Commercial Disk Duplication

How Publishers Actually Duplicated Their Disks

If copy protection was so effective, how did publishers make thousands of copies of their own games? The answers are stranger than you'd expect. First-hand accounts from comp.sys.cbm (1999) and a 2020 interview with Scott Cronce — technology manager at Disclone, EA's main duplication house from 1985 to 1987 — open a window onto the industry that sat behind every shrink-wrapped C64 game.

Large Publishers: Industrial GCR Duplicators

Major publishers used commercial disk duplicators with autoloaders, or contracted out to dedicated duplication houses. These were PC-controlled machines modified for GCR copying. From someone at Abacus Software in 1999:

The same account describes a single title that could only be produced on a standalone PDP-11-based copier — requiring a three-disk initialization process (format disk, source disk, protection data disk) before the machine could begin stamping copies.

CMD's DIY Autoloader

Creative Micro Designs (CMD) acquired a second-hand 50-disk autoloader intended for PC disks. Rather than pay for commercial modification software, they built their own system from scratch — wrote control software for the Commodore side, replaced the mechanism with a 1541, connected to the drive controller via a parallel cable on the User port, stored disk images on a CMD Hard Drive, and wrote them out via the parallel interface. It couldn't handle most protections, but worked fine for their own GEOS disk protection.

Small Publishers: Modified MSD-2 Drives

Smaller publishers used a modified MSD Dual Drive (a third-party 1541-compatible drive). Basement Boys Software (makers of Fast Hack'em) used a modified MSD-2 for all their production disks. Chip Level Designs sold a hardware modification for the MSD-2 that added 2K of extra RAM to an empty socket on the motherboard, designed specifically for rapid disk copying.

GEOS Wheels: Unique Disk IDs

Maurice Randall's Wheels OS used a particularly clever approach. Disks were initially produced on an MSD-2 from an unprotected master. They were then placed in a 1541, and a special program ran inside the drive to apply protection — asking for the next disk when done, incrementing a unique ID number each time. Each copy had a different ID, making mass illegal copying detectable. The in-drive operation took about 7–8 seconds per disk.

Inside a Professional Duplication House: Disclone & the Formaster System

In July 2020, computer scientist John Aycock (University of Calgary) interviewed Scott Cronce under academic ethics approval, with results published as technical report TR 2020-1117-03 and released under a Creative Commons license. Cronce was manager of technology at Disclone from 1985 to 1987 — the company was EA's primary copy-protection and disk duplication vendor, responsible for roughly 60 protected EA titles across multiple platforms. This is the most detailed primary-source account of professional C64 duplication ever published.

The Formaster Machine — PDP-11 with AMD 2900 Bit-Slice Chips

The heart of professional C64 duplication wasn't a Commodore computer. Formaster was the dominant duplicator manufacturer. Their machines were remarkable: the controller cabinet contained a PDP-11 minicomputer, laid sideways on a hinge and connected via Q-Bus to the system. The actual disk controller — the part that wrote custom magnetic signatures — was based on four AMD 2900 bit-slice chips (4 bits each, 16 bits total), connected in parallel.

The 2900 was originally developed for aerospace and military use. Writing code for it is notoriously difficult — programming is essentially raw microcode, where operations like comparisons must be set up before the relevant math completes. Much of Disclone's microcode was written by Kate Manbert, a friend recruited specifically for this expertise: she had come to Formaster from a classified satellite project that also used 2900-series chips.

The PDG System — Program Disk Generator

Formaster's software ran on RT-11, DEC's real-time OS for PDP-11s. The toolchain included Macro-11 assembler, a Pascal compiler, and the Microtech Meta Macro Assembler for 2900 microcode. The critical runtime was the Program Disk Generator (PDG): a framework for defining and writing protection signatures onto target disks.

The production workflow was:

1. Boot RT-11 from an 800K floppy system disk
2. Mount and run the PDG software from the PDP-11
3. Write a PDG data disk — encoding the protection scheme's parameters: encoding type, sector count, error correction method, and 2900 microcode binary
4. On the Formaster cabinet: load the PDG data disk to configure what signature to stamp
5. Read the master disk into memory (compressed, since there wasn't enough RAM for a full raw image)
6. Begin the duplication run: format → write → optional 100% verify

CLIENT MASTER DISK
        │
        ▼
  PDP-11 (VT100 terminal on engineer's desk)
  RT-11 + PDG System + 2900 microcode assembler
        │
        ▼  PDG data disk (compiled protection params)
        │
        ▼
  Formaster Q-Bus ──── Custom AMD 2900 Controller
                              │
                    ┌─────────┼─────────┐
                 Drives    Drives    Drives
                 Cabinet   Cabinet   Cabinet
                    │
                    ▼
  Mountain Computing Autoloader (preferred over Formaster autoloader)
  → 50-disk capacity, PASS/FAIL bins
  → EA: required host boot test on a % of finished disks

Cronce operated from a Digital VT100 terminal in his office. When a new protection was needed, modifying the PDG hex output directly — poking new sector addresses or signature positions — was faster than regenerating from scratch. "Just go hack in the new sector versus the old one in the hex file, then you're done."

Flaky Bits and Marching Bits — The Core Protection Technique

By 1985, the industry had converged on a family of techniques Cronce calls "flaky bits" (also "marching bits"). The principle exploits the relationship between the disk controller's phase-lock oscillator and the magnetic flux transitions it uses to stay synchronized:

The defining test of whether a new flaky-bit scheme was secure: Copy II PC Deluxe Option Board from Central Point Software. This was an IBM PC expansion card with a custom floppy controller that could read and reproduce many non-standard disk structures. Industry standard practice was for every protection engineer to have one on their workbench. If the Copy II PC Option Board could copy a protection, the scheme needed to be revised.

The Anti-Crack Arsenal — Software Defense Layers

Physical disk signatures were only the first line of defense. Once crackers shifted to attacking detection code directly — finding the check routine and patching it out — the industry responded with increasingly sophisticated software techniques.

The simplest crack: find the protection check's pass/fail branch and replace the branch instruction with a jump to the success path. The response was to remove the protection check from the disk entirely — assembling it only in RAM during load, executing it once, then overwriting the routine with zeros so it was never visible in memory after the fact.

The next level was delayed-effect detection. Rather than crashing immediately on a bad copy, the protection would:

• Allow the game to boot normally and appear to work
• Set a flag deep in memory — invisible in casual debugging
• Trigger the failure at a specific game moment: after reaching level 3, during a save operation, or partway through a boss fight
• Cause a subtle failure: a puzzle that becomes unsolvable, a level that can't be completed, a crash that looked like a bug

The logic was sociological as much as technical. Most crackers of the era were distributors and collectors, not players — they cracked software to build their scene credentials and distribute images, but rarely played the games for more than a few minutes. A failure that didn't manifest until level 5 of a 10-level game might not be discovered for weeks, by which time the bad crack had been distributed to thousands of people. The cracker's handle — their reputation — was at stake every time a game was found to have been incorrectly cracked. Making crackers look incompetent was itself a form of protection.

The PACE Anti-Piracy Connection — Industry Genealogy

The protection Disclone used didn't come from Disclone itself. Allen Cronce (Scott's older brother) had worked at Formaster learning to write PDG disks and format-level code, then moved to XEMAG under Greg Galloway. When Xidex acquired Dysan — the disk manufacturer — and rolled XEMAG into the combined operation, Allen left before the consolidation to found PACE Anti-Piracy. PACE's first two copy protections were for the C64 and Macintosh, and those were the protections Disclone licensed. EA had previously used XEMAG; they switched to Disclone specifically because of PACE.

The interpersonal connections are remarkable. Greg Riker (head of EA Technology) lived across the street from Joe Fontana, another protection programmer who is Allen Cronce's CTO at PACE to this day. Formaster's co-founder was a friend of Scott's family from their monthly wine-tasting group. Scott attended Formaster events for years before getting the job at Disclone. When a legal dispute between Disclone and Formaster required depositions, Scott was greeted warmly at the law office by Gerry Pressman — the Formaster co-founder — rather than by opposing counsel.

Scale and Revenue

By Scott Cronce's records, Disclone protected approximately 60 EA titles across C64, Mac, Apple II, Apple IIGS, Amiga, Atari ST, and PC between mid-1985 and October 1987. The company had over 200 clients total — including The Learning Company, AT&T (Unix distribution), Xerox Desktop Publishing, and ABC Compute Publishing (Compute! and Gazette magazines). Large production runs were in the hundreds of thousands; small runs in the hundreds. Disclone generated approximately $5 million in fiscal year 1986 — more than Formaster itself, despite Formaster being a publicly traded company. The lesson Cronce draws: the per-disk fees from duplication services substantially outperformed selling the duplication equipment.

The Mechanical Reality

Formaster's own autoloaders were a constant maintenance headache. 5.25" disks arrived in bulk cases — roughly 3×2 feet, two rows stacked, plastic-wrapped. When the autoloader mechanism grabbed a disk even slightly off-center, the disk folded in half. The team switched to Mountain Computing autoloaders (which became the standalone Trace duplicators) as soon as possible. Even the better machines required constant improvised maintenance: finding rubber bands from alternative suppliers when the originals stretched out after a few weeks, sourcing softer printing rollers that didn't damage disks in the labeling machine.