> Sinclair products were almost comical in how ingeniously their engineers shaved off any pennies they could.

Yes, that was their rule from day 1.

My favorite: in the early days, Sinclair designed a very cheap transistor radio -- so cheap that it didn't even have a speaker, just an earphone (mono, one ear). Sinclair noticed that the available earphone jacks had a built-in switch that was intended to disconnect the speaker when the earphone plug was inserted. Since there was no speaker, this switch was repurposed to serve as the on-off switch for the unit -- plug in the earphone and it's on, pull it out and it's off. That saved both the cost of a separate switch and the space it would have taken up.

I believe this hack required manually bending the contacts on the jack switch to turn it from NC to NO.

It's a fantastic example of 'good enough', because the alternative was doing this all by hand or using mechanical tools which - while sometimes quite fast - were error prone especially for long and repetitive calculations.

An Olivetti electronic/mechanical calculator for instance took up half your desk, was horribly slow and consumed a ton of power. No way that you would ever take it with you and working with the beast would require a serious investment in time, they weren't cheap either.

These little calculators were complete game changers. The model in the article was one of my prized possessions, there simply was no alternative that I could have afforded.

> "three decimal places of accuracy at best"

To be fair, that's all their competition (a slide rule) was capable of.

If that. I used to be pretty good at reading them and depending on where you were in the range two would already be quite impressive.

>without the ability to set individual pixels in any way

You could and I did. ALas it involved rewritting the screen driver and drawing a diagnol across the screen would take over a minute and all the CPU, now if you turned the display off (ala fast mode polling) then you could improve upon that.

One stand-out thing that puts perspective upon the ZX81 was as you moot - the price for what you did get at the time, a time in which you could pay more for a pack of 8" floppy discs or a keyboard.

It's ironic how you point out the short-cuts - something Amstrad had more a reputation for. With that I'd tend to agree that the case design of the ZX81 was a work of art in design.

How exactly did you set individual pixels on an unmodified ZX81 without extra hardware? As far as I know, that’s not possible, even when rewriting the screen driver. The font came from ROM, and while you could do things like switch where in ROM the font starts with every scan line, that only gave you “Pseudo HiRes” capabilities where not every combination was possible, see e.g. https://perfectlynormalsite.com/zxhires.html

Was few ways and sure changing the ROM, but with RAMpack you had the space to redo the interupt controlling the display and probebly best demoed with https://www.youtube.com/watch?v=EjA-qlDv5Xw Though for this they just pointed the character map to some RAM to do UDG's for the effect.

I'd add though to do true resolution you kinda had to update things a lot of times and why the code I did at the time was in effect doing 8x8 more work - hence was about 2 mins just to draw vertical line. Just not enough processing power to ever be of use. Though the remap the interupt into RAM and char set was good and lost no performance and indeed you would expect it to be slightly faster with the old RAM vs ROM access times, but kinda mute back in those days of when RAM was faster than the CPU's.

Yeah but that was my point, you at the very least need a RAMpack, and often one with modifications. It is not just for having the space to have the custom display code, you actually need the RAMpack for the data lines on the bus. Without a RAMpack, the ULA would fetch the character rows from ROM and ROM only, no way around it. And not all RAMpacks worked for this, a lot (most?) had to be further modified in hardware.

The original device without RAMpack did not let you set individual pixels at all (and even with a suitable/modified RAMpack it was a giant hack obviously).

I got curious whether Manic Miner actually found a way to do actual hires without added hardware, but this answer claims it's also (just) pseudo hires... although a rather impressive example of it, I wasn't sure from the video you posted: https://retrocomputing.stackexchange.com/questions/6134/how-...

“only two screws hold down the PCB to the case“

Nowadays, that’s two more than a product with cutthroat pricing should have. Engineering tolerances, even at low cost, are good enough to clamp the PCB in place.

Are you Munroe Live on Youtube? It's a channel where they disassemble and review the internals of EVs. Any time he comes across a screw (threaded fastener), which happens all the time, he turns red and politely hints how the designers could have done a much better job.

The funny part is that they obviously planned for much more.

Or their manufacturing was so imprecise that they had to make more than two holes to (most of the time) get at least two that were in the right-enough place and where the screw would actually “grip” the plastic.

I expect, though, that the first batch used all holes and that this was a matter of using excess inventory with too many holes after realizing that they could do with fewer ones.

Interesting with the imprecise manufacturing, haven't thought about it like that (and obviously don't know anything about the mechanical part of manufacturing).

I'm better at judging the electronics/computer design, and there it's obvious to me that they went out of their way to stay cheap.

Is this any different than what Chinese manufacturers do today when they knock-off products?

I think so, cheap knockoff products nowadays usually use cheap parts and put them together in cheap but obvious ways. If you dive into Sinclair engineering, it’s contrived and interesting.

“Real artists ship”