As 2021 draws to a close, Yellowstone is very nearly finished, with just a few details left to iron out. I recently tested a new v2.3 prototype with a couple of small PCB changes, and so far it’s looking good. The new PCB fixes an obscure problem with the Unidisk 5.25 and Disk IIc, where one of those specific drives connected as Drive 2 would interfere with another 5.25 inch drive at Drive 1. Initially I’d planned to write this off as a known compatibility issue, but I had second thoughts, and rushed through a new PCB revision.

Parts sourcing continues to be a major problem. The good news is that I was able to find more of the FPGAs that Yellowstone needs, but I had to go through a third-party dealer in Hong Kong and pay roughly twice the normal price. I have enough parts now to manufacture about 500 Yellowstone cards. Beyond that, the outlook is murky. I expect the semiconductor shortage to get worse before it starts getting better, and many parts are currently quoting lead times of a year or more. Given the current environment, Yellowstone should probably be considered a “limited edition” in 2022, with the possibility of some restock in 2023 or 2024.

I’ve received a couple of manufacturing quotes already from CircuitHub and MacroFab, two vendors with automated web-based quoting tools. These are helpful sanity checks and estimates, but they’re not yet fully baked quotes that I could move ahead with as-is. Optional features like beveled PCB edges and gold fingers require a custom quote. Programming and testing requirements are also difficult to factor in to automated quoting tools.

The preliminary information looks like beveled PCB edges and gold fingers would add about 30 percent (combined) to the total cost. I don’t know the cost breakdown between the two features. 30 percent is quite a lot, and I’m debating whether people will be willing to pay 30 percent more for a card with these features. The prototypes have square edges and ENIG fingers, and it doesn’t seem to have been an issue for the beta testers, so I’m leaning against including these extra features.

I talked to the support staff at Circuit Hub, and they were reluctant to discuss using my own-designed Yellowstone tester. They felt that the quality of their assembly process was so high, with automated visual and x-ray testing, that additional functional testing was unnecessary. And they stressed that if an assembly problem were ever found, I could send the board back to them for free rework. I admit to lacking experience here, but this makes me uneasy. It puts the onus on me to actually test each board before it’s sold, instead of having the vendor do it, but that’s exactly the type of work I want to outsource. And it assumes that free rework would make it acceptable to receive faulty boards. The cost of faulty boards is mostly lost time for testing and troubleshooting, not the cost of rework. By the time I’ve tested a board myself, confirmed that it fails, diagnosed the problem, and identified it as an assembly problem, I could probably just fix it in five minutes with a soldering iron. I’m not going to ship a board back for that. But I don’t want to be doing that kind of work in the first place. To be fair, they did say they would do functional testing if I really wanted it, but their process doesn’t seem to be designed for this, and they implied that it would be expensive.

I have the general impression that the automated quoting vendors like CircuitHub and MacroFab are geared more towards low volume prototyping, or simple projects involving common parts. I think they may not have the most competitive prices either, but we’ll see. For anything slightly non-standard, or for the most competitive pricing, I think it may still be necessary to go the traditional route of phoning or emailing a vendor, and having an actual person-to-person discussion about manufacturing options and requirements. I’ll be doing that soon.

One other interesting option is Seeed Studio’s Fusion PCB service. They have a semi-automated quoting tool, but the tool still needed some manual help to finish my quote, and even then it wasn’t entirely correct. However, the total quoted cost was less than half the cost from the two other vendors. They even agreed to include functional testing for free, but other options like beveled edges and gold fingers don’t seem to be available at all. The big downside is the two-way trans-Pacific shipping that would be needed. I would need to ship all the FPGAs to them for assembly, which they don’t really like doing, and which would create extra paperwork hassles and tax/tariff concerns. I would still prefer to work with a local vendor, or at least a vendor in the same country, if I can find one that’s competitive.

Most vendors are quoting lead times of about two months, so if I can get started with manufacturing soon, I could have product ready for sale by March. The lunar New Year holiday is fast approaching though, and that usually shuts down most Chinese businesses for a couple of weeks. That may delay the schedule, depending on what vendor and manufacturing process I ultimately choose. But barring any further unexpected problems, final Yellowstone cards should be available by March or April at the latest. Woohoo!