[Solved] DAC nonlinearity questions

I'm looking at a 16bit DAC for a programmable load.. On possibility is the AD5662, seem to be quite easy to communicate with this DAC and datasheet is very detailed. There seem to be two categories, with +/-16 LSB INL and +/-32 LSB INL.

So my two questions are: a) Do I understand it right that 16LSB nonlinearity is better then 32LSB? b) In general, a 16LSB nonlinearity - is a lot?

Do you understand what non linearity means ?

What is you application ?

AlfaOmega:
So my two questions are:
a) Do I understand it right that 16LSB nonlinearity is better then 32LSB?

NO at least without qualification of what you mean.

AlfaOmega:
b) In general, a 16LSB nonlinearity - is a lot?

No.

aarg: No.

Thank you! I'm glad to hear this!

Boardburner2: Do you understand what non linearity means ? What is you application ?

Yes, I do understand the concept of nonlinearity, but I do not understand the unit 'LSB', at least not in this context. In application notes by TI and tutorials by Maxim Integrated it is quite clear - LSB means last digit on the left and 1/2 LSB means DAC/ADC cant decide whether to put a 0 or 1 there. But when the datasheet says '32LSB INL' for a device that only got 16 bits of resolution... Then it becomes confusing))

As for the application - within a programmable load I use a mosfet in unsaturated mode as a variable resistor. In my current setup I'm experimenting with it MCP4922 (12bit DAC). In general everything works, but I'm curious to try to try something with greater resolution.

In that context LSB doesn't mean the least significant bit but the voltage associated with it. Thus 32LSB is 32 times that voltage and represents the greatest deviation of the output voltage from the ideal linear response.

Russell.

In other words, 32 LSB means 32 counts.

AlfaOmega: Thank you! I'm glad to hear this!

  • LSB means last digit on the left and 1/2 LSB means DAC/ADC cant decide whether to put a 0 or 1 there.

Least Significant Bit , is the last bit on the right not left.

Boardburner2: Least Significant Bit , is the last bit on the right not left.

Ouch, that was a silly error on my part, of course you are right ))

Thank you, Russell and vaj4088 for explanation! Now I understand why a device with 32LSB INL is more accurate.

AlfaOmega: Now I understand why a device with 32LSB INL is more accurate.

. . . you mean less???

Russell