All of the sounds for Pacman appear to be based on arithmetic sequences. Each note is approximately the same number of Hertz distant from the previous one. (The equal-tempered tuning suggested by my notation isn’t perfectly accurate.)
E-Bb-Eb-Gb-A could be seen as 160 – 230 – 300 – 370 – 440, each pitch increasing by 70 Hz. The second half of the sound (the “-cka” in wacka) similarly seems to descend by 70 Hz: 490 – 420 – 350 – 280 – 210.
I don’t know the Namco WSG chip well enough to say for certain why this approach was used. Pitches might be entered as frequency, which would make this method less tedious than using equal-tempered notes, or it might use a sweep function that spits out notes equidistant from each other. It’s become iconic, either way.
The 12th covered by this sound is actually a fairly small interval for this game, and the 50 Hz difference between the first and second half creates some disparity to suggest Pacman opening and closing his mouth.
This one starts and ends with an arithmetic sequence, but switches to a geometric one for the 5 C#’s. Instead of adding or subtracting a constant, a geometric sequence multiplies or divides by one – in this case, it divides C#3 (about 140 Hz) by 2 to get C#2 (70 Hz), and then divides by 2 again to get C#1 (35 Hz). This octave movement creates a big black hole of bass in the sound that gives it some oomph.
If you’re trying to recreate these sounds, it helps to understand that the Namco WSG used 4-bit wavetable-lookup synthesis, meaning it had custom waveforms instead of the more common triangles and squares. To recreate the sounds perfectly would require access to the waveform samples the game used, but I’ve reverse engineered a couple patches in fxpansion’s Strobe that come close. For those who own the DCAM Synth Squad, the Start and Eating Pellets sounds can use this patch – which, for everyone else, is simply a sine wave hard synced at +7 semitones. For all the Pacman sounds, you can get closer to the original by crushing them down to 4-bits!
This sound uses a special arithmetic sequence – the Harmonic Series, also called the Overtone Series. The constant in this sequence is equal to the starting frequency, so we have (approximately) 280 – 560 – 840 – 1120 – 1400 – 1680 – 1960 – 2240.
The waveform sample used is a bit percussive, which gives this sound a güiro-like quality. This Strobe patch comes pretty close to the original. I can’t easily describe this patch for non-Strobe users, but if you render a square wave with a linear decay of about 22 ms at C#4 and use that in a sampler, you should come close.
Another Harmonic Series, this time starting on F#3 (185 Hz).
This is just the “bloop bloop” after the spiraling, crying sound (which I haven’t successfully decoded yet.) It uses a 4-bit sine wave, which gives it a digital sound that contrasts nicely with the more organic cry spiral, punctuating your failure.