Real 60Hz NTSC 48K ZX Spectrum

edited May 2012 in Hardware
Not a 3rd party mod.
The joystick mod was added by the user, but the board and the computer is exactly what was sold in Chile, South America. We might be looking at the NTSC model that was sold in the U.S. via mail-order long ago, or the one that was imported by "English Micro Connection". But since Google isn't our friend in these matters, our only hope is that someone will enlighten us and explain if the U.S. NTSC 48K ZX Spectrum was just a prototype, if a few prototypes were made, or if it was really sold here the same way it was sold in South America.

A gentleman in the TS2068 group posted some interesting technical information after seeing these pictures. I will quote his posts later.

SSA40738.jpg

http://img.photobucket.com/albums/v400/zxspectrum/SinclairChile/SSA40738-1.jpg

SSA40743.jpg

http://img.photobucket.com/albums/v400/zxspectrum/SinclairChile/SSA40743-1.jpg

SSA40745.jpg

http://img.photobucket.com/albums/v400/zxspectrum/SinclairChile/SSA40745-1.jpg

SSA40746.jpg

http://img.photobucket.com/albums/v400/zxspectrum/SinclairChile/SSA40746-1.jpg
Post edited by zxbruno on
«134

Comments

  • edited April 2009
    What would be involved in converting a Pal speccy to NTSC. Dumping in a modulator and 'crystal' thingimabob?...is it as simple as that?
  • edited April 2009
    I also want to know. :)
    I'm not an expert in ULAs, but I remember Piters saying something about the ULA having to be specifically built for NTSC.
  • edited April 2009
    mentioning ULA's

    any idea where i can get my mits on them? (especially for the +2) or do i have to butcher working spectrums?
  • edited April 2009
    See this site:

    http://www.zxdesign.info/book/

    Soon we won't need to get ULAs from other Speccies. :)
  • edited April 2009
    zxbruno wrote: »
    See this site:

    http://www.zxdesign.info/book/

    Soon we won't need to get ULAs from other Speccies. :)


    OoooOooo :) interesting :)
  • edited April 2009
    I suspect Winston would be interested, even more so if it connected to the internet. :p
    I wanna tell you a story 'bout a woman I know...
  • edited April 2009
    Awsome !!

    ZXBruno - If you have any more details on this NTSC'ness, please send them over.

    Basically, an NTSC machine does require a different ULA - one that counts through 265 lines in a frame instead of 312 for a standard PAL machine, at 60 frames a second. This is confirmed by the ULA in the machine you show being a 6C011E and not a 6C001E.

    This is backed up by the two crystal values (14.110 and 3.579545 MHz)

    For the technical among us: The 14.110MHz drives the ULA, which divides this by 2 and produces 443 pixels per scanline, thus: 265 * 312 / (14.110x10^6 / 2) = 60Hz (approx)

    The 3.579545 MHz crystal produces the NTSC colour burst component.

    So basically, what you have here is a factory produced NTSC spectrum....
    It's definitely not a prototype, as the initial labour and fab costs for a ULA was quite expensive.

    By the way the code on the top tells you that the ULA was manufactured between 29th Oct and 4th Nov 1984 (week 44, 1984).

    Cool stuff.
    Any more?

    Chris
  • edited April 2009
    Just been chatting to aowen about this.

    Here are some of my thoughts:
    I've just plugged some figures into the calculator.
    If I were designing this NTSC ULA, then I would use 263 lines at the standard 14MHz clock to give 60.008Hz refresh, as the internal counter logic would be almost identical to the 312 line PAL ULA, and require only a tiny modification.
    At a 14.110MHz clock, 263 lines gives 60.553Hz
    Increasing the lines to 265, the refresh would be a more realistic 60.096Hz, but the modification to the counter logic would be more complicated.

    As this is a real NTSC machine, it would be nice to emulate this properly (and I'd like the details for my book ;-) ).

    For this we need to know what the refresh rate is....

    Bruno, do you have this machine in your possession??

    Chris
  • edited April 2009
    Phil's fusetest program can extract frame length info on 60Hz machines (it works on a TS2068 with spectrum emulator cart). I believe we would get some interesting info if we can run that against this NTSC machine.
  • edited April 2009
    Hi Fred. How're tricks?
    Fred wrote: »
    Phil's fusetest program can extract frame length info on 60Hz machines (it works on a TS2068 with spectrum emulator cart). I believe we would get some interesting info if we can run that against this NTSC machine.

    I did think of this, but convinced myself that it wouldn't work because the time measured would be relative to the clock, which in this case is 14.11MHz.

    Doh!

    If we know the number of T-states per frame, then we can work out exactly how long that takes because we know precisely how long a T-state is, given any arbitrary clock freq.

    Thanks Fred for putting me back on the right track.

    Chris
    (Having one of those days when he should have stayed in bed).
  • edited April 2009
    I'm willing to buy one. I'm going to see if Claudio could source me one (he's been extremely helpful!). If I can get one in my hands I'll loan it to you Chris. I'll even cover the shipping expenses both ways.
  • edited April 2009
    zxbruno wrote: »
    I'm willing to buy one. I'm going to see if Claudio could source me one (he's been extremely helpful!). If I can get one in my hands I'll loan it to you Chris. I'll even cover the shipping expenses both ways.

    Excellent!
    I promise not to get the ULA decapitated and planed down and imaged!
    I'd just need the ULA actually....

    Doing some software tests in the interim would give us an idea though....

    Chris
  • edited April 2009
    csmith wrote: »
    Hi Fred. How're tricks?

    Good thanks, new baby taking up all my time though :)
    I did think of this, but convinced myself that it wouldn't work because the time measured would be relative to the clock, which in this case is 14.11MHz.

    Doh!

    If we know the number of T-states per frame, then we can work out exactly how long that takes because we know precisely how long a T-state is, given any arbitrary clock freq.

    I note that the crystal on the NTSC Spectrum produces a CPU clock that is within 500Hz of the believed value of the TS2068. Given that Timex would have been given the ULA source when making the TS2068 I believe there is a fair chance the TS2068 describes exactly the internal counters used in the NTSC ULA and the timings are the same as the TS2068.
  • edited April 2009
    I've emailed Claudio and asked if there's a way to find one for me.
    Meanwhile, Oscar in the TS2068 group says that the ULA isn't different from the one used on a standard 48K, and that the trick is just in the crystal and fine adjustments.

    I will try to continue to quote things here and there, but it might be better if we keep the topic in just one place.
  • edited April 2009
    zxbruno wrote: »
    I'm willing to buy one. I'm going to see if Claudio could source me one (he's been extremely helpful!). If I can get one in my hands I'll loan it to you Chris. I'll even cover the shipping expenses both ways.


    Ummm if he can get 2 let me know!
  • edited April 2009
    zxbruno wrote: »
    Meanwhile, Oscar in the TS2068 group says that the ULA isn't different from the one used on a standard 48K, and that the trick is just in the crystal and fine adjustments.

    I don't see how this can be true. The ULA generates the interrupts of which there are 50/sec in PAL land and 60/sec in NTSC land. It also has to generate the correct number of lines in the output image; 263 or so in NTSC land vs 312 in PAL land.

    Note this is also how the Timex clones work - the NTSC TS2068 has a different SCLD to the PAL TS2068/TC2048.

    Unlike the ZX81 there doesn't appear to be any external configuration that would allow these differences to be configured in jumpers or the like. It appears to me that also now know that the ULA used in these machines are not the same ones used in the PAL versions.

    I expect that producing an NTSC Speccy involves: correct ULA, correct master clock (14.11MHz for NTSC, 14MHz for PAL), correct colour burst clock and correct video signal encoder for PAL or NTSC.
  • edited April 2009
    Fred wrote: »
    I don't see how this can be true. The ULA generates the interrupts of which there are 50/sec in PAL land and 60/sec in NTSC land. It also has to generate the correct number of lines in the output image; 263 or so in NTSC land vs 312 in PAL land.

    Note this is also how the Timex clones work - the NTSC TS2068 has a different SCLD to the PAL TS2068/TC2048.

    Unlike the ZX81 there doesn't appear to be any external configuration that would allow these differences to be configured in jumpers or the like. It appears to me that also now know that the ULA used in these machines are not the same ones used in the PAL versions.

    I expect that producing an NTSC Speccy involves: correct ULA, correct master clock (14.11MHz for NTSC, 14MHz for PAL), correct colour burst clock and correct video signal encoder for PAL or NTSC.

    What happen if I play PAL ZX Spectrum with NTSC TV by using composite input (not RF)?
    Will it show the picture?
  • edited April 2009
    Fred wrote: »
    I note that the crystal on the NTSC Spectrum produces a CPU clock that is within 500Hz of the believed value of the TS2068. Given that Timex would have been given the ULA source when making the TS2068 I believe there is a fair chance the TS2068 describes exactly the internal counters used in the NTSC ULA and the timings are the same as the TS2068.
    I agree. However, given that the ntsc spectrum uses a ULA, the easiest minor tweaks would have been made to the UK ULA fabrication mask to create the NTSC chip, avoiding extensive routing re-work. The Timex SCLD was however produced from scratch, potentially whilst thinking about both markets, with extra display modes, so the design options would have been more flexible.

    It'll be great to confirm that.
  • edited April 2009
    zxbruno wrote: »
    I've emailed Claudio and asked if there's a way to find one for me.
    Meanwhile, Oscar in the TS2068 group says that the ULA isn't different from the one used on a standard 48K, and that the trick is just in the crystal and fine adjustments.

    I've also e-mailed Aowen to ask for help understanding what needs to be done by Claudio, so that I can translate it to Spanish if necessary.

    That can't work, or the display will be rubbish. As fred points out, the master crystal controls all the timing, and ramping up the clock may give you 60Hz at 312 lines, but not 60Hz at 263-265 lines.

    Claudio needs to run phil's fusetest and get us a screen photo. That'll give us a load of info.
    Fred, anything else of use?
  • edited April 2009
    Fred, can you give me more info regarding the 500Hz difference? What crystal is in the timex machine?
  • edited April 2009
    superfo wrote: »
    What happen if I play PAL ZX Spectrum with NTSC TV by using composite input (not RF)? Will it show the picture?

    Only if the TV is a multi-system one. The best option in my opinion is a TS2068 with Spectrum emulator cartridge, or an ultra-rare NTSC Spectrum. Second best seems to be to use a TV card in your PC with composite input, as the TV cards tend to be multi system anyway. There are a range of other options that can also work but are more complex (e.g. Spectrum 128k RGB port -> CGA -> VGA).
  • edited April 2009
    csmith wrote: »
    The Timex SCLD was however produced from scratch, potentially whilst thinking about both markets, with extra display modes, so the design options would have been more flexible.

    True, note that they still used distinct SCLD parts for NTSC land vs PAL land, and the PAL SCLD.

    To support your point, the PAL SCLD is very slightly different to the PAL ULA (contention starts 15 cycles earlier).
  • edited April 2009
    csmith wrote: »
    Claudio needs to run phil's fusetest and get us a screen photo. That'll give us a load of info.
    Fred, anything else of use?

    I think that is the place to start, once we confirm some initial data it should be possible to tweak fusetest to handle the NTSC 48k and come up with a contention tester to check the contention start and pattern from Phil's contention testing program.
  • edited April 2009
    csmith wrote: »
    Fred, can you give me more info regarding the 500Hz difference? What crystal is in the timex machine?

    The Timex technical manual states that the master clock in that machine is 14.112MHz, while the clock we can see in the pictures provided of the NTSC speccy is 14.11MHz.

    The only picture of the TS2068 PCB I've seen conceals the master crystal and I'm not game to open mine to have a look!

    As a result the TS2068 CPU clock is theoretically 3.528MHz (confirmed by framelength from fusetest and by evaluating the timings of tape samples from the machine), while the NTSC speccy is 3.5275MHz.
  • edited April 2009
    Fred wrote: »
    Only if the TV is a multi-system one. The best option in my opinion is a TS2068 with Spectrum emulator cartridge, or an ultra-rare NTSC Spectrum. Second best seems to be to use a TV card in your PC with composite input, as the TV cards tend to be multi system anyway. There are a range of other options that can also work but are more complex (e.g. Spectrum 128k RGB port -> CGA -> VGA).

    I try connect my spectrum (48K, +2/+2A and +3) with composite out to NTSC TV,
    All work fine except that it show in black and white, no color.
    Does it mean ULA does not do any thing about PAL or NTSC?

    Also TC2048 or TC2068, they use LM1377 for video/color circuit, there is a pin for select PAL/NTSC, does it mean their ULA does not do any thing about PAL or NTSC?
  • edited April 2009
    superfo wrote: »
    Also TC2048 or TC2068, they use LM1377 for video/color circuit, there is a pin for select PAL/NTSC, does it mean their ULA does not do any thing about PAL or NTSC?

    The ULA is colour-encoder agnostic. It just supplies YUV colour difference signals containing a number of line syncs and a field sync. For NTSC this is approx 263 x 60Hz, for PAL 312 x 50Hz. Thus the ULA knows about PAL and NTSC in terms of the number of lines it needs to generate, not how the colour burst is encoded.
  • edited April 2009
    Fred wrote: »
    The only picture of the TS2068 PCB I've seen conceals the master crystal and I'm not game to open mine to have a look!

    Please tell me what to do and I will gladly take a picture. Just keep in mind my photographic skills suck and the auto focus doesn't always help me. :)
  • edited April 2009
    zxbruno wrote: »
    The result of the first test (kindly sent by Claudio):
    Fantastic. I'm not sure what that reported frame rate translates to, as I'm out and about right now. I'm sure someone will reply before I get home ;-)
  • edited April 2009
    NTSC Speccy
    frame length 0xe700 is 59136 tstates
    CPU speed = 14.110MHz/4 / 59136 = 59.65Hz
    224 tstates/line = 264 lines

    vs TS2068
    frame length 0xe540 is 58688 tstates
    CPU speed = 14.112MHz/4 / 58688 = 60.1145Hz

    224 tstates/line = 262 lines

    Next we need to look for the tstate at the start of the paper area and the contention pattern.
Sign In or Register to comment.