EEVblog #1060 - $35,000 DataIO Unisite Universal Programmer Teardown!

Marty! This is the big one, the one I've been waiting for all my life!
Teardown of the classic Data IO Unisite Universal Programmer from the 1980's, with a typical configuration selling for $35,000
It stayed in production for over 20 years.

UPDATE: From one of the original designers!

[kelleecrisafulli]

Great tear-down Dave I really enjoyed the review. My team developed the mother board the Toshiba Gate Array and the Ram board and we were responsible for the Pin Processor design you may not have noticed. All the Toshiba Gate arrays (one per quad pin driver) work together with a dual ported interface to the 68000 to form a variable width instruction word for the distributed associative Pin Processor. The UniSite is still likely capable of programming most parts that exists from a hardware perspective. It can run single ended or differential signals, any logic level, any current level, and any voltage level within reason and can generate any required pin waveform within reason up to several MHz using the Pin Processor. If I recall correctly the only part it didn't program is the old 1702 because of the crazy high voltage that part required. The Pin processor instruction word grows wider as more pin drivers are added and it can grow to many hundreds of bits wide. The Toshiba Gate array was pin configurable to support either 4 universal pins or 16 logic only pins for gang programmers. Designers that worked with the DIP version of the 68K will likely guess the reason for the ferrite bead soldered to the 68K CPU strobe line is due to internal ground bounce in the 68K which occurs under heavy bus loading. It was decided the cost of a board spin was too high to justify a new PCB layout, not sure if that was ever fixed. I also designed a 68020 upgrade for the UniSite but the board was never layed out. At the time we designed the UniSite I received a lot of questions about why I would put a "Workstation class" CPU into a programmer. About two years later the 68K was already low cost so I had no regrets and my manager always supported the 68K choice. The software group blew through the first 640K of ram before we could release the UniSite and I had to scramble to build them a memory expansion board which was released and running two weeks later. The 68K was one of the few CPUs available at the time that could address above 1M of RAM without paging. There were a great many design challenges creating the UniSite and we had an amazing team of mechanical, software, digital, analog and SI engineers to devise solutions and make them work. We also created a couple lower cost versions of the UniSite which programmed fewer parts: the 2900 & 3900 and also a really low cost version initially called Chip-lab. All of these version used the same programming algorithms and the same Pin Processor design. A newer version of the gate array chip was even developed that quadrupled the number of pins supported and integrated many of the analog pin driver functions into a single full custom chip. All ancient history at this point :)

[NathanaelNewton]

These 'random niche super expensive custom thingies' tear-downs are my favorite.. The giant ancient hard drive comes to mind.. This was an awesome video, must have been so much work to edit, thanks!!!

[bostedtap]

Incredible Electronic Engineering, "Universal" is easy to state, but far more difficult to truly achieve. Looking back can be both demoralising and satisfying, but always an achievement. One of the best architecture execution I have seen. Many thanks for sharing this Mr Jones.

[kostaskritsilas]

We had one at Philips electronics in Montreal. The gold sheet conducts only in the vertical direction (through the thickness of the sheet) and was a significant wear item for PLCC packages. A sheet would wear out every 5000-7000 programming cycles. Our unit was originally dual floppy; eventually went to the hard drive version, which stored all of the operating system, and device models. Ours were used to program mostly PALs, GALs, and EPROMs. Unit was about $20K Canadian, full pin cards, but with only the DIP module. PLCC module, SOIC modules, and the PQGP modules ere extra, as were the adapters for each package. Service contract was $4-5K/year, which got you all the new device models, updates to existing device models, OS updates, and hardware service (we carried spare driver cards, and just sent the bad driver boards back for repair). The memory expansion board was needed for more complex devices as time went on (more complex GALs, FPGAs, high density EPROMs). In our facility, FLASH was programmed on the ICT after they were soldered to the board due to the FLASH pins being so delicate (SOIC .020"), and because the Genrad 2287 was as fast as the Unisite. It was eventually replaced by the Data I/O 3980, which had a higher pin count, and could be used with automated IC handlers.

[herbertsusmann]

I remember using a smaller version of this programmer back in the 80s. They were the gold standard for sure. Absolutely reliable and bullet proof.

[ctoforhire]

Thank you for the nostalgia. Data IO was just down the street from an electronics surplus company I worked for as a kid. I could practically smell that programmer when you cracked the lid open ;)

[greenvm]

For anyone wondering: "The Unisite was the flagship model of the Unifamily line, selling for over $35, 000 in a typical configuration"

[TheGreaseySpoon]

DataIO will actually buy back any of the old programmers to re-furbish and sell on as they are still used and in demand but they do not manufacture them any more.

[mikeselectricstuff]

Those Toshiba parts are not programmable - they are "sea of gates" ASICs where just the final metal layer(s) are customised for the desired function, so cheaper than a full-custom ASIC.

[Rob-wcjw]

I remember when they first came out, I used to calibrate the DATA I/O devices as I was an engineer at an Office in Nashua NH in the States I would still be there but they shut the office down and move it closer to the customers which was NC Triangle Research Park where all the East coast high tech companies lived, equivalent to Silicon Valley. This unisite and the gang programmers were state of the art along with the other company they bought which was a lazer label company that marked the devices with labels using lazers instead of the white paper labels which would wear out over time. Great company they always kept up with the times always thinking ahead. My first employer out of college, fun times there. Enjoy you video.

[electrofan]

Damn Dave, this programmer is absolutely badass! So complicated engineering here! I'm really impressed. You definitely must to fix and try it in action!

[alanholt]

We still use one where I work (avionics repair).

[sleibson]

H&R is Hunter and Ready. They developed the original commercial RTOS (VRTX) and essentially created the commercial RTOS industry.

[electronicsNmore]

Great video Dave, and congrats for passing 500K subscribers!

[WestCoastMole]

Thanks Dave for the trip down memory lane. You brought a tear to my eye.

[mikeselectricstuff]

The real fun stuff was for bipolar PROMs and PALs that specifed all sorts of wierd voltages as well as rise/fall times - EPROMs and MCUs were trivial in comparison

[kennethryan]

Notice the jacks "terminal" and "remote". Not only could you hook up your VT100, but the unit can sit inline to the terminal server to your VAX. You could get the programmer ready then log in to your account on the VAX and download the hex file directly to the programmer, then switch back to local operation to burn your chip. Very convenient!

[pocoapoco]

I have to say, those bodges are some of the most skillfully done that I've seen in all of your teardown vids.

[Bushougoma]

17:26 those relays are made by Fujitsu.

[qzhk]

We wore those programmers out keeping up with changes back in the day, Y2K planning and manufacturing systems across the continent and pond. Who knows AGV AVL? Thanks for the memories.