Cloning a Tandy TRS-80 Model 1

[GK]

Now that my Commodore Pet 2001 clone is done and dusted, it's on to the next computer clone project, so here is a new thread......

Like the PET clone, this is going to be a functional recreation entirely in current production parts, eschewing any TTL. Due to public holidays, Digikey aren't going to ship my Z84C00 until Wednesday the 3rd and I won't be seeing it this week, but that will give me time to complete the PCB layout for the video generation circuitry.

So far I've completed the breadboard verification of the video generation circuitry - the schematic is attached. The only things not currently shown/included in this schematic are the '244 !RW/!WR buffer pair which will isolate the video circuitry data bus from the MPU system data bus, a jumper to select either a 50 Hz or 60 Hz field frequency, and the switchable Electric Pencil lower case modification. All chips are 74HC(T) CMOS. Any device with an input driven from a data output pin of the AT28C256 (the character ROM) and the AS7C256 (the video RAM) is a HCT device, all else are 74HC.

In the original TRS-80 the designers skimped on memory, providing only seven 1-bit static RAM chips for the video memory. Bit-7 controlled whether a memory location defined either a graphics character or a ROM-generated "text" character. Bit-6 was the missing memory chip, leaving only six bits (bits 0 through 5) for selecting the character. So, that gave 64 ROM-defined "text" characters and 64 graphics/block characters (the latter giving chunky, pseudo bit-mapping). Only 64 text characters didn't leave enough room for lower case letters of the alphabet. Lower case letters, in addition to several other characters, were actually defined in the second 64 character page of the TRS-80's character ROM, but the original hardware just couldn't access them without bit-6.

The breadboard photo shows the compete video generator. For testing purposes only, I temporarily have a PIC16F874 plugged into the (currently) mostly vacant breadboard on the lower left. In lieu of the, currently absent, Z80 MPU system, this PIC is wired to the video circuitry address/data and control lines and programmed to load a test screen and character-set dump into the video memory.

The 64 "text" characters are sequentially dumped onto the very top character row, and the 64 "graphics" characters on the next. The remaining video memory locations are filled with #. I made a single byte of error in both ROM characters 1 and N. Also the left justification of 1 and perhaps : and ; probably doesn't look 100% right either. The ROM chip is currently kinda buried beneath wire links at the moment, so I think I'll put off re-flashing it just for the time being 

In the low-resolution video mode, every second (odd) video memory location becomes redundant and the even characters are doubled in width by a halving of the pixel serial-shifting clock frequency. The vertical resolution remains the same.


 
High resolution mode:


Low resolution mode




This schematic is shrunk to fit your screen by the forum software. Save and load it into something else to view full size.

[wilfred]

I don't know what character set the original TRS-80 used but the 2513 character chip had the "narrow" characters centered in the matrix. You have the "1" ";" ":" and so on over on the left. It is hard to see the eventual effect from the screen shown but the "1" is abutting the "0" a little too close. Also the 2513 had the blank row of dots above the character. That doesn't mean the TRS-80 did, or had to mimic that, but are you sure you have it right?

[GK]

You have the "1" ";" ":" and so on over on the left. It is hard to see the eventual effect from the screen shown but the "1" is abutting the "0" a little too close.

Yes, as I mentioned in my opening post. The revision of the ROM binary file won't have these characters left-justified.

Also the 2513 had the blank row of dots above the character. That doesn't mean the TRS-80 did, or had to mimic that, but are you sure you have it right?

Yes, I have this right. TRS-80 used 12 lines and six horizontal pixels for each character. The four bottom lines were always blanked, as was the left-most pixel of each horizontal row. This was hardwired. This left a ROM-definable grid of 5x8 pixels for each character. Most characters only used 7 of those horizontal lines - the upper ones, giving a total of five blank lines at the bottom of each character.

The mask-programmable Motorola ROM used in the TRS-80 was called a "5x7 character generator", even so the actual definable character field was 5x8. It was called such because in use you would typically always leave either the very top or the very bottom horizontal row of pixels blank so as to leave a single vertical pixel of space between character lines. In the TRS-80, the usage was different as four permanently blank lines were added to the bottom of each character. Some characters (punctuation and lower-case letters) in the TRS-80's set therefore actually made use of the 8th row (despite the technical literature actually being ambiguous in this regard, erroneously stating that there are 5 hardware-blanked lines at the bottom of each character instead of the actual 4).

[LaserSteve]

A friend has a complete TRS80 model 100 Service Manual with full schematics.  You want to go portable next?   

Steve

[Tomorokoshi]

A friend has a complete TRS80 model 100 Service Manual with full schematics.  You want to go portable next?   

Steve

Hey! I have that too! Along with the DVI!

[PA0PBZ]

Tandy sold a lower case upgrade kit (26-1104) for the model 1 which had the missing ram and an updated character generator. You probably want to implement that in your design.

[firehopper]

A friend has a complete TRS80 model 100 Service Manual with full schematics.  You want to go portable next?   

Steve

that I'd like to see, I actually have a model 100 with only 24 k ram. should try and find a upgrade I guess.

[EEVblog]

Tandy sold a lower case upgrade kit (26-1104) for the model 1 which had the missing ram and an updated character generator. You probably want to implement that in your design.

Yeah, I believe that was pretty standard in the end, everyone had it.

Nice work! 

[MrAl]

Hi,

This is quite interesting.  I have a TRS80 the portable model.  I got it back in the 1980's and ended up making mods to the operating system because the file dates would not go above 1987 :-)  The unit still works as tested about 2 years ago, but i dont use it anymore.

I loved that thing though and used it for several years before moving to the more standard PC.
Brings back many memories.

[nick_d]

I love it.

I bought some Atmel 5V CPLDs for this kind of purpose. Might be easier?

cheers, Nick

[GK]

I love it.

I bought some Atmel 5V CPLDs for this kind of purpose. Might be easier?

cheers, Nick

Kizmit99 mentioned that he/she might start a thread on his/her FPGA TRS-80 Model 1 clone here: https://www.eevblog.com/forum/projects/otp-eprom-programmer/msg2083648/#msg2083648


As per my PET clone, I am doing a traditional traditional hardware construction again with this one. Like the PET, the TRS-80 Model 1 keyboard unit didn't use any propriety/custom CRT controller chips and a plug-in equivalent for its original microprocessor is still being produced. I think this would not apply to and rule out cloning a model 100 similarly. I am seriously contemplating an Apple II next though.

[GK]

Tandy sold a lower case upgrade kit (26-1104) for the model 1 which had the missing ram and an updated character generator. You probably want to implement that in your design.

Yes, I mentioned the Electric Pencil modification in my opening post, which was similar. My clone will include the lower case function - it is just a matter of providing a switch to toggle the MS address line of the character ROM between D6 of the video SRAM and the faux bit-6.

One thing about the Electric Pencil lower case mod is that it didn't come with a ROM upgrade, as the ROM originally had a lower case set programmed into it, even so it wasn't accessible with the standard hardware. The Tandy lower case upgrade kit, as you mentioned, actually updated the ROM as well. I am not sure at the moment in what way the upgraded lower case set differed from the original, but I intend to find out eventually as I am adding a manual ROM page select to my design so that you can choose to run either the original lower case set or the upgraded one.

Anyway, I have already programmed my ROM with a lower case set which I found on the net. It is an additional 32 characters. I just took a break from my PCB layout and changed a few jumper wires on my breadboard. Here is the lower case set dumped to the screen in addition to the upper case set.

Yes, I made a data-entry boo-boo on another ROM character - "g".
 

[kizmit99]

I've noticed quite a few differences between the font your using and the one I have...  I'm pretty sure mine is an accurate recreation of the font rom used in the "official" lowercase mod (with one line descenders).  Although I have to say that I've found conflicting info on the net for a few of the character positions (0x60 and 0x7E specifically).



I've seen 0x60 with the pound-symbol, and 0x7e with a yen-symbol on this site: https://www.kreativekorp.com/software/fonts/trs80.shtml
but those symbols seem far too modern for the TRS-80, so I went with the original ROM glyphs (open-tic and tilda) for those two positions.

I've attached a hex file of the ROM I'm using in case you're interested...
(Note -- the 'graphics' characters are not in this rom-image, I'm generating them dynamically as in the original design)

[German_EE]

My first computer was a Tandy TRS-80 Model 1. I thought I could save money by buying a 4K machine then getting a friend to upgrade it to 16K and install the lower-case mod. He did this but a couple of months later the machine died and he couldn't fix it. Tandy computers used to come with a warranty sticker on the bottom and if that was broken the repair cost was a minimum of four hours labor. Tandy fixed my machine and I got another three years of life from it before the Apple //e came along.

Other happy memories include building my own high-speed cassette interface that loaded and saved about eight times faster and playing my first adventure game through the night. Good times!

[GK]

Quote from: kizmit99 on Today at 03:21:00 am

    I've noticed quite a few differences between the font your using and the one I have...  I'm pretty sure mine is an accurate recreation of the font rom used in the "official" lowercase mod (with one line descenders).  Although I have to say that I've found conflicting info on the net for a few of the character positions (0x60 and 0x7E specifically).



    I've seen 0x60 with the pound-symbol, and 0x7e with a yen-symbol on this site: https://www.kreativekorp.com/software/fonts/trs80.shtml
    but those symbols seem far too modern for the TRS-80, so I went with the original ROM glyphs (open-tic and tilda) for those two positions.

    I've attached a hex file of the ROM I'm using in case you're interested...
    (Note -- the 'graphics' characters are not in this rom-image, I'm generating them dynamically as in the original design)

Thanks for the ROM file and yeah, the issue I am also having is with conflicting info. Last night I watched Dave's TRS-80 tear down blog video from 2014 and noticed that the ROM was originally socketed. It's amazing that no one seems to have just wired one to a 27XX footprint and read the bloody thing. When I have time I might go over to the Tandy/RS section of the VCF forum and see if anyone with an original machine might be kind enough to dump the character set to the screen and make a high-res. photo.

I am also generating the 64 graphics block characters with discrete logic rather than with the ROM, but now that you bring this up I have to smack myself on the head. Doh! WTF am I doing that? By shifting this function to the ROM (which has way more space than I actually need) I can wipe four chips off the schematic. Going back to the breadboard now........

[GK]

OK, now that I've had my light bulb moment  , here is how I am now going to define all of the characters in the ROM. Each character is mapped to an 8x16 grid and will thus take 16 bytes of memory. 64 characters will therefore take 1kbyte of memory.

The 0-11 line counter, bits L1, L2, L4 and L8 will directly address the least significant address bits of the ROM. This makes the last 4 bytes of each character redundant as there are only 12 lines per character. Also bits 6 and 7 are redundant as there are only 6 horizontal pixels per character.

So far I've manually entered the first kilobyte of the ROM which contains all of the (64) graphics characters. Now I just have to finish transferring the "text" characters to the new format. All in all, this will allow me to simplify the video generation circuitry quite significantly. But first I am going to go eat something.

[GK]

OK, I've just finished rewiring my breadboard and successfully testing the revised circuit with all of the modifications required to generate all of the characters with the ROM. I ended up simplifying the design by five logic IC's    All 8 bits from the video ram now address the character ROM and thus all 256 possible character positions are entirely defined in the character ROM.

Yay, now I can get back to the schematic entry and PCB layout.

[GK]

BTW, thanks for posting this image as I can now see the correct sequential order of the individual characters. I had to edit a couple of my previous posts, including the opening post, as I incorrectly believed that data bit D7 needed to be set high to select a ROM "text" character. The technical manual states:

"Delay bit 7* is sourced from latch Z27............... when this input is high, data in Z63 is low, which defies and alphanumeric character"

On my first reading I glossed over the *, which means inverted. Sure enough, checking the schematics, there is an inverter (Z42F) that changes the sign of D7. So in my ROM I currently have my graphics and alphanumeric characters sequenced the wrong way around.  Also the two chunks of 32 alphanumeric characters from "@" to "_" and from "SPACE" to "?" appear to be back to front. Funny enough this is the opposite way that they are put in sequence in the ASCII tables and the character set glyph table that I was using as my reference. I also did not know that the lower case page contained a duplicate of "@" through "_" in the first 32 locations, but now that I have the sign of D7 correct it is obvious why this needs to be so. Without the lower case mod., the original hardware addressed the character ROM with a fake bit 6 generated from NORing bits 5 and 7. Bit 7 is low to select a "text" character, so bit 6 simply becomes the inverse of bit 5. This means characters "@" through "_" will be access from locations 64 through 95. With the lower case mod. applied characters "@" through "_" will be accessed from locations 0 through 31.

Now have a look at not-so-helpful reference, for example. It gives the correct binary values for the characters but plots the "@" to "_" and "SPACE" to "?" chunks in the reverse order and completely omits the duplicate block. Argh:

[biff]

It's amazing that no one seems to have just wired one to a 27XX footprint and read the bloody thing. When I have time I might go over to the Tandy/RS section of the VCF forum and see if anyone with an original machine might be kind enough to dump the character set to the screen and make a high-res. photo.

I actually found out about this topic over on the VCF Forum (someone posted a link to here) and since I'm interested in adding lower case with true descenders to my Model 1s, I came over to here looking for a hex dump of a char gen with descenders.  (I saw one was posted upstream - and I'll definitely be looking at it)

Anyhow - Ian Maverick over on the VCF Forums, is actually selling a plug in char gen upgrade that consists of a small PCB which converts the char gen pinout to a 27XX footprint (I assume).  His solution is nice, except that who ever came up with the ROM data substituted a small graphics block for the underscore - i.e. cursor.

I'm hoping to wire up the adapter board so I can read it using my TL866 - and then modify the graphic block to a proper underscore - and write it to a 27XX and swap out the masked ROM.

Assuming the TL866 supports reading it - I'd be happy to read my original chargen from my Model 1.  Probably wouldn't happen till Sunday though.

Jim

[davefiddes]

If you are looking for official character generator ROM dumps the xtrs emulator has the best collection I'm aware off: https://github.com/TimothyPMann/xtrs/blob/master/trs_chars.c There is quite a storied history behind these roms.

A recent(ish) find on the VCF board was two character generator EPROM replacements which can be used as an improved lower-case mod. They offer 1-line or 3-line decenders depending on how much you are willing to mod the video generation circuit from the Tandy original. The thread with ROM dump links can be found here: http://www.vcfed.org/forum/showthread.php?58665-GENDON3-improved-character-generator-for-the-Model-I-Discussion

If I'm following your schematic it seems that you are preserving the delay-base mechanism for horizontal sync generation. A recent article in the TRS8BIT newsletter discusses the issues with the original circuit and offers some suggestions on how to fix the resulting character wobble which can develop with the original chips: Page 48 of http://www.fabsitesuk.com/tandy/trs8bit1202.pdf

Hope this helps. Looks to be a fascinating project. I always figured my TRS-80 will keep going long after any of the other computers I've used because it used such basic logic building blocks. This promises to keep things going for even longer!

[biff]

I
A recent(ish) find on the VCF board was two character generator EPROM replacements which can be used as an improved lower-case mod. They offer 1-line or 3-line decenders depending on how much you are willing to mod the video generation circuit from the Tandy original.

Yup - this is the mod being sold by Ian which I mentioned earlier.
The 1 line descender mod (which I installed) has the underscore replaced with a funky graphic block, and the 3 line version - while it has a true underscore, looked too different from what I was used to on my Model 1.

Jim

[MrAl]

Hi,

Not sure what you guys are going for and it does sound very interesting, but just to note the TRS80 was based around the Z80 CPU.  I would think you would want to use that unless you plan to emulate that too.  Z80's were cheap 20 years ago, not sure anymore though.
Planning to expand the memory too?  That would be cool.
What kind of disk drive, if any?
USB stick capability?

[GK]

If you are looking for official character generator ROM dumps the xtrs emulator has the best collection I'm aware off: https://github.com/TimothyPMann/xtrs/blob/master/trs_chars.c There is quite a storied history behind these roms.

A recent(ish) find on the VCF board was two character generator EPROM replacements which can be used as an improved lower-case mod. They offer 1-line or 3-line decenders depending on how much you are willing to mod the video generation circuit from the Tandy original. The thread with ROM dump links can be found here: http://www.vcfed.org/forum/showthread.php?58665-GENDON3-improved-character-generator-for-the-Model-I-Discussion

If I'm following your schematic it seems that you are preserving the delay-base mechanism for horizontal sync generation. A recent article in the TRS8BIT newsletter discusses the issues with the original circuit and offers some suggestions on how to fix the resulting character wobble which can develop with the original chips: Page 48 of http://www.fabsitesuk.com/tandy/trs8bit1202.pdf

Hope this helps. Looks to be a fascinating project. I always figured my TRS-80 will keep going long after any of the other computers I've used because it used such basic logic building blocks. This promises to keep things going for even longer!

Thanks a lot for the links, I'll check them out. For the sync generation delays I am using proper monostable chips (74HC4538) which are rock stable, unlike the original design which looked a bit dodgy to me as it used some funky combination of 74C04 inverters and RC networks.

[GK]

I'm hoping to wire up the adapter board so I can read it using my TL866 - and then modify the graphic block to a proper underscore - and write it to a 27XX and swap out the masked ROM.

Assuming the TL866 supports reading it - I'd be happy to read my original chargen from my Model 1.  Probably wouldn't happen till Sunday though.

Jim

That would be awesome. I don't see a problem with wiring either your adapter unit or the original MCM6670 masked ROM to a 27XX footprint and reading it with your TL866. I use the same programmer.
Here is how you'd wire it. I'd use a machined-pin IC socket in place of the 2716, and plug that directly into the TL866. It's easy to solder wires to the pins from the top.

EDIT: oops, don't connect anything to A10.
 

[davefiddes]

Yup - this is the mod being sold by Ian which I mentioned earlier.

Sorry. Our posts crossed mid-flight. 

For the sync generation delays I am using proper monostable chips (74HC4538) which are rock stable, unlike the original design which looked a bit dodgy to me as it used some funky combination of 74C04 inverters and RC networks.

The original circuit is definitely a bit dodgy. After 38 years my TRS-80 needs fixing as the 74C04 has probably "gone off". The approach in the TRS8BIT article is to use a 74HC164 to derive the required timing from the main clock signal. Your approach seems simpler to be honest.