Garth: Awwwwww that would've been perfect! Thank you for the thought, though.

MichaelM: I'll have to read through the whole user's manual to see what the deal is with this. I agree, though, this chip would've been a great design platform. It's a shame the 68K line (this processor in particular) never really caught on. It seems superior in most ways to the kludge that is the x86 ISA.

barrym95838: You know, I've been trying to do this, but I can't seem to find any schematics either. The source would most likely have to be an old Amiga, so maybe I'll jump over to their forum and see what I can find.

BigEd: I'll have to review these notes and take things from there. Are you saying I should make the computer 3.3V throughout? The only problem is I don't know of any logic family which runs at 3.3V to use for glue. Maybe ECL?

Edit: If I'm reading this correctly, ECL would not only be able to run at 3.3V but also features faster switching speeds. A win-win.

I think 74HC series would be the usual choice. But I don't have a go-to reference for logic families - wish I did.
See http://electronicsclub.info/ics.htm#logic

(Yes, I'm thinking 3.3V throughout would be most straightforward.)

Cheers
Ed

There are several families that are 3.3V powered that provide suitabl drive and compatibility with 5V TTL logic. The family that BigEd referenced is one, but you may want to look at the LVTH family. It is a derivative of the 5V ABT logic family, so it supports signal drive levels of the magnitude that you may require for high speed operation of your design.

The ECL logic family is not appropriate for your project. There may be certain logic functions that benefit from the non-saturating, differential nature of this family. However, I would say those functions can be more easily obtained using standard saturating logic families such as HC, LVT, LCT, LVTH, etc.

As others have already said, with high speed comes the need for good bypassing and power distribution, and signal impedance control.

That link was very helpful to me. Aside from gate speed and power consumption, I never knew the different series in the 74 family had differences like that. So it looks like the 68060 version will be 3.3V throughout with 74HC glue.

Thanks, Michael, that sounds good. Where can I find more info on the LVTH family?

I generally pull a datasheet from digikey. TI and several other vendors, see digikey, make these families.

I found another link, but Michael knows a lot more than I do on this.
http://www.interfacebus.com/Design_Logi ... ction.html

Very helpful, guys! But before I go much further (on the 68060 board, anyway - I think I'll have the 68000 one pretty much ironed out) I need to make a few decisions... with some input from you all, of course.

Each logic family has its strong points. I've noticed the faster glue logic is usually harder to find and more often than not is surface mount. Granted, I am trying to build a speed demon here, but I also want this project easily duplicated at home, which means using easily found parts and through-hole mounting. Maybe it's better to sacrifice a bit of speed in the name of ease of assembly and parts sourcing? Also, the '060 has a sizable cache and - as someone pointed out before - this should help speed dramatically. Maybe I could go with glue that's a little on the slow side and just make the system use a wait state here or there if it has to. Would this speed penalty be worth the increased ease of assembly? What would you guys want more? Speed or the use of easily found, commonly understood parts?

Right now - and don't think I haven't appreciated everyone's input when I say this - I'm kinda leaning toward making the entire board run at 3.3V and using 74HC glue which is a bit slow (logic gates run around 12ns, yielding a top speed of around 66 MHz) but will run adequately on the aforementioned 3.3 Volts. I'm still looking into the alternate logic families that were suggested, and if I can find a suitable one this all may change, but for now I think this may be a good starting point.

Also, I'm starting to wonder if it's really a good idea to jump right from the 68000 to the 8-processor 68060. It may do me well to make a single processor 68060 board first, then maybe a dual- or four-processor version before finally creating the 8-chip model.

What do you all think?

74AC is about three times as fast as 74HC, has much greater drive strength, and is commonly available in DIPs. 74ABT is even faster, but the selection (especially in DIPs) is slimmer, being limited mostly to bus-oriented parts, ie, not individual NAND or NOR or inverter gates.

Is there a way you can make the computer modular, so you don't have to start over to go from one to two to four to eight processors?

From the brief glance I took on DigiKey and Mouser, it looks like 74AC works at 3.3 Volts also, which is good. The speeds I'm seeing there would allow my glue logic to hit around 80 MHz... not too shabby. Good find!





I don't even know why I went down that road of thought... my tiredness must be kicking in. lol

From the beginning, my plan has been to give the board 8 ZIF sockets so you can load the ones you want and leave the rest empty. Voila, simple modularity. I have no clue know how I apparently totally forgot about this... *facepalm*

Ask yourself this: would you rather have a working system at 4MHz which you'd like to run faster, or a non-working system at 20MHz which you'd like to get running at all? Which is to say, don't prioritise speed: prioritise simplicity and robustness.

Cheers
Ed

I like that philosophy. What good is the computer if the glue runs as fast as the processor, yet the whole computer is unstable?

So far it looks like 74AC is the winner: it runs on 3.3 Volts, it's pretty fast and it's available in DIP packaging.

If I do decide to go with the 74AC family for the 68060 board, I wanted to use 100 MHz for the processor clock and 80 MHz for the RAM... but it looks like it would be too difficult to run at dissimilar speeds like that. So these are my options:

1 - Run the entire system at 80 MHz and have a 1:1 ratio between processor and memory. This design wouldn't prioritize speed, as per Ed's suggestion.

2 - Run the processor at 100 MHz and use a flip flop like this to provide a half speed memory clock at 50 MHz, as some Macintoshes did.

3 - Try to find some counter which could run fast enough to successfully pull off the 100/80 mixture.

As usual, I don't know which to go with. Although the simplicity of a 1:1 ratio makes the first option mighty attractive. What do you guys think?

Run them at the same speed, for sure. You can always think up a more complex clocking scheme after you've got it working.

Same for wait states - run slow enough that you don't need any.

I did some work on two fronts today. First, the dev board is ready (in theory, at least) to accept it's first bit of test code! This will be a simple three or four lines of assembly I'll hammer out in the morning which will continuously output a character to the serial port so I can see if this thing is actually going to work.

Second, I did a bunch of research for the '060 board's timing system. Let me emphasize that. A BUNCH of research... literally most of my day was spent on this. I'm tired, it's now 4 AM... you get the idea.

Anyway, I searched through a TON of clock generating ICs on various sites. No dice. Either they don't run fast enough, or the voltage isn't right, or they're some funky kind of QFN, DFN or ZOMGBBQ packaging or, or, or...

Next, I tried oscillators. Again, no luck. Did you know they offer programmable oscillators? That's a new one to me.

Finally, after around five hours of searching, I found this (datasheet here) and I think my problems are solved. I can run this straight or through a variable divider to yield the desired clock rate. What do you all think?