IBM 2260 acoustic delay line

[William Donzelli]

OK - I was told they were mercury, but frankly, I have been sort of
skeptical that mercury delay lines would still be used in new digital
designs that late in the game. Wire makes far more sense.

--
Will

On Sun, Dec 13, 2015 at 1:09 AM, Brent Hilpert <hilpert at cs.ubc.ca> wrote:
> On 2015-Dec-12, at 7:13 PM, Eric Christopherson wrote:
>> On Sat, Dec 12, 2015, Jon Elson wrote (in the big top posting thread):
>>>
>>> Later they got some
>>> IBM 2260's, which were Zenith 9" TV sets and a keyboard connected to an
>>> interface box in the machine room.  Very primitive, but very interactive,
>>> great for quick program editing and submission.
>>
>> I'm reading about those terminals and find it just fascinating how they
>> used acoustic delay line memory to remember the pixels. But I have lots
>> of questions:
>>
>> 1. Did the cables connecting the 2260s to the display controller
>> actually contain the delay lines themselves, over the whole length; or
>> were the delay lines just inside the controller and then some electronic
>> signal was sent out to the terminals?
>
> The delay lines for the 2260 systems were magnetostrictive (not mercury) acoustic delay lines, contained in the controller.
>
> Magnetostrictive delay lines were a somewhat common memory technique in the 60s, they were used in other early CRT display terminals and in some electronic calculators of the period.
> They were an improvement over the mercury delay lines of the first stored-program computers, easier to work with and not as temperamental.
>
> Speaking generally of the technique (I never worked on a 2848 controller, I have worked on them in calculators):
>
> Magnetostriction is a characteristic of some materials in which the material will physically expand or contract slightly in response to an applied magnetic field.
> Thus a solenoid with a core of magnetostrictive material going through it acts as a transducer of electrical energy to or from mechanical (acoustic) energy.
> IIRC, nickel or a nickel alloy was the commonly-used magnetostrictive material.
>
> Generated acoustic pulses are entered into one end of the delay medium (a metallic wire line) to be picked up at the other end by another transducer, so converted back to electrical energy, electrically amplified and restored to account for losses and distortion from the traverse of the delay medium, changed as appropriate (read/write operations) and re-entered into the delay medium.
> Some implementations sent the acoustic pulses longitudinally through the delay wire (expansion/contraction or push/pull of the wire), some sent them torsionally (twisting/rotating the wire).
>
> The whole point of course, is the much slower speed of acoustic pulses relative to electrical pulses results in a much longer latency through a delay medium of a given length and hence more (time for) storage.
>
> The delay wire is formed into a spiral/coil loop for compactness.
>
> If you're interested in some specifications and physical data of a 1960s magnetostrictive memory, in this instance in a calculator, see the commentary part way down the page here, beside the picture of the coiled delay line:
>         http://www.cs.ubc.ca/~hilpert/eec/calcs/Sony2500.html
> These are from measurements and calculations I made on that delay line.
> Or in summary: the acoustic pulses travel at 8140 Km/h, taking 1.6 mS to traverse a delay wire 3.62 m long, providing storage for 1024 bits, each bit occupying about 3.5 mm in the line.
> This calculator uses longitudinal pulses, while the 2260/2848 apparently uses torsional pulses.
>