APU Noise: Difference between revisions

From NESdev Wiki
Jump to navigationJump to search
(correction to table (pitch $1 for PAL is 8, not 7))
(Mixer nonlinearity caveat)
(7 intermediate revisions by 5 users not shown)
Line 2: Line 2:
The [[APU|NES APU]] noise channel generates pseudo-random 1-bit noise at 16 different frequencies.
The [[APU|NES APU]] noise channel generates pseudo-random 1-bit noise at 16 different frequencies.


The noise channel contains the following: [[APU Envelope|envelope generator]], [[APU Misc|timer]], [http://en.wikipedia.org/wiki/Linear_feedback_shift_register Linear Feedback Shift Register], [[APU Length Counter|length counter]].
The noise channel contains the following: [[APU Envelope|envelope generator]], [[APU Misc|timer]], [[wikipedia:Linear_feedback_shift_register|Linear Feedback Shift Register]], [[APU Length Counter|length counter]].


<pre>
<pre>
Line 12: Line 12:


{| border=1
{| border=1
| '''$400C''' || <tt>--le.eeee</tt> || '''[[APU Length Counter|Length counter halt]]''' and '''[[APU Envelope|envelope]]''' (write)
| '''$400C''' || <tt>--lc.vvvv</tt> || '''[[APU Length Counter|Length counter halt]]''', constant volume/envelope flag, and volume/'''[[APU Envelope|envelope]]''' divider period (write)
|-
|-
|colspan=3| &nbsp;
|colspan=3| &nbsp;
Line 43: Line 43:
# Bit 14, the leftmost bit, is set to the feedback calculated earlier.
# Bit 14, the leftmost bit, is set to the feedback calculated earlier.


This results in a pseudo-random bit sequence, 32767 bits long when Mode flag is clear, otherwise 93 bits long (the particular 93-bit sequence depends on where in the 32767-bit sequence the shift register was when Mode flag was set).
This results in a pseudo-random bit sequence, 32767 steps long when Mode flag is clear, and randomly 93 or 31 steps long otherwise. (The particular 31- or 93-step sequence depends on where in the 32767-step sequence the shift register was when Mode flag was set).


The [[APU Mixer|mixer]] receives the current [[APU Envelope|envelope volume]] except when
The [[APU Mixer|mixer]] receives the current [[APU Envelope|envelope volume]] except when
* Bit 0 of the shift register is set, or
* Bit 0 of the shift register is set, or
* The [[APU Length Counter|length counter]] is zero
* The [[APU Length Counter|length counter]] is zero
Within the mixer, the DMC level has a noticeable effect on the noise's level.


On power-up, the shift register is loaded with the value 1.
On power-up, the shift register is loaded with the value 1.


The earliest revisions of the 2A03 CPU ignored the Mode flag, treating it as always 0. These CPUs were used in the first batch of Famicom consoles that were recalled, in Vs. System boards, and in the arcade games that used the 2A03 as a sound coprocessor.[http://nesdev.parodius.com/bbs/viewtopic.php?p=58046#58046]
In the earliest revisions of the 2A03 CPU, the Mode flag was nonexistent and the shift register always used bits 0 and 1 for feedback. These CPUs were used in the first batch of Famicom consoles (which were recalled), in Vs. System boards, and in the arcade games that used the 2A03 as a sound coprocessor.[http://forums.nesdev.org/viewtopic.php?p=58046#p58046]
 
The 93-step sequence is about a quarter tone (50 cents) sharp relative to A440 tuning. The approximate frequencies and pitches (in [[wikipedia:Helmholtz pitch notation#Variations|LilyPond's variant of Helmholtz notation]]) are as follows:
<!-- Generated with https://github.com/pinobatch/pently/blob/master/tools/noise_freqs.py -->
{| class="wikitable"
|+ Pitches of 93-step noise on NTSC
! Period setting || Sample rate || Fundamental || MIDI note || Pitch
|-
| $80 || 447443.2 Hz || 4811.2 Hz || 110.41 || <nowiki>d''''' + 41¢</nowiki>
|-
| $81 || 223721.6 Hz || 2405.6 Hz || 98.41 || <nowiki>d'''' + 41¢</nowiki>
|-
| $82 || 111860.8 Hz || 1202.8 Hz || 86.41 || <nowiki>d''' + 41¢</nowiki>
|-
| $83 || 55930.4 Hz || 601.4 Hz || 74.41 || <nowiki>d'' + 41¢</nowiki>
|-
| $84 || 27965.2 Hz || 300.7 Hz || 62.41 || d' + 41¢
|-
| $85 || 18643.5 Hz || 200.5 Hz || 55.39 || g + 39¢
|-
| $86 || 13982.6 Hz || 150.4 Hz || 50.41 || d + 41¢
|-
| $87 || 11186.1 Hz || 120.3 Hz || 46.55 || a#, + 55¢
|-
| $88 || 8860.3 Hz || 95.3 Hz || 42.51 || f#, + 51¢
|-
| $89 || 7046.3 Hz || 75.8 Hz || 38.55 || d, + 55¢
|-
| $8A || 4709.9 Hz || 50.6 Hz || 31.57 || g,, + 57¢
|-
| $8B || 3523.2 Hz || 37.9 Hz || 26.55 || d,, + 55¢
|-
| $8C || 2348.8 Hz || 25.3 Hz || 19.53 || g,,, + 53¢
|-
| $8D || 1761.6 Hz || 18.9 Hz || 14.55 || d,,, + 55¢
|-
| $8E || 879.9 Hz || 9.5 Hz || 2.53 || d,,,, + 53¢
|-
| $8F || 440.0 Hz || 4.7 Hz || -9.47 || d,,,,, + 53¢
|}

Revision as of 21:41, 11 October 2020

The NES APU noise channel generates pseudo-random 1-bit noise at 16 different frequencies.

The noise channel contains the following: envelope generator, timer, Linear Feedback Shift Register, length counter.

   Timer --> Shift Register   Length Counter
                   |                |
                   v                v
Envelope -------> Gate ----------> Gate --> (to mixer)
$400C --lc.vvvv Length counter halt, constant volume/envelope flag, and volume/envelope divider period (write)
 
$400E M---.PPPP Mode and period (write)
bit 7 M--- ---- Mode flag
bits 3-0 ---- PPPP The timer period is set to entry P of the following:
Rate  $0 $1  $2  $3  $4  $5   $6   $7   $8   $9   $A   $B   $C    $D    $E    $F
      --------------------------------------------------------------------------
NTSC   4, 8, 16, 32, 64, 96, 128, 160, 202, 254, 380, 508, 762, 1016, 2034, 4068
PAL    4, 8, 14, 30, 60, 88, 118, 148, 188, 236, 354, 472, 708,  944, 1890, 3778
 
$400F llll.l--- Length counter load and envelope restart (write)

The shift register is 15 bits wide, with bits numbered
14 - 13 - 12 - 11 - 10 - 9 - 8 - 7 - 6 - 5 - 4 - 3 - 2 - 1 - 0

When the timer clocks the shift register, the following actions occur in order:

  1. Feedback is calculated as the exclusive-OR of bit 0 and one other bit: bit 6 if Mode flag is set, otherwise bit 1.
  2. The shift register is shifted right by one bit.
  3. Bit 14, the leftmost bit, is set to the feedback calculated earlier.

This results in a pseudo-random bit sequence, 32767 steps long when Mode flag is clear, and randomly 93 or 31 steps long otherwise. (The particular 31- or 93-step sequence depends on where in the 32767-step sequence the shift register was when Mode flag was set).

The mixer receives the current envelope volume except when

Within the mixer, the DMC level has a noticeable effect on the noise's level.

On power-up, the shift register is loaded with the value 1.

In the earliest revisions of the 2A03 CPU, the Mode flag was nonexistent and the shift register always used bits 0 and 1 for feedback. These CPUs were used in the first batch of Famicom consoles (which were recalled), in Vs. System boards, and in the arcade games that used the 2A03 as a sound coprocessor.[1]

The 93-step sequence is about a quarter tone (50 cents) sharp relative to A440 tuning. The approximate frequencies and pitches (in LilyPond's variant of Helmholtz notation) are as follows:

Pitches of 93-step noise on NTSC
Period setting Sample rate Fundamental MIDI note Pitch
$80 447443.2 Hz 4811.2 Hz 110.41 d''''' + 41¢
$81 223721.6 Hz 2405.6 Hz 98.41 d'''' + 41¢
$82 111860.8 Hz 1202.8 Hz 86.41 d''' + 41¢
$83 55930.4 Hz 601.4 Hz 74.41 d'' + 41¢
$84 27965.2 Hz 300.7 Hz 62.41 d' + 41¢
$85 18643.5 Hz 200.5 Hz 55.39 g + 39¢
$86 13982.6 Hz 150.4 Hz 50.41 d + 41¢
$87 11186.1 Hz 120.3 Hz 46.55 a#, + 55¢
$88 8860.3 Hz 95.3 Hz 42.51 f#, + 51¢
$89 7046.3 Hz 75.8 Hz 38.55 d, + 55¢
$8A 4709.9 Hz 50.6 Hz 31.57 g,, + 57¢
$8B 3523.2 Hz 37.9 Hz 26.55 d,, + 55¢
$8C 2348.8 Hz 25.3 Hz 19.53 g,,, + 53¢
$8D 1761.6 Hz 18.9 Hz 14.55 d,,, + 55¢
$8E 879.9 Hz 9.5 Hz 2.53 d,,,, + 53¢
$8F 440.0 Hz 4.7 Hz -9.47 d,,,,, + 53¢