cancel
Showing results for 
Search instead for 
Did you mean: 

12bit DAC performance

oleg
Associate
Posted on February 04, 2014 at 15:28

Hi there,

In my application I need precise control of applied voltage. I use 12bit DAC with 2.5V reference. The mcu is stm32f107

I was going to do error budget analysis but was confused by the datasheet. It gives +/-4LSB of INL (which is quite high) but then there is another term - Offset(difference between measured value at 0x800 and the ideal value = Vref/2). And its max value can be +/-10mV!!! This offset is more than 16 LSB.

So my question is what to believe? the datasheet seems to contradict itself. Or it's me?

Cheers,

Oleg

#error-budget #dac #dac-performan
3 REPLIES 3
ilmars
Associate II
Posted on February 04, 2014 at 16:03

So my question is what to believe? the datasheet seems to contradict itself. Or it's me?

I do not see any contradiction. Offset and INL are different errors of ADC. Perhaps this will give you idea (sorry, did not find any ADC error article by ST):

http://www.maximintegrated.com/app-notes/index.mvp/id/283

oleg
Associate
Posted on February 04, 2014 at 17:42

thank you for the reply ogden,

although INL and DNL for ADC and DAC have the same nature, I was talking about DAC. I believe I understand INL - simply it's difference between generated voltage and straight line (using ''best fit'' approach or end points). In both cases I can't see why would the offset error exceed INL that much (10mV=~16bits vs INL=+/-4bits)

have a look at attached specs.

It looks like to find the ''worst case'' error I would have to add INL to Bias Error?

________________

Attachments :

specs1.png : https://st--c.eu10.content.force.com/sfc/dist/version/download/?oid=00Db0000000YtG6&ids=0680X000006I1Mo&d=%2Fa%2F0X0000000bly%2Fos.5OUat3q9OhxEO_1wK3o7VqghuOe5adSBQzOx8NQ4&asPdf=false

specs2.png : https://st--c.eu10.content.force.com/sfc/dist/version/download/?oid=00Db0000000YtG6&ids=0680X000006I1Mj&d=%2Fa%2F0X0000000blz%2FPjZ.JPCn7RXUQwP9q47OJv6YynEue_z1bCNUEzTzIGg&asPdf=false
chen
Associate II
Posted on February 04, 2014 at 18:10

Hi

Not sure if I should join this conversation because I admit I am out of my depth here.

''I use 12bit DAC with 2.5V reference.''

If you do the basic calculation of V/bit - you get 0.0006V/bit or 0.6mV/bit

You would need a high end multimeter (most multimeters are only 3.5 digit ie can only measure 0.001V - 1mV not taking into account the DMM errors!) to measure that.

''In both cases I can't see why would the offset error exceed INL that much (10mV=~16bits vs INL=+/-4bits)''

In my experience 10mV noise in a real electronic system is a pretty good system.

Basically, the electrical background noise is going to swamp the LSB change in 12 bit mode! They reckon that 10mV noise is 4bits - sounds about right.

''So my question is what to believe? the datasheet seems to contradict itself.''

No - the datasheet is just saying that noise obscures the accuracy of 12bit mode.