APU Pulse: Difference between revisions

From NESdev Wiki
Jump to navigationJump to search
m (Added note about open-bus behavior)
 
(2 intermediate revisions by one other user not shown)
Line 1: Line 1:
[[Category:APU]]
[[Category:APU]]
Each of the two [[APU|NES APU]] pulse (square) wave channels generate a pulse wave with variable duty.
Each of the two [[APU|NES APU]] pulse (square) wave channels generate a pulse wave with variable duty.


Each pulse channel contains the following:  
Each pulse channel contains the following:  
Line 23: Line 22:
</pre>
</pre>


== Registers ==


'''Note''': the addresses below are ''write-only!'' Reading from these addresses exhibits [[Open bus behavior|open-bus behavior]].
'''Note''': the addresses below are ''write-only!'' Reading from these addresses exhibits [[Open bus behavior|open-bus behavior]].
Line 55: Line 55:
|colspan=2| Side effects || The sequencer is immediately restarted at the first value of the current sequence. The [[APU Envelope|envelope]] is also restarted. The period divider is ''not'' reset.[//forums.nesdev.org/viewtopic.php?p=186129#p186129]
|colspan=2| Side effects || The sequencer is immediately restarted at the first value of the current sequence. The [[APU Envelope|envelope]] is also restarted. The period divider is ''not'' reset.[//forums.nesdev.org/viewtopic.php?p=186129#p186129]
|}
|}
== Sequencer behavior ==
The sequencer is clocked by an 11-bit [[APU Misc|timer]]. Given the timer value ''t = HHHLLLLLLLL'' formed by timer high and timer low,
this timer is updated every APU cycle (i.e., every second CPU cycle), and counts ''t, t-1, ..., 0, t, t-1, ...'', clocking the waveform
generator when it goes from 0 to t. Since the period of the timer is ''t+1'' APU cycles and the sequencer has 8 steps, the period of the waveform is
''8*(t+1)'' APU cycles, or equivalently ''16*(t+1)'' CPU cycles.
Hence
* f<sub>pulse</sub> = f<sub>CPU</sub>/(16*(t+1)) (where f<sub>CPU</sub> is 1.789773 MHz for NTSC, 1.662607 MHz for PAL, and 1.773448 MHz for Dendy)
* t = f<sub>CPU</sub>/(16*f<sub>pulse</sub>) - 1
'''Note:''' A period of ''t < 8'', either set explicitly or via a sweep period update, '''silences the corresponding pulse channel'''. The highest frequency a pulse channel can output is hence about 12.4 kHz for NTSC. ('''TODO:''' PAL behavior?)




Line 73: Line 88:




'''Sequencer behavior'''
''' Implementation details '''


The sequencer is clocked by an 11-bit [[APU Misc|timer]]. Given the timer value ''t = HHHLLLLLLLL'' formed by timer high and timer low,
The reason for the odd output from the sequencer is that the counter is initialized to zero but counts ''downward'' rather than upward. Thus it reads the sequence lookup table in the order 0, 7, 6, 5, 4, 3, 2, 1.
this timer is updated every APU cycle (i.e., every second CPU cycle), and counts ''t, t-1, ..., 0, t, t-1, ...'', clocking the waveform
generator when it goes from 0 to t. Since the period of the timer is ''t+1'' APU cycles and the sequencer has 8 steps, the period of the waveform is
''8*(t+1)'' APU cycles, or equivalently ''16*(t+1)'' CPU cycles.


Hence
{| class="wikitable"
 
! Duty || Sequence lookup table || Output waveform
* f<sub>pulse</sub> = f<sub>CPU</sub>/(16*(t+1)) (where f<sub>CPU</sub> is 1.789773 MHz for NTSC, 1.662607 MHz for PAL, and 1.773448 MHz for Dendy)
|-
* t = f<sub>CPU</sub>/(16*f<sub>pulse</sub>) - 1
| 0 || <code>0 0 0 0 0 0 0 1</code> || <samp>0 1 0 0 0 0 0 0</samp> (12.5%)
 
|-
 
| 1 || <code>0 0 0 0 0 0 1 1</code> || <samp>0 1 1 0 0 0 0 0</samp> (25%)
'''Note:''' A period of ''t < 8'', either set explicitly or via a sweep period update, '''silences the corresponding pulse channel'''. The highest frequency a pulse channel can output is hence about 12.4 kHz for NTSC. ('''TODO:''' PAL behavior?)
|-
| 2 || <code>0 0 0 0 1 1 1 1</code> || <samp>0 1 1 1 1 0 0 0</samp> (50%)
|-
| 3 || <code>1 1 1 1 1 1 0 0</code> || <samp>1 0 0 1 1 1 1 1</samp> (25% negated)
|}


 
== Pulse channel output to mixer ==
'''Pulse channel output to mixer'''


The [[APU Mixer|mixer]] receives the pulse channel's current [[APU Envelope|envelope volume]] (lower 4 bits from $4000 or $4004) except when
The [[APU Mixer|mixer]] receives the pulse channel's current [[APU Envelope|envelope volume]] (lower 4 bits from $4000 or $4004) except when
Line 99: Line 114:
If any of the above are true, then the pulse channel sends zero (silence) to the mixer.
If any of the above are true, then the pulse channel sends zero (silence) to the mixer.


 
== Pulse channel 1 vs Pulse channel 2 behavior ==
'''Pulse channel 1 vs Pulse channel 2 behavior'''


The behavior of the two pulse channels differs only in the effect of the negate mode of their [[APU Sweep|sweep units]].
The behavior of the two pulse channels differs only in the effect of the negate mode of their [[APU Sweep|sweep units]].


 
== See Also ==
== Implementation details ==
* [[Pulse Channel frequency chart]]
 
The reason for the odd output from the sequencer is that the counter is initialized to zero but counts ''downward'' rather than upward. Thus it reads the sequence lookup table in the order 0, 7, 6, 5, 4, 3, 2, 1.
 
{| class="wikitable"
! Duty || Sequence lookup table || Output waveform
|-
| 0 || <code>0 0 0 0 0 0 0 1</code> || <samp>0 1 0 0 0 0 0 0</samp> (12.5%)
|-
| 1 || <code>0 0 0 0 0 0 1 1</code> || <samp>0 1 1 0 0 0 0 0</samp> (25%)
|-
| 2 || <code>0 0 0 0 1 1 1 1</code> || <samp>0 1 1 1 1 0 0 0</samp> (50%)
|-
| 3 || <code>1 1 1 1 1 1 0 0</code> || <samp>1 0 0 1 1 1 1 1</samp> (25% negated)
|}

Latest revision as of 04:08, 6 October 2022

Each of the two NES APU pulse (square) wave channels generate a pulse wave with variable duty.

Each pulse channel contains the following:

                     Sweep -----> Timer
                       |            |
                       |            |
                       |            v 
                       |        Sequencer   Length Counter
                       |            |             |
                       |            |             |
                       v            v             v
    Envelope -------> Gate -----> Gate -------> Gate --->(to mixer)

Registers

Note: the addresses below are write-only! Reading from these addresses exhibits open-bus behavior.

Address Bitfield Description
$4000 DDlc.vvvv Pulse 1 Duty cycle, length counter halt, constant volume/envelope flag, and volume/envelope divider period
$4004 DDlc.vvvv Pulse 2 Duty cycle, length counter halt, constant volume/envelope flag, and volume/envelope divider period
Side effects The duty cycle is changed (see table below), but the sequencer's current position isn't affected.
 
$4001 EPPP.NSSS See APU Sweep
$4005 EPPP.NSSS See APU Sweep
 
$4002 LLLL.LLLL Pulse 1 timer Low 8 bits
$4006 LLLL.LLLL Pulse 2 timer Low 8 bits
 
$4003 llll.lHHH Pulse 1 length counter load and timer High 3 bits
$4007 llll.lHHH Pulse 2 length counter load and timer High 3 bits
Side effects The sequencer is immediately restarted at the first value of the current sequence. The envelope is also restarted. The period divider is not reset.[1]

Sequencer behavior

The sequencer is clocked by an 11-bit timer. Given the timer value t = HHHLLLLLLLL formed by timer high and timer low, this timer is updated every APU cycle (i.e., every second CPU cycle), and counts t, t-1, ..., 0, t, t-1, ..., clocking the waveform generator when it goes from 0 to t. Since the period of the timer is t+1 APU cycles and the sequencer has 8 steps, the period of the waveform is 8*(t+1) APU cycles, or equivalently 16*(t+1) CPU cycles.

Hence

  • fpulse = fCPU/(16*(t+1)) (where fCPU is 1.789773 MHz for NTSC, 1.662607 MHz for PAL, and 1.773448 MHz for Dendy)
  • t = fCPU/(16*fpulse) - 1


Note: A period of t < 8, either set explicitly or via a sweep period update, silences the corresponding pulse channel. The highest frequency a pulse channel can output is hence about 12.4 kHz for NTSC. (TODO: PAL behavior?)


Duty Cycle Sequences

Duty Output waveform
0 0 1 0 0 0 0 0 0 (12.5%)
1 0 1 1 0 0 0 0 0 (25%)
2 0 1 1 1 1 0 0 0 (50%)
3 1 0 0 1 1 1 1 1 (25% negated)

Notice that a few Famiclone units have swapped APU duty cycles, as 12.5 [0], 50 [1], 25 [2] and 25 negated [3] instead.


Implementation details

The reason for the odd output from the sequencer is that the counter is initialized to zero but counts downward rather than upward. Thus it reads the sequence lookup table in the order 0, 7, 6, 5, 4, 3, 2, 1.

Duty Sequence lookup table Output waveform
0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 (12.5%)
1 0 0 0 0 0 0 1 1 0 1 1 0 0 0 0 0 (25%)
2 0 0 0 0 1 1 1 1 0 1 1 1 1 0 0 0 (50%)
3 1 1 1 1 1 1 0 0 1 0 0 1 1 1 1 1 (25% negated)

Pulse channel output to mixer

The mixer receives the pulse channel's current envelope volume (lower 4 bits from $4000 or $4004) except when

  • The sequencer output is zero, or
  • overflow from the sweep unit's adder is silencing the channel, or
  • the length counter is zero, or
  • the timer has a value less than eight (t<8, noted above).

If any of the above are true, then the pulse channel sends zero (silence) to the mixer.

Pulse channel 1 vs Pulse channel 2 behavior

The behavior of the two pulse channels differs only in the effect of the negate mode of their sweep units.

See Also