Retrocomputing, part 2

[slburris]

Attached are the latest schematics and a
PCB layout with a test routing run.

Still have some work to do, I think I'm about 90% done.
We're missing bypass capacitors, some net connections,
and a few components.

Once I'm done, I'll get all the parts together, and then
print out the PCB at true size for a test fitting.  I can't
tell you how many times I've caught little problems
doing this.

Once that passes, I'll send the gerbers off to a board
house and build a prototype.  Hopefully it will work the
first time!


Scott

[FreeThinker]

I learned (to a basic level) machine code on a zx80, Yes it was possible . You have stated that you will use a soft core t80 which I assume is an emulator.Will it fully support the code set of the Z80 or only the "Official" codes?.As I remember the Z80 had a set of "sercret" opcodes that were not supported (actually did not work as expected) but were used a lot in games software and "Protection systems" because normal disassemblers would not handle them correctly.Probably being a bit anal bringing it up but I feel that it should work as originally built warts and all.Just a thought .
PS. As to the wall wart problem, have you considered a bridge rectifier after the jack? Then the board would work no matter what polarity the tip is. 

[slburris]

I learned (to a basic level) machine code on a zx80, Yes it was possible . You have stated that you will use a soft core t80 which I assume is an emulator.Will it fully support the code set of the Z80 or only the "Official" codes?.As I remember the Z80 had a set of "sercret" opcodes that were not supported (actually did not work as expected) but were used a lot in games software and "Protection systems" because normal disassemblers would not handle them correctly.Probably being a bit anal bringing it up but I feel that it should work as originally built warts and all.Just a thought .
PS. As to the wall wart problem, have you considered a bridge rectifier after the jack? Then the board would work no matter what polarity the tip is. 

I'm working on two separate designs.  The "classic" design is the schematic a couple posts ago
and that uses a genuine Z80.

The "FPGA" design uses the T80 core from opencores:

http://opencores.org/project,t80

It supports all undocumented Z80 instructions and is heavily used by
the retroarcade community. See for example:

http://www.fpgaarcade.com/platforms.htm

So I have pretty high confidence that it works!

Scott

[FreeThinker]

I learned (to a basic level) machine code on a zx80, Yes it was possible . You have stated that you will use a soft core t80 which I assume is an emulator.Will it fully support the code set of the Z80 or only the "Official" codes?.As I remember the Z80 had a set of "sercret" opcodes that were not supported (actually did not work as expected) but were used a lot in games software and "Protection systems" because normal disassemblers would not handle them correctly.Probably being a bit anal bringing it up but I feel that it should work as originally built warts and all.Just a thought .
PS. As to the wall wart problem, have you considered a bridge rectifier after the jack? Then the board would work no matter what polarity the tip is. 

I'm working on two separate designs.  The "classic" design is the schematic a couple posts ago
and that uses a genuine Z80.

The "FPGA" design uses the T80 core from opencores:

http://opencores.org/project,t80

It supports all undocumented Z80 instructions and is heavily used by
the retroarcade community. See for example:

http://www.fpgaarcade.com/platforms.htm

So I have pretty high confidence that it works!

Scott

How cool is that! . Not just me then Lol.

[slburris]

I've placed an order with Digikey for the parts I don't have in stock, so
I should get them Wed or Thur I think.

So next weekend I should be ready to do the test fitting of parts to
a printout of the board.

While I wait for parts, it's time to get back to the FPGA design.

Scott

[slburris]

Woohoo, my Digikey order has arrived, so I'm going to try to
finish up the schematic and PCB this weekend so I can do
a test layout on paper with the components.

Scott

[slburris]

The board is going to look something like the attached test layout render.

Scott

[slburris]

Have I mentioned yet that I hate autorouters?  I've been screaming at the Eagle
autorouter because it's making stupid routing decisions.  Come on, can't this thing
see this nicely aligned data bus and just connect everything up there first before
messing it up with other traces?  No, it can't!

I've even tried the freerouting.net autorouter, which I think generally does a better
job than Eagle's built in router.  Still, neither autorouter gets to 100% .

It's only a 4Mhz design, so layout really isn't critical, and there are zillions of connections,
so I don't want to route this by hand.

Next up, I'm going to try some old tricks, namely writing a script file that autoroutes
the "obvious stuff" first, then let's loose on the whole board.  If that fails, I may have to
at least partially hand route this board.

Scott

[Zad]

There is only one way to learn really. Go and do it. With the time you spent looking for alternative autorouters and shuffling around horrible looking routes you could have done it by hand and ended up with a decent looking board that you know inside out. Get the power buses and data lines in and take it from there.

[TechGuy]

Next up, I'm going to try some old tricks, namely writing a script file that autoroutes
the "obvious stuff" first, then let's loose on the whole board.  If that fails, I may have to
at least partially hand route this board.

Route in groups, don't route the entire board at once (assuming eagle has a option to route a selected group of components - I use Diptrace). Then fix the problem ones manually and do another group, fix the problem traces, move on to another group. Not perfect but better than other options.

[slburris]

OK, we've reached the PCB stage of things. 

I'm starting to shop around board makers for a roughly 25 sq in board, qty 25, 2 layer.
This is too large for some of the real deals, like ITead Studio, or DorkbotPDX, or BatchPCB,
so time to quote some of the standard places I've used before.

Scott

[slburris]

Here are the latest schematics, at least until I find another error :-)

Scott

[slburris]

Oh bah, I realized I made two part changes fairly late in this process
and so the SD card connector has the two side holddowns in a different
spot than the one I found in the Eagle library.  Plus I ordered a 5 pin mini USB-B
connector instead of a 4 pin which is on the PCB.

Back to more corrections, reroute, stare at it some more, then send it off
to be fabbed.  I think this board is going over to pcbfabexpress.com

Scott

[slburris]

I added headers for the address/data/control buses for a logic analyzer
connection, just in case I need it.  And now I'm having a hard time getting
everything re-routed....

Plus I discovered I made a mistake in the PCB symbol design for the compact
flash socket.  I thought I'd be clever and modify an existing design rather
than start from scratch, but I discovered that I computed some coordinates
incorrectly (measured from where the solder mask ended rather than the
pad itself) and a few things are off by a fraction of a mm.  So off to fix that
symbol.  Have I said how much I hate creating new parts in Eagle?

Then, I'll stop making changes and just send the boards off to be manufactured.

Scott

[slburris]

OK, I ordered 5 boards from pcbfabexpress.com yesterday morning and they
let me know in the afternoon that my uploads passed their design rule check
so they are in process.  I'm not in a hurry, so they are being made on a 10 day
turn.

The boards are rectangular, about 30 square inches.  It sure takes a lot of chips
to get anything done using those big Zilog chips of the 80's!

Scott

[slburris]

Here are the latest schematics and an image of the board.

No, I didn't route the board by hand.  At 4 Mhz, I figure the
circuit will work in spite of any craziness done by the autorouter.
I did wirewrap a Z80 circuit in the 80's and it worked just fine.

That being said, I didn't use the Eagle autorouter to do this
board.....

Scott

[RayJones]

LOL, a true retro project would be using the ubiquitous 1488/1489 RS232 interface chips and the associated +- 12V rails

Nah, only joking. This is the stuff I cut my teeth on, and it is good clean fun and amazingly capable.
More importantly they boot in seconds!

[slburris]

LOL, a true retro project would be using the ubiquitous 1488/1489 RS232 interface chips and the associated +- 12V rails

Nah, only joking. This is the stuff I cut my teeth on, and it is good clean fun and amazingly capable.
More importantly they boot in seconds!

Yeah, and a 40lb transformer with multiple windings to provide the power.

And none of this DB9 stuff, RS232 uses DB25's, darn it! (yes, I know
it's really called DE9 and DE25 but nobody seems to use those terms).

And 8" floppy disk drives....

Scott

[slburris]

And now I'm on day 4 of my 10 day turn for the prototype boards.

Did I just order another set of green coasters?  Or will the boards work?
Or at least be able to be fixed by cutting traces and using jumpers.

Stay tuned!

Scott

[baljemmett]

And none of this DB9 stuff, RS232 uses DB25's, darn it! (yes, I know
it's really called DE9 and DE25 but nobody seems to use those terms).

Nah, DB25 is right, but what's commonly called DB9 is really a DE9.  The letter indicates the size of the shell; hence VGA uses a DE15 connector, since it's the same size shell but with more pins (not to be confused with DA15 for Ethernet AUI connectors!).  All good fun, and probably why lots of people just call them "9-pin D" etc.

Looking forward to seeing the boards when they arrive!  Have you decided yet how you're going to present the storage to CP/M?  I seem to recall you were thinking about partitioning schemes a while back.

[slburris]

Ah, there's misinformation about DB vs DE then:

http://en.kioskea.net/contents/elec/connecteur-prise-db25.php3

I haven't really thought more yet on how the CF card is going to
work.

My most immediate goal is to get an EPROM burned that initializes the
serial port and prints "Hello World!".  Then I'll build a little monitor
program that can read a bootstrap off the CF card.

I think I'm going to need a CP/M system to build the BIOS and BDOS
for the CF card.  I'm a little vague on the details of how this hangs
togther, since it's been 30 years when I last customized a CP/M system.
Something about using SYSGEN to write a diskette that you can boot from?

What I think I'll do is bring up an Altair z80 simulator running CP/M
and use that to build the right images for my system:

http://simh.trailing-edge.com/
http://www.schorn.ch/cpm/intro.php

It will be good practice for the next project, the one I originally
proposed, which was to use a soft core Z80 in an FPGA talking to
a VGA connector, mouse, and keyboard.  For that design, there will
be more memory, as my intention is to let you boot either CP/M or
MP/M.

And then of course as I was rummaging through my chip collection,
I came across some 68000 series chips.  I have 68000's, 68020's,
and a 68040.  If I haven't burned out on this stuff, maybe I'll do some
retro 68000 thing. 

Scott

[RayJones]

LOL, a true retro project would be using the ubiquitous 1488/1489 RS232 interface chips and the associated +- 12V rails

Nah, only joking. This is the stuff I cut my teeth on, and it is good clean fun and amazingly capable.
More importantly they boot in seconds!

Yeah, and a 40lb transformer with multiple windings to provide the power.

And none of this DB9 stuff, RS232 uses DB25's, darn it! (yes, I know
it's really called DE9 and DE25 but nobody seems to use those terms).

And 8" floppy disk drives....

Scott

Let's not forget that having a 10MB ST-506 HDD would be *very* extravagent

[slburris]

The PCBs have shipped!  Picked up this evening from Sunnyvale, CA.
Tracking says will be delivered by Dec 3 4:30pm.  Hmm, they just have
to move a couple hundred miles down the state into southern California --
even if Fedex drives them down by bicycle, it shouldn't take 3 days!

Scott

[slburris]

The PCBs are here!

I'll try to post a high resolution picture of one in a bit.

Scott

[slburris]

The build has begun!

First up, I put together the switching power supply and immediately hit
a small problem.  For some reason, the outline of the input jack is incorrect
and crowds the polyfuse and the first Schottky diode.  Not enough to be
a showstopper, but if you look closely I accidentally touched the jack
with the soldering iron and melted it slightly.

Second problem was with the National Semiconductor switcher chip.
I used a pinout with a staggered pin arrangement.  But I only have an
inline arrangement type in my parts box (about a hundred from an Ebay
purchase).  OK, no problem, just bend the pins with a long nose pliers.

I also didn't use the exact values of the capacitors I specified on the
schematic.  I had a bunch of 100uf caps, so I just used three of those.
I figure the exact values aren't too critical, and going bigger shouldn't
hurt anything.

The LM2575 is only a 52Khz switcher, and the datasheet seems to
say you can use regular electrolytics and not need to get into the
low ESR tantalum or ceramic types.  Yeah, probably at that switching
speed nothing's too critical.

So I built the power supply first because when I have a new board,
I like to verify that it's going to output the proper voltage *before*
I put a bunch of rare or expensive ICs on the board.  It's probably not really
necessary for this board because I have high confidence in this
Simple Switcher design and there's very little to go wrong.  But
I tend to be paranoid about letting the magic smoke out.

Connected up a 12v unregulated wall wart, fired up the DMM,
and immediately remembered why I shouldn't work on this stuff
late at night.  I was using my UNI-T meter instead of a Fluke, and I
had the leads plugged in to measure current, although the meter
was set to measure voltage.  I measured the 5 volt output
and got nothing.  Hmm, puzzled I put the probe on the polyfuse
and got a small spark.

Luckily with a switcher that is short circuit protected and a polyfuse
to handle the raw input voltage, nothing bad happened.  Fixed my
test leads and sure enough, 5 volts out!

Now in looking at the PCB more closely, I spotted another mistake.
The CompactFlash card socket has two tabs near where the card is
inserted.  The tabs are supposed to be soldered to the PCB.
But when I made the part in Eagle, I forgot to block out the solder
mask over the pad!  Oh well, I can scrape the solder mask off with
an X-Acto knife I guess.

So where am I?  I've installed the CPU, the 4Mhz master clock
oscillator, and some sockets.  I've put my Rigol on the clock and
verified we're getting a good rail to rail 5V CMOS swing on the
waveform, and the Rigol says it's 4Mhz to boot.

So far so good, and I haven't yet had to cut any traces.

Scott

[slburris]

Here are some links to high resolution PCB images.

There are each about 5MB in size:

front - http://www.electroscott.com/images/P1000918.JPG

back - http://www.electroscott.com/images/P1000919.JPG

Scott

[slburris]

Here's the first serious errata.

The symbol for the MCP131 voltage monitor is backwards.
So as installed, it holds the RESET line low permanently and
the processor doesn't operator.

The fix is to ignore the silkscreen on the board and turn
the TO-92 package around.  Then it's connected properly.

I did that and now my Z80 CPU free runs.  Monitoring M1,
I can see machine cycles being run.

Next up, I need to make a diaignostic EPROM of some sort and
program the GAL, then test the outputs of the GAL to make sure
they are doing the right thing.

Scott

[slburris]

So here's what I'm going to do for diagnostics.

I'm going to read from each of the SIO, CTC, PIO, and Compact Flash
ports in a loop.  The first port I will read once per loop, the second port
I will read twice, and so on.

Then, by monitoring both IORQ and a particular chip select line, I should see
the correct burst of 1, 2, 3, or 4 selects per cycle, if the GAL is programmed
correctly.  If not, I'll have to figure it out.

If needed, I'll pull out the Tektronix 1240 logic analyzer, but I'm hoping this will
go well enough to make that unneeded.

Here's the program I'm going to put into the EPROM:

0000                1   org     0
0000                2
0000                3 siobase   equ     $00
0000                4 ctcbase   equ     $20
0000                5 piobase   equ     $40
0000                6 cfbase    equ     $60
0000                7
0000                8 loop:
0000 db 00          9   in      a,(siobase)
0002 db 20         10   in      a,(ctcbase)
0004 db 20         11   in      a,(ctcbase)
0006 db 40         12   in      a,(piobase)
0008 db 40         13   in      a,(piobase)
000a db 40         14   in      a,(piobase)
000c db 60         15   in      a,(cfbase)
000e db 60         16   in      a,(cfbase)
0010 db 60         17   in      a,(cfbase)
0012 db 60         18   in      a,(cfbase)
0014 18 ea         19   jr      loop
22 bytes code generated and 5 labels defined

Scott

[slburris]

Picked up a bunch of 27C256 EPROMs from my collection and put them
in the EPROM eraser.  Can't remember how long they are supposed to be
in there, so let's try 1/2 hour.

I'm having flashbacks to how I built stuff like this in the mid 1980's! :-)

Scott

[slburris]

So I programmed an EPROM, programmed a GAL, plugged them
in, and ..... nothing.

No IORQ cycles at all, which means not even the part of the GAL
dealing with memory address decoding is working properly.

Back to debugging...

Scott

[slburris]

Uh oh, I grabbed a Lattice GAL20V8 instead of an Atmel ATF20V8 and
I made the fuse map with Atmel's version of WinCUPL.

It's been so long since I programmed GALs, that I can't remember if the
fuse maps are identical between different manufacturers...

Scott

[slburris]

Poking around with my trust Rigol scope, it appears that the GAL isn't working.
I can't seem to find any Atmel 22V10's, but I have plenty of Lattice Semiconductor
22V10's.  So I'm now downloading their "classic" software and I'll have to figure out
how to generate a fuse map all over again.

Turns out even though the software is free, they *still* require a license.  So I'm
awaiting a license key in email.

Maybe I should have done all the memory and I/O decoding in discrete logic, but the
board already has so many chips on it to do so little.

When I built my first CP/M computer from scratch after the Altair, I wirewrapped a
ton of chips together.  It worked, as long as you didn't breathe too heavily on the board.
I learned over time if the system didn't boot to press just so in a certain area on the
board.  Must have been an intermittent wire wrap.

This PCB should be much more reliable, once I get past these early startup problems....

Scott

[slburris]

Well, that was pretty stupid.  I actually pulled a GAL20V8 instead of a
GAL22V10 and programmed it with the fuse maps of a 20V8.  Uh, that's
not going to work....

I picked a 22V10 and now things are looking much better.

Remember how I was going to do one I/O read from the SIO, two
from the CTC, and so on?  Here are capatures of !siosel, !ctcsel, !piosel,
and !cfsel.  See how the groupings increase each time?

This verifies that the GAL is at least mostly working, plus the EPROM
is working, woohoo!

Now I have to get the serial port to print "Hello World!" :-)

Scott

[slburris]

Here's a shot of the power input jack that's too close to the polyfuse
and input diode.  Kinda hit it with the soldering iron....

Scott

[slburris]

My attempt to solder the tiny FTDI chip.  It looks slightly
off, but should still work, I hope.

Scott

[slburris]

And I'm currently stuck trying to get the USB chip to be
recognized.  For awhile I had convinced myself that I had
wired the mini USB-B connector backwards, but comparing
to some commercial products, it's OK.  I did have a bad
solder joint connecting the GND pin, so that had me
scratching my head for a bit.

But when I plug it in to a computer, the FTDI chip seems to
immediately go into SUSPEND mode.  The computer doesn't
see it as a USB uart chip, and FTDI's EEPROM setting utility
doesn't see it either.  I'm following the self-powered diagram
in the FT232R datasheet, but no joy.

I must be doing something stupid wrong, but I'm not seeing
it tonight.  Maybe tomorrow will be a better day.

On another note, I had forgotten how hot these NMOS chips
got.  And the Lattice GAL chip is the hottest of them all, even
though I'm using an LP (low power) version.  Even the LP version
draws 90ma at 5v, that's almost 1/2 watt!

Scott

[slburris]

Well, things are still not going well in USB land.

I found a bad solder joint at the USB connector such that the
USB D+ line wasn't making contact.  That's now been repaired.

So when I plug the board into windows, it now says that's
it's detected a new device, yea!

Then it says this USB device is broken, boo!  I can't seem
to see the device with FTDI's programming utility either.

Am I going to have to capture what's going on with the USB data
lines to figure out what's going on?  I hope not -- this would take
me far off my field of expertise.  The whole point of using the FTDI
chips is to not have to really know anything about USB.

Scott