Apple's backwards design mistake and the reversed capacitor Did Apple really fit a capacitor backward on the Mac LC III? A multimeter-wielding retro fan has confirmed that, yes – somebody made a mistake decades ago, and a capacitor ended up installed the wrong way. Doug Brown outlined his findings in a lengthy post in which he stripped one of the machines of its original capacitors. …
The good thing is that this didn't affect people. First, a brand new capacitor is likely to form at least OK, if not well, from a reversed polarity, as opposed to a properly used capacitor that's then introduced to the opposite polarity. Second, this just acts to clean the voltage from the power supply, so it's strictly not needed. Third, Macs of that age don't need negative voltage for anything important, so long as the expansion card doesn't need it. Sound works, as does RS-422 and LocalTalk, although it's possible there are machines / devices to which serial can be connected that wouldn't be happy with the lack of proper negative voltage.
I have quite a few 68030 and 68040 Macs that I've built in to various cases, and none need negative voltage, although it's really not hard to add in most instances.
I am coming to believe that when a scifi protagonist of some kind makes a "reverse the polarity" comment, it really should be regarded as just a short way of saying "I have an idea far too complicated to explain to idiots like you, but can I just get on with it anyway?"
. . that knowing what is actually happening electricity-wise is still something that needs to be taught.
To think that we're going toward a future where electrical components will be designed by computers . . . if ever there is a glitch, there will be one hell a big puff of smoke.
The words "measure twice, cut once" come to mind.
Sounds more like whoever was drawing up the silkscreen didn't notice/forgot that the supply was minus five volts (perhaps because they'd not labelled that bit yet), and just labelled it the same as the two caps next to it.
ie just a simple fuckup, not a lack of knowledge.
I remember that a bunch of graphics artists where I did a short term contract all had the 'magic mouse', the sign of a charging cable in the mouse meant that the occupier of the desk was either in the toilet doing No 2's or in the break room, having a coffee (all drinks were banned from anywhere near the computers). That was all it needed to get enough charge for another few hours of work.
I do remember on manage who would complain loudly (so that everyone shared in his pain) when he came into work in the morning and found it hadn't been charged overnight. No one had the nerve to tell him that it was always unplugged by the last artist to leave the office. He was not liked and this was one way to get one over on him..
Those were the days.
Apparently that was done intentionally, i.e. to not be able to use the mouse while it was charging. Goes to show that you might be a great designer, but if you don't actually consider the end user, your design ends up being a pain in the arse rather than being sensible and usable.
This post has been deleted by its author
I've been doing board design and layout for more than two decades. And if there is anything I am guaranteed to screw up on revision A through about C... It's something on the silk layer. Either a reference designator ends up close to the wrong component or a polarity gets switched. Usually a diode, though, not a cap.
This almost seems like a physical law... Regardless of the effort you put into automation and review... The silk will have an error. Guaranteed.
AC, good question. Three answers-
1) You're right in that a reversed cap on a negative rail is probably an error in the schematic. It's very easy to put all your big decoupling caps in a row, beautifully aligned with the positive bar at the top and curved negative side pointed down. Then you slap a bunch of GND at the bottom and slap in the power net labels at the top... Forgetting that one is negative and needs to have the cap flipped. Sails through review because everyone focuses on the "hard" sections and it just looks right.
2) The pick n place robot is responding to a centroid file that specifies reference designator, side, x/y coordinates, and rotation. If your parts library has inconsistencies in, say, whether pin 1 of a diode is anode or cathode - fail. Early 90s tooling had little or no ability to automate management and review of parts libraries
3) In the early 90s pick and place was a two stage process. Very light components were more or less put in place by a high speed chip shooter. Gantry head stays still while the board is (violently) moved underneath. Big components like the electrolytics are too large to place this way because they'd go flying. So the precision moving head 2nd stage would be used... But computer vision didn't work well back then so there was a lot of man-in-the-loop for precision placement and a tech will follow an incorrect silk every time. Crude machine vision wasn't really great at tall components...
"It's very easy to put all your big decoupling caps in a row, beautifully aligned with the positive bar at the top and curved negative side pointed down. Then you slap a bunch of GND at the bottom and slap in the power net labels at the top... Forgetting that one is negative and needs to have the cap flipped. Sails through review because everyone focuses on the "hard" sections and it just looks right."
Yes that's my take on this too. Has anyone found a copy of the schematics yet?
As the +16V rated capacitor had -5V across it and there was little load on that circuit it wouldn't have been picked up in Design Verification.
The "fun" begins when you swap the electrolytic for a tantalum. These are much more fussy about polarity and I have had the through hole version vaporise the orange body (with a suitable "bang!") on power-on, leaving you with the challenge of finding the two legs poking out of the PCB with no tell tale scorch marks to give you a clue that they aren't just a couple of test pins!
"The "fun" begins when you swap the electrolytic for a tantalum. These are much more fussy about polarity..."
Not just that. Tantalum caps are notoriously limited in the in-rush current they can handle. Even correctly polarized they had a tendency to explode on power-up simply because the initial current surge coming from the PSU (a heavy one given the power consumption of the era's TTL logic) was too great for them. I remember several series of Chinese XT clone motherboards with that particular problem. I made some quick and easy money then during summer school holidays repairing mobo's that had gone bang.
This was a few years before the dreaded capacitor plague that caused regular electrolytic caps to start bulging and stop working. I remember buying a very nice printer for 25 bucks on a flea market intending to strip it for parts because I expected it to be broken, being offered at that price (in the early 1990s when the list prices for that make and model were at least fifty times that) and indeed it was dead as a dodo when I switched it on. But upon careful examination I noticed two bulging caps in the power supply. I replaced them and the printer worked fine. I've been using it for about ten years thereafter.
Acorn interchanged NPN and PNP transistors in the video output of their Atom computers, making it a small miracle that anyone got a TV picture. Transistor notwithstanding the botched mono output was at US 60 Hertz (for UK TVs expecting 50) and glitched white spots whenever the 6K ‘hires’ display RAM was accessed by the CPU during pixel output. Space invaders looked like a snowstorm even after you adjusted the vertical hold and assuming a signal leaked through the reversed transistor. Still, the BBC adored them ;-(
Schematics were for a pair of NPN transistors, which are in an configuration that works - I've used the 6847 a couple of times as an easy video display for embedded systems where the output wasn't too demanding and borrowed that circuit to get a video signal.
The glitching white spots weren't to do with the 50Hz/60Hz discrepancy as such, but because the dot clock on the 6847 ran on both edges of a 3.57MHz square wave (chosen because 3.57 NTSC colour burst crystals were cheap, and if I remember correctly, the 6847 would insert a few cycles of 3.57MHz after line sync for NTSC colour burst), while the 6502 ran at 1MHz.
When directly accessing video RAM, the 6502 took no notice of where the 6847 was in its pixel reading process and just enabled the bus buffers connecting video RAM data/address to the CPU bus, and switched off the address output of the 6847 for ~0.5us (possibly up to 1us, I can't remember whether the logic gated off the whole cycle or just the active part of the 6502's phi2). Hence the 6847 would momentarily output whatever the 6502 was reading/writing in video RAM, rather than what it was expecting to read/write.
Writing characters to the text screen didn't cause snow because the software polled the frame sync signal and only wrote text when it knew the 6847 was in a non-image part of the display, but doing that generally for bit mapped graphics would have slowed things down too much, so it was left down to software designers to decide whether or not to sync.
Possibly the worst culprit was the unexpanded Atom where the 1K of screen memory had the upper 512 bytes for program storage, so running a program in that configuration caused loads of screen 'snow'.
The 6809 in the TRS-80 colour computer runs at 0.89 or 1.78MHz (1/4 or 1/2 of the 3.57MHz) and I believe CPU access to video RAM is interleaved with the 6847 to prevent snow. I have a feeling you could remove the snow on the Atom by rewiring the clocks to run the 6502 at 0.89MHz, but I never tried - I always wanted the raw speed over a clean screen! (actually mine was quite happy being overclocked to 2MHz).
I never knew about the NPN/PNP transistor mix-up though - I'll have to check mine to see if they put the right ones in - it worked though!
It's amazing what you could get away with, back then. TTL logic was fairly forgiving compared to today's CMOS nonometre technology. I remember a Chinese PC-XT clone motherboard that had a 74LS04 soldered in backwards. It was part of the 8MHz clock pulse generator. Which, unsurprisingly, didn't work. However, enough of the 4.77MHz clock pulse leaked through, so when you switched the system to "Turbo" it still ran, not on 8MHz but on half the 4.77MHz clock frequency.
Those were the days...
They *are* different pictures. The perspective of the two connectors changes from one to the other.
I think the damage is visible in the first picture (if you look at the high re copy), but the corrosion has run between the copper and the green solder resist. The silk screen is on top of that so appears undamaged.
The solder resist and silk screen probably came off when the board was brushed with flux remover after desoldering the caps.
Many years ago I picked up a very cheap eMac from the well-known online auction site for my then young kids to play around on. It crashed constantly and after a quick search on the Apple discussion boards, it was clear that there had been a rogue batch which had been recalled.Mine must have missed out and it was too old to go back to Apple. I got some excellent advice from one of the posters on there and bought the replacement parts online. I have a very technically capable pal who did the soldering stuff & we got a few good years of trouble free use out of the machine.
The creepy LinkedIn bit came later when they popped up a suggested contact to the guy who had helped me on the Apple discussions board. We had no other online connection whatsoever and work in unrelated industries. I binned LinkedIn shortly afterwards.
I worked for a TV manufacturer in the 70s, before they were all made in China. The component auto-insertion machines (before SMT) were clever enough to check each diode polarity and rotate by 180 if it was wrong. That's fine until a batch of mis-marked diodes passed through. Of course the TVs worked fine but we'd occasionally get a repair engineer phoning us, wondering if the reversed diode was causing the fault he was chasing. It wasn't, and DON'T REVERSE THE DIODE we'd emphasise.
That reminds me of when I was a student in the 1970's, working part-time in a small TV repair business. On Saturday afternoons, work would stop when Dr Who came on, when we'd all watch it on one of the sets we were working on - testing it, of course! One Saturday, the set chosen was an ITT model and it was an episode with a lot of alien action (I don't recall which one) but we were well into the episode before we noticed everyone had green skin. Once the programme was ended (we just ignored the colour error) I got the set to fix properly. It took a fair bit of chasing down via the schematics and an Avo, but it was a blown diode in the board that decoded the colour from the PAL signal - in that state, it was a 50:50 chance that red/green would be correct or switched when the set was powered on. That was why the fault hadn't been noticed during previous soak testing. Replace the diode and it was good to go - after some on/off cycling to test, of course.
And that dive into the memory also reminds me of the Philips dual-standard (405 line mono and 625 line mono or colour) TV we once had in for repair. It was a monster and had 36 valves (aka vacuum tubes) - it took two of us to move it around on the bench (and a couple more to help get it on and off). I can't remember what fault it came in with, but I do recall that it took a lot of troubleshooting and we were glad when we got it back to the customer...
When you were testing the TVs by watching Dr Who, did you get Maureen Lipman appearing on them all saying “Hungrrrrryyyy…. hungggrrrryyyy….”? :)
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
