Let’s start the New Year with a tale of redemption. We have two bedrooms in our condo and in each bedroom you will find what eBay might call a “vintage” LED clock. One is a Timex T133T electronic desk clock with a jumbo 7-segment LED display and a trio of “Nature Sounds” for its never-used alarm. The other is a Radio Shack Chronomatic-290 AM/FM desk clock radio. Both clocks were purchased at some indeterminate time in the distant past (it looks like the Radio Shack clock radio is from the mid 1990s based on a fax-back document I found online) and both recently started to run noticeably fast. In fact, both clocks started gaining minutes per day. Now this is a pretty strange failure for plug-in digital clocks that need do nothing more than count power-line cycles to keep accurate time.
Because of the coincident failures, I suspected that someone in the condo had started to do something funny with the power line in the building. Perhaps they were running power-line Ethernet signals that caused enough EMI to fool the counting circuits in the clock. With that suspicion, I dug out an old EMI extension strip that I’d cobbled together using an industrial EMI power-line filter, a line cord, two deep-drawn aluminum enclosure halves, a duplex outlet, and a 6-outlet power-tap expander. It was easier to start with this approach rather than ripping into the clocks. I know that I built this EMI unit in the 1970s because it has a unique push-on, push-off illuminated power switch that I could only have gotten from lab stock while working as a lab engineer at Hewlett-Packard’s Desktop Computer Division in Loveland, Colorado.
I decided to tackle the Timex clock first. I plugged the filtered power strip into the wall outlet, plugged the Timex desk clock into the filtered power, and set the time. Then I waited a day. The Timex kept on ticking all right, but it was still ticking fast. Wrong guess.
The next step was to pull the Timex apart and look for obvious signs of component failure. I pulled an unbelievable nine screws to open the case of the Timex and was pleasantly surprised to find an isolation transformer, which meant that troubleshooting was going to be a lot safer than I’d feared. Then I took a closer look at the innards. The Timex was a cabling nightmare! No way could it be built these days for any reasonable amount of money. Too much hand assembly.
There was a main circuit board with a clock chip on it, a small circuit board for the snooze bar, another small circuit board with three signal diodes and the time-setting push buttons, and yet another small circuit board with slide switches to set the alarm sounds (birds, ocean, and brook) and to dim the display (which is why we bought this particular clock model). There was also a small daughter card soldered directly to the slide-switch board at a right angle, which appeared to carry nothing more than the birds/ocean/brook sound chip. The sound board was assembled using chip-on-board technology and the sound chip lay under a lump of epoxy.
All of these circuit boards except for the snooze button and sound daughter card were joined by hand-soldered, solid-wire ribbon cables. The main board also joined to the large LED display with yet another hand-soldered, solid-wire ribbon cable. The three boards and the LED display were all bolted in place with more than a dozen screws of at least two sizes. The amount of hand-soldering required to make this clock startled my eyes, which have become acclimated to 21st-century electronic assembly techniques. It’s becoming harder and harder to remember that we once built everything this way.
Unfortunately, there were no visibly obvious component failures. I first looked for swollen electrolytic capacitors because there was a rash of bad, counterfeit capacitors right during the time these clocks would have been made. The telltale indicator of electrolytic failure is a bulging or burst top but nothing appeared amiss.
Time to take one more step up the troubleshooting ladder. It was time to get serious.
The Timex main board had but a single integrated circuit on it so that had to be the clock chip. What a strange chip! It was a 28-pin through-hole DIP (dual-inline package) package but the pins were not on the standard 0.10-inch centers. They were closer together than that. A shrunken DIP! Unfortunately, the clock chip’s part number was obscured with some sort of brown crust that looked like solder flux. A wipe down with an alcohol-soaked Q-Tip swab took care of that problem and the part number was revealed. It was an LM8560. Google told me that an LM8560 is a digital alarm clock chip and then coughed up an 8-page data sheet with some handy reference-design schematics.
Lordy, this is a PMOS clock chip! I haven’t seen such a beast in decades. It was a good thing I was using a line filter from the 1970s because that’s precisely where you will find PMOS—back in the distant past. PMOS was big in the heyday of calculator chips and that was four decades ago. It was the original process technology for LSI devices. PMOS was supplanted by NMOS and then CMOS back in the 1980s, yet hobbyists still seem to be using this clock chip judging from what Google dredged up on the Internet.
With no obvious physical problems in sight, I had no choice but to break out my Fluke Scopemeter 97 portable scope. Naturally, it wouldn’t turn on. I opened the Scopemeter’s battery compartment and found that the NiMH batteries I’d installed a few years ago had started to salt up. Nasty. I cleaned out the batteries and the battery box and left it empty. Then I plugged the Scopemeter’s power brick back into the wall and old ’97 finally powered up. It was time for some signal probing.
A check of the clock chip’s clock input showed some reasonable—if ugly—60-cycle clocking. (At 60Hz, your clock can be plenty ugly and still work just fine.) According to the LM8560 data sheet, the power supply for this device is anywhere from -7.5V to -14V. A check of the negative power supply rail showed about -12V with some ripple—actually a bit more noise than I’d expect to see even for this simple, unregulated power supply. No doubt the 470µF, 16V electrolytic power-supply filter capacitor had started to fail. The capacitance had probably fallen and the ESR probably jumped up, but running fast seemed a funny symptom for a slightly noisy power supply rail. Perhaps the clock chip was reading the power-supply ripple as clock pulses or it might be taking the odd power supply level as a signal to use its free-running, uncalibrated 900Hz RC oscillator, which allows the clock to keep time during a brief power failure in conjunction with a 9V backup battery.
Whichever it was, my next step was to get a replacement electrolytic capacitor and install it. There are only two retail establishments where you can buy an electrolytic capacitor in 21st-century Silicon Valley on a late Sunday afternoon: Fry’s Electronics and Radio Shack. I checked on the Web and quickly found out that Radio Shack’s in-store capacitor inventory isn’t what it once was. I could order a 470µF radial capacitor from Radio Shack central and get it in a few days (I can do that on eBay, Amazon, Digi-Key, Jameco, and Element14 too) but I couldn’t drop by and purchase a 470µF capacitor at any local Radio Shack that evening. Apparently, there are now fewer capacitors in Radio Shack stores to make room for more cell phones. That’s probably a smart move on the part of Radio Shack’s management. After all, how many people still visit Radio Shack to purchase electronic components, even here in Silicon Valley? Answer: Not many.
So the choice on a Sunday evening narrowed down to just one: Fry’s Electronics.
Twenty minutes later, my wife and I were at the Mayan-themed Fry’s on Brokaw Road—every Fry’s has a theme (Mayan, Egyptian, Old West, Museum, etc.)—and we found ourselves looking at a somewhat spotty collection of blister-packed electrolytic capacitors hanging on pegs and arranged in no discernible order (like by capacitance value, which would have made a lot of sense, at least to me). By getting down on my hands and knees with my head near the floor to look at the bottom row of pegs, I finally found a radial 470µF, 25V electrolytic capacitor. The original electrolytic capacitor’s 16V rating seemed a bit low to me considering the -12V supply voltage. I usually derate electrolytic capacitor working voltages by 50% to prevent premature failures, but then I’ve never designed consumer products so I haven’t had to pinch pennies in a design. Dollars, yes. Pennies, no. I plucked the $1.69 capacitor from its peg and got back up off of the floor. We made our big purchase for the evening and returned home.
I installed the new electrolytic capacitor (checking the polarity twice), folded it over and packed it on top of two other capacitors just as the original capacitor had been positioned. I then buttoned up the Timex desk clock by reinstalling about two dozen screws in the display, various circuit boards, and case parts and then plugged the clock in for a smoke test. The clock was still ticking (figuratively, not literally) so I hadn’t let the magic PMOS smoke escape during the repair. Twenty-four hours later, the Timex was showing the right time. One problem solved; one to go.
Propelled by my first success, I cracked open the Radio Shack clock radio (only four screws this time instead of the nine in the Timex!), did a visual inspection, and got a two unexpected surprises. First, there was the very same 28-pin shrink DIP that I’d seen in the Timex clock. Yep, another PMOS LM8560 clock chip. This was too easy! The second surprise was the use of a real electronic connector to connect the switches in the top of the clock radio’s case to the circuit board. Usually in this class of product, you’ll see direct point-to-point wiring and hand soldering. Connectors generally increase the BOM (bill of materials) cost, which is a problem in many consumer-class products.
To make it even easier for me, I spotted a congealed black puddle next to the 1000µF, 16V filter capacitor that looked very much like cooked-out electrolyte to me. Consequently, I didn’t feel the need to confirm my analysis of the problem with the Scopemeter this time because I had a pretty good idea of what the problem was based on my experience with the Timex.
I rechecked Radio Shack’s Web site and, as it so happens, 1000µF electrolytic capacitors can still be sourced at your friendly neighborhood store. I bought a new 1000µF, 35V capacitor at Radio Shack for another $1.69 during lunch time the next day (appropriate for a Radio Shack clock radio, no?) and took it home after work. Certainly, I could have found a less expensive electrolytic capacitor in one of the endless cardboard parts bins at Halted Specialties just off Central Avenue in Santa Clara but gasoline costs being what they are, it was cheaper, faster, and easier to visit the Shack just down Union Street in South San Jose.
After dinner that evening, I again opened the Radio Shack Chronomatic-290 clock radio, pulled the switch cable loose, and unbolted the circuit board from the lower case. After that, it took me just a couple of minutes to desolder the old 1000µF capacitor and solder in the new one. Another few minutes and I had the clock radio reassembled. I plugged in the clock radio and it came to life. It seems that it too had retained all of its magic smoke. Twelve hours later, with sufficient time to gauge the timekeeping accuracy, I knew I had a second successful fix.
You might well ask why I spent so much time resuscitating two old LED clocks. They could easily be replaced with LCD versions for less than $20 each and my time is worth money (at least in theory). However, clocks with LED displays are increasingly rare and increasingly expensive and we prefer them in some places for high-contrast readability. Also, there’s some satisfaction in keeping yet another two pieces of “lead-tainted” equipment out of the world’s immense waste stream—at least for another few years. Besides, it’s an easy and interesting way for me to be green.
Sadly, we are increasingly unable to repair such household electronic items. That is the price we pay for the rapid uptick in capabilities we can deliver using high-density components and semiconductors fabricated using advanced lithographies. Neither of my 1990s-era clocks contained surface-mount components. Had they been newer, I might not have repaired either of them because the replacement components just can’t be sourced locally (or anywhere at all for components greater than a certain age) or because the components are shrinking to a size that can’t be easily handled with my pre-SMT-era Weller soldering station and my ancient solder sucker.
For another take on this topic, see Ira Feldman’s post on capacitors, warranties, and responsibilities.