Questions:
- What considerations should we have regarding a minimum sample rate due to leakage in the MIM caps?
- Leakage should not be a big problem
- How to generate a Vcm for the comparators
- It is a big consideration to generate Vcm
- Take Vcm to a package pin
- Consider using split capacitor
- Change comparator to accept Vcm
- Use a "oh no bit" to choose a different Vcm
- It is a big consideration to generate Vcm
- What sort of considerations should we look into for the comparators? We were thinking auto zero triple tail
- Autozeroing could go wrong.. consider oh no bit
- Could increase accuracy by 1lsb
- Be very considerate of auto-zero capacitor, could be troublesome from mismatch
- buffer the clock!
- the clock buffer here may be easiier to implement without standard cell
- Are there any recommendations for sampling technique? For now we have conventional bottom plate sampling that is differential.
- Do we scale down our analog switch width in accordance with the capacitor size within the array?
- How much of a problem is kickback? How do we tell if our kickback is acceptable or not?
- it will exist, and simulate it
- Kickback could could confuse comparison
- If kickback causes a diode to get forward biased, voltage will be lost. This is very bad. (small kickback should be ok?)
- Kickback could from parasitic gate drain capacitor.
- Check your cascode
- Maybe have that node not clocked
- How do you recommend we do layout for the CDAC? It seems like that would take a lot of time.
- CDAC is very challenging → hand layout is typical
- 8 bit should be quite easy
- Create a pattern
- Routing is a problem – ensure no crosstalk
- Put a big sheet of metal on M5
- Put a big sheet of metal on M2 and pop in vias.
- Protect your capacitors!
- This introduces parasitics, but these are static
- Consider aggressors and victims (what nodes)
- CDAC is very challenging → hand layout is typical
- Have a separate AVDD and AVSS could be useful
- Make sure that things that cross between digital and analog boundaries are cleaned up
- Make sure Ir drop across vdd is not too much
- Vref+ and - being cnnected to the CDAC is OK.
- Boostrapped switch is ok, but may not be necessary. Also be careful of bootstrapped extra voltages.
- Bootstrapped switch is good for Ron linearity
- not particularrly necessary
- Have to be a bit careful about substrate to source/gate
- gate to body should not be a big deal
- Vgs and Vds are a big deal
- Think about startup and turnoff behavior
- Grab some extra bits for trim bits from digital for risk mitigation
- Think about testing/safety
- Calibrating charge injection away is a possibility
- Can protect the body of the transistor for the deepnwell
- Break before make
- Look for some standard cells that provide delay (dly cell?)
- Concentrate as much digital as possible in as few blocks as possible in layout
- P and notP should be driven by 8x inverters
- FSM logic
- Should not be too bad