If you ever connected to the Internet before the 2000s, you probably remember that it made a peculiar sound. But despite becoming so familiar, it remained a mystery for most of us. What do these sounds mean?
(The audio was captured by William Termini on his iMac G3.)
As many already know, what you're hearing is often called a handshake, the start of a telephone conversation between two modems. The modems are trying to find a common language and determine the weaknesses of the telephone channel originally meant for human speech.
Below is a spectrogram of the handshake audio. I've labeled some signals according to which party transmitted them, and also put a concise explanation below.
The first thing we hear in this example is a dial tone, the same tone you would hear when picking up your landline phone. The modem now knows it's connected to a phone line and can dial a number. The number is signaled to the network using Dual-Tone Multi-Frequency signaling, or DTMF, the same sounds a telephone makes when dialing a number.
The remote modem answers with a distinct tone that our calling modem can recognize. They then exchange short bursts of binary data to assess what kind of protocol is appropriate. This is called a V.8 bis transaction.
Now the modems must address the problem of echo suppression. When humans talk, only one of them is usually talking while the other one listens. The telephone network exploits this fact and temporarily silences the return channel to suppress any confusing echoes of the talker's own voice.
Modems don't like this at all, as they can very well talk at the same time (it's called full-duplex). The answering modem now puts on a special answer tone that will disable any echo suppression circuits on the line. The tone also has periodic "snaps" (180° phase transitions) that aim to disable yet another type of circuit called echo canceller.
Now the modems will list their supported modulation modes and try to find one that both know. They also probe the line with test tones to see how it responds to tones of different frequencies, and how much it attenuates the signal. They exchange their test results and decide a speed that is suitable for the line.
After this, the modems will go to scrambled data. They put their data through a special scrambling formula before transmission to make its power distribution more even and to make sure there are no patterns that are suboptimal for transfer. They listen to each other sending a series of binary 1's and adjust their equalizers to optimally shape the incoming signal.
Soon after this, the modem speaker will go silent and data can be put through the connection.
(UPDATE 02/2013: Due to numerous requests, I made this into a 42-megapixel poster that Redbubble is selling. Some $4 per poster is directed to the poor nerd who made this.)

Thank you for illuminating one of the mysteries of my past. Very interesting stuff.
ReplyDeleteYou're welcome! Thanks for reading!
Deletehmmm, maybe dialup could speed up with higher processing modems, since homes phones are going to be manufactured as conference phones,think i planned that when i invented dialup and broadband 96
DeleteNo, the speed of modems has been capped at the sample rate of the digital phone infrastructure. I don't know what the other stuff in your post was supposed to mean.
DeleteGreat read, thanks.
Deleteno, actually, the telephone lines had to actually be thickened which means the small telephone line could actually give you more bandwith than you think, of course however the signal strength of the phone services has something to do with it as well, remember the modem processes a certain amount of data, (decrypts(receives), and encodes(sends),information at a certain rate, which means the low 56k modem is not taking full advantage of the simple telephone network
DeleteIt really doesn't work that way, at least because the telephone network is no longer simple. (Note: US numbers below, others may vary...)
DeleteOnce the signal hits the telco switch, it becomes 64kbps digital data. You can't squeeze more than 64kbps through a 64kbps pipe, no matter how far equipment improves.
Getting ~56k through this mixed analog/digital ordeal is/was quite a feat.
(That said, it's still possible to order an ISDN phone line with a couple of honest-to-goodness 64k digital channels and the ability to call other ISDN phones elsewhere in the world, but why would you want to?)
Thank you for this amazing amount of effort to explain that handshake sequence! I can't say it's been keeping me up at night, but the detailed diagram showing each part of the handshake sequence is beautiful!
DeleteI still use dialup everyday to support my Unix clients
DeleteAfter hearing this I had to go scour the internet to find the AOL "Welcome...You've Got Mail" sound!
DeleteTo the previous Anonymous posters arguing about what actually capped throughput -- you're *BOTH* right. The PSTN caps audio sampling at a rate which limits 56Kbps throughput (T1s only sample at 64K 8-bit samples/s), *AND* cabling needed to be enhanced in order to prevent mutual cross-talk between adjacent lines in a trunk in order to get the SNR high enough to be compatible with an 8-bit sample size. ADDITIONALLY to that, the head-end of the service you were dialing needed to be pure digital -- that is, an analog 56Kbps modem could never achieve a perfect 56Kbps connection with another analog modem -- the signal degradation was just too much. It _had_ to be analog at the customer premises, and digital at the head-end. Otherwise, the most you could expect to achieve was closer to 40kbps.
DeleteTo anonymous #2, you're right in that normal PSTN is capable of a lot more throughput than most people realize. This realization is what gave birth to the DSL industry. However, look at the changes telco's needed to make internally before this technology could be rolled out to the curb, and how expensive it was to deploy.
It's so odd... I used to hate this sound when I got my first PC. I used to try and find ways to either turn it down or turn it off altogether... of course, you couldn't. Not really. But as time went on, I got used to it. Kind of like when you sign on to your computer and the OS greets you with a noise or a song. It was like the Pavlov's Dogs, and I got used to listening for those sounds. You could actually tell if your connection was going to be good or bad (ie, none at all) by the way the sound was modulating.
DeleteThat sound, to me, is like the sound of an old school typewriter. It's primitive, it's annoying, but I do so miss that kind of feed-back in the here and now! Classic stuff.
&ATL0
DeleteJason George: Turning off that sound was not only possible but was relatively easy. The fact you think you couldn't means you simply didn't know what you were doing at the time, which was common :) The command to turn it off was
DeleteATM0
simple :)
In regards to Samuel, ISDN is used fairly frequently in broadcast installations since the end result is higher quality then IP codecs most of the time.
DeleteYou make modems sound so polite. Please this, please that. ;)
ReplyDeleteMust be because the drivers were written in INTERCAL.
DeleteI'm going to have to start hard coding please and thank you into my software from now on...
ReplyDeleteThank you so much for this post, I have always wondered.
ReplyDeleteA walk down tech memory lane... this time with a guide :)
ReplyDeleteThanks for this smile.
So, would you like to sell a poster of that? Because I'd like to buy it.
ReplyDeleteQuite possibly :)
DeleteI would buy it too.
DeleteI would also be interested.
DeleteMe also
DeletePlease notify me as well! I want this poster!
DeleteT-shirt, make the dial string 231337
DeleteJoshua: Me too on poster. What about direct or thru Think Geek?
DeleteRedBubble is selling it now.
DeleteIt would be really cool if you could post a larger version of your diagram. The text is difficult to read at 100% zoom with the current dimensions.
ReplyDeleteMany sites have limitations on image size. >:( But the one at Dropbox is pretty large. Apparently it can be viewed in full by clicking the "Download" button.
DeleteSo interesting! Thanks for the insight into this forgotten part of the Internet!
ReplyDeleteThe spectrum diagram is missing parts of the audio file: the diagram doesn't contain the funky sounds starting in the audio file at 21.5 second. Oona, can you also include explanations in the diagram?
ReplyDeleteThe scrambled data looks very uninteresting in the spectral domain; it sounds and looks like white noise. I also left it out to save space. But perhaps!
DeleteWhat does normal dail-up data look like? Does it have a pattern or does it just look like static?
DeleteHi Oona again! I don't mean the white noise part at around 0:24. I mean the sounds just before that which go like "nnnnnjjjuuuuuuiiieeeeek pling"
DeleteThat part missing part is the V.90 56k testing, looking for see is the far end is a digitally connected modem, allowing up to 56k (well 64k, theoretically) downstream. Would love to see the spectral footprint of that and explanation as well.
DeleteThe part where the whole spectrum is full is probably some actual digital data exchange in the V.42/V.42bis packets framing and compression, and the request to test for V.90 capability (which triggers the "pling" sound) Turns out there are three major variants of that sound, X2, K56 and V.90. I should dust off some old rockwell and USR modems and record the pre-standard sounds. It was a crazy time leading up to the V.90 standard. It was a big deal prior to the spread of broadband.
This is a great graph, I'd love to see it as a poster.
But why can't the modem be silent from the handshake itself ?
ReplyDeleteOf course I could tell if it's really going to connect by hearing the noise but still...
Modems can be entirely silent, it all depends on how you initialize the modem and the settings you choose to use. By default, the speaker is generally on during handshake, so you can hear if there are any problems. Changing the configuration to disable to speaker is just a matter of changing the speaker setting in your initialization string.
Deleteatm - speaker off
Deleteatm1 - speaker on until carrier detected
atm2 - speaker always on
atm3 - spkr on until carrier detected except during dialing
Some very talented people can do useful diagnostics by listening to the noises, but I think the speaker mainly stays on so you can hear things like busy signals, failure to answer, and wrong numbers (i.e. if a human picks up.)
DeleteIt is also useful if you dial a wrong number. You know something is wrong when your modem starts speaking to you :)
DeleteBeautiful. Absolutely, fantastically, nostalgically beautiful!
ReplyDeleteThanks for this article, I really enjoyed it and the accompanying image!
ReplyDeleteNostalgic post of the year!
ReplyDeleteI thought i was listening to Skrillex.. Oh well..
ReplyDelete"When humans talk, only one of them is usually talking while the other one listens." Tell that to my wife and mother-in-law.
ReplyDeleteYeah, it's not universally true culture-wise, thus "usually"
DeleteThat's all I ever wanted to know.
ReplyDeleteWhat program did you use to generate the spectrogram?
ReplyDeleteBaudline by SigBlips. Also Gimp and other image postprocessing.
DeleteI thought I recognized that particular shade of baudline-green. (:
DeleteWhat an awesome signal analysis job you have done. You broke it down, demodulated parts, decoded bits, and understood a crazy handshake protocol. It must of been a lot of work. I am beyond impressed. Wow!
Thank you! It was a fascinating journey indeed.
DeleteVery nice! Great explanation and visualization of the handshake.
ReplyDeleteI'm curious as to what software you use to generate the spectrogram?
The only software I've toyed with so far has been baudline and it's not bad but I've never gotten the quality or resolution I've hoped for out of it.
Thanks!
Looks like you answered someone else asking the exact same question just as I hit submit. Thanks for that.
DeleteAny other tools you're fond of? Resources for anyone who might be interested in learning a bit about signals at home?
It's always a trade-off between frequency resolution and time resolution. With just the right windowing and FFT length you can do stuff like this. But a lot of compositing was still needed.
DeleteI suggest reading The Scientist and Engineer's Guide to
DeleteDigital Signal Processing http://www.dspguide.com/pdfbook.htm
Oona, do you think it's at all possible to create a spectrogram for DSL modems? They're not audible, but they've a handshake process too at... well, above 25KHz.
ReplyDeleteA spectrogram would probably look quite uninteresting; I'm guessing they utilize the bandwidth very efficiently, which means the data will look just like white noise.
Deleteif your from the era of that sound then you had a great childhood.
ReplyDeleteVery cool!
ReplyDeleteNice article. Finally i can see that irritating chrxxsdxxsadxsaxasx piiip ppppiiiiiip nxaskxnalskncasc!
ReplyDeleteThanks for making this. I really enjoyed being able to match the "names to faces," so to speak. I made a simple jsbin version of the spectrograph with a "playhead" overlaid so you can see what is being played when, and click around as well. http://jsbin.com/amunug/17
ReplyDeleteVery cool mashup, just what I was hoping for! Thanks for sharing!
DeleteModem A: hey babe, you dtmf?
ReplyDeleteModem B: u know it
Modem A: what u up 4 2nite? wanna v.8?
Modem B: i wanna ack u like my daddy net2phone use 2 ack me
Modem A: um ok... v.8 then
Modem B: lol jk, u comin?
Modem A: brt just gotta turn off echo suppressors n cancellers
Modem B: ok i wait
Modem B: my pcm is so modulated
Modem A: lol rly? u think u can handle V.90/V.92?
Modem B: D/A?
Modem A: ...D?
Modem B: wtf no, im not into that
Modem A: lol jk we can do V.42 LAPM if u want im down 4 nething
Modem A: up to 3429 o/c
Modem A: u know i give as good as i get, ne way u want it, loud or soft, high or low, fast or slow, i got all the time in the world 4 u babe, my clock source is internal
Modem B: of course no 3429. and same 4 me. except i might lose track of time, lol
Modem B: and honey if u with me we gon be makin sum NOISE
Modem B: 6db at LEAST u know how i like it
Modem A: lol i hear ya, 3200 all nite long, the way u get me goin maybe we even go 2 4800 lol
Modem A: set ur pre-emphasis filter params n put on that 1920 hz carrier frequency i got u
Modem A: im here baby
[SCRAMBLED]
-ultimat142
http://www.reddit.com/r/programming/comments/17jwoi/dialup_handshake_explained/c86evsf
This is golden.
DeleteAwesome.
DeleteGiggled my ass off!
Deleteunbelievable :D
DeleteVery cool! I'm wondering, did you just look up the conversation data or did you use something to sniff the data off of the modem line?
ReplyDeleteI wrote a few Perl scripts to demodulate the FSK and DPSK, then manually decoded the bits based on the relevant ITU-T standards.
DeleteGreat work! Can I trade my brain for yours? ( It's only very slightly used )
DeleteMy brain isn't that great :D
DeleteIs your code available somewhere?
DeleteUnfortunately not. The scripts aren't useful per se.
DeleteExceptionally cool. Thank you for the work and sharing this!
ReplyDeleteVery nice work! Takes me back to the 90s when I first started using dialup :-)
ReplyDeleteTakes me back to the mid-80s when I could distinguish connect speed by ear (at least 1200, 2400, and 9600). Ahh, the days of listening to Telemate auto-redial the busy BBS number. ;-)
DeleteThank goodness we no longer need to worry about line noi230t9uc!%!f3333
NO CARRIER
Awesome. I remember working phone support in the 90's for dialup internet users. I would sometimes have people hold their phone up to the modem so I could listen to the handshake, but I never knew exactly what the various parts were for. Thanks!
ReplyDeleteHas someone the V.flex tones as well?
ReplyDelete(You analysed V.90. That was the most common standard.)
Here is what V.flex sounded like:
http://030.co/vflex
Actually, It it was really a USR modem, that would have to be X2.. The competing Lucent/Rockwell K56plus/flex was more like this one: http://www.dialupsound.com/
DeleteThere's a distinct difference between the one posted above and the one linked by you.
DeleteListen to it carefully: The V.Flex has a sound that is like "gong strokes".
Hi,
ReplyDeletegreat! Thanks for sharing this. What about doing something similar for DSL line handshaking:
"The ITU-T has recently standardized a handshake and activation method for xDSL modems. This method is
contained in the ITU-T Recommendation G.944.1 (formerly known as G.hs). Several of the features of
G.994.1 came from the voiceband handshake Recommendations V.8 and V.8bis,"
Very nice! Thank you for sharing.
ReplyDeleteThank you Oona from a mother of an Ouma - that was fun to read
ReplyDeleteThis sounds like US signaling. We had totally different in Europe and former USSR.
ReplyDeleteIt is indeed a U.S. dialtone.
DeleteThanks for making me feel old...
ReplyDeleteawesome
ReplyDeleteThanks for such a nostalgic read! I think I just played the sound clip about 20 times. I would love to print a large format poster of the spectrogram with all your notes and put it up in my office. Would you mind sending me a high-res version that would print really clear? A vector version if you have it is best?
ReplyDeleteThanks in advance!
Tony
You just got BoingBoing'd... stand by for bandwidth uplift....
ReplyDeleteAny chance of you posting a higher resolution version of this? I'd love to do a wallchart of it.
ReplyDeleteDear Oona Räisänen
ReplyDeleteyour lastname must sound a bit like the krrrBliipkrraaa sounds of my good old modem of mine carefully chucked in a box in the cellar.
But what an interesting and great article !
thank you !
Edualc
Thank you, Oona. This brought back fond memories.
ReplyDeleteMy first modem use was in 1979 with 300 Bps and then in 1981 with 1200 Bps. In the late 1980s I used a 2400 Bps modem for the first time and loved it. The move to actual Kbps speeds and then digital modems, and then routers were all just normal parts of an evolving network. But I realized I missed the modems' sounds and control/handshake sequences, which allowed an engineer to more closely follow and understand the data transmission process.
We've coma a long way, and we should capture this kind of thing to remind us of the journey.
Again, immense thanks.
My pleasure!
DeleteMuch the same, until in 1999 I was trouble shooting faxes and modems, mostly by ear, although I have access to some great services which would provide the customer printouts.
DeleteNeat read... thanks!
ReplyDeleteJust wanted to say thanks, very informative!
ReplyDeleteWould love to have that without the explanation, it'd make an awesome wallpaper!
ReplyDeleteCan you link me to that please!!
You can use any audio software to make a custom spectrogram from the sound file :)
DeleteBeautiful. Thank you!
ReplyDeleteI remember when I got my first 2400 baud modem @ an amiga show in NY, it was between that or a digiview video digitizer. Best purchase I ever made. I also remember when my friend bought a 14.4 baud modem, those days it was crazy fast! ah the memories BBSing late into the night, of board games and war games dialers.. those were the days..
ReplyDeleteYeah, 33.6 was a superhighway compared to 2.4. I couldn't have imagined text flowing so fast.
DeleteJust absofreakingly beautiful... thanks for lifting the myth and especially for providing the tone, I was looking for a sample all over the web...
ReplyDeleteYou're welcome :) The tone was not recorded by me though, it's from Wikimedia Commons.
DeleteTerrific! The infographic is great - those conversation 'balloons' really help to understand what's happening. Thank you!
ReplyDeleteI have created a moving-pictures version of the above here:
ReplyDeletehttp://youtu.be/qEPIoyYB8hc
Thanks for the inspiration.
Great! Also there's this one http://www.youtube.com/watch?v=dSxgCVc6wcM
DeleteFabulous resource Oona, many thanks. Have you considered doing the Ethernet Auto-Negotiate version? (Good starting-point for a poster series!)
ReplyDeleteThanks for the idea! I'll definitely see into it. Won't be a spectrogram though :)
Deletethanks for share...
ReplyDeleteFound this from a post on imgur. I actually use this same sound clip as my morning alarm haha. I alaways wanted to know what it all meant. Thanks formaking this. :)
ReplyDeleteImgurians unite!
DeleteJust read about this on HAD, really interesting stuff. I actually have this as my ringtone. Leads to a few strange looks but also gets some smiles from the people old enough to remember dialup internet. :) . I really wish there were a lot more girl geeks out there. Thank-you for a really enjoyable read. :) Jon W
ReplyDeleteThanks for reading! I also had it as a ringtone for some time, gave me some true old-skool street cred.
DeleteThis is a really good read for me. I just found your website a few days ago and I have been reading through it regularly. Thanks a lot for enjoying this beauty blog with me
ReplyDeleteThank you, made me happy!
DeleteI recently came across your blog and have been reading along. I thought I would leave my first comment. I don't know what to say except that I have enjoyed reading. Nice blog. I will keep visiting this blog very often.
ReplyDelete