The software communicates with the card by directly reading and writing to I/O addresses of the card. No modern OS or virtualization would ever allow that. It's why you can't have sound in old MS-DOS games unless you boot DOS or Win9x, or emulate the complete system.

If you need any kind of specific hardware other than a 3.5" floppy drive, an ordinary VM probably won't work easily, if it can be gotten to work at all.

Even with something as simple and ubiquitous as a parallel port, buying an old computer is probably going to be simpler and more reliable.

With Windows, it's not as simple as emulating an S100 bus. Windows 95 had about fifteen million lines of code. That's before you add in whatever kludged, cowboyed, and cobbled together code your hardware and application actually need.

Basically, every old windows hardware driver is a corner case and everything had a hardware driver. Often one ported from MS-DOS.

Nope. Sorry.

Thanks for playing.

This is what our IT people had set up to run Windows XP, in VMware for a couple older scientific instruments they had transferred from their "modern" main lab before I got to my final employer about 10 years ago.

Those two were not pretty but all my other instruments were also a shitshow on various vintages of the original bare metal, which were within reach so I set to work wiping and properly reinstalling Windows XP on rarely-supported NIB 2014 PC's, along with the vintage proprietary software, and productivity was multiplied manyfold.

Took loads of hours over a couple years but somebody needed to do it and IT's hands were full. I'm just a science dude but isn't computer science still a science ;)

Their VM's worked like a real dog but they weren't putting any effort into it and it was too encumbered with licensing for me to do anything equivalent on bare metal, plus like a few other things the company was using it like that before I got there, so I needed to strongly neglect those by not touching it. They also still were using VM's for their office machines and still imagined they might have been a lot more sensible than it turned out to be, and they were not yet in condition to be reasoned with otherwise.

So it was not my fault at all when the robotic arm destroyed itself under virtual control :\

After that I had more complete co-operation from IT than would have ever been possible, and got those two PC's out of their hair and onto bare metal finally.

Regardless of how fast modern PC's have gotten, the simple stuff like mouse movements & file manager have gradually become worse than W95 was on truly low-spec PC's by comparison. Anything more complex and you've got to be kidding. Compared to today, DOS and Windows 9x were still like a "real-time" OS since they had no Hardware Abstraction Layer, unlike like the NT forms of Windows which introduced HAL to the masses and ended up prevailing in the 21st century. Virtual is just not as good as the real thing, and with a HAL in the way for most of the 21st century it's not as real as it used to be with the same instruments.

Which really loved it on DOS, but virtual is asking way too much.

It's not really obsolete electronics, it's a people problem which evolved from extinct ancestors so long ago you couldn't retrace your steps if you tried :)

The IT operators need to keep up with networking, office machines, security, and stuff like that, that's where they are advanced and their hands are always going to be full.

The instrument operators themselves aren't going to be writing software or even manipulating it much, nor networking or keeping up with emerging security threats or other moving targets. Not unlike the CNC lathe operators, they just want the mission-critical machines to keep on doing the routine work the same old way they did when first purchased. People have "grown" to expect IT to handle "the electronics" to some degree or another which logically includes the software. But sometimes there are electronics and software that are so far out of the ball park compared to where the IT talent really lies, that things are not getting better by design.

A dependency on a type of IT that never really existed has developed, and things are moving right along rapidly and more nonideally than more advanced people during vintage times would have ever imagined :(

It's just too bad that so many different proprietary devices were either first computerized in the 1990's, or at that time began to rely on what are by comparison cheap office machines ever since, and sometimes on OS and/or software which has an even shorter lifetime, by design.

And just like the lathe operator, the expensive, truly well-built, long-lasting analog hardware at some point became dependent on a particular generation of office machine which is cheap junk by comparison, and even cheaper vintage OS when you do the math, even if the proprietary drivers and software from the industrial manufacturer is quite expensive on its own, and far more rare and "irreplaceable" to begin with.

Why should it not only be more complex under the hood, but also more of an ordeal, and more dependent on "support" from outside the "workshop" for it to end up less reliable, more frustrating, and require more time consuming effort from more different people more often under emergency conditions to all do their job without fail 24/7 ?

Just so you can get a spectrogram almost as reliably as you could before computers, and PC's were supposed to be an unqualified improvement.