Image Map Image Map
Page 1 of 2 12 LastLast
Results 1 to 10 of 13

Thread: Overclocking a Tandy 1000 EX: ever been done?

  1. #1
    Join Date
    May 2011
    Location
    Outer Mongolia
    Posts
    1,262

    Default Overclocking a Tandy 1000 EX: ever been done?

    Recently I ordered a couple NEC V20 CPUs from China which are supposedly rated for up to 16mhz. (I'm painfully aware that the listing may be an elaborate tapestry of lies, but from poking around before I ordered it appeared that at least some people on this forum ordered parts from a similar listing and ended up with CPUs that were the real deal.) I'm hoping that the V20 will provide something like a meaningful upgrade all by itself for my Tandy 1000 EX; I like the humble little box but my experience so far with software that actually uses the 320x200x16 video modes hasn't been particularly positive.

    Just for fun I decided to look and see if there was any simple way to goose the clock speed of this era Tandy 1000. (I've seen the thread referring to the "FrankenTandy SL", but it looks as if that machine differs in some substantial ways from the EX.) Clock generation on the 1000 EX is handled by a chip called the "light blue", U16. According to the tech manual this chip generates a pile of independent clock signals by applying various dividers to two oscillators running at 16 and 28.3 mhz. The schematic sheets in the manual are a little difficult to follow, but it looks like the only things directly connected to the CPUCLK output of U16 are the CPU (via a resistor), and the CLK line on the expansion bus (also via resistor). It does *not* appear as if the video circuitry, which also controls the onboard 256k of RAM, uses this signal directly; it seems to run off its own 28mhz input signal.

    My ignorant question at this point would be: if one were to hijack the clock line at the CPU socket and feed it a faster clock would that "work", as long as the peripherals that are driven by the CPU's control signals can respond fast enough? (Granted there may not be much headroom in the design, which I suppose would render this moot.) Or is this just flat-out not going to work because there's a requirement that the CPU in the Tandy 1000's clock has to run synchronously with one of the other clocks?

    For a faster clock source I was specifically drawn to this design for a simple clock doubler:

    https://www.maximintegrated.com/en/a...ex.mvp/id/3327

    In theory at least inserting this between the CPU socket and the CPU would make the "slow" speed 9.54mhz and "fast" 14.32mhz. No doubt that latter speed would almost certainly be too fast for the RAM and other components in the machine, but perhaps "slow" might be usable? Alternatively, is there a simple way to monitor the control signals on the CPU and perhaps switch the clock doubling on and off depending on whether the CPU is working internally verses doing I/O? I'm sure that would negate at least a significant chunk of the speed boost, but perhaps still provide some benefit?

    I know the Tandy RAM card with the DMA chip relies on the CLK signal on the expansion bus to generate its control signals, but my machine doesn't have that. (I'm currently running an experimental homebrew RAM card sans DMA.) Does anything else rely on that signal?

  2. Default

    You have my attention There were ALLOT of "turbo" 8088/8086 motherboards out there toward the end of the XT lifespan and you could always look and see how they managed the various clocks, the bulk of the cards seemed to work fine on those.

    I was just thinking about an old daughterboard I had that installed in the CPU socket and held the CPU and a 8087 so you could add a math copro to a system without a socket, those copros could clock pretty high as well. I wonder if the HX/EX would play well with a copro, it might be a neat PCB to have made up in volume. A copro and overclock might be handy and worth doing just because.

  3. #3

    Default

    Following. My only issue with my HX was the slow performance for the games I enjoy. I'd love to make it more useful.

  4. #4
    Join Date
    Aug 2006
    Location
    Chicagoland, Illinois, USA
    Posts
    5,945
    Blog Entries
    1

    Default

    Search this forum for posts by "carlostex" who has done much of this work and posted about it.
    Offering a bounty for:
    - The software "Overhead Express" (doesn't have to be original, can be a copy)
    - A working Sanyo MBC-775, Olivetti M24, or Logabax 1600
    - Music Construction Set, IBM Music Feature edition (has red sticker on front stating IBM Music Feature)

  5. #5
    Join Date
    May 2011
    Location
    Outer Mongolia
    Posts
    1,262

    Default

    Quote Originally Posted by Trixter View Post
    Search this forum for posts by "carlostex" who has done much of this work and posted about it.
    He's the owner of the "FrankenTandy" SL, correct? I did see the thread for that (mentioned it at the start), but it didn't look like it was directly transferable because the SL uses a different timing chip. Those threads are a little confusing to follow but the TL;DR seems to be that the SL derives its 8mhz CPU from a separate 24mhz crystal than the 28mhz crystal that's the master for a pile of other timebases, and his hack involves replacing that crystal with a faster one.

    From what I can tell from the technical manual the CPU clock on the EX is a division of the 28mhz crystal, which feeds both "Light Blue" and "Big Blue". I assume all the video timings are set up by Big Blue so *maaaaybe* you could cut the line and feed Light Blue from a different oscillator, but it looks like doing that would mess up the NTSC colorburst clock, the Keyboard clock, and possibly a few other things. So I don't think the same technique is going to work here, unless I'm missing something.

    Edit: There is a second oscillator in the EX, a 16mhz one, and its output is divided down into 8.000 and 4.000mhz clocks for the floppy controller. I wonder if it would be a valid experiment to hook a jumper onto the 8mhz clock and feed the CPU's clock pin with it. I don't have a DMA controller to worry about, so maybe it'd work as a proof of concept?

    Edit #2: Specific post in the FrankenTandy SL thread where it's applicability or lack thereof to the 1000HX is mentioned.
    Last edited by Eudimorphodon; August 8th, 2019 at 02:12 PM.

  6. #6
    Join Date
    May 2011
    Location
    Outer Mongolia
    Posts
    1,262

    Default

    My NEC V-20s showed up in the mail, so one of these days I need to get around to installing one. I wish I didn't have to tear all that plasticized tinfoil off the motherboard to get to the CPU socket, but I guess that's what you sign up for with a 1000 EX.

    After looking at the service manual and scratching my head for a while I've been wondering if a glorious (but almost certainly doomed) experiment to try would be to tie the CLK pin of my supposedly 16mhz V-20 to the buffered OSCY signal on the expansion bus, which runs at 14mhz, in phase with the 7.16mhz fast clock. The reason I'm thinking this is probably doomed to failure (beyond a general issue of a 100% overclock simply being too much for various components to tolerate) is on page 24 of the technical manual, per whit:

    The READY circuit synchronizes the system "ready" signals with the CPU clock and generates the CPU input READY. If a function needs one or more "wait" states added to its access, it must set the RDYIN line low. From the main logic board, RDYIN is set low by the sound IC for 32 extra "wait states" and the video/system memory sets RDYIN low for typically one or two "wait" cycles. The READY circuit of the Timing Control Generator (U16) is operated in the non-asynchronous mode; i.e. two sequential edges of clock (a rising edge first) are required to set the READY signal true. RDYIN is pulled-up by R30.
    This specifically says that READY runs synchronously with the CPU clock. I don't know what the implications would be for having the CPU executing two ticks for every one; the timing diagrams imply that READY is pulled low synchronously with the falling edge of CPUCLK, which I suppose would still happen at a simple clock doubling. If the CPU simply isn't going to care about exactly how many cycles go by when it's asserted (IE, one wait state just turns into two) then maybe it'll work okay, at least for the components that assert wait states?

    I'm also wondering if there's something significant about the fact that when it's running in "slow" mode the CPUCLK duty cycle is 33/66% and in fast mode it's 50/50%, IE, the "high" pulses are always the same width, that being 1/4 of the 28mhz master clock's pulse width. Pulling the clock from the 14mhz source would halve that and I'm wondering if something is vitally dependent on that wavelength.

  7. #7
    Join Date
    May 2011
    Location
    Outer Mongolia
    Posts
    1,262

    Default

    Quote Originally Posted by Eudimorphodon View Post
    My NEC V-20s showed up in the mail, so one of these days I need to get around to installing one.
    Apparently I couldn't wait, so as soon as the work whistle blew I got to it. And...

    As I anticipated, the idea of just tying to the 14mhz clock didn't work. To try it I *slightly* bent the CLK pin of the CPU (19) out so it didn't engage the socket, connected a test clip to it, and then tried the following:

    Control Test: Clipped the other end to the CLK pin on the expansion bus, B20, which is the same clock fed to the CPU clock. Machine made it through self-test fine and booted.

    14mhz Test: Clipped to "OSC", pin B30 on the expansion bus. Hard black screen on power on.

    I think I'm getting cold feet from going any further, as I don't have any Tandy spares for the proprietary clock chip, etc. More reading of the manual enlightened me that the clock chip doesn't just generate clocks, it's tied to a couple of the CPU control lines so it, among other things, controls the '245 buffers on the data bus. I suppose it's *possible* the black screen could have just been from the onboard RAM being too slow and replacing it might get me further but... man, I wish spare Tandys were cheaper. Used to be people couldn't give a 1000 away.

    Good news is the $2.50 V-20 from eBay seems to work fine. That was probably a worthwhile investment based on what Topbench is telling me.

    If someone smarter than I am with a closet full of Tandy 1000's wants to take this up...

  8. Default

    I just PM'd you and forgot to comment on this as well. I DO have a few spare 1000HX motherboards, one I think is bad and a few are good. I would be happy to send you one as a test platform for this project, I would love to see success on this.

  9. #9
    Join Date
    May 2011
    Location
    Outer Mongolia
    Posts
    1,262

    Default

    I can't make any promises about success with the overclocking (although if I had a beater board I would be willing to try a few things like faster RAM and maybe some experiments with separate crystals or maybe that PLL/doubler idea if I can work out the details), but if you're swimming in boards it might be hugely useful to have an HX board (if I can work out how to run it on the bench) for compatibility testing my RAM board and upcoming CF (and hopefully serial) card projects on. Right now I can only definitely say the prototype works on *my* EX.

  10. #10

    Default

    Hi there.

    I think you'd be better off to instead look at the schematic of the motherboard and find the trace that leads from the Oscillator to the light blue chip.

    Which OSC frequency is sent to the CPU? For the Tandy EX i assume its a 14.318MHZ signal, which /3 gives 4.77 and /2 gives 7.15Mhz.

    So to maximize the V20HL potential you need to feed the light blue chip with 32MHz. /3 will give a slow speed of 10,66 and /2 will give full CPU 16Mhz potential. The issue is if the light blue chip has a single 14.318MHz input, patching a higher clock signal into that will screw up the other signals. Check the light blue chip pinout.

    Patching a clock signal directly into the CPU pin, is a bigger recipe for disaster, you need to assure duty cycles and everything else is in synch.

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •