Saturday, 29 November 2014

BBC Micro Issue 3

I recently won a BBC Micro issue 3 board and case on ebay. One of those situations where the seller decides they could make more money if they split it up and sell the parts separately. Not my favourite tactic. I didn't see the power supply on there, so it may have suffered from exploding suppression caps. A common issue on BBC micros, and an easy fix.
I was going to insert a link to my previous article about BBC power supply repairs. I've done loads of them, but it seems I've never written a blog post about it. I'll make a note to do that with the next one.
So this is what I got, the case and the main board. The aren't many of the very first BBC computers, issues 1 and 2. The issue 3 was the first to appear in quantity.
After that, there were lots of issue 4's. Issues 5 and 6 weren't released. The issue 7 was the final and most prolific release, and most of the BBCs I see tend to be issue 7's, with the occasional issue 4. I don't have a complete issue 3 machine, so now is the time to make one.
This is one of the machines produced by ICL, so has an ICL serial number, rather than the standard Acorn one.
The main board appears to be in reasonable condition.
It would have been a model B, as it started with all the connectors fitted and 32K of RAM, as 16x Fujitsu MB8118-10 (4116 clone) DRAMs.
Well, I say 16, 15 of them were like that. One has been replaced by a Hitachi HM4816A. Based on the date codes, it would have been produced in early 1982.
Continuing the industrial archaeology, it would appear that the system 6522 VIA was replaced some time after 1988 with a Rockwell R6522. Below that, it look like in 1984, it had an Intel 8271 floppy disk controller upgrade fitted. The 8271 used the older FM (single density) encoding, so had a maximum capacity of 400K for a double-sided, 80-track disk, accessed as two separate sides. Soon after the release of the BBC, add on kits became available which used the cheaper and faster WDC1770 floppy disk controller, which also had double the capacity using  MFM (double density) encoding (thanks to Alex Taylor for the correction).  The 8271 is one has the one it was originally designed to take, and even the much later model B+ has been designed to take either the 8271 or the 1770. (yes, model A, B and B+ may sound familiar to Raspberry Pi fans, but the BBC was first by about 30 years).
The BBC has 5 ROM sockets. The one on the left is the 16K OS ROM. The other 4 are paged 16K ROMs, which one is selected is based on a register. The BBC can access any one of up to 16 ROMs, although only 4 sockets are provided. With a simple wire modification, it is possible to use 4 64K EPROMs, each with 4 images on to get all 16 ROMs. Again, I thought I'd written a blog post about this. I'll add that to the list as well. The board underneath two of the sockets looks suspiciously clean, the ROMs in there were probably sold off separately as well.
The one on the right is BASIC. There should have also been a DFS ROM to go with the 8271 floppy disk controller, so I've burned one.
Since this is going to be an assembly of parts, I've decided to keep it as an issue 3 machine. The 6502 CPU and the user port 6522 VIA are the usual Synertek branded chips, and the original system 6522 would have been the same. So I replaced the Rockwell R6522 VIA with a Synertek SY6522 to match.
The Ferranti video ULA has also been replaced with a VTI VC2069, as would have been used on issue 7 boards. Again, I have decided to roll this back to the it's original state.
The Ferranti ULA is less efficient than the later one, and needed a heatsink.
With fresh hearsink compound applied, that's the board sorted, now it needs a power supply.
Later BBC micros had switch mode supplies, but the early BBC micros had these black linear power supplies. This is one I've had for years, I was given this as it's owner replaced it the a switch mode one. The linear ones do not have a disk drive power connector, and the owner wanted to power an external drive. Only 20 years later, I finally have a suitable case for it!
The power connector for the BBC main board is unusual. Rather than one plug and thick tracks going around the board, they have three pairs of 0V and 5V  connectors which attach at various points across the board. The red and black are marked 0V and VCC.
The brown is -5V, but the label seems to be missing the - sign. That's one to watch out for.
That's the power installed, time for a keyboard. Some of the early ones had a DIP switch installed bottom right. I have one of those somewhere, but the PCB was poorly etched and it's a bit intermittent. This is a spare one, I think from an issue 4.
There are a few differences at the back of the case, compared to later versions. The wording is different. Here Allen Boothroyd is credited. On later machines, that is moved to the bottom of the case. The ventilation slot is a lot larger, it was reduced in height later on to stop small fingers getting inside. This looked like it might have been taped up at some point,but that cleaned up ok.
Note also the reset switch is labelled. All machines have space to install a reset switch, and a hole in the back, but I don't think there were ever fitted. Only the early ones have it labelled on the back, the rest have it covered up. Like the Econet port, this would have have a plastic cover, but it has been punched through at some point. Most of them have. I've put a piece of black card to cover the holes.
The BBC was well designed to be expanded and serviced. There was always one annoying part which was the BNC video jack. That had to be desoldered to removed the board, and often didn't get reattached if the owner was using an RGB monitor.
I was surprised to see this version had a removable connection, using the same spades as the power supply. Shame they discontinued that in later revisions.
The board appears to be laid out to take a PCB mounted BNC socket, but I've never seen one fitted.
Underneath is the usual collection of IDC expansion connectors. The power out is marked, but there is no socket on the linear power supplies. There is also an ICL serial number sticker matching the board.
The lid was a bit dirty, but cleaned up alright. The other thing that usually gets pushed out is the slot on the side of the keyboard. This was designed for plug in modules to expand the speech chips, and some third party add-ons use the hole to add a ZIF socket connected to one of the ROM sockets. Again I covered it was a bit of card. I also replaced the perspex cover strip with a slighly clearer 'BBC Microcomputer' legend. Later BBC's said 'British Broadcasting Corporation Microcomputer'.
It looks a lot yellower in the photos. It doesn't look too bad. So there it is, an assembly of parts from various sources, but something like an issue 3 model B.
Oh yes, and it works. BBC's usually do.

Thursday, 27 November 2014

Commodore 8032 Main Board Repair #1

Following on from the repair work on my Commodore Pet 4032 earlier in the year, I've got a few more Pet board repairs coming up. This first one is from a blog reader in the States.
It is an early 8032 board, unlike the later universal boards (like the one I started with from an 8032-SK), this one is 80 column only.
It would have been manufactured in early 1980 (based on the chip dates), and contained the first version of one of the BASIC 4 ROMs (901465-19). This was changed shortly afterwards due to various bugs (I can't find any information on what the issues actually were).
Later boards had this chip replaced with 901465-23. I have a similar 8032 board which has a ceramic version of this, from later in 1980.
The fault is the usual nothing on the screen, no startup beep. So to start with, I remove any socketed chips and check the board for shorts on the power rail. I then power up without the main chips and check voltages and that nothing is getting hot.
Next I put back the minimal number of known working chips (6522, 6502, 6545, keyboard 6520), using the ROM / RAM replacement board to bypass the onboard RAM and ROM chips. Optimistically connecting a keyboard and the video to the 4032 monitor via an extension lead.
That's a good start. That's using replacement chips, bypassing the onboard ROM and RAM, but it does prove the power, clock, reset, video RAM, character ROM etc. are all working.
I checked the ROM chips in an EPROM programmer, and found that the 901465-22 Kernal ROM was not working, so replaced that. I also temporarily substituted a 901474-04 (80 column, normal keyboard, 50Hz editor) for the 901474-03 (80 n 60Hz), but later found it worked well enough with the 60Hz chip. With those replaced, I could turn off the ROM part of the ROM/RAM board and it ran OK.
Next was the RAM. 16x TMM416D-4 (4116 clones), and unusually, it all seemed to be working.
The final step was to the 5 main chips. I replaced my test chips with the originals one by one. The 6522 VIA and 6502 CPU were fine. It wouldn't boot with one of the 6520 PIA's in. The top (or left in that picture) 6520 is only for the GPIB port, so isn't needed to boot, so I removed it. The other 6520 PIA was OK, and finally the 6545 CRTC. That didn't work. These are a bit difficult to get hold of, so I replaced with a pin compatible 6845 CRTC.
The 6520 was replaced with one of the new WDC 65C21-N chips, again, a pin compatible replacement. With those in, it was running standalone for the first time in a while.
The ceramic chip and the chip with what looks like a price sticker on are expansion ROMs. The purple ceramic chip marked '8032' has a string table in. It appears to be 'Command-O' (c) 1980 Rob Chang. This is a BASIC extension that adds commands such as renumber etc. I couldn't find any further documentation on it. To enable it, type SYS 36864 (or SYS 9*4096 if that's easier to remember). It's renumber function was quite useful, so I took the opportunity to renumber a few of my test programs that have evolved a bit messily over time. It doesn't seem to work without the other chip, although that is a 1K PROM with no string table?
As a burn in test, I left it running a simple program
   10 print chr$(142)
   20 print chr$(205.5+RND(1)); 
   30 GOTO 20
This just generates a random maze from / and \ graphics characters (line 10 puts it into graphics mode).
It seems a good test as it repeatedly does it bit of maths, and you can see if it gets it wrong. It also fill the screen buffer so you can check that. It's also quick to type in, testing all that side of things as well. This failed after a while, jumping into the machine code monitor.
Reinstalling the ROM / RAM board in ROM mode, it ran for half an hour without a problem. Switching out of ROM mode and back to the onboard ROMs, it only ran for a few minutes again. After trying various things, I noticed the 901465-19 ROM was getting rather hot. I didn't have another of those, so I went for the later 901465-23 which fixed the ROM bugs (maybe the bug was it crashed after generating random numbers for five minutes?). With that replaced. it ran that maze test program from about 4 hours without a problem, so I think that's fixed. For further testing I run my memory test program a few times, and all passed with no problems.
Finally, I tried a number of games from tape. Many of the games do not work well or at all on 80 column screens, which is why I converted the board in my 4032 back to 40 columns.
The titles from Revial Studios do work, but just use the left hand side of the screen. Still very playable and Down is quite addictive. All in the name of testing of course.
And there it is, ready to go back to be reunited with it's original case.

Sunday, 9 November 2014

ZX Spectrum Plus Keyboard Restoration

I seem to be doing quite a lot of keyboard cleanups recently, the Pet, C128 and BBC that I've written up, and the usual collection of Commodore 64's and Spectrums I seem to do quite regularly. This is one such Spectrum, and it strikes me just how many parts there are in the thing.
So you start with you standard Spectrum Plus, complete with 30 years of dust and dirt, and in this case, missing a key. When opening these up, the membranes tails tend to be very fragile, and if they haven't already failed, opening the up without due care and attention tends to snap them. This seems to have happened to most of the 'untested' ones on ebay, usually just before or just after they went up the loft.
This is one of the better ones, these clear membranes seem to be a lot more resilient than the brittle opaque ones.
With the tails carefully removed, you get to the main board. This is the same board as used in the original rubber keyed ZX Spectrum, with the addition of the reset switch (wired across a capacitor in the reset circuit). The big lump of metal at the top is the heatsink.
The keyboard is stripped down and the keys washed and cleaned.
The case is also given a clean up.
With the keys drying, they do seem reminiscent of a certain game.
Maybe a new logo?
Maybe not. However, as a side note, when designing a new TARDIS console for the 20th Anniversary Doctor Who story in the 1983, they did the reverse of this, and stuck on upside down scabble tiles to look like buttons.....
The process of putting this back together shows the number of layers involved in the Spectrum Plus's keyboard.
So first you have the case, then the keys, then the plungers, slowly being repopulated (and the broken one replaced).
Once they keys are all in place, there is a rubber mat.
After that comes the membrane.
This is a multilayer membrane, and the layers need to be squeezed together by clamps. Sometimes a keyboard which doesn't seem to be working can be fixed by tightening these up. Although usually it means a visit to SellMyRetro to buy another RWAP replacement membrane.
After the membrane, comes a sheet of cardboard, or sometimes plastic.
And after that a sheet of metal and then it can finally be screwed in place.
With everything all put back together, the final product is a much nicer looking Spectrum Plus.
If the Spectrum is going to be used a lot, I recommend replacing the voltage regulator with a modern switch mode device. I usually have a number of refurbished computers available, contact me if you are looking for one.