16 Nov
2012
16 Nov
'12
2:20 a.m.
On 14 Nov 2012, at 23:09, Dave Caroline
<dave.thearchivist at gmail.com>
wrote:
I'm very excited about Dave's scans, because the circuits look right and
the date of 1967 would allow for both Richard's experience of building this
at school and for the author to have published as a book. He states the
intention of using the design as a teaching tool. Unfortunately there's no
named credit in the scanned pages (anything in the contents list Dave?).
But we do have "The computer will be on show at the R.S.G.B exhibition."
The one-transistor NOR gate and the toggle-flop design are both as I
remember them. Likewise the fact that it's bit-serial rings a bell.
Dave: can I assume you're happy for me to re-publish some or all of your
scans? I want to clip out at least some diagrams and tables.
[Does this list allow for attachments or MIME emails?]
Fourpence-ha'penny for a diode!
Cheers
Ed
On Wed, Nov 14, 2012 at 8:36 PM, Dave Caroline
<dave.thearchivist at gmail.com> wrote:
Date: Thu, 15 Nov 2012 18:30:43 +0000
From: Richard Smith <mole42 at gmail.com>
Hi,
I'm interested in what you have found, I'd love it to be an article about
the computer. It's a long time ago and I have only vague memories of
soldering the logic units - some of the older boys got to assemble the
actual machine!
Best regards,
Richard On Wed, Nov 14, 2012 at 8:04 PM, Ed Spittles
<ed.spittles at gmail.com>
wrote:
Now found and scanned but due to the state of the paper are scanned in
greyscale
The article above refers to it continuing into 1968 but the stuff I
have found ends in December 1967
pm me for url as the files are large and cannot handle a download swarm.
Dave Caroline On 29 Apr 2012 at 18:46, Richard Smith wrote:
> This thread reminds me of a computer we built at school from discrete
> transistors. Each transistor was a NOR gate with three resistors on
> the base and a collector resistor. All soldered onto squares of tag
> board. We put a bunch of them together to build a shift register with
> small laps as output. That would be about 1969 or 1970. Does anyone
> remember any more? It must have been a published design somewhere.
Richard, I think I read the book this project was based on - in the
school
library, mid-to-late 70's. I've been looking for it, but my
recollection
is so vague I haven't found it yet. I think it may have kicked off
with
some physical computing based on wood and ball bearings, but anyhow it
worked up to a full serial CPU. My searches have been based on the
recollection that the author was Wilkinson (but maybe Wilkins, Watson,
Wilson, Watkins, Watkinson, ...) and, of course, it might not even
have
starte with W.
How about the magazine wireless World
http://www.fano.co.uk/history/WWcomp.html
I do believe I have the set of articles here somewhere
16 Nov
16 Nov
2:50 a.m.
[conversation about Wireless World articles - a
digital computer built from reject germanium transistors]
Aha: here's an online version of a pamphlet reprinting that article
series, which gives credit to Brian Crank:
http://www.smrcc.org.uk/members/g4ugm/Manuals/wirelessworldcomputer.pdf
Cheers
Ed
See also http://www.classiccmp.org/pipermail/cctalk/2001-November/241361.html
3:13 a.m.
On 11/16/2012 12:50 AM, Ed Spittles wrote:
How come this was never posted when you still could get germanium
transistors. :)
Ben.
[conversation
about Wireless World articles - a digital computer built from reject germanium
transistors]
Aha: here's an online version of a pamphlet reprinting that article
series, which gives credit to Brian Crank:
http://www.smrcc.org.uk/members/g4ugm/Manuals/wirelessworldcomputer.pdf
Cheers
Ed
See also http://www.classiccmp.org/pipermail/cctalk/2001-November/241361.html
3:26 a.m.
ben wrote:
How come this was never posted when you still could
get germanium
transistors. :)
You can still get germanium transistors.
It appears to me that 2N3906 silicon transistors should work in place of
the 2G371 with only minor changes to the circuits. However, you should
test individual circuits before you build the whole computer and
discover that it doesn't work.
12:29 p.m.
On 11/16/2012 1:26 AM, Eric Smith wrote:
ben wrote:
I was thinking 2N2222's. I took a quick look and nothing seems to need
ge transistors.
I still have to check what is PNP or NPN's so 2N2222's may not work.
Ben.
BTW how much is 1p in today's money, in US dollars.
How come this was never posted when you still
could get germanium
transistors. :)
You can still get germanium transistors.
It appears to me that 2N3906 silicon transistors should work in place of
the 2G371 with only minor changes to the circuits. However, you should
test individual circuits before you build the whole computer and
discover that it doesn't work. 
12:50 p.m.
On Nov 16, 2012, at 12:29 PM, ben wrote:
On 11/16/2012 1:26 AM, Eric Smith wrote:
2N2222s should work, but they're heavy hitters. Seems like
overkill. Also, I haven't read the article in question
yet (need to dig through the thread to find the link), but
if Eric was talking about 2N3906s, those are PNP (the 2N2222
is NPN). You'd probably need to modify the schematic, if
only a little bit.
2N3906 and 2N4402 are the PNP complements of 2N3904 and
2N4401, respectively; those are probably the cheapest
general-purpose BJTs you're likely to find. I have a
general preference for the 4401 over the 3904, but it
probably doesn't make a huge difference.
- Dave
It appears to me that 2N3906 silicon transistors
should work in place of
the 2G371 with only minor changes to the circuits. However, you should
test individual circuits before you build the whole computer and
discover that it doesn't work.
I was thinking 2N2222's. I took a quick look and nothing seems to need ge
transistors.
I still have to check what is PNP or NPN's so 2N2222's may not work. 
6:14 p.m.
On 11/16/2012 12:50 PM, David Riley wrote:
On Nov 16, 2012, at 12:29 PM, ben wrote:
The 2N2222A is a great device but a bit heavy duty for this application
and more costly.
However for high current buffer/inverter/driver it could be a good choice.
The winner is PN2222 (plastic 2n2222) and 2n3904.
I use 2N3904s a lot for switching and RF applications and it's a good
all round device.
Typically when I buy them I get 100 at a time as I use them for Zeners
(base-emitter
reversed is about 7V), Varicaps (base-collector reverse biased), High
conductance diode
(base shorted to collector) and of course as a transistor. Generally I
pay about 3-4 cents
(US) for known vendor parts.
Another device that is newer than the old bipolar transistor days are
the MOSFETs
specifically those in the 2N7000 and similar class as switches. They
offer very low
on resistance and high noise margin with the gate turn on voltage in the
3-8 Volt range.
Any one using them in quantity would do well to create a go/no-go tester
for leakage
and gain as a project like that will consume a lot of transistors and
marginal ones are
a big nuisance.
For diode the 1n914 and the 1n4148 are good choices for DRTL style logic.
Allison
On 11/16/2012 1:26 AM, Eric Smith wrote:
2N2222s should work, but they're heavy hitters. Seems like
overkill. Also, I haven't read the article in question
yet (need to dig through the thread to find the link), but
if Eric was talking about 2N3906s, those are PNP (the 2N2222
is NPN). You'd probably need to modify the schematic, if
only a little bit.
2N3906 and 2N4402 are the PNP complements of 2N3904 and
2N4401, respectively; those are probably the cheapest
general-purpose BJTs you're likely to find. I have a
general preference for the 4401 over the 3904, but it
probably doesn't make a huge difference.
- Dave It appears to me that 2N3906 silicon transistors
should work in place of
the 2G371 with only minor changes to the circuits. However, you should
test individual circuits before you build the whole computer and
discover that it doesn't work.
I was thinking 2N2222's. I took a quick
look and nothing seems to need ge transistors.
I still have to check what is PNP or NPN's so 2N2222's may not work.
6:59 p.m.
On 11/16/2012 4:14 PM, allison wrote:
I use 2N3904s a lot for switching and RF applications
and it's a good
all round device.
Typically when I buy them I get 100 at a time as I use them for Zeners
(base-emitter
reversed is about 7V), Varicaps (base-collector reverse biased), High
conductance diode
(base shorted to collector) and of course as a transistor. Generally I
pay about 3-4 cents
(US) for known vendor parts.
What I was wondering was how much it would cost today ( single sided pcb )
and modern parts, compared with the original product. I wonder what
speed leds
work at, and use them for switching diodes. The advantage is a nice visible
logic gate. Hmm, A blue transistor when on would be nice too!
Another device that is newer than the old bipolar
transistor days are
the MOSFETs
specifically those in the 2N7000 and similar class as switches. They
offer very low
on resistance and high noise margin with the gate turn on voltage in the
3-8 Volt range.
Any one using them in quantity would do well to create a go/no-go tester
for leakage
and gain as a project like that will consume a lot of transistors and
marginal ones are
a big nuisance.
For diode the 1n914 and the 1n4148 are good choices for DRTL style logic.
I am planning at the momement using 2901's, a nice 10/20 bit computer.
20 address bits, 4.77 mhz clock , byte access bus, and front panel - The
PC* killer . :)
1976 ish era mini-computer, rather than micro-computer with a micro-chip.
(Of course if any wants to put this design in silcon, feel free to do so) :)
Allison
Ben.
* At the moment a design with static memory, now for dynamic memory
refresh cycle
I would have to check if it will fit in the micro-code rom.
10:10 p.m.
On 11/16/2012 06:59 PM, ben wrote:
4.77mhz is about as fast as 2901s will run without pipelining.
I did an 8/16 bitter with 2901Cs and was profoundly disappointed.
Add length like 20 or 36bits makes for problems like devices like disks
as most available will be 8 or 16bits. Then again with large disks
half word splits will fit. Makes it troublesome to do 12 or other odd by
current standards lengths for words.
As to PC killer, that was never needed, the pc was killing itself with a
cpu that
was 8080 with a bag on the side. I always felt that the 8086/88 was a
market successful while failed design. Maybe why I used PDP11s and Z80s
till 1991 when 32bits made the scene at least interesting. The PC was an
appliance, not the latest computing tech.
Allison
On 11/16/2012 4:14 PM, allison wrote:
Go two sided for ground plane and easier layout. the cost difference is
small.
Leds make crappy diodes and are very slow (large die). Leds are slower
than transistors.
Building them in as state indicators on board for testing and
troubleshooting
has merit. However even small Leds eat power and are as costly as a
transistor.
I use 2N3904s a lot for switching and RF
applications and it's a good
all round device.
Typically when I buy them I get 100 at a time as I use them for Zeners
(base-emitter
reversed is about 7V), Varicaps (base-collector reverse biased), High
conductance diode
(base shorted to collector) and of course as a transistor. Generally I
pay about 3-4 cents
(US) for known vendor parts.
What I was wondering was how much it would cost today ( single sided
pcb )
and modern parts, compared with the original product. I wonder what
speed leds
work at, and use them for switching diodes. The advantage is a nice
visible
logic gate. Hmm, A blue transistor when on would be nice too!
Another device
that is newer than the old bipolar transistor days are
the MOSFETs
specifically those in the 2N7000 and similar class as switches. They
offer very low
on resistance and high noise margin with the gate turn on voltage in the
3-8 Volt range.
Any one using them in quantity would do well to create a go/no-go tester
for leakage
and gain as a project like that will consume a lot of transistors and
marginal ones are
a big nuisance.
For diode the 1n914 and the 1n4148 are good choices for DRTL style
logic.
I am planning at the momement using 2901's, a nice 10/20 bit computer.
20 address bits, 4.77 mhz clock , byte access bus, and front panel -
The PC* killer . :)
1976 ish era mini-computer, rather than micro-computer with a micro-chip.
(Of course if any wants to put this design in silcon, feel free to do
so) :)
Allison
Ben.
* At the moment a design with static memory, now for dynamic memory
refresh cycle
I would have to check if it will fit in the micro-code rom.
11:57 p.m.
On 11/16/2012 8:10 PM, allison wrote:
4.77mhz is about as fast as 2901s will run without
pipelining.
I did an 8/16 bitter with 2901Cs and was profoundly disappointed.
How ever in hindsight, memory speed and bus width ( 8 bits ) was the
limiting factor in speed back then.
Add length like 20 or 36bits makes for problems like
devices like disks
as most available will be 8 or 16bits. Then again with large disks
half word splits will fit. Makes it troublesome to do 12 or other odd by
current standards lengths for words.
Disks don't care what bits they get. With my limited knowlage, interface
chips where made for the upcoming 8 bit micro's, other machines still
relied on racks full of equipment.
As to PC killer, that was never needed, the pc was
killing itself with a
cpu that
was 8080 with a bag on the side. I always felt that the 8086/88 was a
market successful while failed design. Maybe why I used PDP11s and Z80s
till 1991 when 32bits made the scene at least interesting. The PC was an
appliance, not the latest computing tech.
The computer market, did need a machine
with more the 64K and only Motorola
and Intel had a product that could sell in personal computer. All the
other chips
came out just too late.
Since when is the latest, allways the best. I still favor analog audio
media,
and film for photos, and REAL IBM keyboards.
Allison
Any how since I am doing this computer as hobbie,I can play with the design
as "what if" and get a feel why computers ( crappy as they be in some ways)
are the way they are from that era in time.
Ben.
17 Nov
17 Nov
10:39 a.m.
On 11/16/2012 11:57 PM, ben wrote:
Actually they didn't but in practical terms you cant use/buy them much any
more so your limited to building now with devices that are based on 8/16 bit
interfaces.
As to racks, no the most of the interfaces are less than a fraction of a
rack
like about 4-7u.
Yes, true. save for the 8088/86 was still a 16bit program counter with
an embedded
MMU at no time could it address more than 64K of instruction space or
data space.
Even that was doable with then z80. The PDP11 had the MMU that was
worth emulating.
On 11/16/2012 8:10 PM, allison wrote:
No, the system I was building ran 2167s a 55ns parts not a problem.
The time it takes for the 2901a to do things like propagate a carry and
other internal operations that involve anythingthat ripples through is
the limit.
Then you have to allow for bus timing.
4.77mhz is about as fast as 2901s will run
without pipelining.
I did an 8/16 bitter with 2901Cs and was profoundly disappointed.
How ever in hindsight, memory speed and bus width ( 8 bits ) was the
limiting factor in speed back then. Add length like 20 or 36bits makes for problems
like devices like disks
as most available will be 8 or 16bits. Then again with large disks
half word splits will fit. Makes it troublesome to do 12 or other
odd by
current standards lengths for words.
Disks don't care what bits they get. With my limited knowlage, interface
chips where made for the upcoming 8 bit micro's, other machines still
relied on racks full of equipment.
As to PC killer, that was never needed, the pc
was killing itself with a
cpu that
was 8080 with a bag on the side. I always felt that the 8086/88 was a
market successful while failed design. Maybe why I used PDP11s and Z80s
till 1991 when 32bits made the scene at least interesting. The PC
was an
appliance, not the latest computing tech.
The computer market, did need a machine
with more the 64K and only
Motorola
and Intel had a product that could sell in personal computer. All the
other chips
came out just too late.
Since when is the latest, allways the best. I still
favor analog audio
media,
and film for photos, and REAL IBM keyboards.
The keyboard is ok. But most of my analog media (1/4in tape and vinyl)
are degrading.
Any how since I am doing this computer as hobbie,I can
play with the
design
as "what if" and get a feel why computers ( crappy as they be in some
ways)
are the way they are from that era in time.
Ben.
Yes you can. I only give the benefit of been there and did that as in I
did build fairly
complex transistor logic, TTL based systems and even a few things with
2901s. The
frustration then was more bits were expensive and everything was lower
than I'd
have liked. However some of the architectures were inspired and I
still see the
PDP-8 as an amazing machine for what it could do with so little actual
hardware.
Allison
16 Nov
16 Nov
3:45 a.m.
On 16/11/2012 08:13, ben wrote:
On 11/16/2012 12:50 AM, Ed Spittles wrote:
How come this was never posted when you still could get germanium
transistors. :)
Ben.
There was a dire shortage of the very rare round tuits during that
period of my life, plus no access to a decent scanner. I actually wrote
my first program when I was 17 or 18 years old ( so 1971/2) on a
germanium transistor computer that had been built by my school derived
[conversation about Wireless World articles - a digital computer
built from reject germanium transistors]
Aha: here's an online version of a pamphlet reprinting that article
series, which gives credit to Brian Crank:
http://www.smrcc.org.uk/members/g4ugm/Manuals/wirelessworldcomputer.pdf
Cheers
Ed
See also
http://www.classiccmp.org/pipermail/cctalk/2001-November/241361.html
from the wireless world design. I was taught by the
school Music
Teacher Hector Parr (http://www.hectorparr.freeuk.com/) who learnt to
program so he could calculate tables of Organ Pipe lengths to assist in
re-building the School pipe organ.
I can't remember how many memories it had, or the word length, but I do
remember that it was composed of flip-flops constructed from surplus
"red spot" transistors. Each as bit was constructed on a small paxolin
panel with a screw-in flash lamp bulb to show if it contained a "1" or a
"zero". The whole assemble was mounted on the wall on a panel about 6
or 8 feet square, with a small console attached by a cable. It was a
serial design so each "bit" was loaded separately by pressing a couple
of door bell pushes labelled "0" and "1". The first thing you did
before
using it was to load it full of "1"s so you could check no bulbs had
worked loose or blown.
I do remember it would play NIM if you used the final HALT instruction
as a "0" data byte.
We didn't write much code on it, once we had written a successful
program we were allowed to code Fortran II on an IBM1620 at a nearby
college. We posted (or mailed for the US readers) our coding forms off
on a Wednesday and with a bit of luck got the answers back on Friday, so
sometimes you got two runs a week. I wasn't very careful and needed the
two runs to get things working...
--
Dave Wade G4UGM
Illegitimi Non Carborundum
P.S. Al, if you want to take a coipy of the scan, or of any of the other stuff on that
site please feel free.
4:12 a.m.
Very interesting.....
I was at college between 1964 and 1969.
My major subjects were Computing and Industrial Electronics.
This article bears a strong resemblance to my lecture notes.
17" long Veroboard would warp like heck if you did not put a stiffener top
and bottom.
I can't see why other than cost and availability they used germanium
transitors.
?
Rod Smallwood
?
-----Original Message-----
From: cctech-bounces at classiccmp.org [mailto:cctech-bounces at classiccmp.org]
On Behalf Of ben
Sent: 16 November 2012 08:13
To: cctalk at classiccmp.org
Subject: Re: discrete transistor computer
On 11/16/2012 12:50 AM, Ed Spittles wrote:
http://www.classiccmp.org/pipermail/cctalk/2001-November/241361.html
How come this was never posted when you still
could get germanium
transistors. :)
Ben.
[conversation
about Wireless World articles - a digital computer built
from reject germanium
transistors]
Aha: here's an online version of a pamphlet reprinting that article
series, which gives credit to Brian Crank:
http://www.smrcc.org.uk/members/g4ugm/Manuals/wirelessworldcomputer.pdf
Cheers
Ed
See also
4246
days inactive
4247
days old
13 comments
7 participants
participants (7)
-
ajp166＠verizon.net -
bfranchuk＠jetnet.ab.ca -
dave.g4ugm＠gmail.com -
ed.spittles＠gmail.com -
eric＠brouhaha.com -
fraveydank＠gmail.com -
rodsmallwood＠btconnect.com