Building the BBC Micro (The Beeb) - Computerphile

[max666]

Interesting interview with Professor Steve Furber, who was one of the designers of the BBC Micro.



I really enjoyed hearing about the reasons behind the power supply decisions and other fun stories about the development of the BBC Micro. But since today’s youth seems to have an aversion to spoilers, I'll shut up and let you watch the video yourself.

Also very interesting, "How to fudge a 6 MHz clock from a 16 MHz clock" 

[Howardlong]

That was indeed how it was back then.

The comments about getting 6MHz out of 16MHz don't surprise me. I remember the dodgy RC networks on many designs, particularly for DRAM timings which were infamous in their difficulty in attaining the right RAS/CAS/address timing setup and holds. In fact, it was a bit of a joke in those days on the Z80 that it had integrated DRAM refresh, because to use it you had to add a fair bit of glue logic and timing, and of course RC delays.

I do remember the interleaved video/CPU on graphics. Around that time either you had to implement wait states or suffer "snow", which was when the CPU prioritised its access over the graphics controller. The graphics controller standard back then was either roll your own from TTL or use the 6845 which became the de facto standard for many years, even in the IBM PC era. Imagine a graphics chip lasting that long these days?

Good stuff, it really took me back to a time when an individual could design and understand an entire computer and its software. Thanks for posting.

[SeanB]

6845 was about the only Motorola 68 series part that was designed into a part with an Intel part, and the nice kluges you had to do to get them to talk to each other reliably.

I remember some logic where you HAD to use a plain standard 74 series IC in some locations, a LS, S or anything faster would not work reliably, and it worked well with the delay of the original 74 series, and even was happy with the 74L series, with the much higher propagation delays, though there it would not always carry on working over it's full temperature range. The fun days of playing mix and match through the big box of cards till you came out with a set that would both run reliably at room temperature ( 40C), and which would still run in the ATE room ( 20C from 4 big console units in a line on the wall) while still passing the final test there of being enclosed in a wooden box till it reached 100C case temperature ( operating environment emulation) while the last pass was running on it. On the test bench you could use the DUT as a pie warmer, or keep coffee piping hot on the top of the case. Good thing TTL is so reliable, as there changing an IC was something you had to do with a fine file, a ground down hacksaw blade and patience, seeing as the solder joints were totally covered on each IC by the DIL parts mounted on the back side along with the front. You cut the legs off, pulled out the dead ceramic package ( almost all were CERDIP) and filed down the replacement plastic case slightly to get the leadframe to fit in the space, then cleaned the legs, snipped off half the pin and slid it into the hole and soldered it to the stubs. Most common IC to fail was a 5474 dual FF that did clock recovery on input digital data lines.

[AndyC_772]

Interesting video, thanks for posting.

I grew up wanting, and eventually owning, a BBC Micro. I owe my start in EE to that machine, it was superbly expandable compared to the alternatives available at the time, BBC BASIC was an excellent language, and having a 6502 assembler included was fantastic.

It was a reliable piece of kit too, provided it wasn't abused too badly. I bought my 'model B' second hand and it needed a fair bit of work to get everything working, but I had a lot of help (they were common in academic circles, thankfully) and I ended up with a really solid computer.

I was reminded of the 'Beeb' a couple of weeks ago, writing a chunk of low-level driver code for a microcontroller. Poking around with vector tables, entry points and interrupt routines reminded me of where I'd first learned those things, in 6502 code on a BBC Micro with a 'sideways RAM' expansion card. "Sideways" because that's how the memory map was drawn when the machine had up to 16 ROM or RAM banks, each 16k in size, which could only be paged into memory one at a time in the region between the top of display memory at &8000 and the start of the Acorn OS ROM at &C000.

(You quickly learn a lot about memory mapping when paging your code in has the effect of paging out the command line interpreter you're using!)

Then you also got "shadow RAM" which was another 20k, located from &3000 to &7FFF which was the same memory region that the display occupied. By poking registers you could choose whether to read/write the display RAM or the shadow RAM, and if you had a BBC model B+ or Master series machine (but not an upgraded model B) you could choose to display either bank too.

As for the logic ICs on the board, for the most part you could leave well alone and the machine work work OK - but if you had a RAM expansion card, you'd need to replace the 74LS245 on the board with a socket and a 74ALS245 instead. Only the 'A' part had the drive strength to cope with the extra loads on the address & data buses of a bunch of extra RAM chips and sideways ROM sockets.

Then we got the Archimedes, with its ARM 2 processor (8 MHz) and GUI... but you could still press F12 to get the OS's * prompt, and type BASIC to get into BBC BASIC V which included a full ARM assembler.

I still miss my A5000, but I do have a BBC Master 128 in a cupboard. Still works fine. So do all my old 5.25" discs...

[PA0PBZ]

Oh that brings back memories. I should still have one of those 'export' models somewhere with the extra shielding required for the US market. And the weird joystick known as the 'bit stik', the Z80 second processor and even the 10MB 'Winchester'. Now if I could remember where the box is...

[mikeselectricstuff]

That first PSU was known as the "fried egg" supply, for obvious reasons.
This is another interesting talk by Steve

[GK]

Another part of the Beeb where the engineering was on the edge is in the video generation circuitry, particularly the part for generating the chrominance component of the composite video output. I had an NTSC motherboard but converted it back to PAL. It wouldn't work properly until I fudged in a 10pF cap to replace my CRO probe, that magically caused the colour to lock reliably whenever it was attached to a specific spot. The discrete logic and resistor and capacitor design is both ingenious and awful at the same time! I really don't know why they didn't use a dedicated chip for this function instead. RGB-to-composite PAL/NTSC video encoder/convertor/generator chips have been around for donkey's years - e.g. MC1377.
   

[miguelvp]

Interesting interview with Professor Steve Furber, who was one of the designers of the BBC Micro.

@5:03 of the first video he talks about they left the z80 as an option to run CPM <sic>80
But whoever put the video together follows with CP/M-86 footage 

[max666]

Continuing the interview with the ARM Processor. Most of it has already been covered by the video Mike posted, but for those of you who are unable to watch anything other than HD 

You think powering a chip from it's inputs is a novel idea? I'm afraid you're decades too late.

[Rasz]

And the other side of equation:

[Mechanical Menace]

BBC BASIC was an excellent language, and having a 6502 assembler included was fantastic.

BBC BASIC was indeed excellent, definitely on a par with Commodore BASIC even if they had some different approaches. This idiot though bodged together a very crappy assembler before I found out about the built in one :/

It was a reliable piece of kit too, provided it wasn't abused too badly.

My heavily upgraded B+ is still kicking, and I was playing with that since I was about 4/5, so I've got to say the BBCs abuse threshold was amazingly high anyway.

[jeroen74]

The BBC micro, the computer that started it all for me At one point we had three BBCs at home, networked, with a harddisk; and even one with a 68008 based extension board from Cumana that ran OS-9.

I still have a couple of books on the machine.

[AndyC_772]

This idiot though bodged together a very crappy assembler before I found out about the built in one :/

Lol  Muppet!

I wrote a disassembler... never found one of those buried in the OS, so I put it in an EPROM along with various other utilities and system tools.

More than 20 years later, there's still no better memory editor than MZAP. It was almost a standard; several commercial firms produced a variant, and I like to think my implementation was the fastest around