Nix McRetro
The real Avtek MegaPlus V.32bis CD950 dialling 8218 3400. Eight digits, two frequencies for each one.
There’s a particular sound from the dial-up internet days that’s been stuck in my head for more than twenty years. Not the modem handshake itself, although I remember that rather well too. It’s the telephone number being dialled before the connection even gets started.
I can still sing it. Well, I can sing what I think it sounded like, which is an entirely different problem. Rather than leave it alone, I decided to see whether the melody might tell me which Sydney dial-up number I was hearing back in the late 1990s and early 2000s. This became quite the rabbit hole.
Two Numbers, One ISP
I remember using TIG/iHug in Sydney, and having two numbers to try when the lines were busy. Windows Dial-Up Networking handled the dialling, but if a connection failed we’d have to start it again. I can still remember doing that far too often.
After a trawl through the Wayback Machine, the two Sydney numbers turned up on the archived iHug access list from March 2000: 02 9373 5100 and 02 8218 3400. They belonged to the same ISP. By April 2002, the archived list showed only 02 8218 3400 for Sydney, although that doesn’t tell us exactly when the other number stopped working.
Both TIG/iHug numbers reconstructed using the same settings measured from my later Avtek recording. These are synthetic tones, not two historical recordings.
There’s also the question of whether Windows actually dialled the 02. With the local calling rules configured for Sydney, it could dial just the eight-digit number. That gives us two possible starting points before we’ve even looked at the music. Lovely!
Singing the Thing Back
I made six recordings of my best attempts at remembering the dialling sounds. Recordings #1 through #4 and #6 came out as individual little notes. Recording #5 was more of a continuous tune. Apparently my brain has spent two decades working on its own arrangement.
Recording #1 through Recording #6, in order. The bright horizontal stripes are voice harmonics. Nothing has been pitch shifted or time stretched.
Even with the variations, the shape of the melody kept coming back. A little drop around the third sound, a higher group in the middle, then another drop towards the end. Recordings #1, #2 and #3 are especially consistent, despite my singing pitch wandering around a bit.
The same melodies viewed as relative pitch rather than coloured frequency bands. Recording #5 is shown separately because it was sung as a continuous phrase.
We measured the time between the shorter sung notes as well. Across Recordings #1, #2, #3, #4 and #6, which have identifiable note bursts, the median was about 0.225 seconds. My real Avtek modem dialled at about 0.200 seconds between tones. That’s not too shabby for something I last regularly heard more than twenty years ago.
The five recordings with distinct notes have a similar cadence to the Avtek dialling. Recording #5 was a continuous melody, so it is not included in this comparison.
The Real Avtek Modem
Here’s where the old hardware helped. I had a video of my Avtek MegaPlus V.32bis CD950 attempting a connection, published in 2022. I manually entered 8218 3400 after finding the number on the Wayback Machine, so it wasn’t a number I’d recovered independently from memory. But it did give us a genuine modem recording to examine.
The eight dialling tones decoded perfectly as 82183400. Each digit produces two frequencies at the same time, which makes the spectrogram look like a tidy little ladder. The two eights match one another, and the final two zeroes are identical too.
Measurements taken from the real recording. Every digit contains a lower and an upper DTMF frequency.
After the digits, the recording continues with a series of repeated calling tones at about 1,300 Hz. Those begin around 2.3 seconds apart, rather than every 0.2 seconds. It turns out that the quick notes in my memory have a lot more in common with the dialling part than the waiting part.
One genuine recording, two very different sounds. The eight dialled digits are on the left, followed by the longer calling tones.
Two Frequencies Versus One Dodgy Voice
The catch is that a telephone keypad doesn’t play a conventional single musical note for each digit. DTMF combines two frequencies. When I sing the melody back, I’m producing a vocal fundamental with harmonics above it, which isn’t the same thing at all.
My voice on one side, the real Avtek DTMF on the other. It’s easy to see why a remembered melody doesn’t decode neatly into telephone digits.
We tried comparing the relative rises and dips, differences between DTMF frequencies, and spots where the same digit should make the same sound twice. Some results looked rather good for 9373 5100. Others favoured 8218 3400. The awkward part was that moving the eight-note window by just one or two positions could change the winner.
The earlier comparisons, all shown together. A different starting note can change which number looks best, so I’m not going to pretend we’ve decoded it.
Still Stuck in My Head
So, which one was it? I still don’t know for certain. Both numbers are genuine TIG/iHug Sydney access numbers, and I do remember switching between them. The rhythm of my recordings is surprisingly close to actual dialling, while the general melodic shape has survived better than I would have expected.
What started with me humming an old dial-up number has turned into a collection of archived ISP pages, six rather questionable singing performances, a real modem recording and some fantastic-looking graphs. I’m rather pleased with the spectrograms, even if they haven’t coughed up an exact telephone number.
Perhaps an old setup guide will turn up something useful one day. Until then, I can still sing the darned thing. I just can’t tell you exactly what number it was.
Sources
- TIG access numbers, 29 January 1999
- iHug access numbers, 6 March 2000
- TIG access numbers, 15 April 2002
- TIG access numbers, July 2003
- TIG access numbers, January 2004
- TIG access numbers, August 2004
Related Posts