ΣSigmaStudio Ask AI Русский Log in

Question about ADC frontends...

SigmaDSP & SigmaStudio mdot Jul 5, 2011 00:00 6 replies Original on forum ↗
Thread Summary

Thread Summary

The user is designing a prototype PCB for a project using the AD1940 or AD1941 SigmaDSP and AD1938 CODEC. The final answer discusses the impact of using a two-layer PCB instead of a four-layer board, suggesting a star grounding scheme for two-layer boards to maintain performance. It also advises on capacitor selection, preferring ceramic caps for decoupling and poly-caps for coupling, and emphasizes the importance of correct coupling capacitor values to avoid low-frequency response issues. The user considers a modular approach with separate analog and digital daughter boards to isolate signals, and another engineer confirms the importance of proper grounding in this setup.
AI Generated Content
Subscribe
mdot Question
Jul 5, 2011 00:00

Guys,

Not even sure if this is the right forum to ask this question. If not, just point me in the right direction and I'll ask it there.

But, since it has to do with me interfacing to a SigmaDSP, I figured I start here.

After using the evaluation board for the ADAU1701 for a few months now, I've decided that I want to base my little project prototype on either the AD1940 or AD1941 and use a AD1938 CODEC as the analog interface. I figured the easiest way to get up and running would be to base the prototype PCB on one of the evaulation boards, that brought up a few questions:

1.) In the datasheet for the AD1940 evaluation boards, it says that the actual board architechture is four layers with dedicated ground and power planes. Of course, four layer boards can get pretty expensive for prototyping. My question is, does anyone know what the affect on performance would/could by using a two layer PCB instead?

I'm sure there is some app note, or article addressing this line of questions, I just haven't found anything that is both "on subject" and "up-to-date".

2.) The evaluation board has mix of different capacitor with different chemistries. How does one go about deciding which chemistries are used where, in the analog input buffer? Also, why is it that it seems electrolytic caps are only used either as the first component on an analog input, or the last component of an analog output?

3.) Does anyone have any suggestions of how to possibly reduce some of the component count of the ADC frontend without completely ruining soudn quality? Is there such a thing as an ADC frontend IC for audio?

I think that's all I have for now, as if that's not enought, if anyone has any input or, has some references that could point me to, it would be greatly appreciated.

Eventhough I have a degree in Electrical Engineering, I have been writing software since I graduated from college. So, I have never had to design a circuit this complex. This little project is really showing me the difference between what I learned in school and "the real world" of circuit design. *laffs*

ad1940

Audio Signal Processors

ad1941

ad1938

Integrated/Special Purpose A/D Converters

SigmaDSP Audio Processors

Audio CODECs

adau1701

Show More

KKSL Reply
Jul 11, 2011 00:00

Let me try:

1. Dedicated ground planes is an easy way to avoid tracks having to share both analog and digital currents and also to avoid ground loops. As a simple rule, if a circuit has analog and digital circuit on the board, then no track should share analog and digital current.

Star type grounding can be used on two layer boards to achieve this, but one needs to route the board manually and should have some experience on this to keep the critical trace lengths short.

Dedicated power planes also allow for short traces especially the de-cap traces, where the trace inductance will affect the EMC performance.

One thing, however, can be debated: Do we really need a seperate analog ground and digital ground planes? In my opinion, no. I cannot justify the two planes connected by a tie that is properly located. My question is: Will the tie offer lower impedance than a plane? If the answer is no, which is correct answer, then, we should be in a good position to get as good noise performance using a single plane layer for analog and digital ground signals.

What I am saying is that, provided we have a dedicated layer for ground, the analog and digital grounds need not be seperate.

2. Apart for the capacity/ value and tolerance, we need to consider the effective series resistance the capacitor will offer and the life of the capacitor. Lower the ESR, the better. I would avoid series electrolytic caps as they would introduce noise once the electrolyte dries-up with time.

My preferred choice would be ceramic caps for de-caps and filters and and poly-caps for coupling.

3. Coupling capacitors serve a primary function of keeping the signal free from and DC. Audio signal riding on DC can saturate the succeeding stages.

The value of coupling caps should be correctly calculated since it is likely for form a high pass filter with the input impedance of the next element in the signal chain. This high pass filter could become a culprit of reducing your low frequency response.

Whether you want to have a passive ADC front end or a buffered front end depends on the source. If the source is unknown, a buffered front end will be preferred as it will enable maximum signal tansfer from source to load and can also avoid loading of the source.

Note: My views may differ from those who are more learned and experienced. I would encourage you to kindly express your reactions as design comes more with experience than from the books. Agreement will help all of us to follow certain guideline while it could otherwise better it.

KKSL.

mdot Reply
Jul 11, 2011 00:00

Outstanding!

This is my first mixed signal type design so it appears that the considerations are growing exponentially.

I think that I am going to take a "modular" approach to the mixed signal layout. I think that I will use analog and digital "daughter" boards to help with isolation. I figure that if I isolate the CODEC on one board, the only digital signals on the entire board would be the outputs to the digital board containing the SigmaDSP and the supervisory microcontroller. That way, the digital board will be pure digital signals.

What do think about this approach, instead of using the multilayer board with dedicated plane?

ADIApproved Reply
Jul 11, 2011 00:00

Thanks very much KKSL. I agree with all of the points you listed.

ADIApproved Reply
Jul 11, 2011 00:00

mdot, with a modular approach you have to be aware that there are still considerations for grounding that need to be addressed. For example, if you have multiple boards in the unit and there are wires connecting signals from one board to another, there will be a return current for that signal on the chassis of the unit.

A star grounding scheme is a good idea for multiple PCB units. You basically need to pick a singular ground point and make everything referenced to that.

mdot Reply
Jul 11, 2011 00:00

BrettG and KKSL,

You guys have me regretting my choice to pursue embedded software instead of electrical engineering after college less and less...*laffs*

ADIApproved Reply
Jul 11, 2011 00:00

I'm still learning a lot about PCB design. I'm reading a book called "The Circuit Designer's Companion" by Tim Williams and it has been a good reference so far.