Dynamic spectrogram of dial-up modem handshake sounds
February 10, 2016 9:37 AM   Subscribe

Short-but-pretty SLYT. So that's what was going on ...

Some good comments too.
posted by carter (25 comments total) 27 users marked this as a favorite
 
The thrilling details.
posted by theodolite at 9:45 AM on February 10, 2016 [10 favorites]


The first time I heard that sound, I couldn't believe it. In a way, I still can't believe it.
posted by WalkingAround at 10:04 AM on February 10, 2016 [1 favorite]


I remember that tune!
posted by jim in austin at 10:04 AM on February 10, 2016


The sound of my youth.
posted by fordiebianco at 10:08 AM on February 10, 2016


This is why I love science. Thanks!
posted by Captain Chesapeake at 10:09 AM on February 10, 2016


I'm not wistfull for dialup speeds, but that sound represented THE FUTURE once, and it still thrills me a little bit. The thing I hated was the busy signal at AOL.
posted by Devils Rancher at 10:33 AM on February 10, 2016 [1 favorite]


The thrilling details.

Neato.

I always wondered what the "Pssssshhhhhhhhhhhhhhhhhhhhhhhhhhhhh" meant.

But shouldn't DTMF be the termination of the session, not the beginning?

Thanks folks, I'm here all week.
posted by mandolin conspiracy at 10:35 AM on February 10, 2016 [2 favorites]


But shouldn't DTMF be the termination of the session, not the beginning?

You're thinking of the second revision of the standard, DTMFA.
posted by Talez at 10:42 AM on February 10, 2016 [6 favorites]




This is why I love science engineering.

FTFY
posted by Confess, Fletch at 11:18 AM on February 10, 2016 [2 favorites]


Obligatory.
posted by BigHeartedGuy at 11:49 AM on February 10, 2016


The reason there is this whole sequence of tones is that older standards could only do so much. V.21 signaling (mostly Bell 103) could handle full duplex at 300bps. Bell 202 and V.23 defined 1200 baud. Okay, how do you determine when you make the call that your using V.21 or V.23?

V.8 -- the sequence of tones at the beginning culminating in a 300bps data burst from both ends that says "Hey, I can do this." The two endpoints then pick the best one for the job.

Well, then we get V.32bis, which allows connections *up to* 14,400bps. Unlike V.21 or V.23, you could connect at a number of speeds. So, the second phase of the connection comes up, which is the two endpoints saying "Let's try V.32bis -- let's see what the connection is like and determine how fast we can go."

As we get faster rates (and get close to the maximum data rate for a POTS circuit) we found that the tests that V.32bis were doing weren't enough, so there's a second set of tests using random noise. That's phase three, and figures out how close to 33.6kbps you could get, and then the V.90 sweep starts, which tests 56Kbps. In almost all cases, the calling modem starts, and the answering modem either replies-in-kind (and they figure out what to do) or doesn't, in which case, they drop back to the last agreed signaling and make the connect. Also in there, but hard to see because it's very quick, is requests for things like error correction and compression standards like V.42bis, but they don't really show up in the sync, they're negotiated after the sync and can be switched on-or-off during the connection, but the process is very similar -- the caller sends a block saying "Here's what I can do," the answering modem sends a block with "Here's what I can do" and they try to switch to the best common system.

So, basically, the modems are having a chat. Let's listen in!

Calling Modem (CM): (calls receiver)
Answering Modem (AM): (answers, sends the V.8 tone saying "I'm a modem")
CM: Oh good, I'm a modem too! (sends V.8 answer) (raises DCD on serial port)
CM: Here are my capabilities. I can also do full V.8 (send V.8bis info block.)
CM: (repeat 5 times in case the line is noisy.)
AM: Hey, let me tell you what I support (sends V.8bis info block.) V.8 is good! (raises DCD on serial port)
AM: (repeat 5 times in case of line noise.)
CM: Okay, ending V.8bis, entering V.8
AM: (disables echo guard) I'm a'looking for some V.8!
CM: How about a V.8! Here's what I can do blah blah blah
AM: Have some V.8 as well! Here's what I can do blah blah.
CM: (at 1200) Hey, I can deal with these frequencies. Let's see how good the line is!
AM: (at 1200) Hey, I can deal with these frequencies. Let's see how good the line is!
CM: (Sends tones across the spectrum)
AM: (Sends tones across the spectrum)
CM: Hey! I can accept the following symbol rates at the following frequencies (lists) Oh, and don't get any quieter than- Xdb less than you are now and I'll be able to hear you.
AM: Hey! I can accept the following symbol rates at the following frequencies (lists) and don't get any quieter at all. Please use carrier frequency X, I'll be using frequency Y, please send a noise burst to confirm line quality.
CM: (Sends noise burst)
AM: (Sends noise burst)
CM: (switches to 4800-9600bps) I can talk this fast, I can hear you up to Xbps. BTW I can do V.90
AM: (switches to 4800-9600bps) I can talk this fast, but you can't hear that so I'll limit to Y. I can hear you to X, so we're good there.
AM: V.90? Hey, let's try that. Sending hi-res tone sweep now. (Sends hi-res tone sweep)
CM: Sending hi-res tone sweep now. (Sends hi-res tone sweep)
AM: Hey! V.90 Works! I can accept the following symbol rates at the following frequencies (lists). Please use carrier frequency X, I'll be using frequency Y, let's sync at Xbps.
CM: Hey, V.90 Works for me too! I am syncing at Y baud rate. Let's call it a connection.
(speaker turns off.)
CM: Hey, now that we're in sync, I can do the following error correction systems blah blah.
AM: Nice. Please used error correction X. (sends sync)
CM: (Implements error correction X) Reliable bits are reliable! I can also use the following data compression modes blah blah blah.
AM: Awesome sauce! Please use compression mode X (sends compression X header)
CM: (Sends compression answer, enables compression.)
AM: (enables compression)
CM: Great, going to tell the system we have a connection here. (raises DSR)
AM: Nice talking to you! (raises DSR)
User: Downloads low res porn.

Basically, at the point where one end doesn't answer a request, the process then bails out and they work at the last speed they negotiated. That's why things kept falling back to 300baud, 1200baud or 9600 baud -- because those were already negotiated. 14.4kbps and 28.8kbps used to appear where V.90 appears in this example, but V.90 can negotiate those rates as well as higher rates, and looks like V.32.bis, so if you can only do 14.4, you just respond old-style with that and the first modem goes 'Okay, 14.4kbps it is!)
posted by eriko at 12:11 PM on February 10, 2016 [23 favorites]


That is a fantastic visualization. Loved the two-tone DTMF codes.
posted by GuyZero at 12:43 PM on February 10, 2016


This is why I love applied science.
posted by Captain Chesapeake at 12:46 PM on February 10, 2016


I do not miss calling a number and getting a modem signal. How do you say, "I'm not a modem" in modem?
posted by Splunge at 1:01 PM on February 10, 2016 [1 favorite]


I was peripherally (ho ho) involved with the latter stages of modem development, and boy was there a lot of politics between different groups trying to get their standards adopted. In the early 90s, the market went crazy from a combination of interconnected factors as Moore's Law opened things up - home PCs were popular and capable, online access hit critical mass with services like AOL and Compuserve (and, later, proper ISPs), and better digital signal processing drove the ability to do higher bandwidths down analogue phone lines. Modems stopped being uber-nerd boxes of weird, and started to become domestic necessities.

The absolute end point was 64kps, not because that was how much you could actually push through copper lines (ADSL showed that) but because the phone companies used 64kbps PCM internally (analogue phone lines were really nothing of the sort) and that was the hard limit.

For various reasons, the fastest practical POTS speed was around 56k, and there were two groups offering this technology, one led by US Robotics with its x2 spec, and one by Rockwell/Lucent with K56 Flex.

Modems from the two camps could interoperate at lower speeds, because all the stuff mentioned upthread about negotiating the highest attainable mutual speed was enshrined in international standards, but not at full speed - and with dial-up, speed was king.

Normally, each new boost in modem speed happened when a new standard was agreed - you could buy non-standard modem before then, but the market in general waited to be sure that what they were getting would work in general, would be around for a while, and that there would be multiple suppliers who could interoperate reliably. Modem manufacturers tended to have patent portfolios which contained stuff important to the final standard, which had to mix and match between IP owners, so the standards process was half getting the engineering right and half smoothing over the business of coaxing competing companies to share their IP. In general, people understood that standards created markets and the horse-trading was reasonable.

At 56k, the normal ITU business of creating a standard got bogged down by cash-crazed warfare, US Robotics attempted to jump the gun by launching its stuff and trying to pre-empt the market, which meant the other group had to do the same.

Part of the deal with ITU standards is that if you own some IP in a standard, you have to let anybody license it at a reasonable rate. No ITU standard, you kept full control of your IP, and neither group would let anyone have a license if they already had one from the other group. So you could not make a 56k modem that was guaranteed to work with all other 56k modems at 56k. Which rather negated the point of buying one.

This led to such nonsense as early ISPs having to maintain two separate banks of modems with two different sets of dial-up numbers, although both camps had very generous (or free) offers for service providers in an attempt to get the upper hand. Users, if they wanted to be sure to be able to connect to more than one service, had to buy two modems and somehow work out how to switch between them. Not trivial in the world of Windows 95, PC serial port management and differing AT command set implementations.

Users wanted very little to do with this, so it really mucked the market up - by the time agreement was reached in 1997, it was estimated that the war had held back the market to a third of it potential. And, by the time it was all sorted out, ADSL was on its way (ISDN having helpfully being held under water in a pillowcase by the telcos, who figured that keeping people on slower links was just gravy for them, while cheap universal digital connectivity would eat away at their very profitable leased-line businesses) and a lot of money just never got made, a lot of users were pissed off, and online had bogged down for a while.

Which also means some other nice things that were planned never happened - for example, it had been spotted that you could start passing data between users at a very early stage in the training and negotiation sequence, albeit at a slow rate, and that this was exactly what you needed for the normally slow, low-bandwidth business of logging on and asking for a particular service. So why wait for the fuil connection? But by the time everyone was talking to each other again, the glory days were over and the money started to go away.

As did US Robotics, Hayes, Rockwell and many others. Let that be a lesson about regulation, the free market, IP and progress, kids.
posted by Devonian at 1:40 PM on February 10, 2016 [12 favorites]


All that extra negotiation takes time, too. This means that a 1200 baud modem* connecting to another 1200 baud modem is significantly faster, since the host end doesn't bother asking about more advanced capabilities.

* Yes, I had one. Yes, it was slow, even back then. It took fifteen minutes to transfer a 100kb file at 1200 baud! A 14.4k modem, the fastest available for a good while, took only about one minute to transfer the same amount. I know the bitrate isn't quite 15X faster, but that was my experience. I think data compression helped a little bit.
posted by neckro23 at 1:51 PM on February 10, 2016 [1 favorite]


Credit card machines that are (were) on dialup used 300 bps I believe because the negotiation time vastly outweighed transmission time, even at slow speeds.
posted by GuyZero at 2:05 PM on February 10, 2016 [2 favorites]


Devonian: For various reasons, the fastest practical POTS speed was around 56k, and there were two groups offering this technology, one led by US Robotics with its x2 spec, and one by Rockwell/Lucent with K56 Flex.

Timewarp!

It's been so long since I paid attention dial up modem technology, but hearing 2x and Flex was a flashback, which lead me to thinking of shotgun modems, the legend and lore of my geeky high school friends. Imagine the potential!

... home PCs were popular and capable ...

Except some weren't. We actually had a PC that was the bottleneck for our limited internet experience, before we upgraded to the Gateway monster that played Ultima [something - it was a friend's copy] at lightspeed. Travel by ship, which was once felt like it was a real-time endeavor, now was comically fast! (And that was the first PC we named, which required looking in my parents' book collection, as finding inspiration on the internet was less of an option for us then. Orpheus, you were a good computer for your time.)
posted by filthy light thief at 3:05 PM on February 10, 2016


This is good
posted by asok at 3:45 PM on February 10, 2016


Ah, the good ole days of logging on to local bulletin boards to download a pic or play a text game. I pleasantly reminisce until I remember the constant dropped connections resulting in swearing and throwing things.

I was a 2400 baud modem user.
posted by Muncle at 4:27 PM on February 10, 2016


How do you say, "I'm not a modem" in modem?

ATH. H for Hang up.
posted by scalefree at 4:59 PM on February 10, 2016


So that's what was going on ...

Was?
posted by Thorzdad at 5:09 PM on February 10, 2016


I loved that sound circa..1993. I was a little late to the game, but damn that brings back great memories.
posted by Benway at 5:44 PM on February 10, 2016


How do you say, "I'm not a modem" in modem?
I used to whistle. That at least got it to shut up.

The first modem I ever saw (around 1973) was a Bell 103. About the size of a Betamax.
(Oh, you probably don't remember what a Betamax was.)
I got out of teleprocessing just before the 19.2KB modems appeared. Most of what I worked on was Bi-Sync full-duplex at 4800 or 9600.
QAM on a leased line was bizarre enough, but how did they get 56KB async to work on unconditioned lines?
posted by MtDewd at 12:53 PM on February 11, 2016


« Older Sexual harassment in science redux: now with...   |   On Female Fuccboi Style Newer »


This thread has been archived and is closed to new comments