it possible to apply a signal that has VCM=1.65 V? I mean there is no DC coupling trouble?

if it is yes, then my signal will be able to swing maximum of 1V-1.65 V = 0.35 V or 0.7 Vp-p, and then as you recommended to use less gain so I will not cross this border


Coming to FvM suggested chip LTC2223, it is indded a perfect suggestion, my leader liked it and he recommended me to order but unfortunately it run out of the market, I have searched it in several distributer like mouser, digikey, farnell and RS components, its out of stock , and by the way it is 80 MS/s , so nice to have.

Junus2012 said:

For ADC with Vref = 2V which means his VCM = 1 V, is it possible to apply a signal that has VCM=1.65 V? I mean there is no DC coupling trouble?

Click to expand...

There is no general answer. It depends on the used ADC and it´s specifications.

Klaus

Yes LTC2223 is 80 MSPS, my fault, useful ADC with 10 MSPS are LTC1225 (12 Bit) and LTC1255 (14 Bit). Availability of the shelf is a different thing of course. Surprizingly, LTC2225 is presently on stock at Digikey and Mouser.

Thank you once again dear Klaus and FvM, I am so grateful to you

I am learning from you solutions alot, even those I am not going for the moment to use but looks very interesting to know,

I found this IC from Analog Devices AD8138, basically, it used as an ADC driver and can solve as a DC coupler with adjustable common mode voltage, if you remember my team leader want to avoide chaining analog amplifiers, but when I looked to the datasheet I impressed by the amplifier bandwidth of 320 MHz at gain = 1.


Here what concern me is that we put an amplifier in front of the ADC, so I think I will loose the advantages of the anti-aliasing filter for filtering in my analogue readout circuitry. Up to my understanding to the noise performance, without having filter, it will be integrated hence to the bandwidth of the amplifier (or limited to the bandwidth of the ADC depending on which is less), therefore I don't feel the advantage of having a filter.

your answer here will help me also to understand the people who put the filter before the amplifier as well.

Thank you

Best Regards

Caution , when you use that amp pay close attention to layout & grounding
section fo datasheet, and bypass caps that you use, their specific ESR and
parasitic R & L. Its a fast amp and a bit of a bear to tame.

BW is both a blessing and a devil...


Regards, Dana.

Thank you Dana for your response, sure what you said is right but that still keep me puzzled with regards the use of the filter, if our analog front ends with amplifier havieng wide bandwidth, what will be the advantague of haveing the filter before.

Up to my understanding the noise will be considered to the last stage before the ADC and will be integrated according to the amplifier bandwidth or might be filtered with the ADC frequency behavoiral depending on which one of them has less bandwidth,

putting a simple RC filter after this amplifier I can understand it, but I feel that I have wasted my 4th order active filter before the amplifier

KlausST said:

Hi,

VCM is comon mode voltage of a differential system.
REFerence is the DC voltage an ADC refers it´s full scale input voltage on. It determines the voltage level for each LSB.

Klaus

Click to expand...

Back to this point, and just to correct it for future readers, the REF might also refer to the common mode voltage which is VFS/2, please see this document:



In particular, go to section 13. by any way ne must read the data sheet of the used ADC to know what exactly mean.

Thank you

Hi,

Junus2012 said:

s, the REF might also refer to the common mode voltage

Click to expand...

I fully agree "might" .. but it is not always true. One has to differ.
The given document exclusively refers to the datasheet of ADC809 on this matter.

Like:
a fully differentail amplifier "might" use VD/2 as VCOM out, it "might" use any other fixed voltage as VCOM or it may have an input to set "VCOM".

or:
The HIGH threshold V_IH of a logic device
* may depend on supply voltage "74HCxxx"
* or may be a fix voltage "74HCTxx"

or:
ADC VRef may
* be fixed to VCC
* or may be independent of VCC

****
So VRef and VCom generally are independent parameters. But there may be ICs that internally "define" VCom with the use of VRef.

****
Thus there can be no other option than:

Junus2012 said:

must read the data sheet of the used ADC to know what exactly mean

Click to expand...

Klaus

Nice to see you Klaus,

Do you have some comments with regards to my posts in #15, #17