On 2016-May-24, at 1:49 PM, Eric Smith wrote:
> On Tue, May 24, 2016 at 1:49 PM, Brent Hilpert <hilp...@cs.ubc.ca> wrote:
>> We were discussing this last year, perhaps I'm being pedantic but I would
>> note that while, as you say, there is commonality of principle in use of
>> induction and the selective weave to represent the data, TROS and core rope
>> (of the sort used in the AGC) also have differences in their principles of
>> operation - they're not just physical variations on each other.
>
> My understanding is that both used drive lines that either went
> through the transformer, or around it, to either couple a drive line
> to a sense line, or not. In the case of CRM, the wires are essentially
> braided with the cores, while in TROS, holes are punched in strips of
> flex circuit to break one of the two paths the drive line can take for
> each sense position, and the transformer core is a two-part
> rectangular thing rather than a little toroid.
>
> If I'm wrong, or missing some fine point distinguishing them, I'd
> welcome corrections or additional information.
Yes, I examined this in some detail last year after mention on the list, and
wrote it up:
http://www.cs.ubc.ca/~hilpert/e/corerope/index.html
The short of it is, schemes like TROS are using simple induction / transformer
principles with a selective weave through the transformer cores to represent
the data. In contrast, (AGC-style) core ropes are using switching cores and
core-logic principles to also do the 1-of-n address decoding within the cores.
The address decoding requires a varied weave of address wires through the
cores, in addition to the selective weave for the data. The read/access
operation also becomes far more complex for the AGC-style core rope.
