This seems like an exceedingly bad hill to die on. Curtis Flowers was convicted and sentenced to death six times for the same murder by juries of his peers, yet he's a free man today.
These boards were basically laid out using masking tape on transparencies. The rounding comes naturally from turning a corner with the narrow tape. The teardrops around pads and other more complex shapes (like footprints) came from pre-fabricated stickers. The layouts were produced in 4:1 or similar scale and were optically reduced during photolithography of the resist on the actual PCB (just like semiconductor masks).
Way back then you also had some concerns about manufacturability because sharp corners tend to be undercut during etching when your process is poor. That's probably why they usually added little fillets to T and similar connections.
Do they actually bury PON components? Because around here they don’t. Fiber runs from homes to their concentrators and those house both the PON splitters and the OLTs. There’s some roadside boxes as well but afaik they’re only for splices, because those aren’t buried, either.
>100 trillion reads per location over 30 years still means you gotta read locations at over 100 kHz 24/7. Not good enough for main memory, sufficient even for frequently accessed configuration values.
Siemens did real mainframes and their mainframe OS BS2000 is still around, it's just part of Fujitsu, Nixdorf appears in that story as well because that's how the Siemens mainframe division ended up at FSC (Siemens acquires Nixdorf, folds its mainframe division into that, then splits it up into the ATM business and sells the rest to Fujitsu).
Nixdorf shut down their mainframe business in 1989, and sold the remnants to Comparex (which started out as a Siemens-BASF joint venture, but Siemens withdrew around the same time as Comparex acquired Nixdorf's mainframe business). So when Siemens and Nixdorf merged in 1990, Siemens did not acquire Nixdorf's mainframe business, only Nixdorf's other product lines (Unix systems, ATMs, etc). But Siemens still had their own mainframe business. Comparex already sold IBM-compatible mainframes, so they didn't continue Nixdorf's mainframes as an independent hardware line, they were primarily buying the support contracts and the customer base.
Siemens mainframes and Nixdorf mainframes had significant differences:
Siemens BS2000 mainframes were derived from RCA Spectra 70. Their ISA was mostly IBM-compatible in user mode (problem state), but significantly different in kernel mode (supervisor state), and their operating system was completely incompatible–the BS2000 operating system was derived from RCA TSOS. RCA sold their mainframe business to Sperry, who then merged with Burroughs to form Unisys. The RCA Spectra mainframes became Unisys' Series 90 mainframe line, and RCA TSOS was renamed to Unisys VS/9. But by the 1980s or early 1990s, the RCA-derived Unisys mainframe line was dead. Whereas, their Sperry and Burroughs heritage mainframe lines (Unisys OS 2200 and Unisys MCP) survive today, although now they are software emulators running on x86-64 servers instead of physical hardware. RCA Spectra/TSOS only survives today in the BS2000 branch, save that Siemens ended up selling it to Fujitsu.
By contrast, the Nixdorf mainframes were more straight IBM clones, and so aimed for instruction set compatibility both at the user application and operating system level, and could run IBM operating systems. They were mainly used with the low-end IBM DOS/360-derived operating systems rather than the high-end MVS operating system family. Nixdorf faced the same problem that Fujitsu and Hitachi did, of IBM closing their operating systems, but they solved it by buying the American software company TCSC, who maintained their own fork of the IBM mainframe DOS, called Edos, which Nixdorf then renamed NIDOS (Nixdorf DOS). TCSC had started Edos when IBM decided to make new DOS versions available only for S/370, not for older S/360 machines, hence Edos was originally a backport of those newer S/370-only DOS versions to the older S/360 machines. When Nixdorf bought TCSC, they renamed it NCSC. NIDOS ended up offering features that IBM DOS/VSE never had, like a Unix compatibility subsystem (PWS/VSE-AF, derived from Coherent) – much latter, MVS (now z/OS) and VM/CMS (now z/VM) ended up getting one, but DOS/VSE (later z/VSE and now VSE^n since IBM offloaded it to 21CSW) never has.
Siemens also once had a lower-end mainframe line, which ran an operating system optimised for smaller machines, BS1000. BS1000 was discontinued long ago, and there is little information about it online. There was a BS1000 compatibility subsystem for BS2000, called SIM-BS1000 [0], but I'd be surprised if anyone is still using it today.
And Siemens also had BS3000 mainframes – like Nixdorf mainframes, these were fully IBM compatible, and designed to be able to run IBM's operating systems – they ran the Siemens BS3000 operating system, which was a rebadging of Fujitsu MSP – Fujitsu stolen version of IBM MVS. Siemens had to enter into a settlement with IBM as a result, although I'm led to believe the terms were relatively lenient on Siemens, who did their best to portray themselves as innocent victims of Fujitsu's dishonesty. But that was the end of BS3000. I think the remnants of the Siemens BS3000 line ended up with Comparex too. Comparex finally shut down their IBM-compatible mainframe business in 2000; they survived as an IT services business until 2019, when they were acquired by SoftwareOne.
And then in 1999 Siemens transferred their mainframe business to the Fujitsu-Siemens joint venture, and in 2009 Fujitsu bought out Siemens, and hence Fujitsu ended up with Siemens mainframe business.
And so today Fujitsu has three totally incompatible mainframe lines – their own Fujitsu MSP mainframes (previously sold internationally but now only surviving in Japan), the ex-Siemens BS2000 (primarily surviving in Germany, although a little bit in the UK and a few other European countries), and the VME mainframes they got by buying ICL in 2002 (I believe the UK government is the sole remaining user, they really want to migrate off them but it is just too hard.) Both BS2000 and VME now run under x86-64, while I believe the Japanese line still has proprietary physical hardware.
There’s a trade off between sharpness and noise, the GFX have an intentionally lowered fill factor to, essentially, produce a sharper image. Meanwhile noise is one of the most important things when marketing mainstream cameras (next to AF), so they go for gapless microlenses etc.
The reason this impacts sharpness is because a lower FF gets you closer to Shannon’s ideal point sample, while a 99% FF is like a pitch-sized box filter.
There is also a tradeoff between sharpness and aliasing, that's a bigger driver for microlenses than just capturing more photons. A point sample is only ideal if your sample resolution is above the nyquist frequency which for the real world it won't be.
Yes, this was nicely highlighted by the GFX50 vs. GFX100. Both are around 50% fill factor and have no OLPF, the GFX50 produces a lot of aliasing artifacts, the GFX100 much less so, because Nyquist moves up some 40%, so diffraction takes more readily care of attenuating these higher spatial frequencies.
The 32 bit variants are accelerated via SHA-NI on most CPUs, which inverts the performance ranking again, making SHA-256 the fastest common cryptographic hash by far.
reply