Count the Real Weeks a Photomask Set Took to Design and Fabricate
# Count the Real Weeks a Photomask Set Took to Design and Fabricate
## 1. Separate Mask Design From Mask Fabrication
A mask-ROM bit pattern was not something an engineer could change by flipping a switch or reprogramming a device. Getting a correct program into a mask ROM meant translating a verified bit pattern into a mask-design database, then having a mask shop pattern that database onto a set of physical glass photomasks, one per fabrication layer the ROM's process required. These were two genuinely separate stretches of elapsed time, done by different people in different facilities, and both had to complete, in sequence, before a single wafer carrying the new program could be started.
Mask design — laying out the row/column matrix geometry and encoding the verified bit pattern into the specific crossings that would or would not get a diode or transistor gate connection — typically ran on the order of one to a few weeks of engineering and drafting/digitizing time, depending on array size and how much of the surrounding decoder/periphery layout could be reused from a previous design. Mask fabrication — the mask shop's own pattern-generation and plate-etching process for the full layer set — typically added several more weeks beyond that, and this time did not compress much no matter how urgent the request, because the mask shop's own equipment and process steps ran at a fixed cadence per layer, not per customer.
## 2. Add Up the Total Commitment Before a Wafer Starts
Stacked together, a full new mask set commonly represented four to eight weeks of elapsed calendar time from a verified bit pattern to masks in hand at the fab, before a single wafer could even begin processing with the new pattern — and that estimate assumes the bit pattern handed to the mask shop was already correct. Nothing about this lead time was a mask-ROM-specific curiosity; it was the real cost that made the tape-out decision in Step 7 the point of no return it actually was, and the reason a single wrong bit discovered after that point, covered in Step 6, was expensive specifically because this whole stretch of weeks had to run again.
Step 6 puts a concrete cost on exactly that: what a mask-set respin actually meant once even one bit was wrong.