Can I convert 2 8bit adc to a 12 bit adc?

[PRAVEEN K]

Hi,

Is there any possible way to convert 2 8bit adc to 12bit adc?

Thanks
Praveen.

 

[ddet2004]

Hi PRAVEEN K,

you can combine these two converters in different ways to get a 12bit ADC. Even 16bit is possible (in ideal case). But you'll need some further blocks.
My first guess is to use a two-step or a subranging (they are very similar) structure.
The first ADC works as a coarse ADC and gives you the higher 8 bits. These bits are converted back to the analog domain using a DAC. This signal is subtracted from the original signal. This signal is amplified by 2 to the power of 8 (or even less to enable digital error correction) and given to the second ADC to resolve the lower bits.

I hope this helps.
regards

 

[FvM]

If the question is related to IC design, the answer would be basically yes, but it involves digital error correction, as ddet2004 mentioned. If it's related to ADC application, I would say, effectively no.

 

[Usman Hai]

I beg to differ here.

You can use 2 8-bits ADC to get 12 bits but it won't be a 12-bit 'ADC'. Basically to get 12 bits you need sampling capacitors of about 2^4 times higher than used in a typical 8-bit implementation to satisfy thermal noise requirements. Also, opamp must have 2^4 higher gain to satisfy accuracy requirements. So, if an adc is designed keeping in mind the requirements of 8-bits then you can not get the effective 12-bits out of the combination.

So, 2 8-bits can be combined to get 12-bits, but effective bits would be less than 8-bits and also have terrible INL and DNL as well.

 

[Usman Hai]

I don't know if you all are satisfied with this answer?

 

[FvM]

It's a verbose form of "effectively no" in my opinion.

 

[matbob]

Hi PRAVEEN K,

There is actually a method to obtain higher resolution using a low resolution ADC. Search for: 'OVERSAMPLING AND DECIMATION' in GOOGLE. You can find some infromation from this app note as well: http://www.atmel.com/dyn/resources/prod_documents/doc8003.pdf

This method is best when you use an ADC with a microcontroller.

 

[FvM]

actually a method to obtain higher resolution using a low resolution ADC

Yes and no. Oversampling can reduce noise, but it can increase linearity only by a rather small amount. Only one-bit ADC, as used in most simple SD-converters, don't suffer from ADC linearity errors. Multibit-oversampling designs, e.g. most recent audio ADC, have to use sophisticated correction circuits to overcome the problem.

This method is best when you use an ADC with a microcontroller.

Averaging or "decimation" is in fact required with many uP to get more than 7 or 8 effective bits from a "10 Bit" on-chip ADC. When suggested by vendors as Atmel, it sounds in my ears like "it's not a bug, it'a a feature".

 

[ALERTLINKS]

For ac signal, percision rectify each half cycle seperatly. Invert -ve cycle. Feed to seperate adcs.Now you have doubled the resolution. Additionally you have information about zero crossing and + and - half cycles for synchronization.