Re: [M100] Switchable clock doubler schematic
Phew, ChatGPT - sounds like a challenge! Well, with REXCPM offering 4MB of RAM... we could have 4MB of optimized 8085 code! Would that do it :-) In all seriousness Steve, no nothing specifically in mind, just interested. I guess the std keyboard polling & auto-repeat speed on M100 is quite leisurely, so double speed wouldn't be too fast. Philip On 15/05/2023 3:39 pm, Kenneth Pettit wrote: On 5/14/23 6:09 PM, Stephen Adolph wrote: Hi Philip, It does work with both REX flavors, yes. Yes the keyboard is polled more frequently, but I have not seen a problem with that. Did you have something in mind? Probably ChatGPT on the M100 or something. :-) Ken
Re: [M100] Switchable clock doubler schematic
On 5/14/23 6:09 PM, Stephen Adolph wrote: Hi Philip, It does work with both REX flavors, yes. Yes the keyboard is polled more frequently, but I have not seen a problem with that. Did you have something in mind? Probably ChatGPT on the M100 or something. :-) Ken
Re: [M100] Switchable clock doubler schematic
Hi Philip, It does work with both REX flavors, yes. Yes the keyboard is polled more frequently, but I have not seen a problem with that. Did you have something in mind? On Sun, May 14, 2023 at 3:34 PM Philip Avery wrote: > Sounds great Steve. Will it work fine with Rex # & REXCPM? Also, the > keyboard would be accessed faster would it not? Does that cause a problem? > > Philip > > On 14/05/2023 1:51 am, Stephen Adolph wrote: > > I continue to hack away at making an easy to install clock doubler > > board for Model T computers. I realize that modifying hardware is not > > of interest to most, but I spend a lot of time on it myself. > > > > The point of the activity is to allow the user to switch from nominal > > 2.5MHz operation to 2x or 5MHz operation using a BASIC command. In > > M100/T102/NEC/M10/KC-85, all these variants have a very slow screen > > response time, and so doubling the clock rate makes the machine more > > responsive. Tandy 200 isn't as slow, but it is still nice to run fast. > > > > 5MHz operation is way outside of specifications, true. None of the > > parts in an M100 say they can run that fast. So far though, I have > > been making extensive use of it and have had no problems. The only > > issues that show up are when there is a timing loop in software that > > does not like being sped up. In situations like that, the user can > > downgrade to 2.5MHz mode. I think switchability is a must for this > > reason. > > > > So far I have successfully upgraded > > Model 100 (several! both older and newer) > > Tandy 102 (a few) > > Tandy 200 (one) > > > > Generally, when speeding up, the computer needs to run both the main > > ROM and the system RAM at 2x speed, and the power increases a bit. > > > > Main ROM speed: > > I have found that in the M100, for early production models, one needs > > to replace the main ROM for a faster one. Some later production use > > ROMs that are already fast enough. Tandy 102 and 200 ROMs seem fast > > enough. > > > > RAM speed: > > When the computer uses RAM that is 250nsec rated, these are often too > > slow. This can occur in M100, T102 and T200. In T200, I have no work > > around. The T200 needs 200nsec or faster RAM. In M100/T102, I have a > > work around that involves a simple cut/strap. > > > > Power: > > Consumption generally increases. For M100, with a nominal current of > > 56mA for the stock computer, a modified computer in 2.5MHz mode will > > see an increase to about 60mA. This is the "tax" paid for speed up > > capability. When operating in 5MHz mode, the current jumps to about > > 75mA or so. > > > > I'm attaching my current schematic for anyone interested in how the > > little board is designed. The board mounts on top of the 80C85, and > > steals the signals it needs directly from the CPU. There are 5 key > > circuits. > > > > 1) oscillator: generates 9.8304MHz > > > > 2) clock divider: generates 2.45MHz for the computer system > > components. Replaces the CLK output of the CPU, so that the > > downstream elements always run based on 2.45MHz. > > > > 3) State: a circuit that captures the user command to operate in > > either 1x or 2x clock speed. User command is OUT 85, 0 or OUT 85, 1. > > (80"85"). In T200, the real time clock RP5C01 is too slow for 5MHz > > operation, so the CL input is the chip select line for that part. The > > CL input forces the clock to 2.5MHz whenever the RC5C01 is accessed. > > > > 4) Clock select: This circuit uses 2 flip flops to generate either > > 9.8MHz or 4.9MHz depending on state. This circuit feeds clock to the > > 80C85. The original on-board crystal can be overdriven, no problem. > > > > 5) Astar signal: In M100/T102 (and likely M10 and 8201) the A* > > signal is used to gate the SRAM chip selects. What this signal does > > is it limits the "on" time of the SRAM, to limit power consumption. I > > was previously unaware of this trick. In 2x mode, the stock A* > > signal limits the access cycle time to the SRAM, and only fast SRAM > > (200nsec or less) can tolerate this. The solution for older machines > > with slower SRAM is to trigger A* earlier. This circuit generates > > "early A*" to resolve this problem. > > > > This implementation is now current and is being used in M100, T102 and > > T200. I'm planning to extend to M10 and NEC shortly, and if I > > encounter more changes I will update the design. > > > > My ultimate goal is to have a board that is as easy as possible to > > build and install, using SOIC parts (which I think are ok to hand > > solder) and through hole parts for the remainder. > > Currently the boar has 6 SOICs, 4 resistors, 4 caps and a crystal. > > > > The schematic is attached as a PNG for now, for the curious. > > > > Comments and questions welcome! > > If anyone is interested in the EAGLE board file, I'm happy to share it. > > > > Steve > > > >
Re: [M100] Switchable clock doubler schematic
Sounds great Steve. Will it work fine with Rex # & REXCPM? Also, the keyboard would be accessed faster would it not? Does that cause a problem? Philip On 14/05/2023 1:51 am, Stephen Adolph wrote: I continue to hack away at making an easy to install clock doubler board for Model T computers. I realize that modifying hardware is not of interest to most, but I spend a lot of time on it myself. The point of the activity is to allow the user to switch from nominal 2.5MHz operation to 2x or 5MHz operation using a BASIC command. In M100/T102/NEC/M10/KC-85, all these variants have a very slow screen response time, and so doubling the clock rate makes the machine more responsive. Tandy 200 isn't as slow, but it is still nice to run fast. 5MHz operation is way outside of specifications, true. None of the parts in an M100 say they can run that fast. So far though, I have been making extensive use of it and have had no problems. The only issues that show up are when there is a timing loop in software that does not like being sped up. In situations like that, the user can downgrade to 2.5MHz mode. I think switchability is a must for this reason. So far I have successfully upgraded Model 100 (several! both older and newer) Tandy 102 (a few) Tandy 200 (one) Generally, when speeding up, the computer needs to run both the main ROM and the system RAM at 2x speed, and the power increases a bit. Main ROM speed: I have found that in the M100, for early production models, one needs to replace the main ROM for a faster one. Some later production use ROMs that are already fast enough. Tandy 102 and 200 ROMs seem fast enough. RAM speed: When the computer uses RAM that is 250nsec rated, these are often too slow. This can occur in M100, T102 and T200. In T200, I have no work around. The T200 needs 200nsec or faster RAM. In M100/T102, I have a work around that involves a simple cut/strap. Power: Consumption generally increases. For M100, with a nominal current of 56mA for the stock computer, a modified computer in 2.5MHz mode will see an increase to about 60mA. This is the "tax" paid for speed up capability. When operating in 5MHz mode, the current jumps to about 75mA or so. I'm attaching my current schematic for anyone interested in how the little board is designed. The board mounts on top of the 80C85, and steals the signals it needs directly from the CPU. There are 5 key circuits. 1) oscillator: generates 9.8304MHz 2) clock divider: generates 2.45MHz for the computer system components. Replaces the CLK output of the CPU, so that the downstream elements always run based on 2.45MHz. 3) State: a circuit that captures the user command to operate in either 1x or 2x clock speed. User command is OUT 85, 0 or OUT 85, 1. (80"85"). In T200, the real time clock RP5C01 is too slow for 5MHz operation, so the CL input is the chip select line for that part. The CL input forces the clock to 2.5MHz whenever the RC5C01 is accessed. 4) Clock select: This circuit uses 2 flip flops to generate either 9.8MHz or 4.9MHz depending on state. This circuit feeds clock to the 80C85. The original on-board crystal can be overdriven, no problem. 5) Astar signal: In M100/T102 (and likely M10 and 8201) the A* signal is used to gate the SRAM chip selects. What this signal does is it limits the "on" time of the SRAM, to limit power consumption. I was previously unaware of this trick. In 2x mode, the stock A* signal limits the access cycle time to the SRAM, and only fast SRAM (200nsec or less) can tolerate this. The solution for older machines with slower SRAM is to trigger A* earlier. This circuit generates "early A*" to resolve this problem. This implementation is now current and is being used in M100, T102 and T200. I'm planning to extend to M10 and NEC shortly, and if I encounter more changes I will update the design. My ultimate goal is to have a board that is as easy as possible to build and install, using SOIC parts (which I think are ok to hand solder) and through hole parts for the remainder. Currently the boar has 6 SOICs, 4 resistors, 4 caps and a crystal. The schematic is attached as a PNG for now, for the curious. Comments and questions welcome! If anyone is interested in the EAGLE board file, I'm happy to share it. Steve
Re: [M100] Switchable clock doubler schematic
Definitely pushing the limits of our old hardware! Great stuff. As to software incompatibilities I think if the ROM and TS-DOS timing loops were fixed that covers most of what we do. One of these days I'd love to see a demo of it. Installing such a board myself seems unlikely. Just this week I found out how bad my soldering (or desoldering in this case) skills really are. I've been taking woodworking classes and one of my professors is really into 3d printing. I decided to give it a shot and bought a Comgrow "amazon return" used Ender 3 V2 for $109. It did have minor missing stuff, wrong adjustments which I fixed or worked around. The one remaining problem is a busted rotary encoder on the user interface screen. I worked around it by using Octopi to control the printer reflashed with Marlin firmware. For completeness I went for replacing the encoder, but it was really difficult and I ended up pulling up a trace and the via barrel too. Sigh. Oh well, octopi works fine. Better for me to focus on software and my other projects. I've already used the 3d printer to make some little things and parts I needed to complete old home projects so the 3d printing side quest is working out. -- John.
