• Please review our updated Terms and Rules here

Anyone know of any cheap TRS-80 Model III overclock hacks?

Eudimorphodon

Veteran Member
Joined
May 9, 2011
Messages
4,877
Location
Upper Triassic
Probably jumping the gun asking now, since I don't yet have the thing in hand to play with, but I'm curious: I remember sources like 80 Microcomputing printing on multiple occasions simple speed-up hacks for the TRS-80 Model I; a 50% speed-up from 1.77mhz to 2.6mhz was a really common mod, and doubling all the way to 3.54mhz was also apparently usually possible with just a little tinkering with some spare gates on the board. (And maybe swapping the Z80 for a Z80A.)

Does anyone remember if similar 50%-100% hacks were easily possible on the Model III as well? I know from flipping through the service manual that it's a little bit more sophisticated of a machine than the Model I, since it includes that WAIT circuit for the video memory (among other things) so maybe it's more locked into the lower speed, but if it's possible to wring a little more out of it on the cheap I might be interested in that.

(The main application I have in mind for this is Orchestra 90; faster is definitely better for that.)
 

danielbooneamerica

Experienced Member
Joined
Mar 3, 2020
Messages
336
Location
Wayne National Forest
Just as a side note.

I once asked Roy Soltoff if TRSDOS was developed on some larger system and cross assembled to Z80 and he told me NO.

All work was done on stock TRS80s that had been modified to run faster (wonder what happened to these historical pieces).

He never told me what those modifications were.
 

Eudimorphodon

Veteran Member
Joined
May 9, 2011
Messages
4,877
Location
Upper Triassic
I guess I'll just start slogging through the back-issue archives of the various TRS-80 magazines to see if there's something out there. Like I said, instructions for doing 50% overclocks on the Model I are a dime a dozen. Maybe there's a technical reason for the disparity, or maybe the Model III was just so much harder to get in and out of it generally discouraged casual experimentation.

(Did find a really simple CP/M modification for the Model III in an 80 Microcomputing that I might have to try... if I ever get disk drives for the thing, that's the first hurdle right there...)
 

danielbooneamerica

Experienced Member
Joined
Mar 3, 2020
Messages
336
Location
Wayne National Forest
My bad, I meant 2nd post.

My shop back in day did model 4 upgrades.

Did not try speed up of model III but we did dozens of model 4s at each school we upgraded. Kentucky schools had hundreds of TRS80S.

I suppose they all went in trash.

We did speed ups, memory upgrades, hi res installs and many other upgrades.

But I came in towards end of trs80 line and only supported model 4.

Some motherboards just would not speed up. Some components must have been marginal and it would not been easy to find problem.

I would think model 3 could be increased just same. If hardware inserts a wait state it will stillnwait one clock cycle or more with upgrade.

Back in day there was a big marketing stink about wait state's. Customers were asking how many wait states their machine had. And yes it does effect performance.

I remember a tech at RS asking why does it matter? It's only ns.
 

Eudimorphodon

Veteran Member
Joined
May 9, 2011
Messages
4,877
Location
Upper Triassic
Started poking around and so far 80 Micro looks like a dry well. I'll look around and see what other archives of period TRS-80-related content I can find.

Although I guess since I have it in hand I should probably say what this is really for, since it will be clear that there's going to be a lot of work before I can even realistically worry about the speed bump. Was poking around eBay the other day and found a guy selling Model III motherboards with a starting bid of $15. Auction was ending in an hour with no bids so I chucked in the minimum bid, figuring I'd be out-bid and if I wasn't there'd probably be roughly $15+shipping's worth of parts. (24 4116s alone about does it.)

I ended up "winning", so now I officially have in hand a bare Model III motherboard and nothing else. I think I should be able to power it from an older ATX power supply I use for bench work, it's one old enough to have a -5v line. I also have a couple Model IV keyboards some guy gave me for free at the ham radio swap meet years ago. (I need to verify the connector pinout is acceptable, but the matrix should be fine, it just has a few extra keys on the last row.) So my demented thinking is if I can throw together a dingus to convert the video output to composite I'll have the perfect makings of a zombie hack-puter. No disk controller, but the plan I'm piecing together is I might be able to build a TRS-IO and use its FreHD emulation plus a burned Model III auto-boot ROM to boot it straight into a pseudo-disk environment. (Or maybe a disk controller will fall out of the sky, one step at a time.)

So I guess it's time to stop searching for accelerator plans for now and figure out the video adapter... wait, I guess the power wiring harness is first...
 

Torch

Experienced Member
Joined
Sep 7, 2018
Messages
76
Location
Indiana
I've been interested in something like this for quite some time but haven't found anything for the III.
Ian has the Sprinter III kit on eBay for a reasonable price, but after reading the manual for the original, I don't think it can physically fit between the motherboard and hi-res board.
 

Eudimorphodon

Veteran Member
Joined
May 9, 2011
Messages
4,877
Location
Upper Triassic
I managed to track down the original manual for the Sprinter III (and a couple other vintage accelerators) and, knock on wood, even though it didn’t have full schematics there was enough theory-of-operation description in the manual I think I might take a shot at making something myself. For another project I was playing with software-selectable clock dividers implemented in a single GAL, with the theory info from the manual I might be able to adapt that.
 

GK2001

Experienced Member
Joined
Jul 17, 2017
Messages
179
The Holmes Sprinter III manual:

http://prof-80.fr/images/TRS_80M3/Overclock/Sprinter_III/Sprinter III (1982 Holmes Engineering).pdf

Funny, the description closely describes the clocking scheme I've been planning for a TRS-80 Model III clone I'm working on (auto detecting I/O operations and reverting to the standard clock rate, etc). Switching clock speeds just requires a little logic to ensure phase-coherent transitions from one frequency to the next, so that the Z80 doesn't get hung up on any runt pulses or part of a period cut too short. It would still be nice to see someone else's schematic for the same function. Does anyone have the Sprinter III schematic or know for sure if the manufacturer ever released it?
 

Eudimorphodon

Veteran Member
Joined
May 9, 2011
Messages
4,877
Location
Upper Triassic
Unfortunately it seems like the schematics for most of these add-ons seem to be secret. Oh well. I sat down and read through the manual for the Sprinter in full, and I think I have educated guesses for how they did the clock speed switching cleanly. The part I’m a little more concerned about coming up with out of whole cloth is the wait state generator. Reference designs for those seem to usually involve a shift register, it’s just all a tad confusing.

There is a passage in the manual discussing enabling and disabling various options, and it basically comes down very negative on the chances of running a Model III faster than about three MHz without adding wait states. That kind of makes me want to see if it *will* handle 3mhz without it, because it looks like subbing a /3 for the standard /5 clock might be trivial with a single GAL.

(Even a mere 3mhz would be a significant improvement for Orchestra 90.)
 

GK2001

Experienced Member
Joined
Jul 17, 2017
Messages
179
How about extending the Sprinter concept (namely a board that replaces the CPU by plugging into the Z80 socket on the motherboard) to additionally include modern and much faster ROM and RAM (along with the necessary address decoding logic)? With minimal and easily reversible fudging to the original hardware, the original (slow) ROM / RAM on the motherboard could be easily disabled by maybe lifting some pins at the address decoder to tie all !CS lines permanently high (or any other valid address decoder tweak that disables the original memory).

Edit: You'd still need to upgrade the video RAM on the motherboard though.
 
Last edited:

Eudimorphodon

Veteran Member
Joined
May 9, 2011
Messages
4,877
Location
Upper Triassic
How about extending the Sprinter concept (namely a board that replaces the CPU by plugging into the Z80 socket on the motherboard) to additionally include modern and much faster ROM and RAM (along with the necessary address decoding logic)? With minimal and easily reversible fudging to the original hardware, the original (slow) ROM / RAM on the motherboard could be easily disabled by maybe lifting some pins at the address decoder to tie all !CS lines permanently high (or any other valid address decoder tweak that disables the original memory).

Edit: You'd still need to upgrade the video RAM on the motherboard though.

So... I'm picturing a daughter board with a 128k (or 512k, why not, could add a couple 74LS670's and make it all swappable in 16k pages...) SRAM, a 128k flash chip (costs about a buck, and you can program it with several different versions of the Model III ROM, including HD autoboot, etc.), a 74HCT245 on the data bus, and maybe just a single 22v10 GAL on it with a jumper wire that goes to the 10mhz source clock. (worst case maybe two GALs?) Set up a programmable clock divider in the GAL to switch between the original /5 divider and whatever your fast speed is (with the right Z80 maybe the full 10mhz would be a remote possibility). Then you'll need to watch the following with the GAL:

A11-A15. (Top five bits, that will give us address 2K resolution)(*)
RD/WR, MREQ, IORQ

And gate them out the other side. Set up an address decode solution that suppresses MREQ/RD/WR to the motherboard and isolates the piggyback board's bus for the entirety of the memory map except for the 2K window between 14K and 16K. (Instead all access to these areas is redirected to the memory on the piggyback.) For memory accesses to the 2K covering the keyboard and screen memory, and all I/O operations (IORQ low), the GAL automatically switches (using a clocked gate to avoid runt pulses) to the stock operating speed.

(* This asterisk is here because doesn't the Model III shadow the status register for the printer port right below the 14K mark? Maybe we need to decode that.)

Anyway, maybe this would work without any wait state magic, as long as you're okay with keyboard, screen, and I/O access going at the stock speed?

It's more complicated than I was envisioning, but it kind of sounds like fun.
 
Last edited:

jltursan

Veteran Member
Joined
Sep 24, 2010
Messages
762
Location
Madrid, Spain
Given all RAM/ROM is going to be replaced, if all I/O is going to be run without speed changes I can't see any problem, the faster the better then!
 
Top