The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Mixed-Signal Design >> Randomizer in Thermometric DACs
https://designers-guide.org/forum/YaBB.pl?num=1277444367

Message started by Mayank on Jun 24^th, 2010, 10:39pm

Title: Randomizer in Thermometric DACs
Post by Mayank on Jun 24^th, 2010, 10:39pm

Hi,

I want to fix some architecture for a Thermometric DAC.
I want to add a Δ-Σ Randomizer for selection of Thermometric Current Cells at a given code.

Can someone Indicate Pros vs. Cons for this scheme ??

What i can think of,
Advantage -- Distributes & Randomizes Mismatch in Current Cells, thus improving SFDR, SNDR & DNL spikes over the entire range.
Dis-Advantage -- More Switching in Current Cells. Introduces more Glitches at the Output. Hence, more noise.

How to outweigh one of them ?? Which one will dominate ?? How to determine whether Randomizer will improve or degrade my performance ??

What is the Industry Standard ?? Do people often use Randomizers in their design ?

--
Mayank.

Title: Re: Randomizer in Thermometric DACs
Post by carlgrace on Jun 25^th, 2010, 2:07pm

Another Con is that you will need more routing and the design will become more complex. If you want to go down that road, look into Data Directed Scrambling. It is similar to randomization in the Con, but gets you better performance. It is worth doing if you are going to do a randomizer anyway.

Carl

Title: Re: Randomizer in Thermometric DACs
Post by vivkr on Jun 28^th, 2010, 4:30am

Hopefully, you ae speaking of an oversampling DAC when you propose to use data-directed scrambling etc. I would think that it would be obvious that you don't want data-dependent noise if you are building a Nyquist-rate DAC.

As to checking the effectiveness of your techniques, simulations using MATLAB might be helpful and reasonably fast, although you should only use this to verify static effects such as mismatch. The full spectrum of nonidealities that impact DAC performance cannot be captured like this.

And yes, randomization is used fairly often in DACs to improve SFDR. PRBS-based dithering is quite popular in most cases as the spurs are removed from the entire spectrum, albeit at the cost of a raised noise floor (so if you can afford the higher noise floor). In oversampling DACs, there are various methods used for data-dependent scrambling, although I am not sure if there is any industry standard so to speak.

Vivek

Title: Re: Randomizer in Thermometric DACs
Post by carlgrace on Jun 28^th, 2010, 4:20pm

The original question specified a delta-sigma DAC, so data-directed scrambling is appropriate here.

Mayank, I thought about it and most likely the best solution is to use dither like Vivek suggested. It is easy to build a PRBS and then you don't have to worry about spending time chopping up your current cell array.

Carl

Title: Re: Randomizer in Thermometric DACs
Post by Mayank on Jun 29^th, 2010, 1:30am

Thanks for your replies.

& Vivek, yes i am talking about Oversampled DACs.

Quote:

And yes, randomization is used fairly often in DACs to improve SFDR. PRBS-based dithering is quite popular in most cases as the spurs are removed from the entire spectrum, albeit at the cost of a raised noise floor (so if you can afford the higher noise floor). In oversampling DACs, there are various methods used for data-dependent scrambling, although I am not sure if there is any industry standard so to speak.

Could you guys provide me some references on PRBS ??

Quote:

Mayank, I thought about it and most likely the best solution is to use dither like Vivek suggested. It is easy to build a PRBS and then you don't have to worry about spending time chopping up your current cell array.

Dither can be introduced in SDM also along with Randomization. Based on your experiences, you are saying PRBS should give better SFDR than Δ-Σ Randomization/Chopping of current cells as you say. ??

Title: Re: Randomizer in Thermometric DACs
Post by carlgrace on Jun 29^th, 2010, 10:38am

One of the good things about dither is that it is easy. Adding randomization at the same time negates that benefit. As to which would provide a better SFDR you will have to simulate it, because it is different in different cases. But Vivek is right, dither will increase your noise floor by spreading out the peaks that are limiting your SFDR.

A PRBS generator is just a string of flip-flops. Very easy. Good "PRBS Generator". The first hit is a good tutorial.

Carl