Based on a fairly limited amount of testing, it does seem that the data
separator circuitry in the Quay 900 might be inadequate for reliable double
density operation. My formatter program does a verify pass after
formatting, and disks formatted in single density on either drive read back
consistently, but double density with an 0xe5 fill byte gets a few read
errors on the inner (higher-numbered) tracks on drive 1, and many read
errors on the inner tracks on drive 0. I've tried writing other data
patterns after formatting, such as 00, 11, aa, 6db6db, 924924, and
pseudo-random data, and some of those patterns cause noticably higher read
error rates, though I haven't yet collected enough data to draw any
conclusions about specific patterns.
I also don't know how much write precompensation the Quay is doing, nor
whether they enable precomp based on TG43.
I suppose it's possible that the electronics of the drive is not performing
as well as it should. Unfortunately the CDC/MPI manual on Bitsavers
doesn't cover the specific variant of drive in the Quay. It has a PWA that
is noticeably different than the ones covered by the manual, with fewer ICs
and more potentiometers. Without matching documentation I'm not about to
start tweaking pots, though it's entirely possible that their values have
drifted to suboptimal settings.
I'm considering tracing out the schematic of the FDC to drive interface
enough to hack in an SMC FDC9216B data separator.