Fail Of The Week: Teddy Top And Fourteen Fails

Last summer, [Quinn] made the trip out to KansasFest, the annual Apple II convention in Kansas City, MO. There, she picked up the most modern Apple II system that wasn’t an architecturally weird IIGS: she lugged home an Apple IIc+, a weird little machine that looks like an old-school laptop without a screen.

Not content with letting an old computer just sit on a shelf looking pretty, [Quinn] is working on a project called the Teddy Top. ‘Teddy’ was one of the code names for the Apple IIc, and although add-ons to turn this book-sized computer into something like a laptop existed in the 80s, these solutions have not withstood the test of time. [Quinn] is building her own clamshell addition to her IIc+, and somehow failing at something she’s done hundreds of times before.

While the IIc+ has an NTSC composite output, the super-special video add-ons for the IIc+ used a DB15 expansion connector. Here, any add-on could access video sync signals, the a sound signal from the audio circuit, and even a +12V line that could drive loads up to 300 mA. It just so happened the display [Quinn] is using for this project runs at 12V, 200 mA. Everything was great, but as a worthy trustee of this computer’s Earthly existence, [Quinn] thought a bit of current limiting should be included in her addon. She designed a circuit around an NPN power transistor, that would allow the display to draw power until the load was around 250mA. After that, the transistor would start dumping excess power as heat. Yes, a fuse would be better. [Quinn] calls this Fail #1. There are thirteen more to go.

BridgeBoard_brd

With the circuit working on a breadboard, [Quinn] needed something a little more permanent to stick in her Teddy Top. It’s a simple circuit and something that would be a good fit for protoboard, but [Quinn] decided to make a PCB. [Quinn]’s home PCB manufacturing is her specialty, able to put 16 mil traces on a board with 5 mil separation. That’s good for a Chinese board house, let alone someone in her kitchen.

The first board neglected to take into consideration how the input header would map to the DB-15 IIc+ expansion port. Another fail.

After so many successes with her PCB process, [Quinn] was apparently getting complacent about the errors that could creep in. The developer didn’t work, and trying to be scientific about it, she played with the developer time, UV exposure time, and copper etch time. Everything failed until she realized the developer should be diluted to 10:1. We’re up to fail #8 here.

With a PCB properly etched, [Quinn] got to the point where she could see her circuit etched in copper. She drilled the board, began the assembly, and realized the board was a mirror image of what it should be. Correcting this somehow produced another mirror image, and triple checking everything a third time produced another inverted image. Put some text on your copper layer in Eagle, people.

With a working board, it was time to assemble the circuit. While the circuit worked on a breadboard, the copper traces had just enough resistance to throw off the experimentally determined resistor values. While the current limiter circuit would keep everything below 250 mA on the breadboard, the limit on the PCB circuit, with the slight increase in resistance in the copper traces, shot up to 310 mA, above what the IIc+ could provide. That’s fail #13.

The final fail, and something [Quinn] should have caught in the breadboard stage, was flipping the collector and emitter on the power transistor. The idea of running video and power through a current limiter was invalid all along.

At this point [Quinn] decided to cut her losses. The LCD board only draws 200mA anyway, and she opted to simply wire it directly to the IIc directly. If the IIc needs protection, a 250mA fuse will be thrown in. Simple enough, and not as infuriating.

With the fixed circuit, the prototype for the Teddy Top works perfectly: a screen, powered by the expansion port on the IIc, is a working display. No extra wires necessary. At the very least, it’s an exercise in minimalism: don’t do what you don’t need to do. If you decide against that little bit of advice, at least you’ll learn something in the process.


2013-09-05-Hackaday-Fail-tips-tileFail of the Week is a Hackaday column which runs every Thursday. Help keep the fun rolling by writing about your past failures and sending us a link to the story — or sending in links to fail write ups you find in your Internet travels.

72 thoughts on “Fail Of The Week: Teddy Top And Fourteen Fails

      1. There’s a series on youtube called extra credits, they cover various video game development topics, and one of my favorite episodes was “fail faster” basically it was the idea of working towards the most failures you could in a given period of time because every failure teaches you something and makes the final product better, fail faster, save money, ship better products, FAIL FASTER!

    1. I like reading about Quinn’s projects. She’s only got two more things to learn:
      Trumpet your Triumphs!
      (bury your failures)

      That way, everyone thinks you’re a genius.
      Thanks for helping us learn, Quinn :-)

  1. “She drilled the board, began the assembly, and realized the board was a mirror image of what it should be. Correcting this somehow produced another mirror image, and triple checking everything a third time produced another inverted image” – I had to laugh… been there.. done that… the look of confusion and the sudden realisation that despite my ego telling me otherwise, I quite obviously am an idiot… yup.. all familiar ground I’m afraid.

    1. Anyone who done etching at home, flipped, rotated or mirrored the mask and stared at it like an idiot :). You should have seen me at my first double sided board, I printed the damn masks 10 times. Anyway, good job Quinn!

  2. A fuse is *not* the right solution – not when the line’s current limit is 300 mA and the expected operating current is 200 mA. 50 mA of headroom? That fuse won’t protect the circuit. It’ll just blow after a while or when it gets hot enough inside whatever case it’s in. A fuse’s working current should usually be around half of its trip current, or more if you need rapid cut (a non-slow blow fuse).

    A power MOSFET is the right solution, but you want a hot-swap controller to go with it so that when the current limit is exceeded you just turn off the FET. It’d be a bit more expensive, but with that little headroom you really need something active.

    If it were a lower voltage, or higher current limit, it wouldn’t even be that expensive (thank USB for that).

      1. Pat is right (though I failed to see it at first too). And a polyfuse won’t be of much help:
        200mA PF has a trip current around 500mA.

        With such low headroom you need an active circuit, monitoring and reacting quickly to shut it down.

      2. It’s a similar problem. If you think about it, fuses are a material that have dramatic changes in resistance around a temperature point. Polyfuses are actually pretty much the same thing – except now the change is (mostly) nondestructive, because it’s basically bunch of metal balls in a polymer that expands when heated. But it’s still a physical change, so it’s slow and the trip point is really heat rather than actually current. Which means that most polyfuses need a similar headroom.

        In actuality, it’s even worse – polyfuses usually have much, much worse temperature characteristics. Over a ‘normal’ temperature range for electronics (0-70 C) that trip point will vary by 50% or more. (It’d also be a pain in the ass to debug, because it’s non-destructive : the thing would just stop working, with no indication that there’s a problem, and then you take it apart, and it works fine… cuz it’s got airflow and cooler).

  3. I write software and every few years this happens to me. The nice thing about writing software is that as long as you keep a confident expression on your face noone can tell you’re in the middle of a 20 car pileup. Then you can go have a coffee, discard your changes, and revert to the previous revision. It would be nice to have Git handle revisions of reality.

  4. I once etched a nearly perfect thru-hole board, drilled it straight for once, only to discover that it was mirrored when I attempted to stuff the first IC. I actually considered soldering the components on the bottom side, before filing the board in the “circular file” and starting over. Now I am a bit obsessive about triple checking that the damn thing is printed correctly before transferring any toner to copper. Now I mostly just make errors in design rather than printing. I tend to say “What the hell was I thinking???” a great deal when reviewing my first run of prototypes after etching and stuffing them.

    On the positive side, I’ve gotten very good with an Xacto knife and spider wire.

      1. I do something similar for SMT boards. It’s way too easy to screw up a custom footprint. I’ve been burned by stock footprints in KiCAD before as well. I print the board out at 1:1 scale and start laying down components on the paper as a gut check. (At least one component per footprint type.)
        The problem I had with my thru-hole board was that it was scaled correctly, so I failed to notice that the pin out was reversed, because the components lined up perfectly (just on the wrong side of the board). Live and learn.

    1. I was tempted to say: “put text on it, so you know you’ve got it the right way round”

      Then I thought: “but I can read mirrored text almost as well as I can read right-way text”
      and realized that my clever solution was no good at all.

      1. In my case it wasn’t my design, but in hind sight I guess I could’ve edited the image and thrown some text on there. I can read mirrored text fairly well, but I’d hopefully notice it. (Probably depending on what time it is and exactly how sleep deprived I am.)

  5. Wasn’t clear from the summary, but this isn’t using the raw video signals from the DB-15 but just the copy of the NTSC video signal that’s available there.

    Does anyone know of any good boards / ways to get a modern signal out of that port. All I can find are vague references to no-longer-made converters.

    1. The Apple IIc’s DB-15 video connector was a strange beast. It had many raw signals, although I don’t think (this is 30-year-old memory) it included NTSC.

      The one commercial product was the Apple LCDisplay, 80 characters x 12 lines, no backlight and fairly low contrast. But it did make the IIc “portable”. At least once I was in the passenger seat of the car writing an article using that display. (cira 1987).

      The Apple IIc and LCDisplay were featured in the 1984 movie “2010: Odyssey Two” with Heywood Floyd using the computer on the beach (with no apparent power supply). That was before the displays were available in retail stores, and many folks wondered whether or not the actual display would be available by 2010!

  6. I can sympathize with your current limiting related pain right now.

    Built a nice board last weekend, draws ~50mA typical. Normally I’d include a PCB-mount GMA fuseholder and fast-blow fuse of several times required load, of which I have plenty on hand. But I was running short on space, so I went out of my way to order some nice 250mA fast-blow SMD fuses instead; then forged ahead assuming they’d work. Well, it seems these SMD fuses are a *lot* faster than GMA equivalents. I’m liberal with filter and bypass caps, and merely charging up 100uF of capacitance on power-up blows the SMD fuse every time. Fail! I’m not sure how to fix it yet, and for now, the board runs with a jumper in place of the fuse..

    I’m obviously no fuse expert. But for [Quinn]’s situation of limiting to 300mA when typical load is 66% of that, I’m not sure any fuse would behave precisely enough. [Pat] mentioned a MOSFET and I’m inclined to agree. Along with a good current sensor. Allegro makes some nice Hall-effect current sensors. Being magnetically isolated from the load, they’re super easy to use for current measurement – high-side, low-side, HV, AC/DC, doesn’t matter. Some also have adjustable averaging time and a latching overcurrent comparator built-in, specifically lending themselves to e-fuse use. They’re a bit pricey, and a shortcut I would normally not prefer to take (you learn little from using a dedicated IC). But for making sure a piece of vintage gear was well protected with a minimum of head-scratching, I might consider it.

    1. I came up with a soft start circuit for large bulk capacitor to meet USB specs for a charger design. What this design is different than traditional soft start is that the current limit only applies to the bulk cap during initial charging up and the sense/charge resistor is bypassed in steady state to eliminate additional the voltage drops.

      A resistor is used to charge a cap, the resistor value is chosen large enough to limit the charge current. As the cap is near full charge, the current tapers off. When voltage drop across the charging resistor falls below 0.7V, a MOSFET connects the cap directly to the power rail bypassing the resistor. That final step will cause a small surge, but it can be designed to happen after 100ms and below limits, which by then the USB would have been configured for maximum 500mA current drain.

      https://tekkieneet.wordpress.com/2013/07/25/battery-analyzer-part-2-of-n-usb-gotcha/

      1. I believe that is the preferred solution to over current protection in some parts of the world… I’ve taken apart a few ebay power adapters, and the contents are always “interesting”…. I’m going to stick with my polyfuse suggestion however.

    2. For these currents you really just want a sense resistor. The Allegro and Melexis Hall effect sensors typically need big currents to get measurements, like amps or more.

      Most hotswap controllers work with around 50 mV or so range, and I can’t imagine that the screen cares about a 50 mV drop.

      The eFuse chips from TI are actually super-awesome for this kind of thing. Current monitoring, everything integrated, and protection out the wazoo – reverse voltage, undervoltage, overvoltage, overcurrent – all for about $1-2. In single quantity. (Thank the proliferation of high-current hot-plug devices, like SATA external disks). Sadly, their current limits are too high for this (usually min. 2 A or so).

      1. Too low voltage – this is a 12V rail (and it costs way under $2, where are you getting them??). 5V current limit/protection chips are ridiculously cheap, thanks to USB devices. I’m always amazed that I don’t see them in practically every hobbyist device.

        12V+ stuff is harder to find.

          1. For 12V, the fancy new eFuse chips (TPS25940) from TI are really nice. Have to get them through TI right now, but they’re not too expensive (~$2-3). It’ll be nice if those get out to distributors eventually, they should be sub-$1 in large quantity.

  7. On a machining forum I frequent, someone posted about a screwup they’d made on a high pressure bulkhead fitting for a deep submersible ROV.

    When they called to inform the client about the error, the reply was that the person assembling that part had drilled the bulkhead through from the wrong side, thus the mirrored hole pattern on the fitting fit perfectly.

    So they just changed the plans to match the goof. Error? What error?

  8. my co-worker’s ex-wife makes $64 /hour on the laptop . She has been unemployed for eight months but last month her pay was $20500 just working on the laptop for a few hours. browse around this website F­ox-Review­s­.­com

      1. by swearing i was swearing at the ********* advertising troll who thinks we believe him that if we ckick some link our money problems will be solved for life. go insult some sheep on some other site, we dont fall for that on this site

  9. While proofing my ability to use a CNC to manufacture a board (my new favorite way but I’ve done toner transfer too), I started going through a ton of copper clad (fortunately I reused the old clad until I got the orientation and alignment right)…then hit upon the idea of using cardboard to mimic the intended circuit board. Works great. For toner transfer the plain paper pinout mentioned above is probably the best idea.

  10. @tekkieneet: How can you be contacted? I’ve been on your WordPress blog and left a comment on your About page with my email address, and would appreciate if you could get back to me. I’d like to talk about an engineering project with you.

  11. it’s only a failure if nobody learns from it.

    getting it right first time is great, but if you get it right first time every time it’s probably a sign that you;re not stretching yourself enough.

    I did the whole mirrored board thing when I was in school, with 16 3″ x 2″ boards I put my stencil in the UV box upside down, for a stereo mixer, time and money didn’t allow doing it all again, so I wound up making the best of a bad situation by bending out the legs on through hole components and surface mounting them instead.

    Funnily enough I learned that it’s a good idea to put a model number or name on the board then, that way you can immediately see if the stencil is upside down. a stupid mistake that I’ll never make again.

  12. The final fail was not quite as described. It wasn’t swapping the emitter and collector of the NPN transistor. The circuit was a standard constant-current circuit, which (with an NPN) requires a load connected between Vcc and collector.By grounding the video part of the interface, she ended up with the load simply connected to Vcc and Gnd. The transistor was by-passed by the ground connection. One way to fix it would be to use a PNP bipolar transistor and connect the load to collector and ground (PNP’s emitter to Vcc).

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.