quantizer question

[Fraser, Colin J]

> -----Original Message-----
> From: Roman Sowa [mailto:rsowa at WizjaTV.pl]
> Sent: 31 August 1998 14:08
> To: Synth-DIY
> Subject: ODP: quantizer question
> 
> 
> 8 bit DACs DO have poor tolerance, as every other converter.
> But ussually it is limited to the few LSBs. So, the wider DAC input,
> the better linearity within MSB range.
>
> 1/2 LSB quantization noise mentioned by Colin has nothing to do with
> DAC linearity. They're made just as precise as needed to maintain
> differential error less than 1LSB (the code higher by 1LSB gives
> output voltage no less than with code smaller by 1LSB). And that's not
> even a rule for higher resolution DACs (14 bits and more).
> That's what makes'em 8-bitters cheeper.

If the tolerance is a design decision, maybe the devices I used are
built to be accurate even at the LSB end ...
I don't see how a R-2R network can't be fabricated on a chip with close
tolerance resistances, regardless of how many bits there are.
Obviously as you increase the number of bits for a given supply voltage,
the voltage level of the lower bits will get lost in noise.
 
> If you had no problem with 8 bit DACs, then maybe you were lucky
> and got a better piece, or used it as a 6 or 5 bit converter.

I've used half a dozen 8 bit dacs in these applications, some with 7
bits used, some with 8, and never had any problem with tuning of
oscillators driven by them.

What order of error would you suggest I should be getting ?


Colin f