14bit pipeline ADC simulink dnl simulation result

[iamxo]

pipeline adc simulink

I construct a 14b pipeline ADC simulink model and use the Maxim website code to simulate the DNL/INL of ADC, but the result confuses me a lot.

could anyone help me with the dnl/inl simulation result whether it's true or not?? Because the dnl result seems quite weird.

Personally, I touch pipeline adc just recently, so please give your opinions, thanks.

 

[erikl]

simulink pipeline

The Y-axes should read "mLSB". So it's an excellent result!

 

[iamxo]

inl dnl adc maxim

Do you mean that I should divide Y-axis by 1000??
But why?

 

[erikl]

dnl calculation for adc

Do you mean that I should divide Y-axis by 1000??
But why?

Exactly. Because it is the normal representation for DNL & INL. For a good converter, the DNL values should stay below 1LSB (only then it is monotonic). Due to offset pbs., the INL values sometimes are up to a few LSB's. So your sim. result shows excellent values. Why the Y-axis capture is not shown in mLSB I don't know. A program bug perhaps. Just ask the company which is responsible for the program.

 

[iamxo]

converter (adc) simulink

Exactly. Because it is the normal representation for DNL & INL. For a good converter, the DNL values should stay below 1LSB (only then it is monotonic). Due to offset pbs., the INL values sometimes are up to a few LSB's. So your sim. result shows excellent values. Why the Y-axis capture is not shown in mLSB I don't know. A program bug perhaps. Just ask the company which is responsible for the program.

The matlab code is got from Maxim Inc. website, it says the code is reliable and could be trusted.**broken link removed**
In the code, the calculation of DNL/INL is not divided by 1000, and I also find no
trace of mLSB as you mentioned.

So, could you take some time to see the link I provide? I would appreciate you much, if you would do me the favour.

 

[erikl]

pipeline adc matlab

Your link returned "Document not found" - and now I'm off on holiday, sorry! Try to ask Maxim themselves; they surely won't acknowledge DNL/INL values of several hundred LSBs for their converters! ;-)

 

[iamxo]

analog to digital conversion simulink

Ok, thank you for suggestions.
Have a wonderful holiday.

 

[Vamsi Mocherla]

calculate dnl adc matlab

Looks like there is an error in the amplitude(magnitude) of the sine wave you have used for your ADC and the magnitude of the sine wave in the code. Once, you get the amplitudes right in dBFS, you will get the right result

Added after 1 minutes:

I am saying this because there is a curvature in your DNL, which I deem weird.

 

[iamxo]

pipeline adc

Looks like there is an error in the amplitude(magnitude) of the sine wave you have used for your ADC and the magnitude of the sine wave in the code. Once, you get the amplitudes right in dBFS, you will get the right result

Added after 1 minutes:

I am saying this because there is a curvature in your DNL, which I deem weird.

My sine wave input is 1.001 that is a bit larger than ideal value 1 to clip the ADC as the application note requires. what exactly do you mean by an error in the amplitude of the sine wave?? And what is the relationship between sine wave amplitude and DNL curvature?
Could you explain more? thanks.

 

[iamxo]

maxim adc simulink

Hi, guys, I emailed the Maxim support center to ask this question and got the following answer, maybe it will guide us simulating ADC DNL/INL.

From Maxim support center:
The reason your simulation result looks like this is that you
don't use nearly enough points for your INL/DNL calculation. A
rule of thumb is to repeat each code at least 40 times, however
for data we put into the datasheet and use for app notes, we go
for 256k, 512k or even 1M data points. You need to increase your
number of datapoints. 4096 are not nearly enough to get a good
representation for INL/DNL data. Once you'll get that squared
away, you'll see that your results will look more like a straight
band of non-linearity errors for DNL and a bow or S-shaped form
for INL. Bow or S-shape depends on whether your even- or
odd-order harmonics are more dominant. Even-order dominance is
turning the curve into a bow and odd-order dominance will result
in an S-shape.

 

[eegchen]

Re: pipeline adc simulink

hi,

What's kinds of output do you get ?

14 bits binary digital output.

or only 1 bit decimal output

I construct a 14b pipeline ADC simulink model and use the Maxim website code to simulate the DNL/INL of ADC, but the result confuses me a lot.

could anyone help me with the dnl/inl simulation result whether it's true or not?? Because the dnl result seems quite weird.

Personally, I touch pipeline adc just recently, so please give your opinions, thanks.