EEVblog #645 - TRS-80 Model I Retro Computer Teardown

[moemoe]

Hey, the oscilloscope on the upper right is upside down, turn it around before all the calibration data falls out

[ludzinc]

The comment about the 1/2 watt DRAM reminded me of when I repaired a Space Invaders (IPM clone actually).

Board on bench kept driving my 5V 3A PSU into current limit.  Took me a while until I realised that there wasn't any fault with the board - the bench supply just wasn't beefy enough.

IIRC the 5V rail on the space invaders was 30W.... 

[VK3DRB]

Dick Smith sold a TRS-80 ripoff called the System-80 (Video Genie sold elsewhere). I bought one in 1979 - the biggest load of crap ever. I was ripped off a whole  month's salary when I bought it but I wish that I had saved a little longer and bought the far superior TRS-80 from a reputable dealer.

- The tape drive was very unreliable. Errors, errors and more errors reading.
- Keyboard bounce was far worse than the TRS-80. The machine was unusable.
- Dreadful interference to TV. You could not watch the ABC when the System-80 was turned on.

It was a toy - the only thing of use was I could write trigonometric table on it. Although SARGON was one impressive program written by two husband and wife geniuses - Dan and Kathy Spracklen. I was a member of the TRS-80 Users Group in Melbourne. It was a subgroup formed from the world famous MICOM (Melbourne Microcomputer Club) - a terrific bunch of blokes who like me were pioneers with home computers.

[made2hack]

Great video Dave, Good on You!!! 

[RupertGo]

In those glorious pre-FCC certification days, all that digital RF leakage could be a feature, rather than a bug. If you put a MW/AM radio next to the computer (not necessarily that close!) and wrote some tight timing loops with memory access in assembler, you could play tunes. Which was something, as there was no other way to get sound out of some of those early micros. (Sometimes you could flip bits on the cassette out port... you could still pick that up on the AM radio, though.)

Advanced users could also use the AM pickup technique to debug software, as with a bit of experience you could hear the distinct tones of subroutines or even put in little blips as checkpoints. There was even an early UK computer - 50s or 60s, I think an Elliot Automation model - that had an audio amp and a probe built into the chassis so you could hear what was going on at various bus points. With low clock speeds, there was a lot you could pick up.

BTW - I've always heard that it was a TRS 80 Model 100, the NEC 4-line LCD AA-powered portable, that had the last code Bill Gates ever wrote in it. It was certainly his pet project, and the 8080 Microsoft BASIC for the Altair was the first thing he and Paul Allen wrote when they started Microsoft. A legendary piece of coding.

But a lovely teardown. I really like those horrible brown ceramic decoupling caps, which remind me of the Hong Kong 5 transistor superhet radios I cut my teeth on as a schoolkid in the 70s. Tiny circuit diagrams pasted in the back of the case, earphone sockets and crappy PP3 battery connectors that failed in microseconds, and that 'fill a shotgun with components and fire it at the PCB' manufacturing technique. Happy days.

[CC58]

You may want to pull off one of the keys from the keyboard.  These switches were springs, and the hardware didn't have any debounce.  There also wasn't any in the ROMs.


Yes, I forgot to mention that, it was a major issues.
I didn't get an key bounce at all on this, but the owner thinks it's had the debounce mod done. Doesn't appear to be any hardware mod, so maybe in the ROM?

The original keyboard had key bounce problems and you could access the contacts to burnish them to fix it.  Later the ALPS keyboard was introduced that had sealed switches and zero key bounce issues.

I worked as an electronic technician repairing TRS80's in Radio Shack's national repair depot in Fort Worth,TX.  Jerry Ballard was engineer I worked for making test fixtures.  Made a few trips to the ivory towers and saw the new things before they were released such as the Model III and color computer. 

[nixfu]

As a ham I had considered saving up for one of those MDO3000s or MDO4000s because a spectrum analyiser would have been nice to have on board a nice modern color scope.

But, holy cow, the performance on that sucks....I think I will stick with an old analog phosphor screen spectrum analyzer that can do real-time updates on the screen.

Sheesh.  You think they would have at least made something as good as what was available 20 years ago.

[EEVblog]

As a ham I had considered saving up for one of those MDO3000s or MDO4000s because a spectrum analyiser would have been nice to have on board a nice modern color scope.
But, holy cow, the performance on that sucks....I think I will stick with an old analog phosphor screen spectrum analyzer that can do real-time updates on the screen.

To be fair, it can actually perform faster than a traditional spectrum analyser sweep on the same settings, and it captures the entire spectrum in one hit, something a traditional analyser can't do. It's the almost locking up on the front panel, missed key presses etc that make it incredibly annoying.

[jcochran]

Brings back old memories. The TRS-80 was the first computer I spent a lot of time programming. Also learned assembly language programming on that little beast.
One project I did was to write my own floating point package in Z-80 assembly. I then tested it on a TRS-80 in parallel with the build in floating point they had for Level II Basic to make sure my routines gave the same answer as theirs. In the process, I found a bug in their floating point. If you still have access to that TRS-80, try the following little program.

10 A=1E19
20 B=1E38
30 PRINT A,B
40 C=A*A
50 PRINT C

What you'll find out is that you're gonna get an overflow error on line 40. The print on line 30 is to prove to you that 1E38 is within the representable range of numbers (the actual limit is about 1.7E38). So the multiplication of 1E19 time 1E19 ought to be perfectly fine, but it isn't.

[Dave Turner]

Yup, sure brings back memories. The best thing was the manuals contained all the circuit diagrams.

[Rory]

In the early 80's, a friend wanted to use a TRS-80 in his hamshack, it was (obviously) unusable for all the hash noise so he brought it to me with a pile of ferrites and an aluminum enclosure and we boxed it up with "RF tight" DIN chassis connectors, bypasses, ferrites on the ribbons, etc. And it still spewed RF garbage from the keyboard, power supply and the video monitor cables.  Even toroids on the cables didn't work.  A small e-field sniffer probe found the sources, we even tried mounting the keyboard inside the cabinet with holes punched for the key posts, and it still couldn't be tamed.  It was a total exercise in futility, but a fun project anyway.

[PA0PBZ]

Ah, memories...  Tandy started to roll out a few computer centers in the late 70's/early 80's here in Holland and decided it would be nice to have a technical service place in the center in Amsterdam. So they put out an ad looking for someone to start this, and to make a long story short I was hired  Not that I knew a lot about computers back then but who did? I remember that I saw my first floppy disc over there... I fixed a lot, learned a lot and finally moved on after maybe 5 or 6 years.
The Model I could only hold 16K, the other 32K was inside the expansion box. The lower case mod was just another cg piggy-backed on the existing one with the chip select connected somewhere. When one of the RAM chips went bad it got really hot, it was easy to spot with your finger but having the text on the chip burned on your finger tip was not so funny. The biggest problem was when one of the buffer chips went bad so I made a board that connected to the expansion interface and to a second Model I. It then single stepped the broken one till the signals on the bus where different from a working model and halted the system. That way you could easily measure the bus and find the defective chip.
Good times...

[nsayer]

I remember these. We used to call them TRaSh-80s. Dave mentioned that moniker as well.

We had a lab at our high school with 20 of these. They had shared 5 MB (yes, MB) Corvus hard disk system. Each machine had a 40+ pin ribbon cable running back to a set of multiplexers daisy chained off the drive itself (a box the size of a toaster oven).

The problem, though, was that the machine wouldn't boot over that expansion bus. You needed to boot from a special floppy with a modified TRS-DOS (or in our case NEWDOS). But to save costs, these machines didn't have floppies.

Instead, there was a cassette based boot loader that could be used to bootstrap. In addition to the ribbon cable mess each had a DIN5 cable for the cassette port running back to a multiplexer hooked up to the ONE floppy machine in the room. It ran a shell script that when you pressed enter would transmit the cassette program to anyone listening. If you had to reset your "workstation" (ha ha), you'd press enter twice (for CASS and Memory Size) and then type SYSTEM and LOGON. You'd then shout "Boot me!" and someone near the master would press ENTER. 30 seconds or so later, you'd get another "*?" from the SYSTEM command and you'd run the load with "/" and log in.

Memories.

[Dinsdale]

My first job in electronics.
FIrst I did burn-in and boxing, load and start tests in the evening, box the working ones the next morning.Then I learned trouble shooting and worked on fixing boards off of the production line. They bought a GenRad  circuit tester and I ran that (boring after a while).  Finally I moved to Tandy Systems Design.  Learned a whole bunch about signal integrity when working on the expansion interface: pretty tough to run RAS and CAS across a foot of ribbon cable!
Thanks, Dave.  Brings back memories.  Radio Shack/Tandy was great for on-the-job training.

[GeorgeOfTheJungle]

I had the TRS80 and an Apple ][. Compared to the ][, despite its 0.7 MHz/1.7x clock speed advantage, it was such a pitiful snail, I swear it. Running a program made glitches appear on the screen. The Apple video image was in comparison amazingly sharp and stable and absolutely glitch-free. The TRS80 had 64 characters/line and upper and lower case which was much much better than the silly and only upper case 40 chars/line of the Apple ][ (had to buy a Videx 80 colums card to remedy that). The 2 kilograms "power brick" was rather ugly and ran hot like crazy. It was ludicrous compared to the switching power supply of the Apple ][ that also ran much cooler and made a lovely whinning noise when overloaded...  because it had 8 expansion slots built in. In the trash80 instead you had to buy that cheapo-looks external "expansion interface", and have that flat cable hanging forever between your "keyboard" and the silly thing. Even to get past 16k you had to buy that, while in the ][ you only had to swap some jumper wires and plug in more 4116s. Preferably the beautiful ceramic µpD416Ds made by NEC :-). In short, nearly everything in the trash 80 was trashy, and nearly everything in the Apple II was amazing. I'm not an Apple fanboy but the only thing that was better, apart from the price (a big difference!), was its BASIC, much better than the integer (!) basic that came with the IIs. And the DOS too, much much better than Apple's "toy" DOS.

[jcochran]

I had the TRS80 and an Apple ][. Compared to the ][, despite its 0.7 MHz/1.7x clock speed advantage, it was such a pitiful snail, I swear it.

Yep. The TRS-80 vs Apple ][ was a very nice example of "clock speed doesn't mean everything".
The Z80 instructions were generally 4 clock cycles per byte of memory touched give or take a cycle for instructions taking a large number of bytes.
The 6502 was 1 clock per byte of memory touched plus 1 clock.
So minimum cycle time for a Z80 opcode was 4 clock cycles. Minimum time for a 6502 opcode was 2 clocks. And generally a 6502 at a given frequency was about equivalent to a Z80 running at twice the frequency. So a baseline TRS-80 was about 85% as fast as a baseline Apple.
But if you programmed at the assembly language level, the Z80 was a LOT nicer to use than a 6502. But on the other hand, the Apple ][ came with a built in assembly language monitor that made debugging easier (although that 256 byte limit on the stack for a 6502 was a real PITA).

[synapsis]

The upside down oscilloscope sets off my OCD something fierce.

I like how the casing for the display has the cutouts in the front (seen from inside) for VHF/UHF knobs.

The Model 3 is the first computer I ever touched. I still remember the keyboard feel.

[GeorgeOfTheJungle]

But if you programmed at the assembly language level, the Z80 was a LOT nicer to use than a 6502. But on the other hand, the Apple ][ came with a built in assembly language monitor that made debugging easier (although that 256 byte limit on the stack for a 6502 was a real PITA).

Oh yes, I remember that, mine in fact did not boot to BASIC but straight into the monitor *. To get the BASIC prompt you had to type ctrl-B. To return to the monitor CALL -151. But apart from the cycles per instruction differences that you mention, I think the huge difference I saw in running speed was mainly due to the floating point BASIC that came built into the TRS80. The Apple was so much snappier probably due to its (Steve Woniak's) integer-only BASIC. It didn't last long though, later II Plus models ditched the integer BASIC entirely in favor of microsoft's FP BASIC.

[miguelvp]

But if you programmed at the assembly language level, the Z80 was a LOT nicer to use than a 6502. But on the other hand, the Apple ][ came with a built in assembly language monitor that made debugging easier (although that 256 byte limit on the stack for a 6502 was a real PITA).

And don't forget the registers, the Z80 wins hands down compared to the 2 lonely registers + accumulator on the 6502.

[Dave Turner]

It was even more complicated than clock speed and clock cycles. If I recall correctly the 6502 was a 2 phase clock whereas the Z80 was single phase and static thus more easily implemented. In general the Z80's plethora of registers (in comparison) made programming at the assembly language level potentially more efficient, particularly because affordable external memory was relatively slow. Nevertheless the 6502's instruction set was more suited to some types of memory addressing; it all depended on what one was trying to achieve.

[Tek_TDS220]

Yes, it's crusty, but it still runs because it was well-built.  I just repaired a $1000 Samsung TV (4 years old) because they put Chinese caps 5 mm from a hot switching transistor on a board with a lot of room.  I'm suspicious that it was purposely built to fail.

[BobC]

I was an impoverished sailor in the US Navy when the TRS-80 came out.  My local RS would let me use their floor unit for limitless hours, so long as I explained it to customers.  I really should have been on commission.

But when I finally had the money, I got an Apple ][+ with the Language Card so I could use UCSD Pascal.  I wound up attending UCSD in the mid-80's because of it!

[aadderoos]

You may want to pull off one of the keys from the keyboard.  These switches were springs, and the hardware didn't have any debounce.  There also wasn't any in the ROMs.

Yes, I forgot to mention that, it was a major issues.
I didn't get an key bounce at all on this, but the owner thinks it's had the debounce mod done. Doesn't appear to be any hardware mod, so maybe in the ROM?

If I remember correctly, the earlier models you had to load a little programm to debounce the keys. Later on there was a rom revision that had that build in. You could tell wich version you had by looking at the free memory. The original reported 15572 bytes free, the one with the debounce routine reported 15570 bytes free. (I hope I remember these numbers correctly)

I think the TRS 80 was the most popular one in the Netherlands of the three major ones those days (TRS 80, Commodore PET and Apple 2). Everyone thinks Apple was number one but that was just good marketing from Apple. And the fact Apple is the only one arouind these day helps too.

I myself have build a clone of the TRS 80, a LNW 80. You had to solder the circuitboard yourself, it was also available ready build but at a much greater cost. I think the total cost was around 2500 dutch guilders, which is 1200 Euro's. It had some extra grafics, 480x192 monochrome and 384x192 color but these were unsupported by the software. It had a nice (heavy) metal case, that also had room for the expansion board, also made by LNW (compatible with the Radio Shack one, with improvements.)

This LNW was the thirth computer I owned, before I had a sinclair MK14 (very rare today) and a Nascom 1 (maybe even more rare now), both were kits.
Here is some info on the LNW: http://fjkraan.home.xs4all.nl/comp/lnw/index.html

[ggchab]

I bought my TRS-80 in 1980, in a Tandy shop. It was a wonderful machine and I spent thousands of marvelous hours working with it. Some years later, I bought the expansion unit and a first floppy disk drive. It was only 90K and a big part of the space was used by TRS-DOS that had to be on the floppy disk. But at that time, the remaining space was still big enough to store programs and other files !! And floppy disks were a lot more reliable and faster than cassettes. Cassettes were sometimes a real nightmare and it was easy to loose hours of work 

When I watched Dave's great video, many souvenirs came back to my mind

[tocsa120ls]

Yes, it's crusty, but it still runs because it was well-built.   they put Chinese caps 5 mm from a hot switching transistor

I think this is analogous to the "Why is the milk in the back of the store?" debate. The caps have to be close to the FET switch. Imagine the RF you would get when you´d have to route that high power trace around the PCB to the cool side. That´s what 105C caps are for.
And the milk is in the back because old fridges had huge compressors that were located outside of the store. Plus that´s close to the delivery entrance, so the cooling chain is not broken.

And to all of the ppl complaining about the upside down scope, it´s in Australia, so wouldn´t be that the only one right-side up?