My MARC-1 SBC

[Arlet]

BigDumbDinosaur wrote:

Something else to consider is that fill planes (aka ground pours) are detrimental to high speed digital circuits.

Having floating copper islands isn't such a great idea, but well connected ground pours work fine. Also, this board runs at less than 4 MHz, so that's almost DC.

[BigDumbDinosaur]

Arlet wrote:

BigDumbDinosaur wrote:

Something else to consider is that fill planes (aka ground pours) are detrimental to high speed digital circuits.

Having floating copper islands isn't such a great idea, but well connected ground pours work fine. Also, this board runs at less than 4 MHz, so that's almost DC.

We have a topic around here in which the discussion centered on internal ground and power layers vs. filled planes vs. nothing. Long, long ago, when I first learned the science (and art) of PCB design, it was inculcated in us that filled planes of any kind were poison in high speed digital work, and best used for audio and DC applications. At the time, "high speed" was around 5 MHz. I've heeded that admonition over the years and have never regretted it.

I built POC V1.0 and V1.1 on PCBs that have no filled planes, and with 0.15mm signal traces on 0.63mm centers. The boards have internal power and ground layers, which together tend to act like a large bypass capacitor. Both units are dead quiet and can run at 12.5 MHz without the SCSI host adapter installed.

[lordbubsy]

BigDumbDinosaur wrote:

BTW, I don't see a bypass capacitor near the EPROM. An EPROM generates a large transient when /OE is asserted.

They’re just shy and hid themselves

Quote:

Something else to consider is that fill planes (aka ground pours) are detrimental to high speed digital circuits.

I routed the whole board except for GND and VCC. After that I chose GND (because of lesser air wires) to route by hand and leave VCC. Then I filled the back and top with a polygon plane, and connected both planes to VCC. There were only three VCC air wires left to be routed. If I’m correct, then every polygon is connected to VCC leaving no floating polygons. Perhaps I interpreted it wrong, but I thought this was a good thing to do. So I’m confused.

BigDumbDinosaur wrote:

We have a topic around here in which the discussion centered on internal ground and power layers vs. filled planes vs. nothing.

That’s the thing, I read so many different views of the topic...

[Arlet]

BigDumbDinosaur wrote:

it was inculcated in us that filled planes of any kind were poison in high speed digital work

For what underlying reason ?

Quote:

At the time, "high speed" was around 5 MHz. I've heeded that admonition over the years and have never regretted it.

I've made several 2-layer 100+ MHz boards with ground pours on both sides, and never regretted it either.

[GARTHWILSON]

a couple of short articles relevant to ground fills from Dr. Howard Johnson who is a high-speed digital design industry guru:
https://web.archive.org/web/20120302190 ... ws/1_3.htm
https://web.archive.org/web/20120302190 ... ndFill.htm

[Arlet]

Of course, multi layer boards with dedicated ground planes are better. But that's not an option for home made boards, so the choice is between boards with a few ground traces, or boards with ground pours. Ground pours have an advantage over ground traces that they allow return current to flow closer to the signal trace, reducing the magnetic field. Of course, this requires that the ground pour is not cut up into islands, but continuous over the board. This can be achieved by using ground pours on both sides of the board, and attaching them to each other wherever you can.

In addition, ground pours work very well as heat sinks, and they require less etchant.

[BigDumbDinosaur]

Arlet wrote:

In addition, ground pours work very well as heat sinks, and they require less etchant.

Not really. Their thermal resistance is actually quite high because little mass is involved.

[GARTHWILSON]

Quote:

Not really. Their thermal resistance is actually quite high because little mass is involved.

They're good for small things where you don't need much area to get rid of the heat. It gets interesting though when you have several ground layers conducting the heat. We had a board like that with a part that we wanted to heat sink, and even though it only had three copper leads soldered into holes, getting it back off by hand required a 250-watt iron or soldering gun. The board was so effective at carrying the heat away that with any smaller iron, you couldn't melt the solder. I think I went overboard on that part of that design.

[BigDumbDinosaur]

GARTHWILSON wrote:

Quote:

Not really. Their thermal resistance is actually quite high because little mass is involved.

They're good for small things where you don't need much area to get rid of the heat. It gets interesting though when you have several ground layers conducting the heat. We had a board like that with a part that we wanted to heat sink, and even though it only had three copper leads soldered into holes, getting it back off by hand required a 250-watt iron or soldering gun. The board was so effective at carrying the heat away that with any smaller iron, you couldn't melt the solder. I think I went overboard on that part of that design.

That's where use of "thermal pads" can help.

[lordbubsy]

I programmed the ATtiny a week ago and checked for all VCC’s and GND’s. then I installed all chips but unfortunately my MARC-1 didn’t work. To be honest, I didn’t expect it to, at least not right away.

OK, all IC’s pulled out and comparing the eagle board layout with the PCB. That way I found *SEVEN* unsoldered connections. ALL of them were at the top layer on IC sockets.

Result, MARC-1 not working.

Then I rebuild the entire computer on breadboard for comparison, which also didn’t work. Now I began to question the design. On a separate breadboard I rebuild my first design with 32k RAM, 16k ROM and I/O at $8000. I had to have something working! This one ALSO did not work. And I worked really careful and slow. After some measuring I found a faulty jumper wire. It was brand new.

At my MARC-1 breadboard it was a very stupid problem, RXD and TXD have to be cross connected to the UART-USB bridge. The same with CTS and RTS. After that both breadboards were working.

After some searching here on the forum I found some debug topics.

NOP Generator by Lee Davison:
http://6502.org/mini-projects/nop-gen/nop-gen.htm

Garth's checking method:
viewtopic.php?f=1&t=1888&p=17653#p17653

Garth's debugging tips:
http://wilsonminesco.com/6502primer/debug.html

as a result:
RES, RDY, IRQ and NMI were high. There was a clock at CPU pin 37 and PHI2 at 39.
I installed the ATtiny, the 74HCT74 and the MOS6502 and connected %11101010 to the data lines of the ACIA. I checked A15-A0 with the scope. They toggled between high and low, each with double frequency. Except for A5 which was high all the time.
I made a scan with 1200 DPI and searched for the short. Seems I shorted A5 with D6 with a very nice blob of solder!

Then I installed the 74HCT00, 74HCT688 and an EPROM with a test program which jumps from $C000 to $FF00 in a 2 second interval. I checked A8, A9, A10, A11, A12 and A13 for toggling.

After that I installed RAM and used a program which jumps from $C000 to $1000 in a 2 second interval. I checked A12, A14 and A15 for toggling.

Then I installed the 74HCT138, the ACIA and connected the UART correctly.
BINGO!!!

After the "week of hell", MARC-1 finally works.

Now I have a reference board and I finally can prototype the replacing of all glue logic with a 9572!

[BigEd]

Great! It's good to have a story of how things go wrong and how you found and fixed them.

[BigDumbDinosaur]

lordbubsy wrote:

I made a scan with 1200 DPI and searched for the short. Seems I shorted A5 with D6 with a very nice blob of solder!

Urk! That's where solder mask comes in handy. Don't know how you'd create one on a homemade PCB though.

[nyef]

BigDumbDinosaur wrote:

That's where solder mask comes in handy. Don't know how you'd create one on a homemade PCB though.

A quick search of the web suggests the use of UV-curable paint and either transparencies or toner-transfer methods for preventing the cure on the solder pads.

[lordbubsy]

During soldering I probably became too eager and careless at some point.
Adding a soldering mask for two sides would be a significant amount of extra work.
A silkscreen however is relatively easy.

[wmmullaney]

Not really knowing anything about it, what would happen if you left the toner on the traces, just scrubbing it off of the solder points and mistakes?