Steep learning curve would have been good. To have a steep learning curve means that you learn quickly. What they had wasn't a steep learning curve. It was a nearly flat one: you learn and learn and don't make much progress at all.

Improving the build quality is an act finely balanced between improvement and profitability. They can't halt everything while they make improvements. They have a production pipeline and can't continuously rebuild in-process launchers because then they'd not be launching for a few more years. What you see is their chosen locally optimal point between latency of a launch vs. launch throughput.

There's no such thing as antibiotic soap. There is antibacterial soap. Those are very different things. The antibacterial compounds can't be and aren't taken internally and have nothing to do with antibiotics.

Ah yes, the AC problem. I concur. Initially I'd restart the car while driving (no need for starter!) and it usually cleared it for another 20 minutes or so. Eventually I wired power through a manual switch directly to the A/C clutch to override the silly climate controller. I only managed to freeze the evaporator once or twice :)

One stereo problem is due to poor mechanical design of the front panel board, the traces crack and it doesn't respond to keypresses anymore. I repaired it twice before the car was totaled.

Shocks and engine mounts, yes, exhaust from the manifold back, yes. What a pain it was.

That 1st gen S40 was a nightmare. I owned it. It saved my life, but I don't miss it at all. All of them are lemons as far as I'm concerned. At least I got it for cheap. I don't think it was US made, though.

I agree, and I've in fact replaced the spark plugs in my 6 cyl. Volvo after 140k miles. The only discernible difference was that after starting it up, the idle was around 1200RPM whereas normally it would be 800 or so. The idle adaptation took care of it in a couple of minutes. Fuel mileage didn't change. The old plugs really looked like crap, with the platinum spikey electrode completely gone and the thin platinum wire being recessed about 0.2mm into the insulator. They still worked fine.

And all of the hundred or so chips that are in the car are made in Asia :) Without those chips you'd have a rather big paperweight :)

The quality of "stuff" made in China is usually exactly what the stateside customer orders. Cheap chinese toys aren't of poor quality because they're chinese, they're of poor quality because they are ordered to be so. If a U.S. customer is paying 10 cents for an injection-molded trinket, they get 10 cents worth of a trinket, no less, no more. Design for some of this stuff is still done outside of China, and if the design is crappy, you can't blame the factory for that. Heck, if the chinese design is crappy, then it's still approved by the stateside customer to be just that - a crappy design. The factory then executes the crappy design, and you get a crappy product. Sure,

Asia is pervaded by a cutthroat culture where there's no problem with lying to your customer if only it'll get you financially ahead. But that's just a small part of the problem. The biggest problem is the corporations that approve, order and sell those poorly-designed and poorly-made products. It's ultimately their explicit choice that things are so. We, the customers, pay for it.

Some of the problem with customers is simply their general disdain for everything technical. It seems to be a badge of honor for people not to be able to do the most basic of inquiry into the quality of what they are buying. A tiny bit of engineering fundamentals could be taught in high schools, for crying out loud, if for nothing else but to make people less sheepish when it comes to purchases.

Yeah, sure, you can be running it bare metal, if that's what you mean.

It's only realtime if you can get that GPU to do its calculations when driven from, you know, a real-time operating system. That is often a big problem, as some GPUs come with laughably incomplete specs and there's no way to use them without relying on OS-specific, non-realtime driver binary blobs. Raspberry PI SOC's reverse-engineer is slowly coming to a state that lets one its GPU to do truly realtime computation. It's one of a very few. Maybe Intel documents their integrated graphics sufficiently for use in a RTOS. Other than that it's probably all at the level where you can do proof-of-concepts and can't go beyond that.

Almost any off-the-shelf logic element that's easy to get those days is very fast and has risetimes on the order of 1-10ns. It is actually harder to buy slower logic families, they get discontinued left right and center. You might be flipping that gate at a 100kHz, but it can have harmonics going past 1GHz. People often forget that little detail and wonder why they get interference between "slow" circuits etc.

While you're mostly right when it comes to cookbook designs, once you are off the cookbook path you can certainly design circuits that have better than 0.5% accuracy over temperature and such. Yes, they are not trivial to implement, but it can be done, and once done it's pretty damn impressive in my book at least. This is somewhat valuable when your analog signal chain is part of a safety-critical system, where the software has to be validated and has such an overhead that it's actually often cheaper to do as much work as possible in the analog domain.

And heck, these days it's no biggie to buy an off-the-shelf op-amp that has worst-case offset spec better than 20 bits in 0-5V signal range. The stabilization circuits inside are discrete-time, but still analog.

Even in the times of 8-bit CPUs you'd see absolutely awful circuit designs and layouts that radiated and conducted hash like crazy, and then when the EMI tests were done they had to work around poor PCB layouts etc., adding shielding, chokes and whatnot. All the while someone who knew what they were doing could have likely designed everything to pass at the first attempt. I've had industrial drive systems pass emissions at the level called for in medical equipment. As an end user, you'd be glad you used those instead of something that passed emissions with a nod and a wink and corrupted the signals from any analog sensor located in its vicinity.

And that's why I do my own data acquisition, because if all I wanted was cookbook designs, I could have just bought a DAQ card. It's only when you need something that is better than cookbooks that you need to look inwards, and if you want to be competitive you often need something better than a cookbook.

Heck, when you're validating high speed interconnects on your PCB, you are also looking at the digitized form of the analog signal present on the differential data pairs. This requires some rather specialist knowledge to be done properly, if for nothing else than not to destroy the multi-$k differential probes used in such setups. Never mind the oscilloscopes that can actually do something useful with the signals the probes feed them. I don't do any bleeding-edge work in this area, but even I have a few probes that were $800 used, and it was an exceedingly good deal for them, too. Oh, and a $500 signal source needed to make sure that the probes actually work, even if in a pinch I could have just probed a few "standard" signals on the motherboard of my desktop.