Talk:Tektronix

I worked as a technician in the MDS division from mid-1981 to mid-1984. I started as a test technician, worked with engineering on the 8086 emulator prototypes, and then came back into manufacturing as an 8086 emulator specialist. This article strikes me as one in which the author did his best and put a lot of work into researching, but had a very hard time finding any substance. MDS was a pretty obscure division even within Tektronix at that time and the market was development engineers, not consumers and fanboys, so it is not surprising that little information survived about the progression of product developments through the life of the division. Unfortunately, I will be of no help because Wikipedia requires cited sources and I have none. Just my aging memory.

I would, however, like to address some things here which may help others to find cite-able information.

1. TNIX. TNIX was the OS that the MDS ran on from the time I started there in 1980. It was so much a part of the system that I cannot imagine that one ever existed without it. It is important to realize that large technology companies were often deeply involved in tech projects while they were still at the university level and also worked together in various development consortiums. For example, we had access to a TEK Magnolia system with the mouse and graphical interface while it was still in development - presumably because our systems were used to help develop it, though I don't know that for sure. The Magnolia was developed in the same timeframe as the Apple Lisa and both as an outgrowth of the Smalltalk project. (We were bummed that TEK never productized Magnolia!)

2. Emulators. Little is said in the article about emulators which is ironic because the entire point of the system was the emulators. Yes, you could develop software on the system without an emulator, but you could do that on any computer. The emulators allowed the developer to actually run his software on a real microprocessor of the target make and model. And it allowed them to physically unplug a Microprocessor (CPU) from a breadboard or circuit board and plug an emulator "probe" directly into the CPU socket. Then the CPU could fetch instructions from MDS memory or from the prototype's memory. It could execute them just as the CPU would, but the developer could set breakpoints or other logic triggers and get traces of what happened before or after those triggers. It could also interact correctly with I/O and interrupts and such. It allowed a developer to rapidly test and tweak and re-test software and see what it was doing right or wrong without having to burn ROMs for every tweak. And it did all of that in real time by creating a sort of onboard parallel version of the target microprocessor using bit-slice microprocessors, very fast RAM and very fast logic.

3. Their demise. The article states "It is not clear at present why Tektronix eventually withdrew from the MDS business, as their products were highly regarded." While this is very flattering (thank you! we tried really hard) it isn't really hard to understand why they closed down the division when they did. Take a look back at my point number 2. Then consider Moore's Law and the exponential growth in microprocessor technology during the 1980s and early 1990s. Our earlier emulators were on single circuit boards which could be plugged into an 8560 chassis along with selected memory and other peripheral boards. But the 8086 microprocessor had a sprawling set of logic and instructions, larger address and data buses, as well as just a lot more pins than earlier CPUs had. And we had to have circuitry in our emulators to support and simulate all of that stuff so that the probe could emulate an 8086. As a result, the 8086 emulator that I worked on required three circuit boards all cabled together and the fastest logic we could get. We managed to design an 80186 (and I think an 80286 after I'd left the company) but that was about it. Microprocessors became too fast and too complex to discretely simulate within the hardware of the emulator boards. Also, when you spread your simulated CPU hardware over multiple circuit boards you introduce the need to run cabling or more rigid buses between those boards and you induce lots of delays and timing problems due to the long distances that signals must travel and to inductive resistance in the connectors and so on. There are just limits to how large and how fast you can go with TTL logic on printed circuit boards and with bit-slice processors running discrete versions of the CPUs' ever more complex instruction sets. At the same time, software which simulated circuitry had become sufficiently reliable that real-time emulation was less necessary. I know they ran on for quite a while making emulators for simpler microcontrollers, but those devices also gained in speed and complexity and the market was smaller, so it just became time to close a no longer profitable division. FatBear1 (talk) 00:10, 19 October 2024 (UTC)Reply