writing part 2 (IELTS here I come yippee!!)

content/blog/blog-aerations.md

@@ -0,0 +1,85 @@
+---js
+const title = "Aerations - Hacking Furnace to do stuff";
+const description = "Some ramblings about timers";
+const date = "2025-09-17";
+const tags = ["atari st", "timer fx", "timers", "demoscene"];
+const draft = true;
+---
+
+[comment]: # (god the depresso is getting to me i'm going to write something to soothe the nerves, but well this is a hidden comment you're not supposed to see this in the final post - 09/17/2025 @ 8:57 pm)
+
+back in July this year, we released "Aerations" at the Sommarhack 2025 demoparty. with it, we
+demonstrated the ability to compose Atari ST music, with classic timer effects like PWM and
+SyncBuzzer, and the ability to export it under a format usable on real hardware (thanks Natt!)
+
+developing the tech required for it was a real pain in the arse... it turns out that the ST is
+janky, and therefore also has janky audio, and we needed to emulate this jankness in the tracker
+
+# timer effects: what? how?
+generally speaking, a timer is a mechanism that lets an action be done at a preset time. on
+computers, it lets the CPU do things periodically, and if it is able to interrupt the CPU, allows
+actions to be executed automatically at regular, periodic intervals.
+
+on every Atari ST ever made is a chip called the *MC68901: Multi-Function Peripheral (MFP)*,
+and this chip provides an USART, an interrupt controller, and most notably, 4 timers. these
+timers are at the heart of *timer effects* (hereby called *timer FX*).
+
+what is *timer FX*? on the Atari ST, it means any effect that messes with the YM2149 sound registers.
+this means it can be used to write to a register at regular intervals, producing
+audible effects (like frequency modulation, amplitude modulation, etc...). although there are a
+variety of effects that can be done with this, for Aerations in particular, we only used 2:
+- PWM (Pulse Width Modulation)
+- SyncBuzzer
+
+## pulse width modulation
+
+PWM is amplitude modulating the YM square generator output with the timer.
+the timer generates a square wave by switching between minimum amplitude and maximum amplitude at the timer
+period, and the resulting wave which is not always in phase with the square output of the YM, and
+when this wave is applied to the YM output by amplitude modulation, the 2 waves being out-of-phase will cause
+one wave to "shadow" the other (this happens when the square edge of either source is 0).
+the following ASCII diagram shows this:
+
+```
+   +------+      +------+      +------+      +------+
+   |   		|			 |			|			 |		  |			 |			|					[Timer-generated square]
+---+			+------+			+------+			+------+			+---
+
++------+      +------+      +------+      +------+
+			 |			|			 |			|			 |			|			 |						[YM2149 square generator]
+			 +------+			 +------+			 +------+			 +------+
+
+   +---+         +---+         +---+         +---+
+	 |   |         |   |         |   |         |   |						[Amplitude modulated, YM+timer]
+---+   +---------+   +---------+   +---------+   +------+
+
+```
+
+of course, you will not always get a stable pulse wave, since the timer has a different clock and a frequency
+that is not always equivalent to the YM2149 frequency. since both frequencies are not equivalent, they will go
+out of phase, causing phasing, and thus causing pulse width modulation because the part of the square wave it
+shadows is constantly changing.
+
+since it works by amplitude modulation, the minimum and maximum amplitude can be any value one desires, allowing
+for less of the YM square wave to be shadowed, making for a very subtle undertone in the instrument (wow this
+sentence sucks).
+
+[comment]: # (ughhhh i have an exam tomorrow - 09/19/2025 @ 7:13 pm)
+
+in most drivers, this is implemented as 2 timer handlers, and these are alternated between every time the timer
+causes an interrupt. timer handler A will change the vector to timer handler B, and timer handler B will change
+the vector back to timer handler A. in theory, this could be extended to an arbitrary length, and this is actually
+how you can make wavetables. however, we do not use this in the song and the effective frequency of the timer-
+generated waveform gets lower every time you add more steps to the waveform.
+
+## SyncBuzzer
+
+SyncBuzzer is rather simple in contrast to PWM described above. the YM2149 (and its descendant, the AY-3-8910)
+have a peculiar behaviour, where if you write any value to register $0D (envelope shape register), it causes the
+envelope to reset from the beginning, and the phase is also reset.
+it is simpler, because we only need one timer handler to just reset the envelope by writing the current envelope
+shape to it again. this on its own isn't particularly exciting, so you need to run your timers at a different
+frequency from the envelope (or vice versa) in order to get a buzzy tone from the envelope.
+
+the bass in Aerations is actually SyncBuzzer, but the timer is running at a different frequency, which is a multiple
+of the current bass frequency.