[sdiy] ADc & DAC voltage dividers.. fixed or trimmer adjustables ??

Roman Sowa modular at go2.pl
Mon Feb 15 12:02:42 CET 2021



W dniu 2021-02-15 o 09:52, Brian Willoughby pisze:

>> you wouldn't believe what people can do for fun with an equipment. I'm telling that from occasional synth-tech perspective
> 
> Yeah, some people have been known to replace every capacitor, even when nothing is broken!

or put piece of paper between power regulator and its radiator.
I have pictures to prove it!


> A lot of modern MCU chips have way more Flash memory than RAM. It's easiest to store fresh calibration values 

we're talking 4k of flash, where RAM is maybe some 256 bytes

> 
>> Many modern DACs are resistor string, not sensitive to jumps of binary code where all bits flip (like from 0x3FFF to 0x4000) but also because such architecture their linearity is bent all over the place without any plan.
> 
> "Because of the physical size, pure string DACs are primarily limited to resolutions of 8 to 10 bits."
> 
> Mike designed with a 14-bit DAC, so I rather doubt it's available as a string of resistors type DAC. I've worked with a binary-weighted current output DAC that is 14-bit, but never checked for this sort of offset error.


And there are non-pure string, mixed string types, with part of the word 
being converted by resistor string, and the other by R-2R, or even 
better, several stages of resistor strings. Actual DAC selected by 
Jean-Pierre uses this multi-string topology. Quoting from datasheet: 
"The four MSBs are decoded to drive 15 equally weighted segments, and 
the remaining lower bits drive successively lower weighted sections. 
Major carry glitch impulse is very low at500pV•sec, CL= 10pF, ten times 
lower than previous DACs of this type."
So not only midscale jump is avoided but it also helps to remove glitch 
impulses around full scale. Look at INL graph, you really can't tell 
where's 7FFF/8000 crossing.

Having calibrated a few of 16-bit string-DAC channels, I have seen all 
kinds of INL graphs, it's like filtered noise frames captured on a scope.

Roman



More information about the Synth-diy mailing list