Cree Mount wire to board

On a sunny day (Fri, 03 Sep 2021 07:24:28 -0700) it happened snipped-for-privacy@highlandsniptechnology.com wrote in snipped-for-privacy@4ax.com:

Bull, that wiring was done in an evening. And you and you layouts in whatever program would still have to have the board made days if not weeks later find out you need to modify it if it works at all !!!!

Quick turn-around? Worked the same day!

You are blinded! Sort of forced into some pee_see_be jacket,

Depends from where in space you look at earth You humming beans are soooo ancient,

Boards (plural!) can be made in a week, and I can do something else meanwhile. My boards usually work first-pass because I check my work.

Reworking wire-wrap can get hairy.

One nice thing about pcbs is that they are repeatable and documented.

No, I get a lot of stuff done. Fast in both senses, picosecond signals on boards that get made quick. Wire-wrap is not suited to fast stuff; signal integrity is awful.

I do simple circuits as dremel'd copperclad, running in an hour and where I can control impedances and get clean sub-100 ps edges.

On a sunny day (Fri, 03 Sep 2021 08:56:59 -0700) it happened snipped-for-privacy@highlandsniptechnology.com wrote in snipped-for-privacy@4ax.com:

True,. But I only needed one.

Re-soldering wires (if ever needed??) takes a few minutes to heat up the iron :-)

Try bending one, mine will usually have no problem with that.

Look again at the links I gave, I am NOT talking about wire-wrap.

The Saturn PCB Toolkit is great and does all sorts of PCB calcs. It's free and seems to be pretty accurate. It does weird cases like asymmetric microstrip and such.

snipped-for-privacy@highlandsniptechnology.com wrote in news: snipped-for-privacy@4ax.com:

Ever see how the original Cray was wired up?

Now your phone beats it in processing power, but it had thousands of point to point interconnecctions. It was actually one of the elements that 'set' (read throttled) it's speed capability.

We used to worry about the number of vias going into a PCB design, because back then a via could be (and often was)a point of failure. It still could, but we have come a long way to that being a very remotely possible point of failure, and nobody worries any more about a via or a via count being an issue. And we are doing way more layers now than we were then, so it is orders of magnitude more reliable now than it was then.

Wire wrap had its time and its place. There was a time when the practically gas tight connection that could be made from point to point with wire wrap was more reliable than a PCB pock marked full of vias.

The design under review connects a single loose element (the power LED) to a printed wiring board; it'll NOT be part of the printed wiring, it's more like a board-to-board connect opportunity. There, wire-wrap is a design winner.

That would be a simple board, but yeah. Larger boards would take a while, but the engineers, like with many things, were not allowed to do the wrapping. That was for people trained to do that work, and paid appropriately.

I had a board made like a wire wrap board but using a 3M insulation equivalent. It could slide into the rack we used for printed circuits without having to leave a slot empty to make room for the pins. I used it for an I/O board prototype for a super computer... at the time anyway. At 100 MFLOPS it was the second fastest machine in the world next to the Cray super computer until Cray came out with their second machine and ours was the third fastest, lol. But we sold a couple of hundred of these machines compared to what, a few dozen of the Cray 1? I think the ST-100 was about a quarter million without memory. Add in various amounts of memory and it could reach a million. The Cray was more like 2 million I think.

It was never a big deal if you designed the wrap order so as to not have daisy chains where you have to take them all off to get to the first one. Instead sequence them like laying bricks, 1 to 2, 3 to 4, 5 to 6, etc. then come back and wrap 2 to 3, 4 to 5, etc. Then you never need to remove more than three wires.

Wire wrap isn't documented???

Duh! That's the reason why it went out of style in the late 80s or so. I remember one company in Mananas, VA that did nothing else. Rather than change with the market the owners were ready to retire anyway and just let it dwindle away so they could retire. Sure seems like it could have transitioned to PWB assembly or something and sold the thing for a few million to retire in style.

Yes, exceedingly simple. The only real limitation to wirewrap is the speed. That's essentially what killed it. It would be another 20 years before PWB fab became low cost enough to use for prototypes without high cost. In the mean time other technologies were tried like stitch weld which had better impedance as it could be routed on and embedded in a glue layer on the board as it was all machine produced. Still not great, but it worked better than wire wrap at speed.

Making protos is a very different world today than it was 40 years ago.

they even had machines doing automatic wrapping from a netlist (probably on punch cards or tape)

Absolutely. Anything at all complex would need to be done at least with machine assistance to reduce the errors because of the difficulty in finding wiring errors on a complex board. I think the outfit in Manassas used a data controlled light to illuminate the pin to be wrapped, but I might be confusing that with chip insertion. I know they did something that was short of full automation because they had a certain amount of loyalty to their workers.

I seem to recall wire wrap was touted because rats nest routing greatly reduced coupling between adjacent traces, but did double sided PCBs have much problem with that really? I don't recall multIlayer PWBs being very common at that time due to the expense. I guess you use what you have to use to make the connections. Some of the boards in the ST-100 were 16 layer with one layer being an "omega" layer which included resistance for the ECL terminations. We would insert the tip of the Tek probes into the vias for probing and of course they would occasionally be broken off. After a while they gave us different probes. Tek parts were never cheap.

I recall in 1969 I was working at Memorex on a replacement for the IBM 3330 hard disk drive. Our version was called the 3670:

The automated wirewrap machine took all the unnamed nets and connected them together. It took weeks to find and disconnect them all.

I always cringe when I see unnamed nets in a design. I think it shows a level of laziness that is unreasonable. It also makes it very hard to talk about a design. I think problems like you saw would not happen because every net is named now, if only made up names like 1234_$ or something equally screaming "Not an intentional name".

On a sunny day (Fri, 03 Sep 2021 12:20:33 -0700) it happened John Larkin <jlarkin@highland_atwork_technology.com> wrote in snipped-for-privacy@4ax.com:

Thanks

Refer to them by the ic/pin. Name them only if they have particular significance. There's little point in giving a name to, say, the net between the R and C in a snubber.

The bad side of naming every net is that it encourages youngsters to draw schematics where each ic is "isolated", and to follow connections you have to search for the net name.

Refer to what by ic/pin??? Each net has at least two points. Naming nets is no different from naming ICs or pins. Do you have parts in your design without names? That's why modern schematic capture programs name all nets even if you don't.

Silliness. Do you yell at the kids to get off your lawn too? Bad schematics don't need encouragement. A great way to produce a bad schematic is to not name the nets.

Why not solder two little bits of flexible wire?

If you solder on wrap posts, you create four excellent ways to break off the pads below.

We had an automatic wrap machine on one company I worked for.

And an automatic Termi-Point machine; tp was Amp's version of wire wrap, using stranded wire and metal clips on the posts. It was actually worse than wire-wrap.

I started making pcb's for backplanes and the machines got dusty.

PADS assigns a net name to every node that we don't name. Something like $$$1234. But they are not visible on a printed schematic. All off-pages have a proper name like SDCLK or something.

I remember those days. In the late 1970s, I worked a company near Baltimore that made a lot pf wire-wrap backplanes (with DIP ICs in sockets) in their products. The worst job in Engineering was translating the schematics into the card deck that controlled the automated wire wrapping machine, made by Gardiner-Denver if memory serves.

All newly-hired EEs were given this job, along with the promise that it would be only for a year, after which the next generation of June Grads would inherit the task. This was necessary because any longer than a year would cause abrupt total turnover.

Joe Gwinn