| Previous journal: | Next journal: |
|---|---|
| 0209-2024-07-23.md | 0211-2024-08-09.md |
I spent some time recently in the maybe branch trying to figure out a better current-steering DAC that actually uses differential (I think?) steering, i.e. pairs of FETs per bit, to 'steer' equal current in opposing directions.
I did 15 commits before pausing again prior to today.
3rd of 3 tests of DC sweep of gate voltage, plus comparing different transistor sizing/ratios:
Kind of a recreation of bdac.sch as described in 0209 (binary-weighted switched current DAC), except: (a) each bit is tested in turn via SPICE instead of testing the full 0..255 range; and (b) focus on measuring/balancing actual path current without conversion to voltage:
Extending 04 (above) to test out full range. NOTE: Extra smoothing cap (hanging off dacout; not in repo) added in this screenshot, to make the lumpy ramp clearer:
- Green: Mprog width is 1.0um.
- Red: Mprog width is 0.9um.
Adjusting 05 (above) to try and get a better curve (tested with more sweeps):
Trying some more variations in sweeps. Looking pretty chunky:
Screenshot below is altered from actual repo, but shows a sort of workable curve. Big pull-down transistors though:
Trying a W2W CDAC. Sort of works, but pretty ugly:
Not much better when using a pull-down resistor instead of FET; slightly more linear but terribly lumpy.
First true (?) current-steering DAC, i.e. differential. Great linearity, but only because of binary clusters of equally-sized transistors. NOTE: Image below is smoothed with caps (not present in repo):
Evolution of 10 (above): Attempt to use sensibly-sized transistors. Not too bad. Some bumps still, but pretty good. Current consumption is ~500uA. NOTE: I separately measured the pull-up resistor loads later; max. was about 435uA on either (needing 4.35um width?), meaning Vbias transistor pair must be drawing ~65uA. Could work with 1k resistors too, but this reduces the voltage swing from ~1.1Vpp to ~0.45Vpp.
NOTE: Resistors could also be replaced with NFETs:
...or PFETs:
What about this (gamma correction maybe?):
- Rote exercise: Reassemble
tt-vga-funusing simple RDACs but new digital control block (approx. at commit12bbe7f). - Use TT08 3v3 template for future experiments: In case we decide to try 3.3V at some point, we're already in it.
- Try unary ('thermometer') coding for 3 MSBs. These COULD be driven just from the digital block, rather than having to implement the decoder in the DAC.
- Would it work to do 3 levels of unary coding? i.e. 2 MSBs for 0-3 big transistors, middle 3 bits for 0-7 medium transistors, bottom 3 bits for 0-7 tiny transistors?
- Behzad Razavi's The Current-Steering DAC
- https://www.analog.com/media/en/training-seminars/design-handbooks/CMOS-DAC-Application-Guide/Chapter3.pdf
- https://keep.lib.asu.edu/system/files/c7/121290/Jankunas_asu_0010N_14295.pdf
- https://nirmawebsite.s3.ap-south-1.amazonaws.com/wp-content/uploads/sites/22/2020/08/78-150_JJP_IJRTE.pdf