from the but-it-couldn't-happen-today-could-it? dept.
There have been some past rumblings on the internet about a capacitor being installed backwards in Apple's Macintosh LC III. The LC III was a "pizza box" Mac model produced from early 1993 to early 1994, mainly targeted at the education market. It also manifested as various consumer Performa models: the 450, 460, 466, and 467. Clearly, Apple never initiated a huge recall of the LC III, so I think there is some skepticism in the community about this whole issue. Let's look at the situation in more detail and understand the circuit. Did Apple actually make a mistake?
I participated in the discussion thread at the first link over a decade ago, but I never had a machine to look at with my own eyes until now. I recently bought a Performa 450 complete with its original leaky capacitors, and I have several other machines in the same form factor. Let's check everything out!
(Score: 3, Funny) by Snotnose on Thursday November 28, @02:56PM (7 children)
I'm old enough that I was an electronics tech when the transition from people installing parts was replaced with pick and place machines. We got in a set of boards (sets of 10) that didn't work. I got the first one and found out that instead of all the address/data lines having pullup resistors they had capacitors. The whole lot of 10 had to be scrapped.
It's just a fact of life that people with brains the size of grapes have mouths the size of watermelons. -- Aunty Acid
(Score: 3, Interesting) by RS3 on Thursday November 28, @03:12PM (5 children)
It's one of my many dislikes of SMD stuff: most parts are unmarked, esp. the very tiny ones. It's very difficult to tell if incorrect ones have been installed. You could have a circuit that pretty much works, but some in a batch might be flaky, or you could have long-term reliability problems. I think parts manufacturers could do more to mark the parts- micro laser etching would get my vote. It would require a microscope to read them, but that's okay, as we always have microscopes around where we make electronics using SMDs. (Yes, I work in that world now.)
(Score: 3, Informative) by VLM on Thursday November 28, @04:03PM (4 children)
You'll need a standard. Three comments:
Tantalum thru hole caps are already famous for the problem where some mfgrs put an indistinct mark on the plus lead and some put a mark on the minus lead. Electrolytics are at least consistent-ish. "Its small and we're in a hurry and we're cheap" will result in behaviors like placing a dot on one lead. Now does that dot mean its plus, minus, or morse code "E" means model designation "E"?
It's easy to look at a data sheet and in my other comment I pulled up a datasheet for Kyocera caps and the plus side is to the sprocket on the tape... however I'd say its 50/50 if other caps have "plus to the sprocket" simply substituting in chemicons for a production run might (or might not) result in all the caps being installed backwards.
You can already assume that X percent of components from China, where X is a substantial percentage, are reman remarked used otherwise fake and your plan would not fix that. There will still be a giant storage crate full of bags of NPN transistors in China where the seller applies your model number label of choice using a sticky label on the bag and ships it out as if its a genuine product.
(Score: 4, Insightful) by bzipitidoo on Thursday November 28, @04:36PM (2 children)
Yes, I have my own story of backwards markings. Circa 1990, built my own 80486 based PC with a motherboard with lots of jumpers. I followed the instructions to set them all correctly, but things didn't look quite right, and the machine would not boot. After much looking back and forth at the instructions and the motherboard, I finally realized the printed diagram in the installation manual had been mirrored! Thanks to that, I had set all the jumpers opposite of what they should be.
Once I corrected the jumper settings, that lead to the next issue: there was no entry for the particular memory configuration I had, 2 sticks of 4M each. I extrapolated from the list they had (the setting for 1 stick of 4M, and the setting for 2 sticks of 2M each), to arrive at an unlisted jumper setting, and luckily, it worked.
(Score: 2) by EEMac on Thursday November 28, @05:53PM (1 child)
I love retrocomputers, but early 1990s IBM PC setup was a special kind of nightmare. Having to tell software/hardware what memory addresses and IRQ assignments to use . . . !
(Score: 2) by Spamalope on Friday November 29, @04:17PM
Late 90s/early 2000 plug and play was worse.
The addressing/IRQ/io mapping limitations hadn't been addressed, some drivers or hardware absolutely required a subset of the settings, and plug and play... well, couldn't be manually configured. Oh, sometimes there were placebo settings panels, but those were ignored regularly - or randomly each boot so you never knew what you're config would be.
If it allowed true 'soft config' such that you could (optionally) hard assign (without jumpers) the formerly jumper settings that would have been progress. You could have set the critical ones and let the rest float.
But... Plug and play didn't even know to set memory windows to be contiguous - which you needed to do so that the low to high memory paging had as large a memory window between 640 & 1024k as possible (3.1-ME all still worked that way...).
I configured a number of system where all 8 card slots were populated so it mattered to me. It didn't as much with a base system.
(Score: 2) by RS3 on Thursday November 28, @04:38PM
It's interesting how e-conversations go. I respect and agree with everything you wrote, but I was just talking about marking the part's electrical value. You know, uF, or ohms, or microhenries, or whatever. Where I currently work we use some "DC blocking" capacitors. They look like other ones, but circuits get very unhappy if someone mixes them up.
Direct experience with tantalums: let's say you get a new batch of them. Test a couple. If you apply voltage in reverse, even at very small current, they will go POP very quickly. Then you'll know the polarity. Wear eye and other protections.
(Score: 2) by Spamalope on Friday November 29, @04:21PM
Young an poor, I was breadboarding a power supply to use a car CB radio inside.
The radioshack rectifier was labeled backwards! (printed on opposite side, so + was -)
The electrolytic caps exploding were... well, sounded like a gunshot and threw 1 bit of aluminum shrapnel across the room.
And then I had a 'but it's not my fault' scene with my friends mom...
(Score: 5, Informative) by VLM on Thursday November 28, @03:47PM (6 children)
There are other cap problems.
You can see they have 16V rated caps on a 12V rail and completely unsurprisingly at least one failed (well, one failed visually with corrosion, others may have failed open but not corroded, etc)
Vishay has a great "whitepaper" on the topic of wet electrolyte tantalum capacitor voltage derating. If there's ever a chance of applying voltage while the can's above 85C (not ambient... the temp inside the can) you need to derate the voltage by 2/3 up to 125C. Then you need to derate AGAIN for realistic ripple peaks maybe 2/3. So if there's any chance it might be run hot, lay in the sun, power up again, you need to derate to 2/3 of 2/3 or in other words those "15V" caps are really only suitable up to a 5V rail for long term reliability.
Of course, if you're designing a product you would like to resell again in a couple years, feel free to put "15V" caps on a 12V rail and enjoy the increase in replacement sales and repair services.
One final note about the dreaded liquid electrolyte tantalum. You'll hear people suggest replacing with MLCCs however MLCCs have a lower ESR than tantalum and SOMETIMES you'll get weird ringing or noise. Usually no big deal. But if you're doing something like an audio preamp, and you recap with MLCCs, you MIGHT experience ringing or oscillation or other weird stuff, strange alteration in the spectrum of noise, etc. ESR of a MLCC can be darn near close enough to zero whereas a typical tantalum might be "low single digit ohms" so its pretty easy to put a 1 ohm or 2.2 ohm SMD resistor in series with a MLCC in an analog circuit to replace an old tantalum. Probably a waste of time with digital work but I'm sure there's an audiophile out there who turned his record player preamp into an oscillator and wonders why.
Also look out for capacity loss in ceramic caps under high voltage. HILARIOUS to put a Y5V cap under high voltage and measure its capacitance as a tenth rated. I'd rather install bigger MLCC caps than have liq tantalum leak all over a board, but keep it in mind...
Tantalums were cool in the 80s but MLCCs are cooler in the 2020s
A 47 uF 15V tantalum can only reliably work long term up to 5 volts not on a 12 volt rail, and you can replace with a MLCC however 47 uF MLCCs are "single digit dollars cheap" but under voltage they act like as little as 4.7 uF sometimes so you actually have to replace the 47 uF tantalum with a 470 uF MLCC. And the essentially zero resistance of the MLCC might blow a fuse or something on inrush current.
I checked mouser because I'm bored and the guy in the blogpost should theoretically have replaced his tantalums with 25V 470uF MLCCs from Chemicon that unfortunately cost $33 each. Since the dawn of time boards have been wildly overspecd so a 47uF using X7R dielectric will probably work, not work as well as designed, but probably be good enough and chemicon sells those for less than $5 each and they're surface mount so can probably be bodged into place more easily.
Vishay has a nice MLCC whitepaper about X7R dielectric capacitance loss where after about 100 hours of operation it'll settle down to around 1/3 capacitance loss if you use nice expensive Vishay caps under ideal temps etc. So those "47 uF" caps will be 15 uF caps after a couple days of operation. And that's probably enough. Probably.
In summary its not as simple as match the specs and swap if you want perfect performance; but in a digital system not running at its limits you probably don't need perfect performance, and OP is probably out about a fiver for each dead cap if he replaces them with the proper-ish name brand MLCC. On the bright side MLCCs don't use liquid chemistry so they will never leak. They may crack or may short or may lose capacitance over time, but they won't leak.
So now I finished my tea and need to put a turkey in the oven as is tradition in these parts...
(Score: 5, Interesting) by VLM on Thursday November 28, @04:24PM (2 children)
I hate to reply to a reply but here's an interesting idea: Capacitor mfrs do NOT put RC constants on their cap data sheets but they probably should.
A higher voltage electrolytic cap (like in a switcher) might have an ESR around a couple ohms, so the time constant is well under a ms so hooked up to a power supply it'll try to charge itself in exactly one AC line voltage peak "relatively smoothly" its not a light dimming dead short it'll be "OK"
A MLCC with a ESR around 20 milliohms, probably rated at a lower voltage admittedly, will have a Tc measured in nanoseconds LOL so that might or might not pop a rectifier or fuse when it tries to charge up like a dead short across the line. Modern large MLCCs really are pretty much indistinguishable from a zero-ohm jumper 'resistor' for the first couple microseconds of operation, LOL.
A zero voltage electrolytic is a substantial but survivable overload for about one AC power cycle peak, no big deal, but a MLCC is essentially a dead short for a couple us which might blow something. Brand new device will probably be fine; for a fraction of a century old computer, well I'd be careful, it'll probably be OK but maybe not.
To give you an idea of the numbers involved, Vishay's datasheet for a 1N4001 claims it'll flow 30 amps for a single AC line peak. The electrolytic in my example can't draw more than maybe "ten" amps even instantaneously. A MLCC will draw whatever the PCB traces will support to a short circuit for a few microseconds; who knows maybe 50 amps. Will a 30 amp peak rectifier survive a 50 amp surge, yeah, probably for awhile. How long? Who knows. I wouldn't do it.
If I was really bored, which I'm not, I'd figure a way to make a tiny SMD hobby PCB as a product to precisely replicate the behavior of older components and sell them. Some collector/hobbyist types would buy them. Not "you can plug it in and sometimes it'll probably work most of the time well enough for awhile" but a precise replication of the component behavior of an olden days component. Even if I'd never want the hassle of selling this the intellectual challenge would be interesting. Could I make a "black box" that replicates all measured behavior of a brand new 1980s liquid tantalum, and if I can, can I make it small enough to be usable? That's a solid maybe. Fun to think about how I'd do it.
(Score: 2) by RS3 on Thursday November 28, @04:56PM
All really good info, thank you for posting. Keeper stuff- saved.
I'm sure you know that many people make various adapters and "interposers" and SMD circuits to plug in and replicate or even enhance what was in the original, usually through-hole parts. I've seen many in the audio world- better, quieter op-amps to replace older stuff. SMD only, but you use the SMD to through-hole adapter, maybe a few extra needed parts on the adapter.
Regarding the various current peaks, I'm sure you know, but for anyone else's sake, it's all about the resistances in the circuit- ESR, copper resistance, etc., and the parts' ability to absorb and dissipate the heat generated, and of course the time involved. A very short spike of very high current might not kill things, but a long lower one will just due to the time of heat flow. (Partial differential equations were not fun in EE school...)
Regarding tantalum working voltage: a few years ago I was involved in a project that required very long term reliability, like 40+ years. Yes people, electronics can be designed and built to last that long. Anyway, had to do quite a bit of research to find out that, very general rule of thumb- tantalums should be rated for at least twice the running voltage. I've had so many fail over the years, including ones in audio circuits (including microphones) that become noisy before fully failing.
In the case of some microphones, I discovered that lowering the "phantom" voltage from the normal 48 to like 24 and the mics worked again and were usably quiet. Looked inside and sure enough, they used 50V rated caps. Not fun finding 100V rated tantalums small enough to fit. Said mics weren't valuable, so project got shelved.
If I had a schematic, it might be easy to drop the 48 down to 24 before hitting the caps. Most phantom powered audio things work perfectly at 24 volts.
(Score: 3, Informative) by corey on Thursday November 28, @10:49PM
Yeah, I've worked on a rocket motor/squib ignitor system (not big rockets) that used a bunch of surface-mount electrolytic caps as energy storage that would dump their energy into the ignitor. But the charging of them was via a electromechanical relay. Although the relay output contacts were rated at many amps, we kept fusing the relay contacts shut and it took a bit to figure out what was going on because the auxiliary contact (which went back to a micro) was re-opening because it was on its own spring. But at t = 0, the caps were a short, as you say, but being electrolytics they still had some ESR of low ohms. However we had a bunch in parallel, so that's a bunch of resistors in parallel which brings the impedance seen by the relay right down. I think, from memory, to fix it I implemented a slow voltage ramp upon closing the relay, and switched the relay for a forced-contact (aka safety) type, which would force the aux contact closed if the power contacts welded. Worked well after that.
Regarding your earlier comments about Tantalum/MLCC, I now am doing work in the space industry so we pretty much only use MLCC for their small size and low outgassing (certainly wouldn't use liquid tantalums). We use NASA derating standards which typically spec 50% for caps. But it can be higher, our gear cycles around -30C (Earth shadow) to 20C (in sun) every 90 minutes. I've had to take a lot of notice of IC packages because QFNs start coming off the board with such temperature flexure, however if they're small it's not too bad. But yeah have had instances where manual placement guys (for prototype boards) have soldered caps in place of resistors and vice-versa. I've come to notice that caps are typically brown (Kemets anyway) and resistors black. But I had an instance where a 4-pin surface-mount inductor on a buck regulator was soldered 90-deg rotated (so it was shorted). No alignment markers, only the text orientation on the top told which way it should go in. This was using pick and place too. Took ages to figure that one out, I thought there was too much capacitance in the circuit and the regulator was hiccuping from it. After removing almost all capacitors, I started looking elsewhere and found the problem. I wish we had a X-ray machine.
(Score: 1, Interesting) by Anonymous Coward on Thursday November 28, @06:01PM (1 child)
"Wet" tantalum caps exist, but Apple didn't use them.
They are conventional electrolytics, and the tantalum deratings don't apply to them.
Tantalum came up in the posting because hobbyists are replacing the electrolytics with solid tantalums,
which DO have the derating and will destroy themselves if installed backwards.
C22 was installed backwards because whoever created the silk screen put the + marking on
the wrong end of the part.
(Score: 2) by Username on Thursday November 28, @07:58PM
>C22 was installed backwards because whoever created the silk screen put the + marking on
the wrong end of the part.
Eh, I think whoever made the CAD had it reversed and that's why the PCB guys had it, the SMT line put it that way, and the inspectors let it pass. Silkscreen never really matters when programming machines, it's all in the CAD file.
(Score: 2) by Username on Thursday November 28, @07:34PM
>Vishay has a great "whitepaper" on the topic of wet electrolyte tantalum capacitor voltage derating. If there's ever a chance of applying voltage while the can's above 85C (not ambient... the temp inside the can) you need to derate the voltage by 2/3 up to 125C. Then you need to derate AGAIN for realistic ripple peaks maybe 2/3. So if there's any chance it might be run hot, lay in the sun, power up again, you need to derate to 2/3 of 2/3 or in other words those "15V" caps are really only suitable up to a 5V rail for long term reliability.
I'm not a parts expert, but the caps in the photos in the article were always referred to at work as aluminum electrolytic capacitors, or cans. I'm not sure if they had tantalum in them? Tantalum caps were always square and yellow colored with brown markings on them. It is interesting to know how they degrade, never really thought about it, but it does make sense. I assume the device was EoL decades ago though.
>experience ringing or oscillation or other weird stuff, strange alteration in the spectrum of noise, etc. ESR of a MLCC can be darn near close enough to zero whereas a typical tantalum might be "low single digit ohms" so its pretty easy to put a 1 ohm or 2.2 ohm SMD resistor in series with a MLCC in an analog circuit to replace an old tantalum. Probably a waste of time with digital work but I'm sure there's an audiophile out there who turned his record player preamp into an oscillator and wonders why.
I assume it's because most ceramic caps are bidirectional, wouldn't it be easier just to put in a diode?
(Score: 2) by Username on Thursday November 28, @07:52PM (2 children)
Looks like this defect was let go since it did not negatively effect the function of the device, and looks like it lived well beyond it's expected life time.
The issue brought up is when a tant cap was used as a substitute, it acted as a resistor. I assume the resistance running backwards was very high and that's why it was pulling almost 2 amps. Probably gave a nice ouch when touched as well. When doing xray inspection aluminum can caps are not usually included, only the yellow tant caps. Tant caps have what is called a "pcap indicator slug" in them which shows up really well, and determines its orientation on the board. The can caps just have one lead which is wider that can be detected in xray, but all other internals cannot be seen like in tant caps. I assume this means, even though it has polarity, it being backwards wasn't as a big of worry as the tant caps.
I'm not familiar with ancient electronics and have no idea why they even need a dedicated -5v rail to begin with. Can't you just place whatever backwards on the 5v rail? IDK.
(Score: 2) by RS3 on Friday November 29, @05:37AM
WAY back, 1974, first generation Intel 8080 and many other logic circuits needed +5, +12, and -5. IIRC early RAM needed those 3 voltages too. It had to do with NMOS biasing. https://en.wikipedia.org/wiki/Intel_8080 [wikipedia.org]
(Score: 0) by Anonymous Coward on Friday November 29, @03:20PM
That works if you never need to connect the "0V" of the -5V parts of your circuit to the "0V" of the +5V parts of your circuit (as this would short out your power supply), and you take a lot of care to keep everything properly isolated. It's often a bad idea to have "0V" mean different things in different parts of the circuit as it can make the design very complicated.
TFA says the -5V was used for the RS-422 serial ports. Nowadays it is more typical to use single-supply driver ICs (these have their own internal circuitry to generate the required negative voltages locally). But in the past it was almost certainly seen as simpler/cheaper to just add an extra winding to the mains transformer for any required voltage levels, and -5V rails were frequently used for biasing so you might have needed it all over the place anyway so may as well use it for your serial ports too.
By the 1990s when this machine was produced it was probably a toss up either way, maybe even in favour of dedicated driver ICs, but portions of the machine design are likely much older, re-used from previous products.