CPLDs

[rogerk8]

Hi Bob!

Something like that was my original thought. I.e the supply should be low-impedance and use thicker wires than AWG30.

Now you are telling me I was right from the beginning? Excellent tip about the braid by the way!

But this means that I have sawed so many PVC-tubes in vain :-)

But I kind of like not having to cut and shave so many wire-wrap wires (5X50 is much less than 530).

Thank you for this tip. I will take it under serious consideration.

As always Bob, I want you to take care of yourself!

Best regards, Roger

[eteam00]

Now you are telling me I was right from the beginning?

Of course you were right!  There was never any doubt in my mind.

But this means that I have sawed so many PVC-tubes in vain

Perfect use:  wind-chimes for those who are irritated by wind chimes.

But I kind of like not having to cut and shave so many wire-wrap wires (5X50 is much less than 530).

I ordered mine pre cut and stripped on eBay.  Very low cost, great time-saver.

One order was a lifetime supply (for me), I do not use them very often.

Best wishes to you, Roger.

[rogerk8]

You crack me up Bob! Wind chimes, what a great joke :-D

I am attaching a somewhat better (I am trying desperatelly to become a better photographer) closer look at the bottom view. As you might be able to see there isn't much space left for any braids. I was suspecting that this close socket spacing would render some difficulties. But I went ahead anyway. Now I kind of regret it but I will adapt to the situation. With the help of your tip I will fold back to the original plan and wire all supplies descretelly with wires of around AWG20 instead.

I will do this in a daisy chain and try to keep the chain not too long before I begin all over again.

Tomorrow I will however take part in the celibration of one of my best friend and collegue who turns 40. His name is also Roger. And can you believe it, my third collegue's name and best friend is also Roger. This means that almost 50% of the employees are named Roger! Ain't the world small?

Take care Bob!

Best regards, Roger

[eteam00]

With the help of your tip I will fold back to the original plan and wire all supplies descretelly with wires of around AWG20 instead.

Roger,

You may be surrendering prematurely.  Try this:

• Thread copper braid for GND underneath all the wire wrap wires, between the rows of wire wrap pins of each column of sockets.

• Similarly, thread copper braid for VCC underneath all the wire wrap wires, between the columns of sockets.

• Connect VCC and GND supply pins (not pullup/pulldown pins, just supply pins) directly to braid with short wire jumpers.

• Add enough solder to the braids to stiffen them enough so that the braids do not short to the columns of wire wrap pins.

Your photo, marked up:

-- Bob Elkind

[eteam00]

Roger,

You think this is too much trouble, it won't work, you're not sure how to implement this, you don't need it, or ... is your mind still open on this?

-- Bob Elkind

[rogerk8]

Hi Bob!

I find your solution very interesting!

One small problem is that I have already mounted(!) all the decoupling capacitors (they are in the way for the GND braid).

My mind is still open on this but it leans on using AWG20 in a daisy chain instead. The tedious problem with this approach is however that I will have to strip(!) so many wires while your fantastic procedure means pure soldering only.

I would of course want my CPU to work properly even for higher frequencies. And you have taught me that even if I will settle for 100Hz the fast SR of the CPLD might render me problems if I don't take proper care of the supply (and terminations).

Honestly, right now I am not sure what to do. I just want to proceed. Proceed in as good a way as possible that is.

Best regards, Roger

PS

I like my decoupling procedure. I have used it before. And it worked then. Why change a winning concept? But I should not refuse better solutions due to stubborness or lazyness.

I think that if I ever get this CPU to work I will not use it above 100Hz. Instead I will use the then verified concept and upgrade to a FPGA with prober CADed PCB. But I am not sure how much money I want to spend on this kind of pointless endeavour (CPU's are available for $1...).

[eteam00]

Roger, in a previous life many many years ago, at a company which shall go nameless (its initials were FPS), we designed a 64-bit parallel bus machine.  Each board in the backplane bus was quite large, roughly 60cm x 30cm. 

Someone had the brilliant idea of providing only 2 GND pins for these boards, one at each corner of the (60cm) board edge connector.  The brilliant idea was to avoid ground loops.

Every time one of these boards pulled the data lines (up to 64 of them) LOW to HIGH, up to 64x20mA of current was being switched in the span of roughly 3nS (or less).  At the receiving end of the bus, this current spike was being dumped to GND, which had only 2 edge connector pins to the backplane GND.

Needless to say, the system did not work at all.  Every time the bus switched, the on-board GND would jump about 1.3V or so above the backplane GND.  Logic thresholds no longer had any meaning, and no register or latch on the board was safe from corruption.

If you daisy-chain the GND connection for a bunch of logic ICs, consider the worst-case simultaneous switching problem causing a current spike aggregated in the wire-wrapped GND connection.  The current spike will be proportional to the slew rate of the IOs, the number if IOs, and the degree to which the switching is simultaneous (in edge-clocked systems, switching tends to be highly simulataneous).

Note that clock frequency does not affect the simultaneous switching current spike problem, either for better or for worse.

-- Bob Elkind

[rogerk8]

Bob, I always laugh when I read your posts. You have a fantastic way of expressing youself!

In this case I however don't quite follow you. This is partly due to my bad english, partly due to my kind of bad signal dynamics understanding.

But I do understand most of what you are saying.

I think this says it all:

Note that clock frequency does not affect the simultaneous switching current spike problem, either for better or for worse.

So what should I do? I am in a stage where it is possible to choose whatever I like regarding the supply (it is not much work to remove the capacitors). And with regard to what you are saying, what will be my final choice?

Because I am kind of lazy I think I will not remove the capacitors. I think I will use the AWG20 wires in a daisy-chain. Starting tomorrow. This choice might not be the best but it is much better than AWG30 wire-wrap wires which you thankfully avoided me from using. The supply impedence might not be the best this way (because braids are of course better) but I think it will suffice. And you can bet your ass (if I may say so) that I will incorporate your braid-solution if I ever get to wire-wrap something modern again!

Always much fun to talk to you!

Goodnight!

Best regards, Roger

[rogerk8]

Bob, I have now started the AWG20-approach. At first I didn't like it. I thought that the different joints, with obvious risk for being cold-soldered, would somewhat be worse than wire-wraping the supply. But then It kind of grew on me. Now I think I have made a good choice. The wires are thicker and actually shorter which makes it overall better.

After a while I even learned to make small hooks mith my new superb Lindstrom plier.

I am attaching a very bad picture of my progress. I really wish I was a better phographer. But maybe it isn't me just the bad cellular camera. Maybe I should invest in a new better camera? Note that I am not just phograhing for you (even though you make me select the best ones). I want to use the pictures in my homepage too.

Take care, Bob!

Best regards, Roger

[rogerk8]

One more thing. I have, due to supply simplicity, changed my original plan for using 3V for the reset pulse generation. I now use 5V. Because the CPLD inputs are said to be 5V tolerant this should not be any problem. Or should it? All PROM's generate 5V directly to the CPLD inputs. But I am thinking of that this very short reset pulse (<1us) might not work if it doesn't have the normal level. Am I worried in vain or is it justified?

//Roger Knopp