Maximum frequency on prototype board

Thread Starter

stevendt

Joined Dec 10, 2012
6
Hi,

I realise that this is a pretty difficult question to answer, but I'll try anyway......

I am building a Z80 based computer, with 0.1" pitch Eurocard sized boards connected to a DIN 41612 backplane. I am hand soldering it and am looking to get an idea of what a realistic clock speed might be. I have put 0.1uF decoupling capacitors close to all of the ICs and a bulk decoupling capacitor on the board. There are no power & ground planes on this type of board - so all of the connections are wired.

I have not found anything on the web that gives any indication of what I might expect, other than quotes such as "a few megahertz" - whatever that might mean. I am 20MHz CPU, but won't be aiming for that, I'd just like an indication of what I might be able to achieve.

So, based on your previous experience of this type of construction, what do you think would be a reasonable target clock speed to aim for?

regards
Dave
 

Lestraveled

Joined May 19, 2014
1,946
No ground plane.......I can not agree with Mr.Chips statement that, "100MHz is fine".

The performance will depend greatly on how you wire the board, and I mean mainly, how you wire the grounds. Power and ground return should be a "star" topology. Make a power and ground buss bars with a lot of good caps to make it low impedance. Run a Vcc and ground wire to each chip. This is your "star" power distribution. Next is your signal ground topology. Follow the outputs to inputs of your chips and connect ground pin to ground pin of the chips following the signal flow. Even though the grounds are electrically connected through the "star" power wiring, they need to be connected directly, chip to chip, to maintain a low, chip to chip, ground impedance.

You put your design at a great disadvantage by not having a ground plane.
 

Sensacell

Joined Jun 19, 2012
3,420
Buy some copper shielding tape and cut and solder your way to a nice ground plane, it's not hard and makes all the difference with high speed signals.
 

RamaD

Joined Dec 4, 2009
328
The problem I had was ram failures. I do not remember the frequency, maybe 24MHz., and is not Z80 though. Small Series resistors of around 33 Ohms, in address and data lines would damp out reflections. After that there was no failure. Check out the input capacitance and decide the value so that it does not introduce too much delay.
 

atferrari

Joined Jan 6, 2004
4,763
All the 18F family PIC micros I used running with a 40 MHz clock did very well in an almost 30yo prtoboard. It was easy to realize that all those hertz were doing its thing inside the micro. No one of them went outside.
 

MrChips

Joined Oct 2, 2009
30,618
Of course, the devil is in the details: construction practice, length of wires, grounding, clock frequency.

Yes, 33Ω series resistors can do wonders to high frequency digital lines.

I believe the OP is asking for sub-20MHz operation and that should be ok.
I just meant to say you can get 100MHz operation on hand wired prototypes if you follow the rules.
Best performance would be achieved with a properly laid PCB with ground planes.
 

Thread Starter

stevendt

Joined Dec 10, 2012
6
Hi folks,

thanks a lot for the replies - yes, I was particularly interested in <20MHz operation.

I do agree that I would be far batter with a ground plane, but this is a prototype - assuming that I can get it to work, I would intend to convert it to a PCB (professionally made)

The information was very useful, particularly the bit about memory as initial testing (at 16MHz) is showing some memory issues, I will look at the series resistors idea in more detail

thanks again
Dave
 

RamaD

Joined Dec 4, 2009
328
All the 18F family PIC micros I used running with a 40 MHz clock did very well in an almost 30yo prtoboard. It was easy to realize that all those hertz were doing its thing inside the micro. No one of them went outside.
In microcontrollers, the signals outside are rarely at clock frequency or even at a tenth of that. When you do have a high freq signal out of the uC, special care is taken for that.
 

Thread Starter

stevendt

Joined Dec 10, 2012
6
The problem I had was ram failures. I do not remember the frequency, maybe 24MHz., and is not Z80 though. Small Series resistors of around 33 Ohms, in address and data lines would damp out reflections. After that there was no failure. Check out the input capacitance and decide the value so that it does not introduce too much delay.
Hi again, when you said "in address and data lines", what would you suggest as the location? Close to the CPU? At the farthest away point? Etc. The board has a CPLD, RAM, ROM and buffers at the interface to the bus, as far as thus boad is concerned, I guess that the buffers are the end of the lines

Regards
Dave
 

Lestraveled

Joined May 19, 2014
1,946
In microcontrollers, the signals outside are rarely at clock frequency or even at a tenth of that. When you do have a high freq signal out of the uC, special care is taken for that.
This is a very wise observation. If you have a microcontroller and can run your I/O at a slower speed, then you reduce the need for a "hard" ground plane. In fact, this idea touched on a design rule of mine. Never run faster than you need to.
 

RamaD

Joined Dec 4, 2009
328
Sorry, I did not see this.
If the address and data lines are buffered, then the location of the resistor packs be in buffered bus. The leakage current is very small in the order of 10uA, so the resistor does not drop any voltage.
As a general rule, all high frequency signals are to be kept short. Parasitics are always working against us! For some reason, the RAMs, both dynamic and static appear to have more failures than the CPU, ROM. Maybe it is my feeling, maybe the RAMs are more in number.
 

Thread Starter

stevendt

Joined Dec 10, 2012
6
Hi again,

thanks very much for the help.

For my current project, the problem has actually "gone away".

The real issue was the counterfeit Z84C0020PECs from China that I was using ! (serves me right I suppose).

After a bit of testing, I'm pretty sure that they are "black-topped" and remarked NMOS Z80Hs, rather than the 20MHz CMOS that they were supposed to be.
They seem to run OK up to ~10MHz
A "real" 20MHz CPU runs quite happily at 16MHz.

regards
Dave
 
Top