Does the cable modem really need to be unplugged for 30 seconds? (Urban Myth or Truth?)

August 9, 2006 11:26 PM

Does the cable modem really need to be unplugged for 30 seconds? (Urban Myth or Truth?)

I need help researching the validity of what I suspect is just an urban myth. Basically it has to do with when you have a cable modem that isnt connecting to the internet and the cable company tech support instructs you to reset the modem by unplugging the power and leaving it unplugged for 30 (sometimes longer) seconds.

I think common sense would dictate that a fraction of a second after you pull the power plug form the device, it would turn off completley. There would be absolutely no difference between a split second later vs 30 seconds vs 2 years! The device will totally reset as soon as you plug it in again. Some people I spoke to argue that it takes time for power to drain out of all the circuits but that doesnt really make sense to me since electricity moves at superfast speeds...close to the speed of light.

What I think happened is that during the course of fixing a cable modem one day, someone ended up fixing it for some completley other reason which happened to coincide with it having been unplugged for 30 seconds and then ended up attributing that to the success of whatever they did.

Is there any creedence to this idea of leaving it unplugged for a set amount of time?

posted by postergeist to computers & internet (25 comments total)

Is it not more about the device at the other end of the line reassigning the IP? (total guess, that).
posted by pompomtom at 11:29 PM on August 9, 2006

They need to be unplugged for long enough for their short term memory to clear, presumably this depends on some capacitors somewhere. It would be a good idea to hold this info for short power losses to reduce the need of resets due to small power fluctuations.

This is very probably a shorter time on average than 30 seconds, but 30 seconds will make sure.

You want to find out how much the capacitors in a TV hold charge for, that is really surprising, maybe even hours.
posted by sien at 11:51 PM on August 9, 2006 [2 favorites]

pompomtom: I don't think the IP layer has anything to do with it. But it could also be about the other end dropping the cable connection entirely rather than trying to retrain or whatever.

But I think it's mostly about RAM in the cable modem getting reset.
posted by aubilenon at 12:05 AM on August 10, 2006

I think Pompomtom is on the right track. Certainly CMOS circuits can remain powered by capacitors for 2 or 3 seconds. Thirty seconds seems excessive for that. It is more likely that the long power-off period is so that the modem back at the cable company can detect the dropped connection and reset itself. When some types of modems connect, they have a handshake protocol in which they select modes and train for optimal speed. Perhaps they want to restart the training protocol and that requires 30 seconds for detection and initialization on the other end.
posted by JackFlash at 12:16 AM on August 10, 2006

Don't know about cable modems, but I can say ADSL CMUXs/DSLAMS will keep trying to re-establish the physical link for 10-15 seconds after the modem drops off, if it didn't send the "I'm going away now" signal.

Apart from that, I'd agree that it's probably to ensure the modem is properly reset. Simple test even a non-electrical-savvy person can do: count the number of small, low-power electronic devices in your house that take a second or two to extinguish the power LED when you unplug them.

Now, an LED typically draw ~20 milliamps, but memory often draws only a few microamps - 1000 times less. Does 10-30 seconds seem unreasonable "just to make sure" now?

(Yup, that's all very simplified - but basically how it is...)
posted by Pinback at 12:27 AM on August 10, 2006 [1 favorite]

sien has it. its the capacitors. I once asked the cable modem tech guy about it on the phone. the answer: capacitors. 30 seconds just makes sure. otherwise the power wont cycle fully and memory might not be erased.

same thing applies to lots of other kinds of electronic devices, by the way.
posted by jak68 at 12:31 AM on August 10, 2006

And the human factor. Tell the more impatient ones 5 seconds and they'll reset in 1, tell 'em 30 and you stand a chance of getting a good few seconds.
posted by scheptech at 12:43 AM on August 10, 2006

My ISP always tells me to unplug for 10 seconds.
posted by gfrobe at 1:32 AM on August 10, 2006

Yes, it's mostly for the power supply capacitors. You can sometimes see this if the power LED on a box dims over time instead of extinguishing immediately.

The other issue I think may be the protocols used over cable or DSL connections. The protocols must be tolerant to small blibs in the lines moving in the wind or whatever. You flipping the power might look like a transient line problem and the box will just resync with the existing settings instead of resetting properly.
posted by todbot at 2:29 AM on August 10, 2006

Cable modems are reasonably low power devices. They are also connected to TWO infrastructures, the cable/internet and the power grid. Resetting/reacquiring IP takes many seconds, and if the modem went off line every time that the power grid got a hiccup, you'd be highly inconvenienced. There is very likely an intentional design feature to bridge short term power interruptions, and the likely nature of this is capacitors. Since the modems are reasonably low power, in order to actually have the modem enter its 'power up reset' phase, it first has to run out of power.

Where I am, I have a my PC backed up with a UPS. I do not have my cable modem backed up. I get power hits all the time that make my UPS squeak, but never lose my ISP connection. Often, I get several hits in succesion. Never do I lose my connection. QED.

On the power up reset, there are several ways of providing this signal. The simplest is a series cap and resistor to V+. When the power is first connected, the node between the cap and resistor is temporarily at V+, and decays quickly to ground potential. It can be used to reset circuits designed for a positive going reset signal, such as 8051-class micros have. There are a lot of different reset strategies, and caps play a part in many. THey have to be discharged, usually by a large value resistor in parallel. This takes time.

These two factors plus a substantial safety interval are probably what recommends the 30 second interval. In practice, you could determine it experimentally for your individual modem by determining how long you can leave it unplugged before you lose your ISP connection.
posted by FauxScot at 2:57 AM on August 10, 2006

It probably has these things called capacitors. They need time to drain - and because may not know what type of modem you have, it's best to be generous in the amount of time to allow.
posted by tommorris at 3:47 AM on August 10, 2006

I think tommorris is wrong.

My guess is that it has something to do with capacitors.

(apologies for the cheap snark...)
posted by melorama at 4:27 AM on August 10, 2006

Back in the day when my cable stopped working on a weekly basis, I'd call the cable co. and would be told to hold the modem reset button for 5 seconds...or 10 seconds....or anything up to a minute. Basically different people told me different crap. I suppose it's a good way of keeping the user on the line without having to talk much, and since the help-desk people get paid by the hour...

My cable company has gotten much better, usually the modem stops responding once every few months, at which point I press the reset button as quickly as possible and that does the trick.

Of course, like others said it my depend on your modem. Why not try doing an increasing time interval to see how long it takes to make it start working?
posted by Sonic_Molson at 4:53 AM on August 10, 2006

The only capacitors of any size are in the power supply, and given that most modems have external PSUs, if you use the power button on the modem itself, it'll immediately disconnect the capacitors from memory.

So urban myth, not capacitors.
posted by cillit bang at 5:55 AM on August 10, 2006

I would say it's bunk as I've never waited thirty seconds to power cycle my cable modem. I will pull out the cord and plug it right back in and it works everytime. There's a couple of reason they might tell you to do this:

They are first-level support people who generally have less knowledge of how computers, networking and electronics work than the average cable-modem customer.

Newer computer systems have all kinds of nifty power-handling features. On most modern desktops you have to press and hold the power button for five second to get a machine to power off. This is by design, and perhaps they think it's necessary for everything electronic?

In other words, when you call your cable company's tech support, you're not usually dealing with the brightest bulbs in the electrical department. Since most problems that come into a first-level queue are generally caused by the keyboard-to-chair interface, they have a script that was written by some management comittee and they follow it exactly. Independant thought is punished, rather than encouraged, so myths like this propegate faster in these environments.
posted by Spoonman at 6:04 AM on August 10, 2006

I would vote for *both* power supply cap draindown *and* protocol timeouts, myself.

That said, I have some Lanier office all-in-ones (400lb copier/printer/fax's) at some clients, and I have seen 30 seconds *not* fix their problem, and a 2 minute (timed by the clock) powercycle do so, so YMMV.
posted by baylink at 6:47 AM on August 10, 2006

The 30 second number comes from a chain of people high-balling the number to make sure the next person gets it right. When I worked for a cable ISP, I'd tell the head of IT that the modems had to be unplugged for 5 seconds to reset them (a quick in-out sometimes left them in a weird state, and 5 seconds is a reasonable wait.) He would tell the customer service managers to unplug them for 10 seconds, to make sure they got at least the 5 seconds. The managers would tell the customer service reps to unplug for 30 seconds, to make sure they got at least the 10 seconds, and so on; I heard a rep at one point suggesting a full minute.

I think this is a natural reaction; each person assumes the number they heard is the bare minimum required for the reset to work, so they add some padding before they pass it along.
posted by pocams at 6:59 AM on August 10, 2006

Sonic_Molson writes "I suppose it's a good way of keeping the user on the line without having to talk much, and since the help-desk people get paid by the hour"

But they are evalutated by average call duration, the shorter the better.
posted by Mitheral at 7:03 AM on August 10, 2006

Last year, I had to unplug/replug mine at least five times a week and sometimes a lot more than that (thanks, RCN!). After a little experimenting, I almost never spent more than about two seconds with the power cut. It always went back to working fine afterwards (until the next day...or hour, if I was writing a paper).
posted by booksandlibretti at 7:30 AM on August 10, 2006

there are always bulk caps and high frequency caps right on any circuit board, even if the main power supply is separate. it takes a second or two for these to drain, but it can be more as there may not be a whole lot of static resistance between VDD and GND.

i've debugged boards in the lab where even 100mV on a power rail caused power-on reset chips to not actually reset the board, causing fpgas to not be configured, etc (= BADNESS). so it is best to wait a while. 30 seconds is probably excessive but pocams has it right; everyone adds padding to their numbers at each level.
posted by joeblough at 8:51 AM on August 10, 2006

The only capacitors of any size...
It's not simply the size of the capacitors, it's the length of the RC time constant. In a circuit containing resistance and capacitance, the voltage will decay exponentially as a function of the exponent (-t/RC). As you can see, large capacitance OR large resistance will cause a slower rate of decay.
posted by forrest at 3:46 PM on August 10, 2006

Some of these devices (notably the typical home router type things) also have a master reset button -- usually recessed that you must press with the end of a paper clip or something. The design typically requires that the reset button be held for some time period (5+ seconds) in order to prevent accidental erasure, after which some copy of firmware in ROM is copied to flash, or the entire flash is zeroed, or whatever "master reset" means for the device.

I'm willing to bet that the "unplug for 'n' seconds" meme grew out of an incorrect application of this knowledge -- that you must hold down the master reset button for 'n' seconds, therefore you must also unplug the device for 'n' seconds. From 'n' it increased to 30 for the reasons already pointed out in the thread.
posted by Rhomboid at 5:21 PM on August 10, 2006

As a former DSL and cable support drone I can tell you that our trainer told us it is for human engineering and electronic engineering reasons. We use 30 seconds because that is long enough for us to get that 5 seconds we really want for capacitors to clear and (in some devices) temparatures to change (cracked solder can be temp sensative failing when it gets warm but working when cool)
posted by Megafly at 6:18 PM on August 10, 2006

Wow.. i'm definitely convinced based on these responses.

Just as an aside..when I posted this question a few of us were just sitting around a room and the issue came up in conversation...I mentioned how I subscribe to ask.metafilter and within a few minutes of posting this at roughly 3am, there were already two responses. This site is fantastic I've gotta say!

Thanks for all those who responded.
posted by postergeist at 10:41 PM on August 10, 2006

Mitheral: Not always.

When i worked for AT&T Wireless, I'd tell customers to power their phones off for up to 30 seconds. Usually it was because people were really impatient, and that 30 seconds gave me a chance to catch up on my notes, drink some water, scratch my head, etc.

(off topic: dunno how it is now, but back then AT&T Wireless didn't care about how long our calls were. They just wanted the CSRs to resolve the issues, not freak out over the clock.)
posted by drstein at 1:26 PM on August 14, 2006