|
|
| (One intermediate revision by one other user not shown) |
| Line 1: |
Line 1: |
| [[Category:iNES Mappers|085]][[Category:in NesCartDB|085]]
| | #REDIRECT [[VRC7]] |
| | | {{DEFAULTSORT:085}}[[Category:iNES Mappers]][[Category:in NesCartDB]][[Category:Mappers with cycle IRQs]] |
| [[iNES Mapper 085]] is used to represent the Konami [[VRC7]] mapper.
| |
| | |
| Here are Disch's original notes:
| |
| ========================
| |
| = Mapper 085 =
| |
| ========================
| |
|
| |
| aka
| |
| --------------------------
| |
| VRC7
| |
|
| |
|
| |
| Example Games:
| |
| --------------------------
| |
| Lagrange Point
| |
| Tiny Toon Adventures 2 (J)
| |
|
| |
|
| |
|
| |
| VRC7a vs. VRC7b
| |
| --------------------------
| |
| Lagrange Point ('VRC7a') and Tiny Toon Adventures 2 ('VRC7b') both operate exactly the same, but are wired a
| |
| bit differently. VRC7a uses $x010 for regs, and VRC7b uses $x008. Registers below are listed as they exist
| |
| on VRC7a. For VRC7b, make the appropriate adjustments
| |
|
| |
| Also, only Lagrange Point seems to use the extra sound. It's unknown whether or not the sound hardware
| |
| exists on VRC7b, as Tiny Toon doesn't use it.
| |
|
| |
|
| |
| CHR-RAM note:
| |
| --------------------------
| |
| Lagrange Point, for some reason I still don't understand, swaps its 8k CHR-RAM around. How this offers any
| |
| functionality is beyond me, but the game does it, so your emu must support it.
| |
|
| |
|
| |
|
| |
| Registers:
| |
| --------------------------
| |
|
| |
| $8000: PRG Reg 0 (8k @ $8000)
| |
| $8010: PRG Reg 1 (8k @ $A000)
| |
| $9000: PRG Reg 2 (8k @ $C000)
| |
|
| |
| $9010: Sound Address Reg (see below)
| |
| $9030: Sound Data Port (see below)
| |
|
| |
| $A000-$D010: CHR Regs
| |
|
| |
| $E000: [.... ..MM] Mirroring:
| |
| %00 = Vert
| |
| %01 = Horz
| |
| %10 = 1ScA
| |
| %11 = 1ScB
| |
|
| |
| $E010: [IIII IIII] IRQ Reload value
| |
| $F000: [.... .MEA] IRQ Control
| |
| $F010: [.... ....] IRQ Acknowledge
| |
|
| |
|
| |
| PRG Setup:
| |
| --------------------------
| |
|
| |
| $8000 $A000 $C000 $E000
| |
| +-------+-------+-------+-------+
| |
| | $8000 | $8010 | $9000 | { -1} |
| |
| +-------+-------+-------+-------+
| |
|
| |
|
| |
| CHR Setup:
| |
| --------------------------
| |
|
| |
| $0000 $0400 $0800 $0C00 $1000 $1400 $1800 $1C00
| |
| +-------+-------+-------+-------+-------+-------+-------+-------+
| |
| | $A000 | $A010 | $B000 | $B010 | $C000 | $C010 | $D000 | $D010 |
| |
| +-------+-------+-------+-------+-------+-------+-------+-------+
| |
|
| |
|
| |
| IRQs:
| |
| --------------------------
| |
|
| |
| VRC7 uses the "VRC IRQ" setup shared by several VRCs. It uses the following registers:
| |
|
| |
|
| |
| $E010: [IIII IIII] IRQ Reload
| |
| $F000: [.... .MEA] IRQ Control
| |
| $F010: [.... ....] IRQ Acknowledge
| |
|
| |
| For info on how these IRQs work, see the "VRC IRQs" section in mapper 021
| |
|
| |
|
| |
|
| |
| Sound:
| |
| --------------------------
| |
|
| |
| VRC7 has additional sound channels! It is a slightly dumbed down version of the YM2413 (aka OPLL). There
| |
| are only 6 harmony channels and no rhythmic channels. Note that some older docs claim it's an OPL2 -- but it
| |
| is, in fact, OPLL.
| |
|
| |
| Due to the complexity of FM-synth ... details will not be covered here. For details, refer to a YM2413 data
| |
| sheet or technical doc.
| |
|
| |
| $9010 is the address port
| |
| $9030 is the data port
| |
|
| |
| They behave just like the address/data ports on the YM2413. Though remember that there's no rhythm, and
| |
| only 6 channels.
| |
|
| |
| Channels can choose from 16 instruments. One is customizable by the game, the other 15 are fixed with the
| |
| following values:
| |
|
| |
| {
| |
| 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, // custom instrument
| |
|
| |
| 0x03,0x21,0x04,0x06,0x8D,0xF2,0x42,0x17, // begin fixed instruments
| |
| 0x13,0x41,0x05,0x0E,0x99,0x96,0x63,0x12,
| |
| 0x31,0x11,0x10,0x0A,0xF0,0x9C,0x32,0x02,
| |
| 0x21,0x61,0x1D,0x07,0x9F,0x64,0x20,0x27,
| |
| 0x22,0x21,0x1E,0x06,0xF0,0x76,0x08,0x28,
| |
| 0x02,0x01,0x06,0x00,0xF0,0xF2,0x03,0x95,
| |
| 0x21,0x61,0x1C,0x07,0x82,0x81,0x16,0x07,
| |
| 0x23,0x21,0x1A,0x17,0xEF,0x82,0x25,0x15,
| |
| 0x25,0x11,0x1F,0x00,0x86,0x41,0x20,0x11,
| |
| 0x85,0x01,0x1F,0x0F,0xE4,0xA2,0x11,0x12,
| |
| 0x07,0xC1,0x2B,0x45,0xB4,0xF1,0x24,0xF4,
| |
| 0x61,0x23,0x11,0x06,0x96,0x96,0x13,0x16,
| |
| 0x01,0x02,0xD3,0x05,0x82,0xA2,0x31,0x51,
| |
| 0x61,0x22,0x0D,0x02,0xC3,0x7F,0x24,0x05,
| |
| 0x21,0x62,0x0E,0x00,0xA1,0xA0,0x44,0x17
| |
| };
| |
| | |
| And here are Disch's [http://forums.nesdev.org/viewtopic.php?p=96856#p96856 extended notes]:
| |
| <pre>
| |
| ----------------------------------------------------------------------------
| |
| ----------------------------------------------------------------------------
| |
| ----- VRC7 Sound -------------------------------------------------
| |
| ----------------------------------------------------------------------------
| |
| ----------------------------------------------------------------------------
| |
| | |
| VRC7 has additional sound channels! It is a slightly dumbed down version of the YM2413 (aka OPLL). There
| |
| are only 6 harmony channels and no rhythmic channels.
| |
| | |
| Strap yourself in. FM-Synth is a beast.
| |
| | |
| | |
| ---------------------------------------------
| |
| Disclaimers:
| |
| ---------------------------------------------
| |
| | |
| Information here is pieced together from the Yamaha YM2413 Application Manual ("YM2413.pdf"), and Mitsutaka
| |
| Okazaki's "emu2413.c" emulator. Anyone whose looked at those sources know they are not the easiest things
| |
| to comprehend without prior experience with FM synth, so here I attempt to explain things in a more
| |
| traditional form.
| |
| | |
| I don't really care about YM2413 (I hate FM synth... I find it extremely ugly), so I only cover items on the
| |
| VRC7 here (ie: no rhythmic information). If you want details about a full YM2413, you'll have to look elsewhere.
| |
| | |
| I am NOT confident about this information being 100% accurate. I made every effort to be as accurate as
| |
| possible, and my implementation based on the below info sounds *very close* to recordings of the real thing,
| |
| but I do hear some subtle differences. I graciously welcome any corrections anyone can offer.
| |
| | |
| Bitwidths of various counters are kind of an educated guess. With the exception of the phase accumulator,
| |
| which is the only counter whose size is hinted at in the documentation... so I'm fairly certain it is in
| |
| fact 18 bits wide.
| |
| | |
| I mention the use of various lookup tables. I do not know if these lookup tables actually exist on the
| |
| hardware, or if the values are calculated at runtime. Likewise the actual size of these lookup tables is
| |
| entirely unknown to me. You can choose your own size in your implementation.
| |
| | |
| | |
| ---------------------------------------------
| |
| FM-Synth basics & other fundamental concepts:
| |
| ---------------------------------------------
| |
| | |
| The basic idea of FM-Synth is you have 2 sine waves (aka, "slots"), a "modulator" and a "carrier". The
| |
| output of the carrier is what you actually hear. The output of the modulator alters the frequency of the
| |
| carrier, effectively acting like a supersonic vibrato. This bends and twists the carrier's waveform into
| |
| a myriad of different shapes, producing all kinds of different sounds.
| |
| | |
| Each of the 6 channels have 2 slots (a Carrier and a Modulator). Each slot behaves independently and has
| |
| its own settings and counters. Note that I will refer to "slots" often in these docs. Do not confuse
| |
| a slot for the whole channel.
| |
| | |
| "ADSR" stands for Attack/Decay/Sustain/Release. These represent 4 phases of amplitude (volume) changes in
| |
| synthesized audio. This is a common technique in all synth audio (not just FM-Synth).
| |
| - Attack is when the tone begins, and you have a rapid increase in volume, increasing to *above* the
| |
| desired output level.
| |
| - Decay is when attack has reached its maximum, and the volume starts to decline to the desired
| |
| output level.
| |
| - Sustain is when the volume has reached the desired level. It holds the volume at that level for as
| |
| long as the tone is to be played. Although sometimes the volume might slowly drop.
| |
| - Release is when the tone is done, and volume gradually decreases until it's completely silent.
| |
| | |
| "Key on" / "Key off" represents the entry and exit into ADSR. You can think of it like a piano or a keyboard...
| |
| when you "key on", you are pressing a key, and when you "key off" you are releasing a key. Effectively,
| |
| this means that when you key on, you enter "Attack", and when you key off, you enter "Release".
| |
| | |
| | |
| | |
| ---------------------------------------------
| |
| Volume and Attenuation:
| |
| ---------------------------------------------
| |
| | |
| VRC7 doesn't really have a concept of an output volume. Instead, it does everything with "attenuation",
| |
| which is basically the opposite of volume. Attenuation is like a forced reduction -- so high attenuation
| |
| means low output. Zero attenuation means the output is as high as possible.
| |
| | |
| All attenuation levels are expressed in decibels (dB), which is a logarithmic (non-linear) scale. VRC7's
| |
| threshhold or maximum attenuation is 48 dB. This means that at 48 dB, output is zero.
| |
| | |
| Note that even though dB are non-linear, you can still work with them as if they were linear. That is,
| |
| 10dB + 10dB is still 20dB. The only thing is that when converted to linear units, 20dB is MUCH
| |
| MUCH more than 2x 10dB.
| |
| | |
| Since VRC7 handles all its output levels in terms of dB, this means you will only need to convert from
| |
| dB to linear units in exactly one place: when determining the linear output of the "slot".
| |
| | |
| Converting dB <-> Linear can be accomplished with the below formulas:
| |
| | |
| dB = -20 * log10( Linear ) * scale (if Linear = 0, dB = +inf)
| |
| Linear = 10 ^ (dB / -20 / scale)
| |
| | |
| 'scale' is an optional factor you can use to scale up dB so that they're in an easier to use base.
| |
| | |
| I recommend using (1<<23)/48 for a scale (this would mean that 1<<23 would represent 48 dB). This will
| |
| make envelope calculations much easier (see Envelope Generation section for details).
| |
| | |
| Remember the threshhold is 48 dB. So if you have 48 dB or higher, Linear=0.
| |
| | |
| | |
| ---------------------------------------------
| |
| Clock rate:
| |
| ---------------------------------------------
| |
| | |
| VRC7 has its own oscillator to drive the clock rate. It's clocked at 3.6 MHz (exactly 2x the NTSC
| |
| NES CPU clock rate), but those clocks are divided by 72, effectively making the rate at which each individual
| |
| unit is clocked 49715.90909 Hz.
| |
| | |
| I find it very likely that clocking each individual unit is done serially across the 72 cycles, but the
| |
| effect that detail has on the generated audio is tiny to the point of being insignificant.
| |
| | |
| To think of this in terms of CPU cycles, you could say that all units are clocked once every 36 CPU
| |
| cycles on NTSC. However, this is techncially inaccurate, as the NES clock does not drive the VRC7.
| |
| And on PAL systems, the clock rate doesn't sync up like that.
| |
| | |
| ---------------------------------------------
| |
| Registers:
| |
| ---------------------------------------------
| |
| | |
| Register descriptions to follow. Details as to what each field actually does will not be covered here
| |
| but will be explained in future sections.
| |
| | |
| | |
| $9010: [..AA AAAA]
| |
| A = Address for use with $9030
| |
|
| |
| $9030: [DDDD DDDD] -- data port
| |
| R:00-R:07 -> Custom instrument settings (see below)
| |
|
| |
| R:1x: [FFFF FFFF] (where x=0-5, selecting the channel)
| |
| F = low 8 bits of F-Num (frequency control)
| |
|
| |
| R:2x: [..SK BBBF] (where x=0-5, selecting the channel)
| |
| F = high bit of F-Num
| |
| B = Block select (or octave)
| |
| K = Key on (1=key on, 0=key off)
| |
| S = Sustain On (poorly named, has no impact on Sustain mode -- actually affects Release)
| |
|
| |
| R:3x: [IIII VVVV]
| |
| I = Instrument select
| |
| V = 'Volume' (poorly named, it's more like "Carrier Base Attenuation Level")
| |
| | |
| (regs R:1x, 2x, and 3x apply to both Carrier and Modulator
| |
| regs R:0x apply differently to each )
| |
|
| |
| | |
| R:00: [AFPK MMMM] (applies to Modulator)
| |
| R:01: [AFPK MMMM] (applies to Carrier)
| |
| A = Enable Amplitude Modulation (AM)
| |
| F = Enable Frequency Modulation (FM)
| |
| P = Disable Percussive Mode (0=percussive, 1=normal)
| |
| K = Key Scale Rate (KSR)
| |
| M = 'MULTI' Freqency multiplier
| |
|
| |
| R:02: [KKLL LLLL]
| |
| K = Modulator Key Scale Level (KSL)
| |
| L = Modulator base attenuation level
| |
| | |
| R:03: [KK.C MFFF]
| |
| K = Carrier Key Scale Level (KSL)
| |
| C = Carrier rectify sine wave (0=full sine wave, 1=half sine wave)
| |
| M = Modulator rectify sine wave
| |
| F = Modulator Feedback level
| |
|
| |
| R:04: [AAAA DDDD] (Modulator)
| |
| R:05: [AAAA DDDD] (Carrier)
| |
| A = Attack Rate
| |
| D = Decay Rate
| |
|
| |
| R:06: [SSSS RRRR] (Modulator)
| |
| R:07: [SSSS RRRR] (Carrier)
| |
| S = Sustain Level
| |
| R = Release Rate
| |
|
| |
| There are 16 selectable instruments (selected via R:3x). Instrument 0 is configurable via regs
| |
| R:00 through R:07. The other instruments are fixed at the below values:
| |
|
| |
| 0x03,0x21,0x04,0x06,0x8D,0xF2,0x42,0x17 // instrument 1
| |
| 0x13,0x41,0x05,0x0E,0x99,0x96,0x63,0x12 // instrument 2
| |
| 0x31,0x11,0x10,0x0A,0xF0,0x9C,0x32,0x02 // instrument 3
| |
| 0x21,0x61,0x1D,0x07,0x9F,0x64,0x20,0x27 // instrument 4
| |
| 0x22,0x21,0x1E,0x06,0xF0,0x76,0x08,0x28 // instrument 5
| |
| 0x02,0x01,0x06,0x00,0xF0,0xF2,0x03,0x95 // instrument 6
| |
| 0x21,0x61,0x1C,0x07,0x82,0x81,0x16,0x07 // instrument 7
| |
| 0x23,0x21,0x1A,0x17,0xEF,0x82,0x25,0x15 // instrument 8
| |
| 0x25,0x11,0x1F,0x00,0x86,0x41,0x20,0x11 // instrument 9
| |
| 0x85,0x01,0x1F,0x0F,0xE4,0xA2,0x11,0x12 // instrument A
| |
| 0x07,0xC1,0x2B,0x45,0xB4,0xF1,0x24,0xF4 // instrument B
| |
| 0x61,0x23,0x11,0x06,0x96,0x96,0x13,0x16 // instrument C
| |
| 0x01,0x02,0xD3,0x05,0x82,0xA2,0x31,0x51 // instrument D
| |
| 0x61,0x22,0x0D,0x02,0xC3,0x7F,0x24,0x05 // instrument E
| |
| 0x21,0x62,0x0E,0x00,0xA1,0xA0,0x44,0x17 // instrument F
| |
|
| |
| | |
| **** SIDE NOTE ****
| |
| Writing to Regs R:00 through R:07 do NOT seem to have an immediate effect on channels using
| |
| instrument 0. Lagrange Point (Track 2 of the NSF) will write to these regs while a channel
| |
| using instrument 0 is still keyed on and audible, resulting in an ugly and very noticable
| |
| "blurp" noise at the end of a note. This is not heard on the real hardware, so instrument
| |
| data must be cached somehow. Perhaps it only takes effect when the channel is keyed on,
| |
| or when R:3x is written to? Don't know exactly.
| |
| | |
| | |
| | |
| | |
| ---------------------------------------------
| |
| Phase / Frequency Calculation:
| |
| ---------------------------------------------
| |
| | |
| Each slot has an 18-bit up counter which determines the current phase (position in the sine wave).
| |
| Each clock, this counter is incremented:
| |
| phase += F * (1 << B) * M * V / 2
| |
| | |
| where:
| |
| F = 9-bit F-num of the channel
| |
| B = 3-bit Block of the channel
| |
| M = see below
| |
| V = vibrato (FM) output
| |
| | |
| R:00 or R:01 specify a 4 bit 'MULTI' value. That MULTI value is run through the below LUT to get 'M':
| |
| | |
| MULTI: 0 1 2 3 4 5 6 7 8 9 A B C D E F (hex)
| |
| M: 1 2 4 6 8 10 12 14 16 18 20 20 24 24 30 30 (dec)
| |
|
| |
| If FM is enabled for the slot (see R:00 or R:01 for the enable bit), 'V' is the output of the
| |
| FM unit. See AM/FM section for details.
| |
| | |
| If FM is disabled, 'V' = 1
| |
| | |
| | |
| Another 'phase_secondary' value is used to actually generate the phase:
| |
| | |
| phase_secondary = phase + adj
| |
|
| |
| For the Carrier:
| |
| adj = the output of the Modulator. Note that slot output is 20-bits wide, but the phase is only 18
| |
| bits wide... this means that the high 2 bits of the modulator output are effectively dropped.
| |
| | |
| For the Modulator:
| |
| R:03 has a 3-bit 'F' value specifying the feedback level.
| |
| if F=0: adj = 0
| |
| otherwise: adj = previous_output_of_modulator >> (8 - F)
| |
|
| |
|
| |
| | |
| The bits of the phase_secondary value are extracted and used to generate the sine wave:
| |
| | |
| phase_secondary: [RI IIII III. .... ....]
| |
| R: rectification bit
| |
| I: index to half-sine lookup table
| |
| ** 'I' may be more or less bits depending on how big your half-sine lookup table is **
| |
| | |
| 'R' determines what to do with output after it's been converted to a linear level. See next section
| |
| for details of this bit, and details of the half-sine table.
| |
| | |
| | |
| ---------------------------------------------
| |
| Attenuation / Output calculation:
| |
| ---------------------------------------------
| |
| | |
| The attenuation level determines the slot output on each clock. Attenuation level is determined
| |
| as follows:
| |
| | |
| TOTAL = half_sine_table[I] + base + key_scale + envelope + AM
| |
| | |
|
| |
| half_sine_table[I]:
| |
| --------------
| |
| The half-sine table mentioned in the previous section does not actually hold the output of the sine
| |
| function. Rather, it holds the attenuation level of the sine function. Example:
| |
| | |
| sin(pi/2) = 1 ~~~> I='0100 0000' ~~~> half_sine_table[ I ] = 0 dB
| |
| sin(0) = 0 ~~~> I='0000 0000' ~~~> half_sine_table[ I ] = +inf dB
| |
| | |
| This table is effectively:
| |
| half_sine_table[I] = Convert_Linear_To_dB( sin( pi * I / (1 << bitwidth_of_I) ) )
| |
| | |
|
| |
|
| |
| base:
| |
| --------------
| |
| For Modulator: base = (0.75 * L), where L is the 6-bit base level (see register R:02)
| |
| For Carrier: base = (3.00 * L), where L is the 4-bit 'volume' (see register R:3x)
| |
| | |
| | |
| | |
| key_scale:
| |
| --------------
| |
| Key Scale Level, 'K', is a 2-bit value (see regs R:02, R:03) that adds attenuation as the pitch
| |
| of the tone increases (ie: higher pitches = quieter).
| |
| | |
| If K=0: key_scale=0
| |
| Otherwise:
| |
| F = high 4 bits of the current F-Num
| |
| B = 3-bit Block (Octave)
| |
| A = table[ F ] - 6 * (7-B)
| |
| | |
| if A < 0: key_scale = 0
| |
| otherwise: key_scale = A >> (3-K)
| |
|
| |
| table:
| |
| F: $0 $1 $2 $3 $4 $5 $6 $7 $8 $9 $A $B $C $D $E $F
| |
| A: 0.00 18.00 24.00 27.75 30.00 32.25 33.75 35.25 36.00 37.50 38.25 39.00 39.75 40.50 41.25 42.00
| |
|
| |
|
| |
| envelope:
| |
| --------------
| |
| Output of the envelope generator. See Envelope Generation section for details.
| |
| | |
| | |
| AM:
| |
| --------------
| |
| If Amplitude modulation is enabled for the slot (see R:00, R:01), AM is the output of the amplitude
| |
| modulation unit. Otherwise, AM=0.
| |
| | |
| See AM/FM section for details.
| |
| | |
| | |
| | |
| | |
| Finally... after all that, we have our 'TOTAL'. This is the total attenuation for the slot.
| |
| | |
| 1) This attenutation is then converted to linear units to get the preliminary output. This is
| |
| scaled up to a 20-bit value
| |
| | |
| 2) If the high bit ('R') of the 18-bit 'phase_secondary' value (see previous section) is set, this
| |
| means we are in the negative portion of the sine wave, which means output needs to be negated.
| |
| However, if we are rectifying to a half sine wave (see R:03), output is zero'd instead.
| |
| | |
| 3) Output is then run through a filter which averages this output with the previous clock's output
| |
| | |
| 4) The result is the FINAL, actual output.
| |
| | |
| Pseudo-code to clarify:
| |
| | |
| total = half_sine_table[I] + base + key_scale + envelope + AM
| |
| prevoutput = output
| |
|
| |
| // 1)
| |
| output = convert_dB_to_Linear( total ) * (1<<20)
| |
|
| |
| // 2)
| |
| if R:
| |
| if halfsine: output = 0
| |
| else: output = -output
| |
|
| |
| // 3)
| |
| FINAL = (output + prevoutput) / 2
| |
|
| |
|
| |
| 'FINAL' is what the slot actually outputs. This is a 20-bit value. The modulator's output will
| |
| be sent to the carrier, and the carrier's output will be audible (though you will want to scale it
| |
| down... 20-bit audio is crazy loud when ouputting 16-bit samples).
| |
| | |
| 'FINAL' is also the value used when calculating the modulator's feedback (see prev section).
| |
| | |
| | |
| | |
| ---------------------------------------------
| |
| Envelope Generation:
| |
| ---------------------------------------------
| |
| | |
| Each slot has a 23-bit up counter (hereon 'EGC') for envelope generation, very similar to the
| |
| 18-bit phase counter. It determines the output of the envelope generator... which adds attenuation
| |
| to the output (see previous section).
| |
| | |
| | |
| The envelope generator operates as an ADSR unit. When the channel is keyed on, both the Carrier
| |
| and the Modulator enter the Attack Phase. When keyed off, they enter Release phase.
| |
| | |
| When the ADSR unit completes a full ADSR cycle, it enters a 5th 'Idle' phase.
| |
| | |
| EGC is incremented every clock. The value by which it's incremented depends on which phase of ADSR
| |
| we're in. Those rates are then adjusted by a 'Key Scale Rate' factor (see R:00, R:01).
| |
| | |
| EGC also serves as the direct output of the envelope generator (except in the Attack phase).
| |
| When EGC=0, output is 0 dB, and whdn EGC=(1<<23), output is 48 dB. Because of this, I
| |
| recommend scaling all units in your emulator to work with dB in this (1<<23)/48 base. Doing
| |
| so results in minimal unit conversion.
| |
| | |
| Formula for determining the rate to increase EGC:
| |
| BF = (3-bit Channel Block << 1) + high bit of F-Num... forming a 4-bit value
| |
| K = Key Scale Rate bit (see R:00, R:01)
| |
| if K: KB = BF
| |
| otherwise: KB = BF >> 2
| |
| | |
| R = base rate (see subsections below)
| |
| RKS = R*4 + KB
| |
| RH = RKS >> 2 (if RH > 15, use RH=15)
| |
| RL = RKS & 3
| |
|
| |
|
| |
| The subsections below will provide a value for R, then will use RH and RL to determine
| |
| the rate by which EGC is incremented.
| |
| | |
| Note that if R=0, then EGC is not incremented at all.
| |
| | |
| Attack:
| |
| -------
| |
| R = slot attack rate (4-bits as written to R:04, R:05)
| |
| EGC += (12 * (RL+4)) << RH
| |
|
| |
| Once EGC wraps, reset EGC to zero and enter Decay phase
| |
|
| |
| Decay:
| |
| -------
| |
| R = slot decay rate (4-bits as written to R:04, R:05)
| |
| EGC += (RL+4) << (RH-1)
| |
|
| |
| Once output level reaches the slot sustain level (see R:06, R:07), set EGC to the sustain
| |
| level (do not reset it to 0!), and enter Sustain phase.
| |
|
| |
| The sustain level is (3 dB * L), where L is the 4-bit value written to the register.
| |
| This means you enter Sustain when EGC >= (3 * L * (1<<23) / 48)
| |
|
| |
| Sustain:
| |
| -------
| |
| If slot is percussive (see R:00, R:01): R = slot RELEASE rate (R:06, R:07, low bits)
| |
| otherwise: R = 0
| |
|
| |
| When EGC reaches (1<<23), output is fixed at 48 dB and enter Idle phase
| |
|
| |
| Release:
| |
| -------
| |
| If channel has "Sustain On" set (see R:2x), R = 5
| |
| otherwise, if slot is percussive: R = slot release rate (R:06, R:07)
| |
| otherwise: R = 7
| |
|
| |
| When EGC reaches (1<<23), output is fixed at 48 dB and enter Idle phase
| |
|
| |
| Idle:
| |
| -------
| |
| R=0
| |
| Output fixed at 48 dB
| |
|
| |
|
| |
| As previously mentioned, the output of the envelope generator is EGC, except in Attack phase.
| |
| In Attack, the actual rate of attack is logarithmic (it also decreases attenuation, rather than
| |
| increasing it).
| |
| | |
| attack_output = 48 dB - (48 dB * ln(EGC) / ln(1<<23))
| |
| (ln = natural log)
| |
| | |
|
| |
| ---------------------------------------------
| |
| Key On / Key Off:
| |
| ---------------------------------------------
| |
| | |
| R:2x has the Key On bit for the channel. This bit only has an impact when its state transitions.
| |
| Upon transition, do the following for both Carrier and Modulator:
| |
| | |
| When being set (0->1): (key on)
| |
| - Reset EGC to zero
| |
| - Reset 18-bit phase counter to zero
| |
| - Enter Attack phase
| |
|
| |
| When being clear (1->0): (key off)
| |
| - If currently in attack, EGC must be set to the current output level
| |
| - Enter Release phase
| |
| | |
|
| |
| ---------------------------------------------
| |
| AM/FM:
| |
| ---------------------------------------------
| |
| | |
| There is one AM unit and one FM unit. The output of these units are shared across all slots.
| |
| | |
| Both units have a 20-bit counter that is increased by 'rate' every clock.
| |
| | |
| sinx = sin(2 * pi * counter / (1<<20))
| |
| | |
| AM unit:
| |
| 'rate' = 78
| |
| AM_output = (1.0 + sinx) * 0.6 dB (emu2413 uses 1.2 dB instead of 0.6, but that sounds way too steep to me)
| |
|
| |
| See the "Attenuation / Output calculation" section for how this output is applied
| |
| | |
|
| |
| FM unit:
| |
| 'rate' = 105
| |
| FM_output = 2 ^ (13.75 / 1200 * sinx)
| |
| (note: '^' is exponent, not xor)
| |
|
| |
| See the "Phase / Frequency Calculation" section for how this output is applied
| |
| </pre>
| |
