Scaling down as per DAC resolution

[Maitry07]

Hello ,

My Digital data range is 14 bit ( 0000 H to 3FFF H). but my DAC resolution is 12 bit. so I need to change the scale to 12 bit. In order to change the scale, what could the best suitable method. Will it be suitable to divide the 14 bit data with 4 to scale down to 12 bit? or any other suitable method can be used in FPGA?

 

[KlausST]

Hi,

what DAC interface?
* on a parallel interface .. just connect the data lines as you like.

what alignment? (mainly for serial interfaces)
* if you are MSB aligned ... it is no problem at all.
* if you are LSB aligned .. you need to care for

If you need to do it for every data by software: Shift right or left. Or multiply / divide

Klaus

 

[Maitry07]

My DAC is DAC2932. which is 12 bit parallel interface. so in parallel DAC , there is no requirement for scaling. I just need to ignore the 2 LSB, Is it like this?

 

[KlausST]

Hi,

basically yes. And in case of an FPGA you don´t need to wire the 2 LSBs to an FPGA pin at all.

Still you have to care about all timings and requirements the DAC datasheet says.

Klaus

 

[kaz1]

Best solution is discard two lsb after rounding onto 12 bits.
Truncation will cause dc bias otherwise.

 

[danadakk]

Best solution is discard two lsb after rounding onto 12 bits.
Truncation will cause dc bias otherwise.

Thanks for posting that. Goggling "dc bias truncation", quite a few resources
about effects and cures.


Regards, Dana.

 

[KlausST]

Truncation will cause dc bias otherwise.

Yes, this is true.

Here the values:
The (average) offset the truncation causes compared to the rounding method is -0.5 LSB
12 bit means 4096 steps full scale, thus one LSB is 0.025% FS ...
--> and 0.5LSB means 0.0125% FS. If we take 1mA as FS, then the error is 0.125uA or 125nA

The DAC datasheet says
the DAC has an Offset error of ±0.001% FS. This is incredibly good. Thus when the DAC output is close to zero the truncation error (of 125nA) dominates.

The gain error is given with +/-0.5% FS typically. Now if we compare gain error with truncation offset error we can say that at 1/40 or 2.5% of output (or higher) the gain error dominates.

Klaus