Andrew Huang's Blog

The Ware for September 2025 is shown below.

Thanks to Michael Dwyer for submitting this ware! I originally contemplated only showing the digital board to make the ware more challenging, but the analog part is so chaotically gorgeous I had to share it out of the aesthetic appreciation.

Despite the size and complexity of the system, there is no CPU. There was a day and age where it was fairly common to design systems without one. In many cases, a ROM-based FSM was more economical, offering better performance and consistent timing. This gave them an edge over MCUs when the flexibility afforded by a programmable instruction set was offset by higher component costs and the difficulty of working with variable-length, multi-cycle instruction execution timings.

The Ware for August 2025 is the Superboard II from an Ohio Scientific Challenger 1P. Congrats to Tibor Bartos for naming it. Email me for your prize!

Here’s a quick tip for people learning how to solder small components: don’t rely on your eyes. Instead, solder by touch: use the force-feedback available through your fingertips. Human fingers can reliably detect bumps on the order of microns in size – much finer than the resolution of sight, even with the assistance of a decent microscope.

I use my eyes to get the components “about there”, and then switch to using fingertips for feedback. Really focusing in and just relying on vision for alignment tends to increase hand-shake. At least for me, the neurological delay on the loop from eye-to-motor control is longer (and more noisy) than the loop on tactile-to-motor control. So, I keep my eyes open, but allow them to wander slightly so I’m able to pay more attention to what I’m feeling with my fingers.

Try this out: drag a soldering iron tip across PCB traces or soldermask while averting your eyes (closing your eyes entirely might be too disorienting without a bit of practice). As you drag across the traces, you’ll notice a “bump” every time you encounter one. Do the same for pins on tiny components. Practice with a cold tip to get a feel for things, then add heat; pay attention to how the textures change with temperature. Eventually, you’ll be able to tell when the tip of the iron has met its mark by how it feels. Sometimes you can tell you’ve made a good solder joint just by how the metal feels as it melts; you can also tell when your tip hits a pin connected to a copper plane because the solder on the tip will freeze briefly on contact with the cold copper.

Likewise, the reason I use fine-tipped titanium tweezers isn’t because it’s better for gripping components – it’s because the stiffness of the titanium gives intimate feedback about the texture of the PCB & components. It’s also the same reasoning I use bare surgical scalpels to cut traces; if you mount the blade in a handle, you lose tactile feedback on what the blade is doing.

The other pointer I have is that efficient soldering happens not at the geometric tip of the iron, but rather at the spot where the solder is liquid on the tip (and if you don’t have shiny, liquid solder on your tip, clean the tip and put a small dot there). I’ve observed a tendency for beginners to use soldering irons like ball point pens, but in reality the surface tension of solder always pulls the liquid away from the point of the tip. If your components and pads are oxide-free (e.g. through the use of flux), the “right” amount of solder from a much larger liquid solder reservoir will disperse onto the solder joint through the laws of physics (via surface tension), assuming that your board was made using a modern solder mask formulation that naturally repels liquid solder.

Above, foreground: knife-edge soldering tip, with a dot of liquid solder near the point; background shows bare surgical scalpels and titanium tweezers.

This is one of many reasons why I prefer to use the “knife edge” soldering tip versus conical tips; the knife-edge’s versatile shape has a natural landing spot for liquid solder adjacent to a thin linear edge, backed by ample heat capacity. The effective geometry of the soldering point can be varied by adjusting the angle of attack relative to the work piece: the very end of a knife-edge is effectively a conical tip, while the blade is effective at delivering lots of heat and solder. It is often the tip of choice at many factories that I’ve toured. In the hands of an experienced operator, one can balance throughput and precision without changing tips, and it’s good for everything from through hole to fine-pitch TQFP, while also being effective at desoldering a range of small components.

The Ware for August 2025 is shown below.

Thanks to Curtis Galloway for contributing this bit of nostalgia! This board has the look of one that was laid out by hand using masking tape or rubylith – back in the day before computers became affordable and powerful enough to regularly use them for making new computers. It also looks hand-soldered, instead of wave-soldered. I only ever designed a couple of boards using tape, but even today I’m still hand-soldering boards – BGAs, 0201’s and all. I do a lot less of it than I used to, but you still gotta fix bugs and hack things the old fashioned way.

The Ware for July 2025 is a Vernier Lab Pro. While researching the ware with FETguy, we noticed that the OEM for the product is probably Inventec, which also made a line of products for TI calculators at the time. That particular OEM design team applied its design language in several products. I’ll give the prize to Jin who got the exact make and model of the board. Congrats, email me for your prize!

The Ware for July 2025 is shown below:

Thanks to FETguy and Renew Computers in San Rafael, California for contributing this ware!

The Ware for June 2025 is a Gentner EFT-900A frequency shifter for radio audio applications. According to Chris Combs, “it lets two parties send and receive ‘full-range audio’, whatever that means, compressed and then decompressed via the usual 300-3500hz phone line”. Even though Per named the ware first, I will give the prize to Kienan because the response addresses the theory of operation of the ware – shifting the audio up/down by 250Hz so that low-end audio isn’t lost on analog telephone lines. I found that bit very insightful, and now I don’t look at the ware the same way as I did before. Congrats, email me for your prize!

Also, this has never happened before, but the designer of the ware (Bill Gillman) made a cameo in the comment sections. I really enjoyed looking at the board, so as a token of thanks for designing such a nice ware, email me and I’ll also give you a prize.

The ware for June 2025 is shown below.

A big thanks to Chris Combs for this handsome contribution! Despite being 80’s vintage, the board is in mint condition.

The Ware for May 2025 is a Facebook (now Meta) Portal. I’ll give the prize to opticron for the quick guess, email me for your prize!

I actually considered redacting the audio DSP, but decided to leave it in because it’s an interesting detail. Despite having a big burly SoC with multiple cores and a GPU, the designers still opted to integrate a dedicated chip for audio DSP. I guess silicon has gotten cheap, and good software engineers have only gotten more expensive.

I’ll also give an honorable mention to Azeta for doing a great “spirit of the competition” analysis. Thanks for the contribution!

The Ware for May 2025 is shown below.

Because I really like to be able to read the part numbers on all the parts, here’s a couple more detail images of portions that didn’t photograph clearly in the above images.

This ware was donated to me by someone in person, but unfortunately the post-it note I had put on it to remind me who it was had long fallen off. My apologies; if you happen to see this, feel free to pop a note in the comments so you can be attributed for the contribution.

I made a point of not looking up the details of the ware before I did the teardown, so I could have a little fun figuring it out. While pulling it apart, the entire time I kept muttering to myself how this ware reeks of silicon valley startup with more money than sense. The hardware engineers who worked on this were clearly professional, well-trained, and clever; but also, whoever the product manager was had some Opinions about design, and incorporated lots of cost-intensive high-tech “flexes”, most of which I’m pretty sure went unappreciated and/or unnoticed by anyone other than someone like me taking the thing apart. For example, the board shown above is encased in a thixomolded two-part magnesium thermal frame with heat pipes, precision-machined thermal conduction blocks and gobs and gobs of thermal paste, which then necessitated a fairly tricky assembly procedure, and some brand-name custom-designed antennae to work around the Faraday cage caused by the metal casing. You’ll also note that despite the whole assembly being stuck in a metal case, each circuit subsystem still had an RF cage over it – so it’s metal cages inside a metal cage. This project must have had one heck of a tooling budget.

This was all for the sake of a “clean” design that lacked any visible screws. I’d say it also lacked visible cooling vents but ironically the final design had prominent ribbed structures, but they weren’t used for cooling – they were purely cosmetic and sealed over with an inner bezel. I feel like most of the cost for the thermal frame could have been avoided if they just let some air flow through the product, but someone, somewhere, in the decision chain had a very strong opinion about the need for a minimalist design that left little room for compromise. I would lay good money that the argument “but Apple does it this way” was used more than once to drive a design decision and/or shame an engineer into retracting a compromise proposal. Anyways, I found this product to be an entertaining case study in over-engineering.