[time-nuts] DAC performance

[Charles Steinmetz]

Attila wrote:

> Charles Steinmetz <csteinmetz at yandex.com> wrote:

>> how about the LTC1650?
>>      *     *     *
>> [it] is nearly 100x (40dB) quieter (30nV/sqrtHz vs. 280) [than the 1650]

> I was about to ask the same question :-)

Note:  I divided 280 by 30 and got "nearly 100x".   D'Oh!  Of course, it 
is ~10x, or 20dB, not ~100x or 40dB as I stated.  Still, a ~20dB noise 
advantage is substantial.

One point I didn't mention previously -- the 1650 is only marginally 
more expensive than the 1655.

> I am sure there are ways to compensate for the charge injection
> by using multiple switches, but I have not been able to come up
> with a good way. Does anyone have a good idea?

I've tried several approaches to cancel the glitch energy of analog 
switches, but I never hit on anything I was really happy with -- partly 
because the glitch energy is more random than you'd like and partly 
because the "kludge factor" of multiple-switch solutions exceeds my 
tolerance pretty quickly.  Even if you accept the high kludge factor, 
you find that like input bias current cancellation, glitch cancellation 
is most effective (only really effective??) if it is done on-chip.

Best regards,

Charles