SAR ADC Reference Voltage Buffer Design

[mixed_signals]

Hello All,


I am designing a reference voltage buffer for an asynchronous 10-bit SAR ADC with 50MHz sampling rate.

The settling behavior of the reference buffer (especially at MSB cap switching) affects the linearity (ENOB). I am thinking of two ways:

1- Increase the bandwidth of the buffer by minimizing the load capacitance and maximizing the power to get the highest BW possible to achieve the minimum settling time (even the voltage undershoot is large during the MSB switching).
2- Increase the output capacitance (by adding large on-chip Decaps) that will minimize the voltage undershoot during the switching however it will reduce the achievable BW by moving the non-dominant pole to lower frequency.

So in summary which is better for the settling time for my target resolution: Higher BW but with higher voltage undershoot or Lower BW but with lower voltage undershoot??

Thank you in advance!

 

[KlausST]

Hi,

I'd say both are valid, even a combination of both.
Each solution has it`s benefits and drawbacks.

A big output capacitor (driven by a series resistor) has the benefit of creating no ringing and the max error voltage is C_Load / (C_load + C_output).
So if the output capacitor is 1000x as big as the load capacitor, then you can expect 10 bit performance.
For sure the series resistor needs to be small enough, so that the voltage_drop x average_current is also smaller than the expected error.
It is a rugged solution and does not need high speed OPAMP.

The first solution, may be more accurate and more precise, but it needs a high frequency OPAMP and a careful design regarding stability, not to create ringing.

Klaus

 

[danadakk]

Higher BW :

1) More power needed.
2) More noise generated.
3) Greater EMI possible.
4) More difficult OpAmp compensation due to strays
and phase margin. Leads one to lower Z's = more power.

Tradeoffs, always tradeoffs. Sim to get some guidance.....

 

[mixed_signals]

The first solution, may be more accurate and more precise

Thank you so much for sharing your thoughts.

I did not understand why the first solution is more accurate. Could you please elaborate more on this?

For sure the series resistor needs to be small enough, so that the voltage_drop x average_current is also smaller than the expected error.

What do you mean by the series resistor? I guess you mean the DC output resistance of the buffer in negative feedback (which should be small due to the negative feedback). Am I right?

--- Updated Aug 29, 2024 ---

Higher BW :

1) More power needed.
2) More noise generated.
3) Greater EMI possible.
4) More difficult OpAmp compensation due to strays
and phase margin. Leads one to lower Z's = more power.

Tradeoffs, always tradeoffs. Sim to get some guidance.....

Thank you so much for your response.

From what you say, I may understand that solution 2 is much better only if I have the required area for the Decaps (but no other drawback). All the reasons you give above point to avoiding the higher BW solution, right?

 

[danadakk]

From what you say, I may understand that solution 2 is much better only if I have the required area for the Decaps (but no other drawback). All the reasons you give above point to avoiding the higher BW solution, right?

If you do the sims for your specific situation to give you some guidance....


Regards, Dana.

 

[KlausST]

Hi,

I did not understand why the first solution is more accurate.

We miss the two schematics from your side ... so we had to guess.

Maybe I misinterprete your description, maybe not. I don´t know for sure.
There is a good reason why we use schematics instead of a textual circuit despription: It is rather clear information and we don´t need to guess.
If you need a deeper discussion: please post schematics.

Klaus

 

[mixed_signals]

Hi,

We miss the two schematics from your side ... so we had to guess.

Maybe I misinterprete your description, maybe not. I don´t know for sure.
There is a good reason why we use schematics instead of a textual circuit despription: It is rather clear information and we don´t need to guess.
If you need a deeper discussion: please post schematics.

Klaus

I am afraid that I can't share the schematics because of the confidentiality of the company where I work in.

However, I want a deeper discussion based on what you think from the theoretical and practical point of views that of course reflect your experience in the industry.

Please advise!

 

[KlausST]

Hi,

honestly ... what should be so "confidential"? It should be rather basic and is already done million times.

Don´t get me wrong, I have my job to do .. so I won´t do your job.
I will help, but you have to do the main job.
In case you expect others to do your job ... consdier to pay them.

Let´s compare it with this:
If you want to hike a mountain. I will be happy to help you to solve the difficult parts. But you have to go on your own, I won´t carry you up the mountain.

Klaus

 

[mixed_signals]

Hi,

honestly ... what should be so "confidential"? It should be rather basic and is already done million times.

Don´t get me wrong, I have my job to do .. so I won´t do your job.
I will help, but you have to do the main job.
In case you expect others to do your job ... consdier to pay them.

Let´s compare it with this:
If you want to hike a mountain. I will be happy to help you to solve the difficult parts. But you have to go on your own, I won´t carry you up the mountain.

Klaus

I did not ask you to design the circuits for me.
I have done my job and proposed that there are two solutions and I wanted to know if there are any advantage of one over the other based on people's experience here. However, again I did not ask you to design for me. Also, there is no thing obligatory here to share your experience and knowledge with others, this forum is just about volunteering to share our thoughts and experiences togeather in the field.

 

[KlausST]

Hi,

to show what I mean:
There is this thread where a member gives a textual description.
Then we ask for a sketch/schematic which (s)he provides.
And then we see that there is much of a difference between text and sketch.
It was a waste of time to discuss about the textual description.
The sketch way more showed the real application.

Now with your post, we miss a lot of details a schematic could show .. but it´s hard to describe with text.
(That´s why there are schematics. Imagine one should describe - and another person understand - a textual description of the circuit of a washing machine controller)

We all want to help. But helping should be focussed and not be a waste of time. This would be no good for nobody.

So if we have a better view of what your two "possible solutions" are .. we can give better assistance.

I don´t force you into anything. This is just a well meant recommendation. And only for your benefit.
You decide.

Klaus

 

[mixed_signals]

Hi,

to show what I mean:
There is this thread where a member gives a textual description.
Then we ask for a sketch/schematic which (s)he provides.
And then we see that there is much of a difference between text and sketch.
It was a waste of time to discuss about the textual description.
The sketch way more showed the real application.

Now with your post, we miss a lot of details a schematic could show .. but it´s hard to describe with text.
(That´s why there are schematics. Imagine one should describe - and another person understand - a textual description of the circuit of a washing machine controller)

We all want to help. But helping should be focussed and not be a waste of time. This would be no good for nobody.

So if we have a better view of what your two "possible solutions" are .. we can give better assistance.

I don´t force you into anything. This is just a well meant recommendation. And only for your benefit.
You decide.

Klaus

Thank you so much for your clarification.

Your initial reply totally makes sense and shows that you clearly understand my question and I also got your answer except one point that I asked about more clarification about it to know what you mean. However, I am sorry for any misunderstanding, I will post two sketches that shows the two options later today.

 

[dick_freebird]

ADC input buffer specific parts are a thing, and presumably
they might be as useful on the reference side. Have you
looked at any such, to harvest design cues?

 

[D.A.(Tony)Stewart]

use a BW 20 times the sampling rate or 1GHz BW