Delay Line Memory

Post by Charles Richmond
Yesterday I saw a video on YouTube about a delay line memory device.
It did *not* use mercury or any liquid to hold and circulate the
memory contents. Instead it had a *wire*!!! The wire was physically
twisted at one end ot input the bit. The other end "decoded" the bit
and also re-circulated the bit.
http://youtu.be/N9cUbYII5RY
I'd *never* heard of anything like this! Is anyone here familiar with
this type of delay line memory??? What machines used this???

Neat. Kinda like the mechanical reverb on an old musical instrument
amp. I've got a piano amp with one of those, if you jar it when the
reverb is on it sounds like an earthquake in a spring factory. Bet the
computer version was very sensitive to mechanical noise, too.

David Wade

2016-02-16 22:18:27 UTC

Post by Charles Richmond
Yesterday I saw a video on YouTube about a delay line memory device. It
did *not* use mercury or any liquid to hold and circulate the memory
contents. Instead it had a *wire*!!! The wire was physically twisted at
one end ot input the bit. The other end "decoded" the bit and also
re-circulated the bit.
http://youtu.be/N9cUbYII5RY
I'd *never* heard of anything like this! Is anyone here familiar with
this type of delay line memory??? What machines used this???

The Ferranti Pegasus used similar delay line stores as its main memory.
This one is one of the short stores...

http://www.zdnet.com/pictures/bletchley-museum-treasures-vintage-tech/6/

.. the larger ones had a longer coil and held a word of memory plus
parity...

I believe early IBM displays used this system as well....

Dave
G4UGM

Dan Espen

2016-02-16 22:36:43 UTC

I'm not familiar with the device, but it looks like I did use one.

https://en.wikipedia.org/wiki/Delay_line_memory

See the part about Magnetostrictive delay lines where they mention
the 2848, the controller for 2260 terminals.

The same article estimates the storage on a device like this as 1000
bits. You'd need more bits than that for a bunch of 2260s.
One would require (24x80x8 = 15360).

--
Dan Espen

Lawrence Statton

2016-02-16 22:57:18 UTC

I had a terminal that used one -- the wire was in a giant coil in the
base of the terminal. I can't for the *life* of me remember the
manufacturer's name.

bert

2016-02-16 23:08:34 UTC

Post by Charles Richmond
Yesterday I saw a video on YouTube about a delay line
memory device. It did *not* use mercury or any liquid
to hold and circulate the memory contents. Instead
it had a *wire*!!! The wire was physically twisted
at one end to input the bit. The other end "decoded"
the bit and also re-circulated the bit. I'd *never*
heard of anything like this! Is anyone here familiar
with this type of delay line memory??? What machines
used this???

Yes, it's a neat idea. A piezo-electric crystal at each
end of the wire converts an electrical pulse to a torsion
wave, and vice versa. Nickel wire delay lines had already
been used in some earlier radar equipment, then in 1952 in
two one-off computers SNARK (short nickel line accumulating
register calculator) and Nicholas, both at Borehamwood in
the UK. The Elliott 401 and the Ferranti Pegasus followed,
Pegasus being in service from 1956-68 approximately.

Pegasus had eight short single-word wires for its registers,
and six longer eight-word wires for blocks of current code
and data. These eight-word blocks were the units of transfer
to and from the main drum memory. Happy memories!

Information condensed from "The Pegasus Story" by Simon
Lavington, published by the (Kensington) Science Museum.
--

Bob Eager

2016-02-17 01:31:36 UTC

Post by Charles Richmond
Yesterday I saw a video on YouTube about a delay line memory device.
It did *not* use mercury or any liquid to hold and circulate the memory
contents. Instead it had a *wire*!!! The wire was physically twisted
at one end to input the bit. The other end "decoded" the bit and also
re-circulated the bit. I'd *never* heard of anything like this! Is
anyone here familiar with this type of delay line memory??? What
machines used this???

Yes, it's a neat idea. A piezo-electric crystal at each end of the wire
converts an electrical pulse to a torsion wave, and vice versa. Nickel
wire delay lines had already been used in some earlier radar equipment,
then in 1952 in two one-off computers SNARK (short nickel line
accumulating register calculator) and Nicholas, both at Borehamwood in
the UK. The Elliott 401 and the Ferranti Pegasus followed, Pegasus
being in service from 1956-68 approximately.
Pegasus had eight short single-word wires for its registers,
and six longer eight-word wires for blocks of current code and data.
These eight-word blocks were the units of transfer to and from the main
drum memory. Happy memories!
Information condensed from "The Pegasus Story" by Simon Lavington,
published by the (Kensington) Science Museum.

I shall have to ask Simon when I see him on Thursday!

--
Using UNIX since v6 (1975)...

Use the BIG mirror service in the UK:
http://www.mirrorservice.org

Charlie Gibbs

2016-02-17 03:42:38 UTC

Post by bert
Pegasus had eight short single-word wires for its registers,
and six longer eight-word wires for blocks of current code
and data. These eight-word blocks were the units of transfer
to and from the main drum memory. Happy memories!

Cache!

--
/~\ ***@kltpzyxm.invalid (Charlie Gibbs)
\ / I'm really at ac.dekanfrus if you read it the right way.
X Top-posted messages will probably be ignored. See RFC1855.
/ \ HTML will DEFINITELY be ignored. Join the ASCII ribbon campaign!

Morten Reistad

2016-02-16 23:25:38 UTC

Post by Charles Richmond
Yesterday I saw a video on YouTube about a delay line memory device. It did
*not* use mercury or any liquid to hold and circulate the memory contents.
Instead it had a *wire*!!! The wire was physically twisted at one end ot
input the bit. The other end "decoded" the bit and also re-circulated the
bit.
http://youtu.be/N9cUbYII5RY
I'd *never* heard of anything like this! Is anyone here familiar with this
type of delay line memory??? What machines used this???

Tat14?

160 colours of 10 gigabits times 6 paris with 72 ms delay; that should
be able to store 160 x 10G x 6 x 72/1000 / 8 = 72 gigabytes.

It stretches from Oud Bevervijk to Long Island.

-- mrr

Elliott Roper

2016-02-17 14:03:56 UTC

Post by Morten Reistad

Yesterday I saw a video on YouTube about a delay line memory device. It did
*not* use mercury or any liquid to hold and circulate the memory contents.
Instead it had a *wire*!!! The wire was physically twisted at one end ot
input the bit. The other end "decoded" the bit and also re-circulated the
bit.
http://youtu.be/N9cUbYII5RY
I'd *never* heard of anything like this! Is anyone here familiar with this
type of delay line memory??? What machines used this???

Tat14?
160 colours of 10 gigabits times 6 paris with 72 ms delay; that should
be able to store 160 x 10G x 6 x 72/1000 / 8 = 72 gigabytes.
It stretches from Oud Bevervijk to Long Island.

Heh! Nice one.

--
To de-mung my e-mail address:- fsnospam$elliott$$ PGP Fingerprint: 1A96 3CF7
637F 896B C810 E199 7E5C A9E4 8E59 E248

Anonymous

2016-02-17 00:14:36 UTC

Post by Charles Richmond
Yesterday I saw a video on YouTube about a delay line memory device. It did
*not* use mercury or any liquid to hold and circulate the memory contents.
Instead it had a *wire*!!! The wire was physically twisted at one end ot
input the bit. The other end "decoded" the bit and also re-circulated the
bit.
http://youtu.be/N9cUbYII5RY
I'd *never* heard of anything like this! Is anyone here familiar with this
type of delay line memory??? What machines used this???

The first program I ever wrote and ran, in 1966, was a Turing Machine
simulator
written for a Ferranti Sirius, which had just such a store.

I remember seeing the wire coils when an engineer was fixing a fault.

The Ferranti Pegasus and Perseus were among others with that store
technology.

--
Bill Findlay

Quadibloc

2016-02-17 19:46:13 UTC

Post by Anonymous
The Ferranti Pegasus and Perseus were among others with that store
technology.

Much later, magnetostrictive delay lines were used on the Packard-Bell pb250, a
relatively compact and inexpensive computer that predated the PDP-8.

John Savard

j***@earthlink.net

2016-02-18 02:18:11 UTC

Post by Anonymous
The Ferranti Pegasus and Perseus were among others with that store
technology.

Much later, magnetostrictive delay lines were used on the Packard-Bell pb250, a
relatively compact and inexpensive computer that predated the PDP-8.
John Savard

Yep, I worked on one of those machines when I was in the Air Force in 1962.

There was also a Royal Precision computer which was to have used delay
line memory - RPC-9000 or something like that. I don't think the machines
were ever produced, but a bunch of the delay lines turned up in the C&H
surplus store in Pasadena.

David Wade

2016-02-18 09:44:58 UTC

Post by Anonymous
The Ferranti Pegasus and Perseus were among others with that store
technology.

Much later, magnetostrictive delay lines were used on the Packard-Bell pb250, a
relatively compact and inexpensive computer that predated the PDP-8.
John Savard

The experimental MU5 computer at Manchester used plated wire store...

http://www.cs.manchester.ac.uk/about-us/history/mu5/

Dave
G4UGM

h***@bbs.cpcn.com

2016-02-17 00:43:55 UTC

As someone else mentioned, there was probably some sort of electrical
component to receive and then pass on each bit. The delay in doing so
allowed for the memory.

Old issues of C&A had ads for "pulse transformers", which may have been
used for delay line memory.

Today, it's hard to believe that core memory was once so expensive that
cheaper alternatives were sought.

When IBM introduced the 650, which used a magnetic drum for memory,
I think core was available, but the drum was cheaper. Later variants
of the 650 had a tiny bit of core for buffers. As just discussed,
the low end 1401 had only 1,400 characters of memory in it.

Question: when IBM introduced its PC, it had up to 640k of memory.
I think smaller amounts were available, with a cheaper price. Did
customers actually want to save a few bucks and go for the cheaper
and smaller PC? (To me, it seemed people couldn't wait for expanded
and extended memory to come out so the PC could get up to 1 meg and
even beyond, so it's hard to believe people would accept less than 640k.)

Anonymous

2016-02-17 01:24:32 UTC

...

Post by h***@bbs.cpcn.com
Old issues of C&A had ads for "pulse transformers", which may have been
used for delay line memory.

Nope, they are logic elements.
See section 6 of
http://www.findlayw.plus.com/KDF9/The%20Hardware%20of%20the%20KDF9.pdf

--
Bill Findlay

Peter Flass

2016-02-17 01:52:36 UTC

As someone else mentioned, there was probably some sort of electrical
component to receive and then pass on each bit. The delay in doing so
allowed for the memory.
Old issues of C&A had ads for "pulse transformers", which may have been
used for delay line memory.
Today, it's hard to believe that core memory was once so expensive that
cheaper alternatives were sought.
When IBM introduced the 650, which used a magnetic drum for memory,
I think core was available, but the drum was cheaper. Later variants
of the 650 had a tiny bit of core for buffers. As just discussed,
the low end 1401 had only 1,400 characters of memory in it.
Question: when IBM introduced its PC, it had up to 640k of memory.
I think smaller amounts were available, with a cheaper price. Did
customers actually want to save a few bucks and go for the cheaper
and smaller PC? (To me, it seemed people couldn't wait for expanded
and extended memory to come out so the PC could get up to 1 meg and
even beyond, so it's hard to believe people would accept less than 640k.)

Memory was expensive, you bought what you needed. Many people bought
whatever Lotus 1-2-3 wanted, maybe 128K?

--
Pete

Michael Black

2016-02-17 03:19:58 UTC

Post by Peter Flass

As someone else mentioned, there was probably some sort of electrical
component to receive and then pass on each bit. The delay in doing so
allowed for the memory.
Old issues of C&A had ads for "pulse transformers", which may have been
used for delay line memory.
Today, it's hard to believe that core memory was once so expensive that
cheaper alternatives were sought.
When IBM introduced the 650, which used a magnetic drum for memory,
I think core was available, but the drum was cheaper. Later variants
of the 650 had a tiny bit of core for buffers. As just discussed,
the low end 1401 had only 1,400 characters of memory in it.
Question: when IBM introduced its PC, it had up to 640k of memory.
I think smaller amounts were available, with a cheaper price. Did
customers actually want to save a few bucks and go for the cheaper
and smaller PC? (To me, it seemed people couldn't wait for expanded
and extended memory to come out so the PC could get up to 1 meg and
even beyond, so it's hard to believe people would accept less than 640k.)

Memory was expensive, you bought what you needed. Many people bought
whatever Lotus 1-2-3 wanted, maybe 128K?

And I seem to recall IBM wanted more for the memory than other companies.

So it often was cheaper to buy a barebones IBM, and then load it up with
memory and even a floppy drive or two, than pay IBM prices. It even got
better as third party companies started selling IBM bits, so you could buy
that Hercules video card that was better than what IBM offered originally.

Michael

JimP

2016-02-17 17:34:08 UTC

On Tue, 16 Feb 2016 18:52:36 -0700, Peter Flass

Post by Peter Flass

As someone else mentioned, there was probably some sort of electrical
component to receive and then pass on each bit. The delay in doing so
allowed for the memory.
Old issues of C&A had ads for "pulse transformers", which may have been
used for delay line memory.
Today, it's hard to believe that core memory was once so expensive that
cheaper alternatives were sought.
When IBM introduced the 650, which used a magnetic drum for memory,
I think core was available, but the drum was cheaper. Later variants
of the 650 had a tiny bit of core for buffers. As just discussed,
the low end 1401 had only 1,400 characters of memory in it.
Question: when IBM introduced its PC, it had up to 640k of memory.
I think smaller amounts were available, with a cheaper price. Did
customers actually want to save a few bucks and go for the cheaper
and smaller PC? (To me, it seemed people couldn't wait for expanded
and extended memory to come out so the PC could get up to 1 meg and
even beyond, so it's hard to believe people would accept less than 640k.)

Memory was expensive, you bought what you needed. Many people bought
whatever Lotus 1-2-3 wanted, maybe 128K?

When we first got desktop computers at the university, we had to buy
them with 384KB and 2 floppy drives for Lotus 1-2-3. Tandy 1000s, SX
compatibles.

--
JimP.

Dave Garland

2016-02-17 02:06:25 UTC

Post by h***@bbs.cpcn.com
Question: when IBM introduced its PC, it had up to 640k of memory.
I think smaller amounts were available, with a cheaper price. Did
customers actually want to save a few bucks and go for the cheaper
and smaller PC? (To me, it seemed people couldn't wait for expanded
and extended memory to come out so the PC could get up to 1 meg and
even beyond, so it's hard to believe people would accept less than 640k.)

Depends on what they were going to do. The people I dealt with mostly
got... 128K? (I remember a board from Symantec that combined more
memory with a parallel port.) The primary uses were 1) word
processing, and 2) spreadsheet. When the clones started appearing,
those were cheap enough so that more memory became common. ISTM that
the Sperry/Corona/Cordata that I bought at a hamfest had 128K or maybe
256K and sockets for up to 512K (I do remember stuffing sockets, and
replacing the CPU with a V20 for a little performance boost).

Lawrence Statton

2016-02-17 02:23:51 UTC

Post by h***@bbs.cpcn.com
Question: when IBM introduced its PC, it had up to 640k of memory.

Not quite. There was a hole in the address space that big, but the most
memory you could order the first edition machine was 64K. The second
rev of the board took 64k x 1 chips, so you could get 256K on the motherboard.

It was a few months or a year before other manufacturers sold memory
expansion (often coupled with a parallel port and a RTC and marketed as
a "Combo Board") that could add another 256K.

Post by h***@bbs.cpcn.com
I think smaller amounts were available, with a cheaper price. Did
customers actually want to save a few bucks and go for the cheaper
and smaller PC?

By "few bucks" you mean "Several hundred", right?

At the time of the PC's release, 256K of memory would set you back about
$500.

Post by h***@bbs.cpcn.com
(To me, it seemed people couldn't wait for expanded
and extended memory to come out so the PC could get up to 1 meg and
even beyond, so it's hard to believe people would accept less than 640k.)

--NK1G

hgww

2016-02-17 03:35:39 UTC

As someone else mentioned, there was probably some sort of electrical
component to receive and then pass on each bit. The delay in doing so
allowed for the memory.
Old issues of C&A had ads for "pulse transformers", which may have been
used for delay line memory.
Today, it's hard to believe that core memory was once so expensive that
cheaper alternatives were sought.
When IBM introduced the 650, which used a magnetic drum for memory,
I think core was available, but the drum was cheaper. Later variants
of the 650 had a tiny bit of core for buffers. As just discussed,
the low end 1401 had only 1,400 characters of memory in it.
Question: when IBM introduced its PC, it had up to 640k of memory.
I think smaller amounts were available, with a cheaper price. Did
customers actually want to save a few bucks and go for the cheaper
and smaller PC?

Yes, some did.

Post by h***@bbs.cpcn.com
(To me, it seemed people couldn't wait for expanded
and extended memory to come out so the PC could get up to 1 meg and
even beyond,

But not everyone did that.

Post by h***@bbs.cpcn.com
so it's hard to believe people would accept less than 640k.)

Plenty did.

Charlie Gibbs

2016-02-17 04:17:45 UTC

Memory was expensive. Lots of IBM PCs and clones had 256K or less.
(Why should they have more memory than the mainframe?) It's easy
to forget how expensive things were back then. Around the time the
IBM PC came out, I was doing a regular Friday pilgrimage to the
local Heathkit store to spend $16 on another 2016 static RAM chip,
building up my S-100 memory board 2K at a time. It was over 5 years
after the introduction of the IBM PC that I laid out $700 for a 2MB
memory expansion for my Amiga 1000. Then I had enough space to set
up a bootable RAM disk so I could get some speed without spending
the $2000 that a hard drive cost back then.

Charles Richmond

2016-02-18 23:14:41 UTC

Post by Charlie Gibbs

I remember in 1980, seeing an advertisement for a 64k memory board for the
S-100 bus. This board had 32 RAM chips, each was 16k x 1 bits, plus
associated interface and addressing chips. And the board was *only* $1000
US !!! I was amazed that much memory could be had so cheaply!!!

Back in 1986, I bought a 20 megabyte 3 1/2" hard drive in a "shoebox" that
contained the power supply and hard disk interface card to the disk could be
connect to an Atari ST. I think I paid $600 US for it, but it may have been
more like $800. Now this drive held only 20 megs!!!

I later replaced the 20 megabyte 3 1/2" drive mech with a Seagate ST-251 40
megabyte drive mech that was "self-parking". The Seagate drive was great
because it was so *noisy*!!! You felt like accessing your information took
real work from the hard drive!!!

--
numerist at aquaporin4 dot com

Lawrence Statton

2016-02-19 01:27:46 UTC

Post by Charles Richmond
I remember in 1980, seeing an advertisement for a 64k memory board for
the S-100 bus. This board had 32 RAM chips, each was 16k x 1 bits,
plus associated interface and addressing chips. And the board was
*only* $1000 US !!! I was amazed that much memory could be had so
cheaply!!!

In fairness - most S100 memory boards used static memory for "good and
sufficient reason" ("D-Ram timing is *hard*" quoth an engineer friend of
mine), and perhaps because some of the earliest dynamic boards were
Unbearably Badly Engineered, buyers were skeptical of them. I had one
stuffed to the gills with 2114 (1K x 4 in an 18-pin DIP) that was 32K,
and was hot as a pistol. Those chips were close to half a watt each,
and sticking 64 of them in a board sucked a few amps off the 8V bus. It
had two LM323 (TO3 five volt regulator rated for 3A), and each of those
had a honkin' huge shunt resistor to keep package dissipation down.

Good times...

A friend of mine had a box with 7 8K boards, that used 64 pieces of good
old 2102, with the same "WW resistor over the regulator" shunt trick,
and when he replaced the memory chips with 21L02, he didn't know to cut
the resistor off. He came by my lab, and we replaced the chips *again*,
and this time I explained "three terminal regulators can only source
current, they can't sink -- if your load current goes DOWN, you will
need to increase the shut resistance, or in this case, remove it
entirely -- the regulator can easily supply the load by itself"

--NK1G

Charlie Gibbs

2016-02-19 04:14:00 UTC

Post by Lawrence Statton

Yup. When I first built my IMSAI I had two RAM-4A boards, each with
4K worth of 2102s. The regulators ran hot. I replaced it with a
64K static RAM board which I bought with 16K and gradually worked
my way up to 62K, one 2016 at a time (half of the other 2K was for
the 2708 holding the boot code and BIOS, and the other for scratch
RAM for the BIOS). Even fully loaded, the board ran cool.

Michael Black

2016-02-20 00:07:17 UTC

Post by Lawrence Statton

Dynamic RAM was a new thing for a lot of the people creating the very
early home computers. I think the Altair board used RC timing for
something on their dynamic RAM board, and apparently that gave a lot of
trouble. It gave a bad name to dynamic RAM, so others avoided it, at
least for a while.

I ended up with a stray Processor Technology dynamic RAM board, I think
it's 16K, I got it about 1990 and no computer for it. I gather it was
okay, but when I checked the ads for this board, it was something like
$695 in 1976 or so.

But somehow things improved, either people got better, or they just
overcame the bias against dynamic RAM. The Apple II of course made good
use of it, including the ability to upgrade when the higher density RAM
came along. So did the TRS-80 that year, though I have a vague memory
that their expansion RAM didn't work so well, in part because it was in a
separate box. At some point it became easier and cheaper to use dynamic
RAM unless you only needed small amounts. My OSI Superboard had only
space for 8K of RAM, so there was no sense in going dynamic.

I also remember when 2K by 8 static RAM came along, in 24pin packages so
they could interchange with eproms. Godbout soon came out with a board
that I think was 64K, still a lot of ICs on that board, but not as bad as
lower density RAM and since it was CMOS, suddenly the issue of power draw
was no longer there.

The odd thing is, those CMOS static RAMs became even denser, but by that
point dynamic RAM was the way, so we generally didn't see anyone using the
denser CMOS static ram in "home computers".

Though I remember 15 or 20 years ago, people suggesting grabbing cache RAM
off "old" motherboards for projects, 32K RAM in a thin 24 or so pin
package, one or two ICs good enough for a "controller" type project.

Michael

Bob Eager

2016-02-17 01:30:36 UTC

Post by Charles Richmond
Yesterday I saw a video on YouTube about a delay line memory device. It
did *not* use mercury or any liquid to hold and circulate the memory
contents. Instead it had a *wire*!!! The wire was physically twisted at
one end ot input the bit. The other end "decoded" the bit and also
re-circulated the bit.
http://youtu.be/N9cUbYII5RY
I'd *never* heard of anything like this! Is anyone here familiar with
this type of delay line memory??? What machines used this???

Dunno, but the EDSAC rebuild in the UK is using nickel delay lines
instead of the original mercury. This is on grounds of cost, as well as
health and safety.

http://www.tnmoc.org/special-projects/edsac/recreating-edsac

--
Using UNIX since v6 (1975)...

Use the BIG mirror service in the UK:
http://www.mirrorservice.org

Joe Morris

2016-02-18 00:11:50 UTC

Post by Charles Richmond
Yesterday I saw a video on YouTube about a delay line memory device. It
did *not* use mercury or any liquid to hold and circulate the memory
contents. Instead it had a *wire*!!! The wire was physically twisted at
one end ot input the bit. The other end "decoded" the bit and also
re-circulated the bit.
http://youtu.be/N9cUbYII5RY
I'd *never* heard of anything like this! Is anyone here familiar with
this type of delay line memory??? What machines used this???

The IBM 2260 CRT terminal, which was introduced in 1964, was a text-only
device (with 240, 480, or 960 character display capacity) that *really* met
the definition of a "dumb terminal"...its display was merely a pattern
controlled by the 2848 Control Unit, essentially similar to the video
circuitry in a traditional analog television set.

This is on-topic because the 2848 used an acoustic delay line - twisted
wire - to hold the text to be shown on each attached 2260 display head.
There was one delay line for each attached head; my recollection is that
each delay line was housed in a metal can about 10" square by 3/4" thick.

An interesting oddity about the 2848 was that it used a magnetic core matrix
to generate the light/dark pattern of dots that formed characters on the
displays. You could look at the matrix and easily see many of the glyphs.

Wikipedia has an article on the 2260, including links to some documents
(including the 2848 FE manual) on Bitsavers.

Joe

Michael Black

2016-02-18 04:13:15 UTC

Post by Joe Morris

Post by Charles Richmond
Yesterday I saw a video on YouTube about a delay line memory device. It
did *not* use mercury or any liquid to hold and circulate the memory
contents. Instead it had a *wire*!!! The wire was physically twisted at
one end ot input the bit. The other end "decoded" the bit and also
re-circulated the bit.
http://youtu.be/N9cUbYII5RY
I'd *never* heard of anything like this! Is anyone here familiar with
this type of delay line memory??? What machines used this???

The IBM 2260 CRT terminal, which was introduced in 1964, was a text-only
device (with 240, 480, or 960 character display capacity) that *really* met
the definition of a "dumb terminal"...its display was merely a pattern
controlled by the 2848 Control Unit, essentially similar to the video
circuitry in a traditional analog television set.
This is on-topic because the 2848 used an acoustic delay line - twisted
wire - to hold the text to be shown on each attached 2260 display head.
There was one delay line for each attached head; my recollection is that
each delay line was housed in a metal can about 10" square by 3/4" thick.
An interesting oddity about the 2848 was that it used a magnetic core matrix
to generate the light/dark pattern of dots that formed characters on the
displays. You could look at the matrix and easily see many of the glyphs.
Wikipedia has an article on the 2260, including links to some documents
(including the 2848 FE manual) on Bitsavers.

And for some reason, when solid state memory came along, it wsa in the
form of long shift registers, though I'm not sure if that was the only
option or in parallel with "random access memory". I know about 1974, Don
Lancaster's first "TV Typewriter" was said to use it, and slow scan tv
converters for amateur radio used them.

I don't know whether somebody hadn't made a leap to memory directly
addressable or there was some reason to go with long shift registers.

But of course, like the delay lines, you had to keep the bits circulating,
and you could only get to each of the bytes by stepping through all that
came before.

Michael

Jon Elson

2016-02-18 04:25:48 UTC

Post by Michael Black
And for some reason, when solid state memory came along, it wsa in the
form of long shift registers, though I'm not sure if that was the only
option or in parallel with "random access memory". I know about 1974, Don
Lancaster's first "TV Typewriter" was said to use it, and slow scan tv
converters for amateur radio used them.

I cloned a Beehive terminal that used this in 1977 or so. Horrible wire-
wrapped monster, but it worked. The memory used charge coupled devices, so
the actual memory chips had very little "active" circuitry on them. Mostly
just a long string of MOS capacitors with drive lines. I guess they had a
sense amplifier at the output end.

Jon

isw

2016-02-18 07:16:05 UTC

Post by Michael Black

Post by Joe Morris

Post by Charles Richmond
Yesterday I saw a video on YouTube about a delay line memory device. It
did *not* use mercury or any liquid to hold and circulate the memory
contents. Instead it had a *wire*!!! The wire was physically twisted at
one end ot input the bit. The other end "decoded" the bit and also
re-circulated the bit.
http://youtu.be/N9cUbYII5RY
I'd *never* heard of anything like this! Is anyone here familiar with
this type of delay line memory??? What machines used this???

The IBM 2260 CRT terminal, which was introduced in 1964, was a text-only
device (with 240, 480, or 960 character display capacity) that *really* met
the definition of a "dumb terminal"...its display was merely a pattern
controlled by the 2848 Control Unit, essentially similar to the video
circuitry in a traditional analog television set.
This is on-topic because the 2848 used an acoustic delay line - twisted
wire - to hold the text to be shown on each attached 2260 display head.
There was one delay line for each attached head; my recollection is that
each delay line was housed in a metal can about 10" square by 3/4" thick.
An interesting oddity about the 2848 was that it used a magnetic core matrix
to generate the light/dark pattern of dots that formed characters on the
displays. You could look at the matrix and easily see many of the glyphs.
Wikipedia has an article on the 2260, including links to some documents
(including the 2848 FE manual) on Bitsavers.

And for some reason, when solid state memory came along, it wsa in the
form of long shift registers, though I'm not sure if that was the only
option or in parallel with "random access memory". I know about 1974, Don
Lancaster's first "TV Typewriter" was said to use it, and slow scan tv
converters for amateur radio used them.
I don't know whether somebody hadn't made a leap to memory directly
addressable or there was some reason to go with long shift registers.

A serial shift register is just a long string of D-type flip-flops. No
real "addressing" needed, just a counter. Plus, in the days when a lot
of chips were laid out *by hand*, a very simple step-and-repeat was very
appealing.

Isaac

Jon Elson

2016-02-18 18:01:34 UTC

Post by isw
A serial shift register is just a long string of D-type flip-flops. No
real "addressing" needed, just a counter. Plus, in the days when a lot
of chips were laid out *by hand*, a very simple step-and-repeat was very
appealing.

Actually, creating a full-screen (1920 bytes) memory out of D-flops would be
prohibitive. Assuming you only needed 6 bits per character, that is still
11520 bits, or 1920 chips of 6-wide D-flops. When these things were built,
2 FF/chip was the max, so that would be 5000 IC packages! Obviously, not
practical.

The Beehive monster I cloned used the 1404 chip, available from Intel and
AMD. 1024 bits by CCD bucket-brigade technology. It needed only 14 of
these chips to hold 2048 7-bit characters. (the screen only displayed 1920
in that particular unit.)

Jon

Jon Elson

2016-02-18 18:10:16 UTC

Post by Jon Elson
The Beehive monster I cloned used the 1404 chip, available from Intel and
AMD. 1024 bits by CCD bucket-brigade technology. It needed only 14 of
these chips to hold 2048 7-bit characters. (the screen only displayed
1920 in that particular unit.)

And, looking it up, they apparently were NOT CCDs, although from the outside
you really couldn't tell that, by the way they worked. Seems the scheme was
a string of "sticky inverters" with capacitors at the input, and
transmission gates between stages. They needed a 2-phase clock.

Jon

Lawrence Statton

2016-02-18 19:46:26 UTC

Post by Jon Elson

Well - building a memory out of discrete 74S74s would be painful, but
the static S/R memory could do that in a few square mils of real estate,
and did exactly that. (And, yes, static S/R was marginally more costly
than dynamic S/R, but was stable with no clock, which makes them
particularly well suited for terminals) ...

It is somewhat noteworthy, if you go look at some of the databooks from
the early 70s, that S/R memories came in such convenient sizes as 80 x
1, 80 x 2 etc. Many of these were available in convenient TO-5 packages,
so as not to frighten the engineering team with those new-fangled DIPs
:)

I don't remember anyone making anything larger than twin chains, but
that could be bit-rot.

A dozen pieces of 960 x 1 for the page buffer, and a half-dozen 80x1 for
the line buffer is all the memory a full 80x24 terminal would need.

--NK1G

Michael Black

2016-02-20 00:13:19 UTC

Post by Jon Elson

There were "quad latches" like the 7475 to go between counters and BCD to
7 segment decoders, needed when the counter was otherwise always counting.
I can't remember if the pinout allowed the four flip-flops to be used as a
shift register, but if it could, that would improve density 2:1. Later
(but of course by then nobody was using shift registers for memory), you
could get 8 latches in the same package, to latch an 8bit bus.

Michael

Quadibloc

2016-02-18 12:49:46 UTC

For a short time, although there were static random access memory chips also available, they had a limited capacity, and the long shift registers were the only chips that could hold a full 1,024 bits on a single chip.

Or so I thought, but looking it up:

Intel 1402/3/4 shift register - 1970
Intel 1103 DRAM - 1970
Intel 2102 SRAM - 1972

So while 1K static RAMs came 2 years later, 1K dynamic RAMs and 1K shift
registers were at least approximately contemporaneous. But a dynamic RAM would
require significantly more support circuitry, so if you could get away with a
shift register - say if you were designing a video terminal or "TV Typewriter"
(vide Don Lancaster's famous Cookbook) - you would. No doubt they were cheaper
too.

John Savard

Michael Black

2016-02-20 00:21:10 UTC

For a short time, although there were static random access memory chips also available, they had a limited capacity, and the long shift registers were the only chips that could hold a full 1,024 bits on a single chip.
Intel 1402/3/4 shift register - 1970
Intel 1103 DRAM - 1970
Intel 2102 SRAM - 1972
So while 1K static RAMs came 2 years later, 1K dynamic RAMs and 1K shift
registers were at least approximately contemporaneous. But a dynamic RAM would
require significantly more support circuitry, so if you could get away with a
shift register - say if you were designing a video terminal or "TV Typewriter"
(vide Don Lancaster's famous Cookbook) - you would. No doubt they were cheaper
too.

That's the thing. It was available "surplus" (I have no idea if that
actually meant "no longer needed" or it was just a distribution channel)
and thus "cheap". Right at that point, there wasn't much need for large
storage in hobbyist circles, but if you needed it, you ended up using
those long shift registers. I can't remember if I mentioned amateur radio
slow scan TV in my previous post, but that was using it up too around
1974, about the only thing besides that TV typewriter that made use of RAM
(it was a blink before microprocessors took off). I remember talk of one
SSTV scan converter, and a dismissal "because now you can't get the long
shift registers for it", which I assumed meant "the surplus stock dried
up".

Michael

Quadibloc

2016-02-18 12:51:17 UTC

Post by Michael Black
But of course, like the delay lines, you had to keep the bits circulating,

That's not true of shift registers in general - although it _was_ true of the
Intel 1402 et al, as it was a _dynamic_ shift register. So it could have a single
transistor per bit instead of a whole flip-flop.

John Savard

Michael Black

2016-02-20 00:16:54 UTC

Post by Michael Black
But of course, like the delay lines, you had to keep the bits circulating,

Sorry. I was speaking in terms of retrieving a given bit. I'm talking
long shift registers, you had to circulate the bit to the output in order
to make use of it, and then have the bit fed back into the input so it
wouldn't be forgotten by the shift register memory.

Michael

Peter Flass

2016-02-17 01:52:36 UTC

Post by Charles Richmond
Yesterday I saw a video on YouTube about a delay line memory device. It did
*not* use mercury or any liquid to hold and circulate the memory contents.
Instead it had a *wire*!!! The wire was physically twisted at one end ot
input the bit. The other end "decoded" the bit and also re-circulated the
bit.
http://youtu.be/N9cUbYII5RY
I'd *never* heard of anything like this! Is anyone here familiar with this
type of delay line memory??? What machines used this???

The IBM 2260 controller used this. I heard stories that people walking by
could cause garbage on the screens.

--
Pete

Charles Richmond

2016-02-20 02:50:43 UTC

Post by Peter Flass

Post by Charles Richmond
Yesterday I saw a video on YouTube about a delay line memory device. It did
*not* use mercury or any liquid to hold and circulate the memory contents.
Instead it had a *wire*!!! The wire was physically twisted at one end ot
input the bit. The other end "decoded" the bit and also re-circulated the
bit.
http://youtu.be/N9cUbYII5RY
I'd *never* heard of anything like this! Is anyone here familiar with this
type of delay line memory??? What machines used this???

The IBM 2260 controller used this. I heard stories that people walking by
could cause garbage on the screens.

I've worked at PPoE's with "programmers" who caused garbage on the screen...
these programmers' programs *were* garbage!!!

--
numerist at aquaporin4 dot com

Ahem A Rivet's Shot

2016-02-20 06:31:25 UTC

On Fri, 19 Feb 2016 20:50:43 -0600

Post by Charles Richmond

Post by Peter Flass
The IBM 2260 controller used this. I heard stories that people walking
by could cause garbage on the screens.

I've worked at PPoE's with "programmers" who caused garbage on the
screen... these programmers' programs *were* garbage!!!

Copro-grammers.

--
Steve O'Hara-Smith | Directable Mirror Arrays
C:>WIN | A better way to focus the sun
The computer obeys and wins. | licences available see
You lose and Bill collects. | http://www.sohara.org/

2016-02-17 03:08:58 UTC

On Tue, 16 Feb 2016 15:21:14 -0600, "Charles Richmond"

Post by Charles Richmond
I'd *never* heard of anything like this! Is anyone here familiar with this
type of delay line memory??? What machines used this???

Olivetti Programma 101 (1964) used this kind of memory for its storage. It
could store ten 22-digits numbers, each with sign and decimal separator, or
something like 120 machine instructions (plus some reserved work space).

http://www.curtamania.com/curta/database/brand/olivetti/Olivetti%20Programma%20101/
https://en.wikipedia.org/wiki/Programma_101
https://ub.fnwi.uva.nl/computermuseum/p101dl.php

See also this beautiful page about the newer Olivetti Logos 240 desktop
calculator (1970s) which used the same kind of storage with quite modern
circuitry: http://www.johnwolff.id.au/calculators/Tech/Logos240/Logos240.htm

HTH, :)
G.

Stan Barr

2016-02-17 08:19:36 UTC

Post by Charles Richmond
Yesterday I saw a video on YouTube about a delay line memory device. It did
*not* use mercury or any liquid to hold and circulate the memory contents.
Instead it had a *wire*!!! The wire was physically twisted at one end ot
input the bit. The other end "decoded" the bit and also re-circulated the
bit.
http://youtu.be/N9cUbYII5RY
I'd *never* heard of anything like this! Is anyone here familiar with this
type of delay line memory??? What machines used this???

I remember a computer that used a large diameter (6-inch or so) two
turn (I think) coil of wire with transducers at each end as memory.
It used magneto-striction to create the pulses IIRC.
Can't remember where I saw it but I recall a 5-foot cabinet with a
stack of these on the front covered by a glass door.

I have often wondered if it was possible to make a copy of an early
computer using the delay lines found in PAL tv sets. (These are used
to delay a scan line by one line period for colour correction.)

--
Stan Barr ***@bluesomatic.org

Jon Elson

2016-02-17 20:57:34 UTC

Post by Stan Barr
I have often wondered if it was possible to make a copy of an early
computer using the delay lines found in PAL tv sets. (These are used
to delay a scan line by one line period for colour correction.)

Yes, you could, but it would take a lot of them. Assuming something like 65
us per scan line, and assuming you could get a bandwidth of a couple MHz
through them, you would have trouble storing more than 50 bytes or so. So,
a 1K byte memory would take about 20 units of those delay lines. And, you
couldn't tie them all in series without degradation of the bandwidth.
Generally, we found you could get dispersion of a few %, so if you tied 20
65 us delays in series, your bandwidth would likely fall to below a MHz.
But, to speed things up, you might want to run them in a bunch of parallel
strings, looking more like the tracks on a drum-memory computer.

Jon

Charles Richmond

2016-02-20 02:43:03 UTC

[snip...] [snip...]
[snip...]
I remember a computer that used a large diameter (6-inch or so) two
turn (I think) coil of wire with transducers at each end as memory.
It used magneto-striction to create the pulses IIRC.
Can't remember where I saw it but I recall a 5-foot cabinet with a
stack of these on the front covered by a glass door.
I have often wondered if it was possible to make a copy of an early
computer using the delay lines found in PAL tv sets. (These are used
to delay a scan line by one line period for colour correction.)

Stan, surely the PAL TV sets use shift registers for "delay lines". ISTM
that old-style VDU terminals used shift registers to hold the characters....
before cheaper designs were able to use microprocessors for terminal
control.

--
numerist at aquaporin4 dot com

isw

2016-02-20 06:13:37 UTC

Post by Charles Richmond

Stan, surely the PAL TV sets use shift registers for "delay lines".

At least some early PAL sets used quartz blocks for delay. The signal
was fairly wide band analog. The technology to convert to and from
digital (so you could use a shift register) was way more expensive than
the precision quartz blocks. Quartz was used, BTW, because of its
thermal stability -- dimensional change would result in a timing change.

Isaac

Stan Barr

2016-02-20 08:17:42 UTC

On Fri, 19 Feb 2016 20:43:03 -0600, Charles Richmond

Post by Charles Richmond

I have a few in the parts bin, both the glass block type and long coils.
https://www.flickr.com/photos/anachrocomputer/2840365262/
Loading Image...

Shift registers came later.

--
Stan Barr ***@bluesomatic.org

Lawrence Statton

2016-02-20 15:17:36 UTC

Post by Charles Richmond
Stan, surely the PAL TV sets use shift registers for "delay lines".

Almost certainly not.

For the majority of the lifetime of PAL color, the cost of a
sufficiently fast flash converter and a suitably sized S/R memory was
easily two orders of magnitude faster than a simple analog 64uS delay.

Even "analog" television in those few places where it still exists, is
handled in any set made in the last one or two decades in purpose-built
DSP hardware.

--NK1G

Lawrence Statton

2016-02-20 15:20:52 UTC

Lawrence Statton <***@senguio.mx> writes:

A stupid diction capture-error

Post by Lawrence Statton
For the majority of the lifetime of PAL color, the cost of a
sufficiently fast flash converter and a suitably sized S/R memory was
easily two orders of magnitude faster than a simple analog 64uS delay.

Last line should read
easily two orders of magnitude COSTLIER than ...

not faster.

500 ml of hot black coffee, stat!

--NK1G

Quadibloc

2016-02-21 02:49:48 UTC

Post by Charles Richmond
Stan, surely the PAL TV sets use shift registers for "delay lines".

I don't think that's sure at all.

It's sure that most old SECAM TV sets had some kind of delay lines in them, but
PAL TV sets didn't really need them, although they _could_ be put to good use in
the fancier models.

John Savard

isw

2016-02-21 05:06:50 UTC

Post by Charles Richmond
Stan, surely the PAL TV sets use shift registers for "delay lines".

I don't think that's sure at all.
It's sure that most old SECAM TV sets had some kind of delay lines in them, but
PAL TV sets didn't really need them, although they _could_ be put to good use in
the fancier models.

Sets that didn't use a delay line were beset by the dread "high
brightness flicker". The final 6.25 Hz interlace which was facilitated
by the one-line delay mitigated it.

Isaac

j***@earthlink.net

2016-02-18 02:23:19 UTC

Post by Charles Richmond
Yesterday I saw a video on YouTube about a delay line memory device. It did
*not* use mercury or any liquid to hold and circulate the memory contents.
Instead it had a *wire*!!! The wire was physically twisted at one end ot
input the bit. The other end "decoded" the bit and also re-circulated the
bit.
http://youtu.be/N9cUbYII5RY
I'd *never* heard of anything like this! Is anyone here familiar with this
type of delay line memory??? What machines used this???
--
numerist at aquaporin4 dot com

There was yet another form of this memory that used a glass prism - sound
waves were reflected off the faces of the prism internal to the glass and
made a number of transits around the device before emerging on a different
face from the one where they went in. Circa 1963.

isw

2016-02-18 07:06:01 UTC