What are the differences between the +2 and the +2A ?

2»

Comments

  • We're getting older and it's taking us longer to reply to threads, etc. Maybe some of the unanswered questions in some topics will be answered 14 years later.
  • Sorry, can you bump this in fourteen years time, then I can answer... :))

    Mark
  • Simple, invest in a 128k+1 (toastrack)!
    I have 2 of them, extrememly reliable, cool looking (like a turbo looking 48k+), loads virtually everything from 16 to 128k games without crashing,...and worth a mint!
    Yes they are expensive but I'd rather buy one of those than the ugly looking later models made my Amstrad.
  • Your definition of ugliness is quite interesting. ;)

    zx-plus-3.jpeg

    Even the good old ruber Speccy is a hundred times more beautifuly designed, than this awfull QL style.
  • MartinUK wrote: »
    There's a pretty long list of =2A/+3 incompatible games. Think the Amstrad-era tape decks were pretty cheapo, although my second-hand +2's worked well - until my kid brother spilled pineapple juice in it, that is. Never quite the same after that, and doesn't work at all anymore, alas.
    Of course there is an equally long list of games not compatible with non +2A/+3 spectrums too. Particularly with later era 128K only titles, there is a better chance of it working properly on a +2A than anything else (purely because later developers often didn't realise there was any difference and +2A's were plentiful).

    Honestly the 128 line was a bit of a mess, the original machine being badly thought out and poorly documented and that becoming compounded by Amstrad "fixing" some things for the +3 (and hence +2A) and assuming that what little Sinclair documentation they had was correct.
  • edited February 2017
    I think you will find that a lot of home computers from the 1980s were not well documented to the levels software and hardware developers would (have) liked.

    The route cause of the problem here, was that the ZX Spectrum 48k model was not really designed with any possible future machines in mind. So when the 128k toastrack was developed, one of the fixes for a problem with the earlier Spectrum models was to have a common clock (crystal) for the ULA, CPU and the PAL colour encoder chip. This changed the timings slightly. The extra RAM was added as two sets of RAM chips. One set replaced the 4116 "lower" RAM which includes the scree, so is of course, contended. The other set replaced the 32k "upper" RAM chips, which is of course uncontended. The Amstrad +2 uses the same ULA as the toastrack, so is very similar, and hence the memory contention is the same.

    When Amstrad developed the +2A and +3, they switched away from using a ULA and used a ASIC instead. This did away with the simple resistor bus separators used in all earlier machines. They also appear to have based the 128k memory layout on the existing (incorrect) documentation ( but I don't know if this came from Sinclair, or somewhere else) rather than the existing hardware design (which they had already copied!) So caused some further incompatibilities. Compare the Amstrad produced manuals for the +2 (chapter 24, page 147), the +2A (chapter 24, page 195), and the +3 (chapter 24, page 189). They all say the same thing about which banks are contended and which are uncontended despite the +2 having the same ULA as the toastrack!
    +2 Manual

    But despite all this, I think the effort put into trying to retain backwards compatibility was worth it, as many games and software titles still ran okay. It could have been so much worse...

    Mark

    Post edited by 1024MAK on
  • 1024MAK wrote: »
    When Amstrad developed the +2A and +3, they switched away from using a ULA and used a ASIC instead. This did away with the simple resistor bus separators used in all earlier machines. They also appear to have based the 128k memory layout on the existing (incorrect) documentation ( but I don't know if this came from Sinclair, or somewhere else) rather than the existing hardware design (which they had already copied!) So caused some further incompatibilities. Compare the Amstrad produced manuals for the +2 (chapter 24, page 147), the +2A (chapter 24, page 195), and the +3 (chapter 24, page 189). They all say the same thing about which banks are contended and which are uncontended despite the +2 having the same ULA as the toastrack!
    From the conversations I had with Cliff Lawson back in the day, the documentation came from Sinclair but it was sketchy at best and they had to do a fair bit of reverse engineering to figure out how the 128K ROM worked, for example, as it was pretty much a black box. The +2 had literally used the same hardware as the old 128s, so they hadn't really known how the parts worked as such, which is why the change in the memory contention layout wasn't spotted in the +3 redesign until far too late. Some of the design "features", such as the resistor bus separation, had Amstrad's hardware guys scratching their heads as they didn't quite believe a design like that could ever have worked reliably.

    It's quite a contrast from Amstrad's own CPC range, which was extensively documented (both the hardware and software) and probably had better compatibility across the range for it. The only place it tends to stumble is where demo coders relied upon specific details of "off-the-shelf" components like the 6845 CRTC, which Amstrad sourced from multiple manufacturers, leading to demo effects which don't entirely work on all models.

  • To be fair to Amstrad i think they switched away from the Ferranti ULA design to an ASIC, simply because by 1987 Ferranti was getting into difficulties, and the technology was not dating well either. The ASIC in the +2A/+3 design is actually still a gate array, just much more modern technology (apparently made by NEC).
  • Amstrad were no fans of Ferranti, dating back to the development of the CPC464 where design problems with the ULA that Ferranti had been contracted to produce for the machine led Amstrad to abandon that line of development and instead went to SGS to develop what ultimately became the 40007 and 40010 gate arrays that powered the 464 and 6128 machines.

    I'd imagine upon inheriting the ULA technology still central to the Sinclair range in 1986 they were in no small hurry to migrate away from it.
    The Spectrum Resuscitation Thread - bringing dead Spectrums back to life
    zx-diagnostics - Fixing ZX Spectrums in the 21st Century (wiki)
    Sinclair FAQ Wiki
Sign In or Register to comment.