some question about Delta-sigma ADC

[carlyou]

site:www.edaboard.com delta sigma adc

Now, I focus on Delta-sigma ADC. I have some question, please tell me, 3x.
First, what is the definition of out-of-band gain? Second, now I use the SD Toolbox of Richard Schreier, it refers to four different structures, but they are all different from which I have decided to use, it looks like there are so many structures. Are they based on one or two basic structures? Third, I am a graduate student of micro electronic, I don't specialize in the system analysis, can I get a model from Richard Schreier's toolbox directly and begin my circuit design, or I must study the course about system first? Please give me some advices. Thank you very much.

 

[analogic1]

i have no idea about out-of-band gain.
for the system design, if u r not targeting at very complicated structure, u can create your own system in Simulink. It is very easy. Read "Overampling ADC" in David Johns book.

Richard Schreier is definitely an expert on Sigma-delta. but after spending one day on his toolbox, i found it hard to understand. so i gave up and create the simulink model by myself.

 

[jerryhuang]

carlyou,how do you use SD Toolbox of Richard Schreier,I know it is a matlab toolbox,do you get its documnte or tutorial?is it free?

 

[jerryhuang]

hi,analogic1,how do you realize the acumulator in the simulink?I also try to bulid a simple one order sigma delta modulator in simulink,but I don't know how to realize it ,such as quantilizer,delayer,and acumulaor,can you give me some advice ?thank you

 

[carlyou]

yes, it's free. You can download it from www.mathworks.com/matlabcentral/fileexchange. It has a tutorial in the package and also ten demos.

 

[jerryhuang]

do you have email,shall we have a chat on the topic,thanks,are u from uestc?hehe

 

[carlyou]

yes, I'm from UESTC, my E-mail is carlyou3656@163.com

 

[skal81]

Hi,
- First in my opinion Schreier's toolbox is only usefull as a start point for the design. Indeed this toolbox is made to give you the optimum loop coefficient for a given BW, OSR and SNR, but it only uses standard loop architecture. So if you custom the loop configuration I think you cannot rely on the toolbox. Also parasitics are not taken into account. For this refer to the article by Malorbeti attached. (It was in the Trans CAS or JSSC [ i forgot], and I think on Matlab web site).

- The out-of-band gain is the gain of the STF outside the BW. So for audio, it would be the gain from 22kHz to the sampling freq. The out-of-band maximum gain is considered to be related to stability. It is commonly accepted (Temes' book for eg.) the an out-of-band gain of about 1.3~1.4 allows you to keep the modulator stable for a large range of input. In fact that depends of the architecture. A feedforward STF present a peak, so this peak should not be to high. For distributed feed-back there is no such peak, so the total gain should be minimized.

- In Simulink, the accumulator is either an 1/s block (integrator) for continuous circuit (long simulation) or a discrete TF with z^(-1)/(1-z^(-1)).
The quantizer is a "sign" block. You usually normilized the signal to +/- 1. For parasistics and more practical models, refer to the article by Maloberti (link up).

Hope it helps.

 

[carlyou]

skal81, can you give a flow of the design of sigma delta ADC, and you say the Schreier's toolbox is only suit foa beginner, can you introduce some?

Added after 17 seconds:

skal81, can you give a flow of the design of sigma delta ADC, and you say the Schreier's toolbox is only suit foa beginner, can you introduce some?

 

[willyboy19]

First, out-of-band gain is the gain of the nosie-transfer-function outside of your band of interest, i.e. fs/2/OSR. It typically refers to high freuqncy gain close to fs/2.

The toolbox is a great tool (since I was involved in its development phase long time ago The important thing is that you need to have some basic DSP background and control system background (such as discrete-time Z transform, state space, etc.) In reality, of course, you need much more analog IC knowledge to get into the design.

 

[lovelonghair]

this is some guy's(named super) design of a first order sigma delta modulator in

simulink ,but the result is not corect, shall we talk about what 's the problem is ?

thanks

 

[carlyou]

why you use the constant to equal the error produced by the quantizer, I think it's not right, it must be equaled to the white noise. I also think that the source you'd better use the function sine and a sample block should follow it. The most important is the sampled time of the discrete filter and delay you must choose -1 for inherently.

 

[skal81]

To carlyou:
- Schreier's toolbox is not only for beginner I think. I wanted to say that it can only be used for standard architecture. However, when designing a DSM you usually custom the architecture. If you have a look at this year ISSCC papers some DSM are far from the standard feedforward.
- As for a flow I would suggest:
1) Choose the loop TF (depends on the SNR you want)
2) Choose an architecture, and find the coeff to do the TF
3) Simulate using the gain, BW, jitter to find the limit parameters, then do the transistor.

To lovelonghair:
- For the quantizer use the "sign" block. Also as carlyou said you should use a sine wave for the input. In fact DSM can have strange behaviour for DC input. Refer to Temes' book for more details.

 

[lovelonghair]

the modulator I posted justnow is used in fractional-n pll,in reality,the sigma delta is implemented with a model which is a little diffrent with the DSM in DAC,its input is a word as 1001 0010 1111 1010,so it will be wrong.but someone can show me a sample in simulink,thanks

 

[carlyou]

I can't upload the picture, you would search the forum, I have uploaded some model, or give me your E-mail, I will send you, lovelonghair.

 

[lovelonghair]

carlyou,i have sent a mail to you,carlyou3656@163.com