Trivial circuits for sbc

[GARTHWILSON]

Dajgoro wrote:

I got 0.1mm copper enameled wire from ebay, and it looks kinda thin. It has about 3ohm/meter, and now i am wondering if it would work well if i used it instead of the ribbon cable that i usually use for building sbc-s.

.004"? That's less than half the diameter of wire-wrap wire. It sounds fragile. What did you have in mind doing with it?

[Dajgoro]

Use it in places where ribbon cable would be impractical for point to point soldering(connecting 4 ram modules in parallel) or for some lines like r/w or phi2. I can always protect it with epoxy.

[GARTHWILSON]

I used that kind of wire on the board in the third picture at http://wilsonminesco.com/6502primer/WireWrap.html:



but I wouldn't do it again. It was too easy to break wires after they were soldered, while trying to access points to solder additional ones. If you're not making a printed circuit board, I would strongly recommend wire-wrap for digital circuits. Besides being less fragile, WW lets you get the wires shorter and straighter, improving performance.

There is a type of circuit boards made with the enameled wire though, by machine, then the wires are varnished down to the surface of the board. I can't remember what it's called to look up a picture to post.

[Dajgoro]

I know that ww is superior, i saw a ww diy demo cpu on my computer architecture class last year, it was old, made by students who knows when, and it still woks. WW sockets are expensive so i won't be doing any ww until i find a way to earn some money.

[BigEd]

Verowire is 34AWG, 0.15mm - it's popular and effective if you want to avoid the expense of wire wrap.
See here.
Unbranded materials are also available, and cheaper. See this post for "roadrunner".
Cheers
Ed

[Dajgoro]

It is thicker than the one i currently have, but in my case could there be any ac consequences from using such a thin wire(in our hi frequency sbc enviroment).

[GARTHWILSON]

There are PCB traces with that width and even less thickness (.0014" for 1-ounce copper), so I think you'll be fine as long as you put the IC shoulder to shoulder (for short connections) and keep the wires fairly straight. Going really small affects the resistance a lot more than the inductance, and in this case mostly because of the skin effect, but it will be fine if none of your connections are more than a few inches long. Dr. Howard Johnson's articles on skin effect in high-speed digital designs are indexed at https://web.archive.org/web/20120302190 ... n%20effect .

[Dajgoro]

Ok. Then i can use it to connect multiple rams in parallel, since using ribbon cable for that kind of thing gets really tedious and ugly.

[BigEd]

(Garth: thanks for saying "a few inches" - it looks a great deal better to me to see quantified advice, where often we see advice like "as short as possible" which isn't much help to someone who doesn't already have the experience to judge what it means. Even vague quantification is better than none! Cheers.)

[GARTHWILSON]

It still felt very vague, but there are so many variables that one cannot give a solid answer. What rise times? What loads? What frequencies? How much do you have to build, which relates to how short the connections can reasonably be made? What future expansion might there be? And from there you can start doing some heavy-ish math, and then we're often brought back full circle to "Just try it and see if it works reliably, and see how fast it'll go," and if it's not good enough, you start over, doing something better. Sometimes you have to be like a beaver building a dam, or a bird building a mud nest under the eaves, which have no concept of math, but a good feel for what's needed, and they do a great job.

[BigEd]

Sure, vague isn't definitive, and behind a judgement there might be a lot of specific knowledge, but for advice to be useful it has to be actionable, and "keep your connections down to a few inches, and see how you get on" is certainly actionable.
Cheers
Ed

[BigDumbDinosaur]

Dajgoro wrote:

Ok. Then i can use it to connect multiple rams in parallel, since using ribbon cable for that kind of thing gets really tedious and ugly.

Aside from issues of tedium and appearance, ribbon cable is problematic for high speed buses of any kind. The tight proximity of the conductors tends to make the cable highly reactive unless correctly applied and terminated. Unless you know the technical details of the cable, you can't be certain about its characteristic impedance. The principle of separating signal leads from each other with a ground lead and the use of Thévenin termination can make it work, as long as the termination is reasonably close to the cable's characteristic impedance. The only problem with Thévenin termination is that the silicon driving the bus has to be able to cope with the current flow required to skew the lines against the termination's quiescent voltage. Given that a lot of ribbon cable has a characteristic impedance in the 100-150 ohm range, you could be faced with an unreasonably high drive requirement.

If you can, try stacking the RAM chips so as to reduce the wiring to the bare minimum. You should be able to tie almost (if not all) of the address lines together, as well as parallel the data lines.

[BigEd]

You can use alternate conductors in the ribbon cable for signals or ground.

[BigEd]

BTW the Beeb's second processor is connected by (short) ribbon cable. As are many disk drives.

[BigDumbDinosaur]

BigEd wrote:

BTW the Beeb's second processor is connected by (short) ribbon cable. As are many disk drives.

Which, as I noted, can be made to work with proper application. Consider that SCSI-SE can be operated at a 20 MHz rate ("fast SCSI 20") over a 6 meter cable. The key, of course, is alternating signal and ground leads, along with proper termination and high drive strength (48 ma per signal lead). I'm not running the SCSI interface that fast with POC V1 (I'm using 5 MHz asynchronous signaling) but it's still 5 MHz, running on a cable that is about a meter in length, with three connected devices (disk, tape and CD-ROM).