Add missing feedback from #350

I had forgotten to force-push, oops :(

Author: ISSOtm

src/Audio_Registers.md

@@ -163,7 +163,7 @@ Bit 3   - Envelope direction (0=Decrease, 1=Increase)
 Bit 2-0 - Sweep pace (0=No Sweep)
 ```
 
-Setting bits 3-7 of this register all to 0 turns the DAC off (and thus, the channel as well).
+Setting bits 3-7 of this register all to 0 turns the DAC off (and thus, the channel as well), which [may cause an audio pop](<#Mixer>).
 
 The envelope ticks at 64 Hz, and the channel's envelope will be increased / decreased (depending on bit 3) every <var>Sweep pace</var> of those ticks.
 
@@ -217,14 +217,16 @@ Like other channels, turning the DAC off immediately turns the channel off as we
 Bit 7 - Sound Channel 3 DAC  (0=Off, 1=On)
 ```
 
-The channel is often turned off just before writing to [wave RAM](<#FF30–FF3F — Wave pattern RAM>) to avoid issues with accessing it; see there for more info.
+The DAC is often turned off just before writing to [wave RAM](<#FF30–FF3F — Wave pattern RAM>) to avoid issues with accessing it; see further below for more info.
+
+Turning the DAC off [may cause an audio pop](<#Mixer>).
 
 ### FF1B — NR31: Channel 3 length timer \[write-only\]
 
 This register controls the channel's [length timer](<#Length timer>).
 
 ```
-Bit 5-0 - length timer
+Bit 7-0 - length timer
 ```
 
 The higher the [length timer](<#Length timer>), the shorter the time before the channel is cut.
@@ -314,13 +316,16 @@ So, from the CPU's point of view, wave RAM reads out the same byte, regardless o
 
 ## Sound Channel 4 — Noise
 
-This channel is used to output white noise, which is done by randomly switching the amplitude between two levels fairly fast.
+This channel is used to output white noise[^not_white], which is done by randomly switching the amplitude between two levels fairly fast.
 
 The frequency can be adjusted in order to make the noise appear "harder" (lower frequency) or "softer" (higher frequency).
 
 The random function that switches the output level can also be manipulated.
 Certain settings can cause the wave to be more regular, sounding closer to a pulse than noise.
 
+[^not_white]:
+By default, the noise will sound close to white; but it can be manipulated to sound differently.
+
 ### FF20 — NR41: Channel 4 length timer \[write-only\]
 
 This register controls the channel's [length timer](<#Length timer>).
@@ -341,7 +346,7 @@ Bit 3   - Envelope direction (0=Decrease, 1=Increase)
 Bit 2-0 - Sweep pace (0=No Sweep)
 ```
 
-Setting bits 3-7 of this register all to 0 turns the DAC off (and thus, the channel as well).
+Setting bits 3-7 of this register all to 0 turns the DAC off (and thus, the channel as well), which [may cause an audio pop](<#Mixer>).
 
 The envelope ticks at 64 Hz, and the channel's envelope will be increased / decreased (depending on bit 3) every <var>Sweep pace</var> of those ticks.
 

src/Audio_details.md

@@ -50,7 +50,7 @@ tl;dr:
 Each of the four "conceptual" channels is composed of a "generation" circuit (designated "channel" in the above diagram), and a [DAC](https://en.wikipedia.org/wiki/Digital-to-analog_converter).
 The digital value produced by the generator, which ranges between $0 and $F (0 and 15), is linearly translated by the DAC into an analog[^digital_analog] value between -1 and 1 (the unit is arbitrary).
 
-The four analog channel outputs are then fed into the mixer[^vin], which selectively adds them (depending on [`NR51`](<#FF25 — NR51: Sound panning>)) into two analog outputs (Left and Right).
+The four analog channel outputs are then fed into the mixer[^vin], which selectively adds them (depending on [`NR51`]) into two analog outputs (Left and Right).
 Thus, the analog range of those outputs is 4× that of each channel, -4 to 4.
 
 Then, both of these two get their respective volume scaled, once from [`NR50`](<#FF24 — NR50: Master volume & VIN panning>), and once from the volume knob (if the console has one).
@@ -103,11 +103,17 @@ Indicated values are under normal operation; the frequencies will obviously diff
 A high-pass filter (HPF) removes constant biases over time.
 The HPFs therefore remove the DC offset created by inactive channels with an enabled DAC, and off-center waveforms.
 
-Enabling or disabling a DAC ([see below](#DACs)), adding or removing it using NR51, or changing the volume in NR50, will cause an audio "pop".
+::: tip Avoiding audio pops
+
+Enabling or disabling a DAC ([see below](#DACs)), adding or removing it using NR51, or changing the volume in NR50, will cause an audio pop.
 (All of these actions cause a change in DC offset, which is smoothed out by the HPFs over time, but still creates a pop.)
 
+To avoid this, a sound driver should avoid turning the DACs off; this can be done by writing $08 to `NRx2` (silences the channel but keeps the DAC on) then $80 to `NRx4` to retrigger the channel and reload `NRx2`.
+
+:::
+
 The HPF is more aggressive on GBA than on GBC, which itself is more aggressive than on DMG.
-(The more "aggressive" a HPF, the faster it pulls the signal towards "analog 0"; this reduces the pops, but tends to also distort waveforms.)
+(The more "aggressive" a HPF, the faster it pulls the signal towards "analog 0"; this tends to also distort waveforms.)
 
 ### DACs
 
@@ -183,10 +189,12 @@ This does not happen under regular operation, but can be achieved by switching f
 ## Game Boy Advance audio
 
 The APU was reworked pretty heavily for the GBA, which introduces some slightly different behavior:
-- Instead of mixing being done analogically, it's instead done digitally; then, sound is converted to an analog signal and an offset is added (see `SOUNDBIAS` in [GBATEK](http://problemkaputt.de/gbatek.htm#gbasoundcontrolregisters) for more details).
+- Instead of mixing being done by analog circuitry, it's instead done digitally; then, sound is converted to an analog signal and an offset is added (see `SOUNDBIAS` in [GBATEK](http://problemkaputt.de/gbatek.htm#gbasoundcontrolregisters) for more details).
 - This also means that the GBA APU has no DACs.
   Instead, they are emulated digitally such that a disabled "DAC" behaves like an enabled DAC receiving 0 as its input.
 - Additionally, CH3's DAC has its output inverted.
   In particular, this causes the channel to emit a loud spike when disabled; therefore, it's a good idea to "disconnect" the channel using NR51 before accessing wave RAM.
 
 None of the additional features (more wave RAM, digital FIFOs, etc.) are available to CGB programs.
+
+[`NR51`]: <#FF25 — NR51: Sound panning>