I think the practice originates with the first digital clocks. (This should be verified.) Such clocks stored each digit separately (a practice called Binary-Coded Decimal). If you want to set a snooze alarm in 9 minutes, you just have to remember one digit: the current number of minutes minus one. If the clock sees that again, it sounds the alarm. For a larger snooze time, say 10 minutes, you need to remember two digits. That is more complicated and ultimately more expensive.
Because nearly all digital alarm clocks use one of a small handful of control chips that were designed decades ago to allow very cheap and simple construction. A few contact switches, power, a display, a tinny beeper… and you had nearly every alarm clock from the $10 one to the luxe designer clock-radio-CD-TV ones.
At their core is an unbelievably primitive digital logic circuit made around the limitations of simplicity and low cost of about 1978.
There is little demand for actually “better” function from buyers, so this same old, basic design persists, wrapped in ever-updated wrappers.
There are a few alarm clocks out there with more sophisticated chips that have easier time setting, more flexible snooze options, ability to store multiple settings without manually setting each change, etc. But they’re rare because people buy for every reason except core features.
Why yes… (Looking at that diagram with all the comprehension of a chimpanzee looking at a typical Formula One steering wheel). Why that’s obviously correct. That binary coding of those decimals explains everything!
Actually, it’s the best explanation I’ve heard. I’ve always figured that once the digital clock was prefected (I mean without those white on black flipping digits), companies just copied what other companies did, and the chips used in the 1970s and 1980s are probably still in use today.
Interestingly, the iPhone – something that has a completely different chipset, and is mainly software driven, still uses that old, unchangeable 9 minute snooze setting.
I have a nice Boston Acoustic tabletop radio that has dual alarms. While it’s not perfect, it was good enough to get me to set aside Project Perfect Alarm Clock (which would/will use a programmable microprocessor as its core).
For example - to set time, press the TIME button and rotate the front knob. To set an alarm, press ALARM A or ALARM B and rotate the front knob. To set the snooze time… press a combination of buttons and rotate the front knob. (Mine is set at a nice, neat 10 minutes; it can go from 1 to 99, IIRC.)
Still a lot of room for improvement, but this is also the ~$100 unit that sounds every bit as good as the 5-6X Bose Wave.
Maybe I’ll finish that project one of these days. Alarm clocks just shouldn’t be so shitty.
Im an electronics engineer working in Silicon Valley. I looked up the MM5370 datasheet that Cecil referenced in this column. The datasheet clearly says that the snooze button gives you from 8-9 minutes before the alarm sounds again. The datasheet doesn’t give an exact schematic but does give a block diagram of the circuit.
The 60hz (60 cycle per second) sine wave from your AC line goes through a pulse shaper to make a 60 times a second square wave, then into a prescaler ( a digital frequency divider ) to generate a one per second pulse (1pps) then into a divide by 60 digital counter to make a one per minute pulse (1ppm). That 1 ppm pulse goes around the clock circuitry to time different things, including the “Sleep Down Counter”
When you press the Snooze button, a circuit called the “Alarm/Sleep circuits” is activated, silencing the alarm until the “Sleep Down Counter” times out and the alarm is reactivated. The most efficient type of counter in terms of transistors used (transistors being a valuable commodity in 1970’s Chip design) is a binary counter, and a 3 stage binary counter would give a delay of 8 ppm pulses, a 4 stage counter would give a delay of 16 ppm pulses (2^3 and 2^4 counts respectively) so the 8 minute count is the most desirable.
Why, though is the actual snooze variable between 8 and 9 minutes? Because you don’t know when you hit the snooze button. If you hit it just after the 1 pulse per minute pulse has occurred, you have to wait 1 whole minute for the 8 minute timer to start, if you hit it just before the 1 pulse per minute pulse, the 8 minute timer starts right away. And that explains exactly why the clock has a snooze timer of from 8 to 9 minutes.
Now the reason for snooze buttons on mechanical clocks being around 9 minutes has something to do with the gear mechanisms and I don’t know nothing about that. But the idea is the same, efficiently mapping a desired snooze time of around 10 minutes to the hardware limitations of the time.
Having said that, here in the year 2015, I don’t think I have ever had a snooze button that didn’t just work on 10 minute increments. Times have changed, transistors are essentially free on modern chips, and I’m sure its no big deal to set your snooze timer down to the millisecond. Welcome to the 21st century, hope you enjoyed the history lesson.
I am not sure anyone’s ever bothered to make an updated alarm clock chip. They’ve been repackaged, etc. but I think clocks on Target’s shelf are using the same 3 or 4 chips that were first released around 1980 or so. Jellybean stuff like the 555 and its variants.
Of all the postulated answers in Cecil’s column, #6 was the closest to correct, the only mistake being that the respondent assumed the snooze counter counted in seconds, which would have been pretty inefficient for timing things on a minute scale. If he had guessed there was a minute based timer in the clock he would have been right.
Re: the “Engineer’s comment: Nice try, bub, but clocks don’t count that way.”
I’d have to say that must not have been a very experienced engineer.
If an engineer proposed using a 555 timer in a production alarm clock circuit, he would get laughed out of the room. They are inaccurate, short duration, multipurpose timing circuits, not appropriate for use in a clock.
Mass produced clocks since the 60’s have relied on ASIC’s.
I didn’t say anyone used a triple five. I said the chips still used were likely those designed thirty years ago or more. I doubt any chipmaker has done more than trivial updates to accommodate new manufacturing or packaging preferences; there is no economic motivation to design a new chip from scratch when makers are satisfied with the limited functionality of what passed for newfangled to Ronald Reagan.
Oh, and kid - most clocks made before the mid 1970s used “motors.” Including the newfangled “digital” kind. I doubt any kind of mass-market clock chip existed much before then, and certainly weren’t common until the early 1980s. I had several examples of very advanced digital clocks before then. The ones that didn’t use little flip-numbers actually used a motor to drive a code switch wheel to make pseudo-digital number segments light up.
I never noticed this column before. I have been reading the Straight Dope since '80. I was 13 in 1977 (I am 50 now {who called me a kid? Thanks}) and even then I had the Texas Instruments Datasheets that I just looked up. I would have easily been able to answer this one in '77 or '80. Oh wait the column was from '99, bit more of an archaeology discussion than engineering then, big LOL.
Later models of Alarm Clock IC, which had lower power consumption and battery backup featers, like the XX8560. Seem to use the same 8-9 minute snooze configuration circuit. But nothing stops people from simply throwing a microcontroller at the circuit these days. All bets would then be off.
Were very expensive at that time (Late 60’s early 70’s) I swiped a couple when the Apollo project was winding down and my local science museum got some of their surplus (I grew up in Florida). They required complicated high voltages and were really only used in test equipment. It would make a lot more sense to make a 7 segment display out of mechanical components or so-called “grain of wheat” incandescent bulbs. 7 Segment LED’s were pretty common and less expensive than Nixies in the early 70’s but the mechanical clocks with the Slo-Syn shaded pole motors turning huge numbers of flip cards were still the commonest way to make a digital clock up until the mid 70’s.
You gotta realize that a nixie takes a lot of power, high voltage and a separate heater. Their main advantage is long life and quick response
About 15 years ago I made a plasma Union Jack display for fun. That required some logic, but the voltages required were less onerous than Nixies (They worked just like 100volt Neons) It was a trade off between voltage requirements and logic requirement. Nixies were never something that I saw in low end electronics back then.
All my research seems to indicate that, even in modern cheap alarm clocks. They still use a 1 minute timebase, and a divide by 8 binary counter, which generates a pseudorandom 8-9 minute delay when you punch the snooze button. Mystery solved.
I first saw nixies in 1965 on an early digital VOM that was being used as an alternate (to an oscilloscope) readout for a plugboard general-purpose analog computer, so I naturally think of them as components of components, and therefore (in my imagination) cheap. They weren’t, huh? (I’m software, not hardware, and I never again dealt with an analog computer after that summer, so it’s all kinda dreamlike now.)
I can’t say for certain, as I left around 600 pounds of chip references and the like behind in California - and I don’t think any of the standard data sheets get far enough into internal logic - but from my experience working with not only those clock chips but many others of the same general design era, some of which I knew down to the transistor structure… The snooze circuit in all or most of them is based on a BCD countdown timer. BCD is a limited form of binary that only counts from 0 to 9, making it marginally more useful in the real world than binary (and once programming was more common, hex) numbering.
When you already have a chip full of BCD logic, an adjunct circuit that counts down from 9 is a trivial addition. I can assure you that no one designing chips in that era was thinking on any sophisticated plane about the sociological/neurological advantages of a psuedorandom 8-9 minute interval.
It’s possible that the countdown was slaved to the condition of the 1’s digit, meaning it could vary from 8 minutes to 9, or from 9 minutes to 10 depending on the exact seconds count at the moment the snooze button was pressed, but I think it far more likely it was an independent circuit BCD’ing its way to zero from 9. And thus nine minutes, with useless precision in the matter.
I am only guessing about a few slight details here - these chips and their many kin are not historical objects to me. I worked with them on a bench, with a wrench, holding them in a clench, and occasionally needing a quench when an error made them burst into flame.
