Posted By
TLC
on 2010-11-19
10:59:16
|  Re: 256K
Hi guys!
gerliczer: just as above...  In a "perfect" version, Csory's extension might have been present. ...Or not. I don't know. There are too many factors.
As said, I started off from something different than just implementing the capabilities of the old memory expansion schemas. That gave a slightly different viewpoint, where expandability would have been a factor. Up to 1M, the needed number of banked address bits fit in both Hannes' and Csorys schemas (there are enough unused address bits for an extension to 1M in either bank registers). Above 1M, you'd see that you definitely need a "leap", because you run out of free extra address bits (ie. if "expanded" in a straightforward way, you'd find yourself banging memory bank bits in two bank registers for a single bank switch operation, where some bits of the respective bank register wouldn't even be part of the address). This would just be ugly. ...There are another problems; in fact, this "simple at first sight" logics of Hannes' expansion, topped by some late additions by Solder plus the overhead of Csory's logics imposes lots of "exceptions", which will cause the control logics to get reasonably complicated (which in turn increases the number of gates needed), especially if the number of extra address bits to-be-controlled rises. Already at 1M, you find yourself in a situation where you could do a "perfect" memory expansion (ie. no damn exceptions, "just" four bank registers for each 16k memory windows of the 64k memory space), in roughly the similar number of components that you'd need for implementing Hannes' + Csorys logics. Since at some point I came to the conclusion that the expanded (1M) Hannes/Solder/Csory logics plus rams won't fit in the allocated space anyway, and after having asked someone whose opinion I trust whether there are _that_ many stuffs exploiting Csory's memory banking logics (according to him, there weren't), I decided not to fiddle with anything but what was purely needed to start up with an expanded machine, and also decided not to implement Csory's banking logics.
It might be possible to add Csory's banking schema, but the circuit would need some modifications. At the moment, the address decoding logic won't decode $FD15 (...because I needed some extra functions in order to replace the original 6529B bank register by TTLs), and some extra gates would need to be added in order to implement the banking logic. As you can see on the photos: room is already tight, yet extending the board would still be possible. Or one could use SMD components, let alone programmable logics to turn down on the space factor; me won't, I find great enjoyment in using obsolete components and technics in respect to obsolete computers , consequently, that one will need to be done by someone else.
Replacing the PLA: well, theoretically possible... But that'd yet again decrease the available space for extra logics on a daughterboard.
Edit: Originally I didn't consider the installation of that lead very critical (ie. that it needs to be soldered), because someone installing the memory expansion would need to do lots of soldering in the machine anyway (removing 64k memory chips, installing sockets, connecting the extra address pins of the sockets together). It's unfortunately not possible to avoid the installing person's capability to solder things, even if the needed lead were to be installed without soldering...
siz: that's very nice that you could find Csory's 256K tools, and /me will definitely take a look at them once available. Many thanks in advance!
Csabo, Luca: actually, I didn't count on too many potential users of this hack either, for obvious reasons; small number of 256k-capable stuffs, complex installation process, ...actually, I was partly happy about that (ambivalent feelings...), because I sincerely didn't intend to encourage people to believe, at some presumable point, with a possibly high number of 64k+ machines available on the scene in some time, that expanded memory was "standard", ie. that they should get advantage of that in "standard Plus/4" democoding, like it was standard Plus/4 "feature", that it really wasn't and never will be. For me, re-implementing and installing Hannes'/Solders expansion logic was fun, and I think I'll also benefit from that later (because, ...from time to time, I'm fiddling with small experimental codes on the machine; it's very direct and "immediate", something that I'm missing using cross assemblers) -- as you can see, these might be pretty special motives. So far I only checked two 256k specific programs myself, ie. Bionics Dream World and SVS' Mega Assembler (also using them as test cases), so the number of available programs was not a factor for me either, so to say...
SVS: siz already answered your question -- this is a "classic" Hannes/Solder expansion, implemented on a daughterboard.
Availability...
I'm gonna still keep on waiting for some time (...also, taking the time to crash-test the prototype; according to past days experiments, it's already obvious that memory address multiplexing is time critical, the timing of the original machine is already on the edge, and that'll need to be addressed in a "production" version). I'd see if Csory's expansion is worth/possible to add to the design. Though, that low number of Csory-expansion specific stuffs must also be possible to hack to work with Solder's schema, as I guess. ...I don't know. As for the "production run"'s size: as things appear now, there won't be many but just a few machines to be modified; a production run of that size is still possible, and presumably not insanely expensive (...the board design is pretty simple, after all (an advantage of using conservative tech), so I can just live with less expensive PCB companies' services). But I'm gonna refer with some more precise estimations at that stage. ...As long as the relatively low number applies, I can and will be happy to install the memory expansions myself, if needed. That imposes different problems (organizing, and/or posting), but those should be resouble as I guess.
|
