APU - Revision history

Thrimbor: Dead link fix

2025-07-22T20:47:32Z

Dead link fix

2019-07-14T16:08:25Z

The 48kHz samplerate is a key-design feature. The 32 sample length is a key-design feature. The 1500Hz (~1/0.666ms) is derived from said features.

2019-07-14T01:21:03Z

Include APU audio frame duration (0.6ms / 666.6us)

2019-02-02T00:10:14Z

Some more basic APU information

2018-12-29T16:03:23Z

2017-07-05T16:37:49Z

‎Usage in DirectSound

2017-07-04T20:45:23Z

‎Envelopes

2017-07-04T20:41:32Z

‎Envelopes

2017-07-04T20:30:06Z

‎Envelopes

2017-07-04T16:42:40Z

‎Voice Processor (VP)

@@ Line 287: / Line 287: @@
 The EP is disabled in this case.
 The data is send to the ACI AC97 using GP FIFO channel 0 (PCM).
-There is no AC3 / SPDIF during the boot animation[http://www.gamasutra.com/blogs/BrianSchmidt/20111117/90625/Designing_the_Boot_Sound_for_the_Original_Xbox.php]{{citation needed|reason=Link to brians gamasutra stuff}}.
+There is no AC3 / SPDIF during the boot animation[https://web.archive.org/web/20190123092902/http://www.gamasutra.com/blogs/BrianSchmidt/20111117/90625/Designing_the_Boot_Sound_for_the_Original_Xbox.php]{{citation needed|reason=Link to brians gamasutra stuff}}.
 == Related notes ==

@@ Line 14: / Line 14: @@
 The APU does only audio-processing but no audio output. Hence, one of the DSPs typically uses programmable DMA to transfer the finished audio to system memory, where it can be read by other components (such as AC97).
-All audio processing by the APU is typically done in 24bit PCM.
+All audio processing by the APU is typically done in 24bit PCM at 48000Hz.
-An APU audio frame is 32 samples long with a duration of 0.{{overline|6}}ms at 48000Hz, giving a total of 1500 frames per second.
+An APU audio frame is 32 samples long. Therefore, there are 1500 frames per second, with a duration of 0.{{overline|6}}ms each.
-The MIXBUF therefore holds 1024 samples (32 channels * 32 samples/channel).
+The MIXBUF holds 1024 samples (32 channels * 32 samples/channel).
 === Glossary ===

@@ Line 14: / Line 14: @@
 The APU does only audio-processing but no audio output. Hence, one of the DSPs typically uses programmable DMA to transfer the finished audio to system memory, where it can be read by other components (such as AC97).
-All audio processing by the APU is typically done in 24bit PCM at 48000Hz.
+All audio processing by the APU is typically done in 24bit PCM.
-An APU audio frame is 32 samples long, so there are 1500 frames per second.
+An APU audio frame is 32 samples long with a duration of 0.{{overline|6}}ms at 48000Hz, giving a total of 1500 frames per second.
 The MIXBUF therefore holds 1024 samples (32 channels * 32 samples/channel).

@@ Line 1: / Line 1: @@
 The [[MCPX]] contains an APU (Audio Processing Unit).
 * SSL = Stream Segment List
 == Frontend Engine (FE) ==

← Older revision		Revision as of 16:03, 29 December 2018
Line 282:		Line 282:
	** [https://github.com/kcat/openal-soft/blob/master/Alc/effects/reverb.c Mentions I3DL2 Reverb]		** [https://github.com/kcat/openal-soft/blob/master/Alc/effects/reverb.c Mentions I3DL2 Reverb]
	** [https://www.midi.org/specifications/item/dls-level-2-specification DLS2 low-pass filter with resonance and dynamic filter cutoff frequency]		** [https://www.midi.org/specifications/item/dls-level-2-specification DLS2 low-pass filter with resonance and dynamic filter cutoff frequency]
		+
		+
		+	[[Category:APU]]

@@ Line 261: / Line 261: @@
 {{FIXME|reason=When does this happen and what happens to stereo? headphones? mono?}}
 The data is then send to the ACI AC97 using EP FIFO channels 0 (PCM) and 1 (SPDIF){{citation needed}}.
-The EP code is loaded by DirectSound. The EP is not programmable using DirectSound.
+The EP code is loaded by DirectSound. The EP is not programmable using DirectSound APIs.
 === Modifications for Boot Animation ===

@@ Line 90: / Line 90: @@
 * 4: Decay = Rate at which the envelope goes from 100% to 0%
 ** COUNT register counts down from DECAYRATE.
 ** The best approximation I could come up with so far is: <code>255*pow(0.99988799,(DECAYRATE*16-COUNT)*4096/DECAYRATE)</code>.
-** When the output level reaches the sustain level the decay section is over (In my example above, this happens when the output level is less than 0.2 away from the sustain level; so if sustain is 0, the voice would switch to the sustain segment when a value below 0.2 is reached)
+** When the output level (not LVL!) reaches the sustain level the decay section is over (In my example above, this happens when the output level is less than 0.2 away from the sustain level; so if sustain is 0, the voice would switch to the sustain segment when a value below 0.2 is reached)
 ** Writes to the LVL register are ignored
 ** Writes to count (any value?) impact the LVL

@@ Line 78: / Line 78: @@
 === Envelopes ===
-There are seperate sections of the envelopes, 2 registers (CUR and COUNT) per envelope keeps track of this and control the envelopes level (LVL):
+There are seperate segments of the envelopes, 2 registers (CUR and COUNT) per envelope keeps track of this and control the envelopes level (LVL):
 * 0: Off = Envelope is not used
@@ Line 90: / Line 90: @@
 * 4: Decay = Rate at which the envelope goes from 100% to 0%
 ** COUNT register counts down from DECAYRATE.
-** When sustain level is reached the decay section is over
+** The best approximation I could come up with so far is: <code>255*pow(0.99988799,(DECAYRATE*16-COUNT)*4096/DECAYRATE)</code>.
 ** The LVL is not linear, it's a curve which drops steep at first and then slowly becomes level[https://docs.google.com/spreadsheets/d/1fNsfkvBfnlRQ9XplNnTgmh8jBaWJ56bvh2AVWn8B_k0/edit?usp=sharing]
 ** Writes to the LVL register are ignored

@@ Line 110: / Line 110: @@
 * 7: Force Release = Unknown still{{FIXME}}
 ** Seems to happen during invalid conditions? Happened to me when modifying ebo during playback{{citation needed|reason=Need to find out why this happens}}
-** LVL and COUNT seem to be ignored during this? Output level seems to stay at 100% ? (I only got repeating 32 samples during this and the whole Xbox crashed shortly after)
+** LVL and COUNT seem to be ignored during this, but writes go through? Output level seems to stay at 100% ? (I only got repeating 32 samples during this and the whole Xbox crashed shortly after)
 All durations are described using unsigned 12-bit times/rates. The level of sustain is stored unsigned in 8-bit.

@@ Line 9: / Line 9: @@
 == Voice Processor (VP) ==
-A powerful voice processor. There can be up to 256 voices [www.gamasutra.com/blogs/BrianSchmidt/20111117/90625/Designing_the_Boot_Sound_for_the_Original_Xbox.php][https://web.archive.org/web/20010410003338/http://www.nvnews.net/previews/mcpx/mcpx.shtml] and 64[http://www.nvidia.com/object/IO_20010530_6177.html] of those can be 3D.
+A powerful voice processor. There can be up to 256 voices [http://www.gamasutra.com/blogs/BrianSchmidt/20111117/90625/Designing_the_Boot_Sound_for_the_Original_Xbox.php][https://web.archive.org/web/20010410003338/http://www.nvnews.net/previews/mcpx/mcpx.shtml] and 64[http://www.nvidia.com/object/IO_20010530_6177.html] of those can be 3D.
 Per-voice settings: