> On Feb 3, 2023, at 1:12 AM, Chuck Guzis via cctalk <[email protected]>
> wrote:
>
> On 2/2/23 21:23, Tom Hunter via cctalk wrote:
>> The actual ferrite core doughnuts do not break down with continued use, BUT
>> moisture or mechanical impact or vibration will damage or degrade the
>> ferrite cores. Otherwise the ferrite doughnut will live and maintain its
>> properties "forever".
>
> Well, I don't know about that. The CDC 7600 had issues with core
> overheating and included a "Duty Cycle Integrator" on core. See PDF
> page 51, page 2-24:
>
> http://www.bitsavers.org/pdf/cdc/cyber/cyber_70/60367200D_Cyber70-76_Jul75.pdf
>
> --Chuvk
That reminds me of a PDP-11 diagnostic -- not usually run -- called the "core
heating test". The way I remember the description is that it would do rapid
memory accesses in a set of addresses that are physically close in the core
memory (obviously this is model-dependent). The idea was to find marginal
memories. One reason for not running that test is that it was very slow.
From the same era I recall that our IBM 1620 mod 2 had a heating system for the
memory cabinet, and that after power up you had to wait 10 minutes or so for
the memory to be warmed up to its normal operating temperature. I don't think
I ever saw an explanation why this was done.
It's puzzling that temperature would matter. Obviously, when you hit the Curie
temperature the data goes away, but for typical magnetic materials that is in
the hundreds of degrees. Does the hysteresis curve shift enough at moderate
temperatures (a bit over room temperature) to matter?
paul
