I've got a BIGGER one!

In days of yore, computer manufacturers would include tons of useful information on the products you bought. Need a schematic? It's in the back of the manual. Have a problem? Comprehensive troubleshooting in the manual too etc etc. One such company was HP. Their electronic equipment was legendary (not just computers) for the build quality and information provided to the user. It's almost as if they understood that you owned a unit they had made and you had a right to understand, use and even repair it as you saw fit. What a novel concept these days (looking at you Apple).

Anyway, they also made some fairly hefty high-end UNIX workstations. And guess what arrived in the garage in the last couple of weeks? Yep. An HP 9000 workstation, model B132L. Alongside the main unit and the absolute UNIT of a monitor (more on that in a minute) came HPUX on CD with many updates - also on CD - along with books and manuals. The only thing missing is a PS2 mouse. 

Ten grands word of 'puter.

It's an old fashioned, proper HP.

First things first. The monitor for this is FECKIN' HUGE and FECKIN' HEAVY. Seriously, this thing must weigh 40kg. I can just about lift it up off the ground and get it on the desk. Put it this way, the back end of my car was rather low on the drive back from pick-up... It's got three BNCs on the back for the video signal, with what looks like component video (it's actually RGB with sync on green). Resolution seems to be 1280 x 1024 which is pretty good for 1996. Oh, and did I mention it's 20 inches? 20 FECKIN INCHES! It's so heavy, it bends spacetime in the vicinity.

Is it the biggest monitor in the world? No.
Is it the heaviest? Probably.

But does it work? Well, the monitor does, yes. The first time I turned it on it scared the cheeses out of me because it did a de-gauss - as CRTs are wont to do - which makes a bloody loud 'BOING' sound. Subsequent turnings on were less exciting. It is a fixed frequency and resolution though so unless what its connected to is spewing out 1280 x 1024 at 72hz, there's not going to be much displaying going on.

Monitor boings and works.

The actual workstation itself is also built like a tank. These things were expensive back in the day though, starting at around $10,000. That's around a truckload of cash today. And that was the starting price.

So what would you have got for all that dosh, in 1996? Well, not as much as you'd think. First up, you get the HPUX OS which is a variation of UNIX. Anyone who has used Linux will be fairly familiar with some of the operation of UNIX although they are not exactly the same (more on that later). From a hardware perspective the unit I have has a PA-RISC processor running at 132Mhz. There's a whole 64Mb of RAM too, to go along with the IBM 2Gb SCSI (HVD SCSI2) hard disk.

So lets get that top off.

The case slides off like a very heavy
metal glove.

And we're in...

Starting at the front right we have a SCSI CD-ROM drive (caddy variety) which needs some mounting brackets (3D printer might come in handy here). I actually have a SCSI CD-ROM from an Amiga setup somewhere which I might put in here as I really don't like the caddies. 

Behind the CD-ROM there is the hard disk. Not much to say about that. It's a spinning rust drive.

Then heading over to the left side there is an additional graphics card. This card is what I've actually plugged the monitor into since the built in DVI connector doesn't work with a VGA-DVI adaptor (which is a REALLY tight fit). There is some signal but it's distorted and flickery. It may be that a standard VGA monitor might work properly. If I can find my missing VGA cable I could try and prove this...

In the middle is the CPU and RAM. Note the tiny heatsink. But also note the fairly hefty pair of fans at the front too.

At this point it occurred to me that there may be a battery somewhere in here. And I was right:

Ah. There it is. 

Battery.

It actually looks OK so I decided to leave it where it is for now. I will try and get a replacement at some point in the future.

After putting it all back together it was time to see if it would boot. After several days of training and a pep-talk from a nearby weightlifter, I managed to get the monitor up onto the bench (the photo above was taken after my epic lift) and get it all plugged in and ready to go.

Did it boot? No.

Why won't you boot?

So, I took the case off and found out that SCSI drives don't work wirelessly. 

You mean I have to plug it in? Sooo 90's..

Case back on. Pep talk. Work out. Monitor lifted. Try again...

Did it boot? Yes. Sort of...

It started HP-UX but then declared that there were lots of file errors that needed to be corrected. It corrected them without me confirming or denying permission and then re-booted. And got stuck in a boot loop. 

Basically, I think that the files that had errors were actually related to the GUI. So it would start up and attempt to start the GUI and fail, reboot and try again ad infinitum..

At least it starts to boot.

Console login but no GUI :(

Fortunately I have the CDs so I can re-install the OS from scratch. I did give this an attempt but I ran out of time. The disks had around 650 packages to install and it took about eight minutes to install three. If that was an average time for three packages it would take a few hours to get this installed so I need to wait for a lull in proceedings and a spare half day or so to do it.

I may be some time...

Did I mention the monitor is FECKIN heavy?

I saw a mouse! Where? There look, there's loads of 'em!

Ah, mice. The ubiquitous human interaction device that was invented far earlier than you'd expect and which gives us something to throw across the desk when the computer does something stupid (thanks, Windows). 

Not that sort of mouse...

Mice have been mainstream since the late 80's and the Amiga was no different. With its vomit inducing blue GUI in Workbench (pre WB2.0), the Amiga needed a mouse to use any 'serious' software, as well as some games, and they came as standard with the platform. Most Amiga users will fondly remember the squarish 'tank' mouse that Commodore provided with the Amiga and I suspect a large proportion of those users will also remember the joy of replacing it with a third party unit that would be more comfortable and responsive to long sessions of Lemmings.

That's just another actual mouse.

And, would you believe it, in the epic pickup I did recently, I had four Amiga mice included. None had been tested for years and all of them needed a clean having been in storage for so long. So after cleaning up (I'll skip past that) which of them would work? Time to break out Amiga Test Kit.

Four Amiga mice.

Naksha

No x/y movement at all. Buttons register OK

Eklipse

No x movement but y movement OK. Buttons register OK.

Commodore

Working OK

Prism 

Working OK

This is a great start. Two working mice! Interestingly, the Prism mouse has quite a chunky design and the back 1/3 of the case is slightly articulated, so if you put some pressure on the back it moves up and down. Perhaps an early attempt at some proper ergonomics?

But we still have two mice that aren't working. The basic principle of the mechanical mouse is that two disks with small slits are mounted at right angles to each other. Infra-red LEDs and receivers are mounted both sides of the disks and a large ball presses against rollers attached to the disks. As the ball moves, the rollers and disks spin and break the beam between the infra-red LED and its corresponding receiver. The resulting signals are processed by some simple logic in the mouse and the x/y movement is passed back to the Amiga through the mouse/joystick port(s).

One common problem that can prevent the x/y movement from working is if the LEDs or receivers lose 'sight' of the device on the other side of the spinning disk. This is fairly common and, in general, simply bending the LED or receiver (or both) back towards the disk resolves the problem.

Wheel and LEDs

Two wheels at 90 degrees
to each other.

First up, the Eklipse. This mouse has an interesting switch on the bottom that allows for this Amiga version mouse to also do double duty as an Atari ST mouse. I'm not too sure of the differences - I shall have to Google it.  But, as expected, there is a simple two disk arrangement inside for the main x/y control. Since the y movement was OK i.e. up and down, I left that pair of LEDs alone and looked at the others. They didn't seem to be out of alignment as such, but I bent them slightly more towards the disk and did the same for the receivers on the other side. 

And that fixed it! Easy! 

Three down, one to go.

The final mouse was the Naskha mouse. This one also has an Amiga/Atari switch on the bottom. 

Using Amiga Test Kit (ATK) I could see no movement at all from the pointer, but the buttons were working fine. Opening it up, the LEDs/receivers all looked to be perfectly straight. To check that the LEDs were working, I pointed my mobile phones camera at the LEDs to see if I could see the infra-red being emitted. Sure enough, I could see the small purple flashes from the LEDs. So, I tried bending the LEDs towards each other and tried again with ATK. This time, the mouse jittered very slightly in both the x and y axis. This was a very good sign. I tried bending the LEDs and receivers even closer to each other but without fouling the disks. With only a couple of millimetres between the disk and the LED/receiver, the mouse now works perfectly!

What a great result! Four out of four Amiga mice are 100% working! Excellent! 

One each to go with the real critters I had to evict from my workspace. 

Number 3.

Number 4.

For any animal lovers out there, all four living mice were safely transported to a copse a couple of miles away and released. They were all field mice after all.

It's BIG.

As part of the pick-up I recently made, there was an Amiga A2000. This was the primary reason for the trip and was the first item offered which I gladly accepted. Everything else was a bonus - but more on that in another post.

The Amiga A2000 is a 'big box' Amiga and it is an accurate description. These things are bloody HUGE! I had expected it to be a fairly typical desktop PC size from the early 90's. How wrong I was. The main case is probably one and a half times the width and height of the Olivetti PC my brother had in the early 90's.

(Stock image from Wikipedia)

This particular A2000 came with a CD-ROM drive, two floppy drives, a memory expansion (size unknown but suspected to be 2MB), IDE controller card and a hard disk. I don't think I will use the hard disk since the Buddha IDE card immediately gives access to the use of CF card adaptors* such as those used in the A1200 and A600. Even though the CF card adaptors use a smaller pitch connector I just so happen to have an adaptor that I salvaged from an A1200 many years ago. Hoarding works. :)

*foreshadowing - CFs don't play well with Buddha

Anyway, this machine, as with everything picked up, has been in storage for a very, very long time and is in need of an initial clean before any serious work can take place. But before even that, you know what we need to do.... Get that battery out!

I see, you little git.

Once that was snipped out I could carry on and see what damage would have to be dealt with. I've done lots of A500+ in a similar condition but with the size of the board in this machine, anything could be affected. 

Clip for the 2Mb MegaChip expansion. Nice!

Top removed - fat gut for scale

With the top off the scale of the motherboard becomes apparent. Again, it's massive! Far bigger than the A500 boards I'm used to. The majority of the real estate seems to be on the left side with all the Zorro II and PC slots. And don't forget that massive power supply.

Side note - I did not attempt to switch this machine for some time yet. The reason will become clear.

After marvelling at the size of the board and case I also had a look at the keyboard. Sadly, it is missing one key. BUT, if you're a long time reader of this blog you will know I have a thing about the different types of keyboards that Commodore used for the Amiga. Well, it just so happens that this one is a real chunk of unobtanium. Firstly, it uses real Cherry MX key switches (on the cheapest, thinnest PCB imaginable - thanks Commodore). Second, you'll see from the pictures that the keyboard has two 'A' Amiga keys. They are both red. Finally, the LED for caps lock is a rectangle at the bottom centre of the key rather than a circle at the top right of the key. All this means - apparently - that this is the earliest of early keyboards and is probably worth as much as the rest of the machine just on its own.

Having been in storage the keyboard needed a bloody good clean but was really simple to do. The unit itself is held together by clips (no screws here - they'd cost too much money - thanks again Commodore). This means it was relatively easy to get apart although I was nervous about breaking the clips, this plastic is probably 30 years old after all. The keycaps were very easy to remove too and only needed a gentle nudge from the back to pop off. I suspect that's how the missing key got lost but it does look like I could get a modern blank key that will be compatible with the keyswitch or even 3D print my own.

Clean-up started - note the red 'Amiga' key..

Tea required..

Following the usual wash of keys, brush off of dust and clean with wipes, the keyboard went back together without too much trouble. One support post had broken off and was rattling around inside but a touch of super glue put that back into place.

Finally, I popped it out into the sun for a light touch of 'light-brighting'. Not sure it made any difference but there you go.

Enjoying the sun - with guest appearance by
an A1200 case

One thing I found is that the keyboard case halves are slightly bent, almost certainly due to being in storage for so long. Not a problem in itself and it's happened because Commodore cheaped out and did not use screws in the two parts of the case. Where the clips lock into place there is enough room for the halves to bend but remain attached to each other. My efforts above with the weights on each end were to try and correct that - it didn't really do much. All I need to do is put the right thickness of feet on each side and it will be fine. :)

Back to the main board.

First thing to do was to remove the damaged CPU and Agnus sockets. This was fairly straightforward and, on first inspection, there didn't appear to be too much damage around the CPU pins which was a surprise (foreshadowing...). The Agnus socket just needed replacing and there was no damage to the board from the battery in this area.

CPU & Agnus sockets removed

Insert new socket here...

Looking over the board around the CPU I spotted a via that looked to be in particularly bad shape. After a quick run over with the iron and some flux it disintegrated... Arse. Still, I have a handy reel of wire wrap wire which is the perfect size for repairing this sort of damaged. The wire has a thick (relative to the wire thickness) coating. Note that, as this particular type of wire will be important later... 

Via repaired. Others are OK(ish).

Next thing, the power supply. This is more akin to a PC power supply and has a large connector to attach to the motherboard along with multiple plugs attached to another cable for the hard disk, CD-ROM etc. I had no intention of just plugging this in. Given the length of time this has been in storage it's entirely feasible that the caps may have leaked or degraded past the point of no return. This sort of thing could lead to unfortunate voltages frying the board.

After another bit of a clean up, the removed power supply was plugged in for an initial smoke test. 

No smoke. Nice.

Next up, the voltage test.

No so nice. The 5v lines were happily pumping out 7v. The 12v line could barely manage 9v and the -12v was around +4v. This is not a happy power supply. If I'd plugged this into the mainboard and switched it on there is every chance I would have killed the board completely. 

I'm not a power supply expert so this one is, basically, headed for the bin. I have an old PC power supply which should suffice, although it is a totally different size and shape. A problem for another day.

Power supply innards - but faulty. Boo!

The only thing salvaged...

First thing to do with the old PC supply is splice the Amiga motherboard connector on to it. I should point out that the old ATX connector had already been removed as the supply had been intended for another, long forgotten, purpose. Anyway, the colours between the ATX and Amiga supplies do not match at all. On the Amiga cable, yellow is 5v, blue is Ground, orange is 12v etc, whereas on the ATX red is 5v, yellow is 12v (remember that one!!), black is Ground and so on.

More useful information on this can be found on various websites such as here or here. After the splice I carefully checked, then double checked, then triple checked I would get the correct voltages out of the supply and at the motherboard.

So, the first big switch on. What did we get? A black screen. This was not a surprise as there was still the possibility of unseen battery damage on the motherboard. Since black screen is an indication of a non-functioning CPU I put together a spreadsheet as a checklist for each of the pins of the 68000.  These checks showed that pins 34 and 41 were not connected. 

Lovely, another spreadsheet.

Bodge wires - not neat but to be 
tidied up once (if) it works

A couple of minor repairs later and it was time to try again, but this time I used the excellent, if slightly unintuitive, DiagROM. This would tell me, either on screen or via the serial port, what the state of the system was at each point during bootup.

To hook up the serial port I used a fairly standard USB to serial converter so I could use a modern PC. The last part had to be done by using individual wires with female to female 'dupont' type connectors. These are just the right size to push onto the pins of the male plugs. Only three connections were required, TX, RX and GND. TX on the PC was connected to RX on the Amiga, and RX on the PC connected to TX on the Amiga (this is pins 2 TX,  3 RX and 7 GND on the Amiga serial connector, and pins 2 TX, 3 RX and 5 GND on the PC side).

This was the result:

DiagROM serial output

This was actually a good result, even though there is apparently no access to memory. It did mean that the CPU was running the software in the ROM. 

Just for giggles, I decided to put the original 3.1 ROM back in and try again. I was surprised at the result:

Woah! I must be close!

This made me think that I was on the verge of getting the machine up and running again. With DiagROM not finding RAM my immediate suspicion fell on the logic that controlled the RAM. This had always been the problem with the A500+ machines I had repaired previously so it would make sense that the same would be true here, despite the slightly different arrangements on the board.

To verify this, I removed the following chips:

U540, U541 - responsible for the address lines

U103, U104, U105 and U106 - responsible for the data lines

Two things came from this. Firstly, one of the 74LS373 chips came out as bad and needed to be replaced. Second, many of the pads on U540 and U541 disintegrated as the chips were removed. Most of the pads affected were isolated pads on the bottom of the board i.e. their connections to other parts of the boards were on top. I assume this would be down to the battery corrosion weakening the adhesion of the pads. Certainly, I did not have this problem on other parts of the board later.

Broken.

So, after the near miss of the Guru, and the discovery of the faulty 74LS373 we must be on the verge of a successful start-up. Right? 

No.

With a new 74LS373 there was no difference in the DiagROM output. Even with the pads repaired on U540 and U541, the system could still not find any working memory. Very, very frustrating.

In my frustration I took a number of paths that proved fruitless. First, I soldered the CPU direct to the board. This made no difference. Then I tried removing the standard Agnus I'd installed to put the MegaChip back in and managed to crack the socket at one corner. F***. Strapping it back up with a cable ties appeared to work but, again, I was still stuck with no chipram detected. 

It was suggested that I needed a later version of DiagROM since 1.2.1 and 1.3 Beta had additional address validation on startup and this might help. So I downloaded the 1.3 Beta and broke out the Dataman S4 and my phone sanitiser (seriously) to wipe and re-write my current DiagROM EEPROM. But I hit a snag...

Dangnabit.

So I had to resort to eBay for the latest version. 

Anyway, it arrived and it told me, unsurprisingly, I had some address errors. 

Error...

These address errors seemed to be pointing at Address bit 11 being stuck. But despite my best efforts and many hours of tracing out traces (there's no A2000 on Amigapcb.org - sadface), nothing I did changed anything. The actual display output of DiagROM was limited to this:

Bad memory...

This screenshot was taken with the standard 8372A Agnus and J101 set to pins 1 and 2, and J500 closed. This should indicate to Gary that there is RAM in the 'Ranger' address space (apparently) hence the bottom of the screen is showing a yellow and red bar. The yellow bar is the first 512Kb of chipram and the second is the 'Ranger' RAM. In any case, it's broken still.

Annnnd, then the serial output from DiagROM just stopped. Having had some dealings with the serial chips in the A500 (MC1488 and MC1489) I knew where to look. Sure enough, after removing and replacing the 1488 - including a socket - the serial started to work again. Until it didn't. I actually worked out that the wires from the power supply to the motherboard were not making good contact. After a quick wiggle of the connector, the serial output came back. Interestingly, this course of action would also restore colour to the video output when it would suddenly decide to go monochrome - although I should point out that I was using an A520 modulator for a quick colour composite output. 

MC1488 in a socket. Not really needed
but, oh well.

By now, several weeks had passed and I had made no further progress. Except I released some magic smoke. At some point during my constant probing, I assume that I managed to short something to ground. Cue a sudden drop in video output, a large puff of smoke and a panicked leap for the power button. It was just a transient thing though, and after a quick repair, I was back to the same position, i.e. no chipram detected.

Magic smoke generator - oops

I decided to remove the sockets for U540 and U541 again as well as the CPU and try removing the repairs I had put in originally. I had reached the point where I thought that maybe I had made a mistake during the repairs.

The CPU repairs were relatively straightforward. I have a few reels of wire wrap wire which is perfect for this type of thing. As mentioned above, it's thin enough to be a trace replacement but has a relatively thick coating which also allows it to be used for 'bodge wires' without any danger of shorts.

The repairs to U540 and U541 were a bit more involved but not too onerous. I had drawn out where the pins should go and then got the multimeter on them. I added bodge wires where the traces did not seem to be making contact. 

U540 and U541 connections - what lovely handwriting..

U540 missing traces

U541 missing traces

I also went all out and replaced the Agnus socket (again) and checked all the traces from all 84 pins. Not a fun couple of hours but I did find a missing line from Agnus pin 83 - bodge wire added.

CPU pin repairs repaired. Again.

Agnus pin 83 missing link
(repaired during new socket installation)

Given that I now had another new Agnus socket I also went for it and re-installed the MegaChip. After many hours of careful, methodical repair, would this now work?

No.

Or, at least, not quite.

Garbled output - but not the usual RAM screen...

DiagROM reached this point and crashed - every time

This is the first signs of life since the Guru so many weeks ago. But what could be wrong? Fortunately, a very helpful chap on twitter suggested it was a crappy Agnus socket. I was initially doubtful, since this socket was nearly new. To eliminate it again, I took out the MegaChip and carefully bent all 84 pins of the Agnus socket out significantly further than they had been before. 

Another switch on attempt.

Oh.

My.

Goodness.

It works.

ROM screen

Early boot screen (hold both mouse buttons down)

The ubiquitous ATK from Keir Fraser

"Colours, so many colours..."

This has been a real labour of love but it's not over yet. The 2Mb zorro card expansion doesn't work and comes up as 'defective' on boot when it's plugged in. The Buddha card works with hard disks but is, apparently a real pain in the posterior to get working with Compact Flash cards. This is slightly annoying but I do still have a couple of hard disks I could possibly use. The original hard disk does still work but is quite small. I also need to hack the PSU into the space of the original without zapping myself or killing the machine...

Of course, I could just remove my PiStorm from the A500+ and stick it in this too - I think it will fit with the PI mounted to the top.

In any case, this behemoth of an Amiga is on the way back to full use. YAY!