ADR-0042: Difference between revisions
showcase seed |
add breadcrumb navigation |
||
| Line 1: | Line 1: | ||
''← [[Decisions|All ADRs]] · [[Main Page|Home]]'' | |||
== ADR-0042: Ahuja over Miller compensation for the ramp-generator OTA == | == ADR-0042: Ahuja over Miller compensation for the ramp-generator OTA == | ||
Latest revision as of 09:17, 4 September 2026
ADR-0042: Ahuja over Miller compensation for the ramp-generator OTA
| ID | ADR-0042 |
| Status | accepted |
| Date | 2026-05-20 |
| Chips | RC2.0 |
| Blocks | ramp-generator, OTA |
| Sources | timeline/weekly/2026-W20.md |
Context
The ramp-generator OTA needed frequency compensation. Two candidates: classic Miller compensation, and Ahuja (cascode) compensation.
Decision
Use Ahuja compensation. It gives better power-supply rejection and removes the feedforward zero that Miller introduces, at the cost of an extra bias branch.
Rejected alternatives and why
- Miller compensation — simpler, but the feedforward zero degrades phase margin at the ramp speeds required, and PSRR is worse on the noisy 28nm supply.
Consequences
- Extra current in the cascode branch within power budget.
- PSRR improved enough to sit on the dirty supply without a dedicated LDO for this block.
What would make us revisit this
A move to a cleaner supply rail for the ramp generator, or a power budget cut that makes the extra branch unacceptable.