Welcome to Inkbunny...
Allowed ratings
To view member-only content, create an account. ( Hide )
Waccoon

Geek stuff: Amiga repair

First of all, please pardon my absence for the last couple of weeks.  Aside from putting up drywall and other construction events, I've been feeling pretty down as of late and I do tend to recede from society when that happens.  I'll be slogging through my much-neglected notes over the next couple of days.

Anyway, this journal is a [promised] summary of the maintenance of my Amiga computers a couple months ago.  If you're not interested in electronics, this is not for you and you may continue surfing for porn.  8)

Last year I was having trouble with my old Amiga 1080 monitor (manf. 1988) where the red gun was not working.  For a while I was suspecting the problem was with my computer, since the composite signal gave me red, but the RGB connector did not.  It was also intermittent, which led me to believe it may possibly be related to thermal stress in the computer.  I figured a total tear-down was in order.  Old, leaking electrolytic capacitors are a major problem with electronics more than 20 years old, especially since the leaking electrolytes can eat through metal traces, so it was high time to tear my baby apart and do some refurb work.  Opening the A1200 is a common thing as I've done some upgrades over the years, so I already knew what kind of parts were inside and I'd have to learn how to do some surface-mount soldering.

Still, I've never removed the mobo from the case, and when I did I got a nasty surprise.  There was a nice burn mark in the bottom of the case and a hole melted through the plastic insulation.  Uh oh.  Turns out, a ceramic cap had gone nuclear and punched a small crater in the mobo.  Amazingly, the machine still worked, but it was an ugly sight and clearly this was going to be a bigger project than I thought.  Dammit.  I downloaded some A1200 schematics and ordered all the replacement parts from DigiKey.  Panasonic and Nichicon 105 degree caps, and Kemet industrial ceramic caps.  The good stuff.

After watching a number of YouTube videos on the matter (mostly Louis Rossman the Apple repair guy, and various Amiga specialists), I did some practice on an old IDE controller card.  I can't say my results were impressive, but it was good enough given the tools I had available.  The hard part is that solder would blob up on my iron and not transfer to the parts, so it took quite a few tries to get the thin wire I used to bridge the new cap with its respective via, as the pad and trace were long gone.  It was tough, but I did it.

Alas, while the machine turned on and worked, I still had no red on the monitor.  So, I fixed a decoupling cap, but had to find out what was causing the video to be bad.  I borrowed my dad's Hitachi oscilloscope and had lots of fun learning how to probe parts and looking at video signals and stuff.  Alas, the video signals appeared to be exactly what they were supposed to be, save for the green channel (later, I found out the Amiga uses sync-on-green, so the voltages were all different on green).  At this point, I concluded that the monitor was to blame for the red not working.  Not what I wanted to discover, as I didn't feel comfortable opening a monitor and all the thousands of volts of residual charge that builds up inside an old CRT.

Anyway, the A1200 was working fine, so it was time to replace the electrolytics.  Good thing, too, because at least four of the caps were green and fuzzy and clearly leaking.  The removal concerned me, though.  To my dismay, the preferred way of YouTubers for removing SMD capacitors appears to be the brutal twist-and-rip method.  I once again practiced on that poor IDE card, and as it turns out, it works far better than trying to de-solder the parts and rocking them back and forth.  I really tore up some of the pads on my test board trying to be "careful" with the rocking method, but there was no damage at all when just giving the caps a quick twist.  So, on with the show.  All 15 of the SMD caps were twisted off without incident.  In fact, it was amazingly easy to do.  The large aluminum caps were more troublesome, though, as I found out quickly that lead-free solder has a much higher melting point than leaded solder, and it took a LOT of time and heat to get those damn things off.  At one point, solder was so deeply webbed inside a through-hole I had no choice but to break out a tiny hand-held drill and grind my way through to get the solder out.  It sounds alarming, but mostly it was just time-intensive.  I put the replacement can caps on first (some high quality Kemet polymer caps that should last forever), but I still didn't feel confident about putting the new SMD parts on.  Solder just didn't want to stick to the parts, and I made a real mess trying to put the SMD caps back on my IDE card.

After watching some more videos, it was clear I made a major goof: trying to solder without flux.  I figured the flux mixed in my reel of solder would be enough, but once you add solder to your iron, the flux immediately burns away (which is what makes all the smoke and ruins your health).  You NEED extra flux on the mobo itself.  So... off to Digikey again to get some flux, plus some extra caps I forgot to order for my power supply.  Yep, that did the trick.  With flux on the mobo, the SMD caps went on without a hitch!  I mean, now my results looked absolutely professional.  I'm quite pleased with my work, and am even tempted to post a picture of both before and after.

Before moving on to the power supply, I got a kick reading through the Commodore schematics.  What a basement operation!  There were several errata in the schematics, including wrong part values and parts that weren't there, despite the schematics matching my board version.  I never knew that the empty U0 socket was for a planned FPU that got cut at the last minute, the machine was supposed to have a real time clock on board (also canned), and it was supposed to have a 16-bit ROM instead of the 32-bit ROM that got used.  Ah, the wild west days of computing, when specs kept changing right up until the end!  Also, there's quite a bit of impedance mismatches on the mobo itself, resulting in noisy clock signals, which may be a problem if I add a CPU accelerator board (I currently have an FPU and memory board).  I did make a correction to my A1200 by replacing the polar electrolytic caps on the audio circuit with bipolar ceramic type, which should last forever.  I also put a heatsink on the video controller because... why not.  Too bad the audio chip doesn't have enough clearance from the RF shielding for a heatsink, since it's the second-hottest chip in the machine.  I may add some additional EMI protection and fix the system clock in the future.

This was a great opportunity to appreciate the genius of the machine.  Being a 2-layer board, it's possible to trace all the signals between the chips and figure out how the whole thing worked -- something you can't do on modern hardware.  The CPU actually talks to the DMA controller via address and data lines.  Modern CPUs have all the bus logic built-in (some are full-blown SOCs) and you can't do squat with them.  On the oscilloscope, you can get an impression of what the machine is doing by just by probing the data lines -- much like how I used to watch and even listen to my Amiga's memory using RAMscanning techniques back in the day, piping raw memory to the display and audio hardware.  What a fun computer.

Oh, and I noticed some really interesting behavior with the audio output.  The oscilloscope clearly shows that the audio chip produces waveforms with a switching power circuit and a charging cap.  The Amiga can play audio at (I believe) up to 29 KHz, and if my testing is accurate there's a very obvious capacitor charging and discharging profile around a 1/20 a volt, repeated at a rate close to the Amiga's max audio rate, probably 32KHz.  No wonder the audio sounds so funky when you play two chip samples on the same sound channel at the same time, and emulators do NOT produce authentic sound.  The audio is mixed with a switching circuit and smoothed with a cap in the audio chip, resulting in all kinds of beautiful aliasing artifacts.  I wonder if old SoundBlaster cards produce similar waveforms.  I'll have to check that with my old SB Awe32 card.

The power supply was a big question mark.  It seemed to work fine, but made some weird noises in the form of regular ticks, and on both the multimeter and oscilloscope, it seemed the brick wasn't producing stable power.  After replacing all the caps, which was a breeze, I simply could not figure out why the power supply was ticking and how the main transformer worked.  I did a full reverse engineering of the power supply and leaned a lot about how switching power supplies work and how they're built, and how caps behave differently under AC and DC power.  Online tutorials weren't much help, though, since they all cover modern power supplies built with ICs.  My power supply is a strictly discrete model built only with analog parts, and they aren't behaving like I expected.  A full 60Hz AC signal was going to the transformer, not the several KHz switching it was supposed to.  Also, the power supply was pumping out some HUGE spikes to ground, which went right to the computer's ground terminal.  I was not liking what I was seeing, and was wondering if the transformer, optocoupler, or main MOSFET was bad.

After struggling for more than a week, I finally figured out that the power supply won't produce current unless it's under load.  It may seem obvious to experienced folks, but to me this was a surprise.  I thought only modern PC power supplies did that.  Turns out, my power supply needs to draw at least 3/4 of an amp on the 5V rail to be stable, and once I wired up some high-wattage resistors to act as a load, the power supply worked as expected.  Yay, I was pulling my hair out for a week for no damn reason.

This all led me up to the main problem: fixing the red gun on my monitor.  After looking up more videos on how to discharge CRTs, I figured I wouldn't accidentally kill myself if I opened my monitor after all.  I rigged up a 20 watt resistor to a screwdriver and did the discharge.  No satisfying "pop" using this method, but less risk of frying the monitor's logic board.  I found out quickly how dirty the monitor was inside and why the pots on the front of the monitor worked so badly.  A good cleaning was in order, mostly by scrubbing the plastic shell in the kitchen sink, and pulling apart the RCA connectors and using a Dremel and automotive clear coat polish to grind off the tarnish.  Add some contact cleaner, and... all bright, shiny and new!  All the pots were de-soldered and cleaned, which took a lot of time, and there were like two dozen trimming pots that needed to be cleaned as well.  Yep, all old-fashioned analog hardware.  Amazing any of this stuff works at all!

Getting to the video input switch required a lot more de-soldering than is should have, as the design was pretty stupid and obviously not made to be easily serviced, but it was clear why the red signal wasn't getting to the logic board.  I thought it might have been a cold solder joint, but the main video switch (for selecting between digital or analog input) was gunked to hell.  The green and blue signals had almost zero resistance, but the red line was spiking like crazy.  With the input switch cleaned, and some lithium grease applied to some moving parts, the monitor went back together.  It looks like new!  Like, seriously, I have never seen the monitor produce an image so clear.  There's over a hundred caps in the monitor, so I won't be replacing any of those parts, but everything seems to be in good shape.  If the monitor does die on me, I'll just get an arcade machine video conversion board and get a modern LCD for the 'ol miggy.

So, yeah, the A1200 is like gold to me, and now everything is running like new and should last a long time.  I learned a LOT about electronics in the meantime, though I'm not sure what use this will be in the future.  Other than light blinkers, what can you really do with electronic knowledge now that there's SOCs everywhere to do all the heavy lifting?  Maybe I'll do some audio work.  That stuff will always be analog (I hope) and require some good old-fashioned handiwork.

Which reminds me, I still need to find out why the power supply in my homemade headphone amp (built by my dad) is apparently sending 6 volts AC to the amp chip along with the normal 8 volts DC.  Maybe the smoothing cap is bad, as it is pretty old.  Or maybe I still don't know how to use my oscilloscope properly.  Maybe I'm also just getting obsessed with replacing aluminum caps!  They're evil, you know.  8)

Anyway, there's no point or finale to this journal.  It was just a bit of an adventure working on my old computer again.  I took the liberty to disassemble and clean out the case of my A1000, though the unit doesn't boot with the original floppy drive since the drive is out of alignment.  I'll have to look into how to align a 3.5" drive sometime in the future.  Next up, I'll try to get my ColecoVision working again, and maybe do a composite/RGA upgrade to it while I'm at it.  I haven't used it since the power supply died 20 years ago, and it would be nice to get it working again.  I could take a look at my C-64, too, though I never used that machine much since I didn't like it... not as fondly as my Amigas, at least.  The C-64 was, in terms of usability, a pretty bad machine, actually.  Oh, those arrow keys and slow floppy drive!

I wonder if the Amiga engineers ever expected people would be refurbishing their designs 25-30 years later.  I sure am proud to own and preserve these pieces of history.  Nothing beats the real thing, and you'll never get me to part with 'em!
Viewed: 27 times
Added: 1 year, 6 months ago
 
ThaPig
1 year, 6 months ago
That was fascinating to read, even when I don't understand much of the technical details. You should have made a video of all the process and post in youtube!

I had an old Amiga in the 1990s and now I wish I had not get rid of it.
Waccoon
1 year, 6 months ago
No webcam yet, and I won't be investing in that until I get a prototype of my puppet finished.
ThaPig
1 year, 6 months ago
You don't have a cell phone?
Waccoon
1 year, 6 months ago
If you mean a smart phone, then, no.  Either they're too expensive, or I'm too cheap (probably both).

My dumb phone can only take still pictures and even then I have to e-mail them to myself using a data plan I don't have.
ThaPig
1 year, 6 months ago
I have one that's a couple of years old but it has a decent camera you can use to make videos and upload them to youtube. You can install apps on it to record and edit video. You could probably even turn it into a webcam by hooking it to a computer with a usb cable. I have no use for it so I can send it to you.
Waccoon
1 year, 6 months ago
Oh, web cams don't cost much, so I'll probably get a decent one when I'm ready.  My laptop already has a webcam built-in, and I can use that for practice.
Mitsozuka
1 year, 6 months ago
Sounds like you had a blast!

Aside from pulling your hair out for a week for nothing... >.>
Waccoon
1 year, 6 months ago
A computer trademark!

Honestly, I don't know how they managed to get anything done in the analog days when they didn't have simulators or proper debuggers.
Mitsozuka
1 year, 6 months ago
As you discovered, I imagine, trial and ERROR. xD
GreenFur
1 year, 6 months ago
i am surprised you didn't pick up "solder wick", very useful strip of braided copper place it on the solder then place the tip of the iron and it cleanly removes the solder form the board, or choose  a vacuum desolder tool.

so now you seem ready to build your own computer! with the new TSR80 Kit computer from Radio Shack!
(yes, back in the day the original TRS80 ws designed as a kit but the company decided to offer them as a prebuilt unit)
my first comp WAS a Kit
 on old IMSAI that you started the boot procedure with flipping panel switches, then went to the paper punch tape reader, then to the cassette tape to load the programs....fun times
--GF
Waccoon
1 year, 6 months ago
Oh, I had solder wick, but for some reason I just couldn't clear the solder out of this one through-hole.  As a last resort, I figured drilling the hole was a less risky option than cranking the heat on my iron too high and potentially melting the solder mask or lifting a trace.

I looked into a kit computer, but most of the things out there are SOC, like the Uzebox.  Not very interesting (and bloody expensive for what goes on them).
Daggett
1 year, 6 months ago
Solder suckers should work for cleaning out through holes. You may have to add solder, then heat it and suck it all out and clean the pad with desoldering braid, isopropanol and/or acetone.

I'd say that's a good example of why electronics are rarely actually repaired now days, it's just easier to chuck it and buy a new one. :)
Waccoon
1 year, 6 months ago
I tried to add solder and dissolve the crappy lead-free stuff, but it was just too stubborn.  I already had the heat on my iron up to 750 and was worried I'd melt the mask or lift a trace.  Drilling seemed to be the more humane option.

Repair isn't realistic thanks to so much thermal death.  I had a video card just give up the ghost and I'm convinced it was the GPU itself since it always ran wicked hot with the stock cooler.  Can't get a replacement GPU!  Most old computers didn't produce enough heat to cook themselves to death.

Don't even get me started on the DRM, or I'll talk your ear off for hours.  8)
Daggett
1 year, 6 months ago
Are you sure the board had lead free solder, that would be unusual for the time? The ability to heat large blobs of solder is not so much the temperature of the iron, but its heat capacity (wattage and tip design). I'm thinking the capacitors you were having trouble with were actually sinking too much heat from your iron.

It's pretty hard to kill a modern video card with temperature since they regulate automatically to protect themselves. They're actually designed to operate at the maximum temperature for long periods of time, so you shouldn't have a problem with that.
Waccoon
1 year, 6 months ago
Now that I think about it, the one cap that was giving me trouble was tied to the ground plane, so it probably just wasn't getting hot enough.  The cap detatched easily, but I just couldn't get the leftover solder out of the hole.  My iron is a Weller with adjustable temps and multiple tips, so it's pretty good quality, but I didn't want to risk burning anything.

Alas, GPU death is very common, partly due to the way power delivery works across the die, and the bonding solder they use for flip-chip packaging which will crack after a certain number of thermal cycles.  It's not really the temp that will kill the chip, but the hot spots and constant cycling between cool and hot.  Sometimes you can fix this by heating the chip and trying to reflow the bonding solder (typical XBox 360 fix), but once a GPU gets flaky, there's not much you can do.  Thermal sensors don't do much to extend the life of these kinds of chips.  You just have to keep them cool all the time.
Daggett
1 year, 6 months ago
Yes a large ground plane can also cause problems with heat. Sometimes it helps to change tips to a thicker one that can hold heat for longer. Beyond that you would probably have to switch to hot air.

You really shouldn't see problems with GPU heat until you get to about 90C. Like I said, they are designed to run hot, unlike CPUs. Typically they run around 65 to 75C, and that shouldn't give you any problems at all for years of regular gaming. Now GPU manufacturing quality in general seems to have decreased somewhat, so they probably aren't as reliable as they once were. But failures like that seem to be random rather than based on heat.
TootTootSonicWarrior
1 year, 6 months ago
Gif 4 u.
Come with us, I see passion in your eyes! by TootTootSonicWarrior
dahan
1 year, 6 months ago
Neat! :) I have some hardware from that era (mainly Apple stuff, like the IIgs), but I haven't powered any of it on in a while. I also have a Power Mac G4 from around 2000 that I occasionally use—its power supply died 3 or 4 years ago, so I found a new old stock replacement. The replacement only lasted a couple months :( So I bought another one and preemptively replaced all the capacitors and it's working great so far :)
New Comment:
Move reply box to top
Log in or create an account to comment.