M7256

This document is an Engineering Change Order Log detailing various problems and corrections for the M7256 (RK11-D Data Paths) module used in PDP-11 family processors.

The log covers several hardware and documentation issues, along with their solutions and implementation guidelines:

1. Write/Write Check Termination (M7256-00001): The RK11-D prematurely terminated write and write check functions for short sectors due to sampling "WORD COUNT OVERFLOW." The fix involved creating and sampling a "FILE DONE" condition instead.
2. Circuit Schematic Errors (M7256-00002): Corrected various print errors found on the M7256 Circuit Schematics.
3. Data Corruption in Parity Systems (M7256-C0003 & M7256-00004):
   • Problem: In RK11-D systems with parity memory, if BUS D16 was asserted during a short sector write, the sector would fill with 'ones' instead of 'zeros' due to parity information entering the FILE IN buffer.
   • Correction: Modified the M7256 with an etch cut and wire addition to prevent BUS D16 from clocking into the FILE IN buffer.
   • Follow-up (M7256-00004): This entry updated the etch design to include jumpers for proper BUS D16 clocking (allowing or preventing it for RK11-E vs. RK11-D), and adjusted the Parts List to ensure only necessary jumpers were installed during manufacturing.
4. Write Race Conditions / Data Destruction (M7256-B0005 & M7256-B005A):
   • Problem: The "COUNT SA" pulse (originally 200 nsec) was too short, potentially causing data destruction during disk writes if the surface select bit changed before the write current stopped.
   • Correction (M7256-B0005): Increased COUNT SA to 500 nsec by replacing a 10 pfd capacitor (C36) with a 56 pfd one.
   • Enhanced Correction (M7256-B005A): Further investigation revealed that extending this fix could also resolve an inherent RK05 write recovery problem. The updated correction specified using a 150 pfd capacitor to lengthen COUNT SA to 1 usec.
5. Documentation Discrepancy (M7256-0004A): Corrected the Circuit Schematic to accurately depict Jumper W5 as installed (solid line) rather than optional (broken line), as per the Parts List update from M7256-00004.

Many of these changes required immediate rework of existing "B" revision boards, modules in test equipment, and all units in production or shipped after specific dates in late 1973 and early 1974. Field reworks were also mandated for M7256 modules in RK11-D systems, especially those with parity memory.