windows-nt/Source/XPSP1/NT/drivers/ddk/wdmaudio/ddksynth/synth.h
2020-09-26 16:20:57 +08:00

1147 lines
47 KiB
C++

// Synth.h
// Copyright (c) 1996-2000 Microsoft Corporation. All Rights Reserved.
//
/*
DLS Design Overview (DLS Level 1 Specification, MMA)
A musical instruments defined by a sound sample is much more than a simple wave file.
In addition to the actual sample data and associated loop information, the instrument
must indicate under what circumstances each sample should be used, and how to modulate,
or articulate, the sample as it plays.
A generic sample-playback synthesis architecture can be broken down into three distinct
subsystems:
Control logic
Digital audio engine
Articulation modules and connections
Control Logic
The control logic receives a MIDI note event and determines which instrument should play
the note, and, within that instrument, which sample and articulation combination to use.
Choosing the instrument is little more than observing the MIDI channel number in the
event and selecting the proper instrument accordingly.
Choosing the sample and articulation to use is not as simple. Almost all samples synthesis
architectures employ some method of organizing samples by note range across the keyboard.
In addition, multiple samples can be used to represent different velocity randges and
multiple samples can be played at once to create a layered, richer sound.
Terms such as layers, splits, blocks, and regions are commonly used in synthesizer jargon
to refer to the management of multiple samples. For the purposes here, we refer to them
as regions.
DLS Level 1 implements a bare bones approach with no velocity cross-switching and no
layering.
Digital Audio Engine
The digital audio engine is certainly the most obvious part of the synthesizer. It is
composed of a playback engine, or digital oscillator, and a digitally controlled amplifier.
The digital oscillator plays a samples sound waveform, managing loop points within the
waveform so the sound can play continuously if needed. And, as the note plays, it responds
to changes in pitch, allowing for real time expression such as vibrato and pitch bend.
The digitally controlled amplirier modulates the loudness of the instrument. Importantly,
this is used to control the amplitude shape, or envelope, of the note. However, it is also
used for other types of real-time expression, such as tremolo.
Pitch and volume control of the oscillator and amplifier are critical because they define
the shape of the sounds as it plays, and allow it to be dynamically responsive in real
time, giving the samples instrument much more expression than the simple digital audio
playback could ever provide. Real-time control of these parameters comes from modules in
the Articulation section which generate a constant stream of changing pitch and volume to
which the digital audio engine responds.
The digital audio path represents the journey the sound takes from the oscillator to the
amplifier to the digital-to-analog converter (DAC). This patch can optionally include
additional modules, such as filters and effects devices, that process the sound as it
flows from oscillator to DAC.
The DLS Level 1 specificaion implements a simple digital audio engine composed of an
oscillator, amplifier, and DAC. The oscillator supports looped as well as one-shot samples.
Articulation Modules and Connections
The articulation modules are a set of devices that provide additional control over the
pitch and volume of the sample as it plays.
The articulation modules include low frequency oscillators (LFOs) to contribute vibrato
and tremolo, envelope generators to defint an overall volume and pitch shape to the sample,
and MIDI real-time controllers, such as Pitch Bend and Mod Wheel, to give the music real-time
expression.
Generally, these modules can be linked in different ways to provide different results. For
example, a LFO might generate a sine wave which modulates the pitch of the sample for vibrato
of the volume of the sample for tremolo. Modules can also receive as well as send control
signals. An envelope generator might use the key velocity to influence the attack time of
the envelope.
Articulation modules can be configured in different ways, and this configuration is an
important part of the instrument definition. In fact, the term patch as used for an
instrument preset refers to the early days when the hardware modules in an analog
synthesizer were "patched" together with cables, which routed signals from module to module.
The ability to configure the modules in important because it yields a flexible approach to
synthesizer design. At the same time, it is important to define the base level configuration
that can be supported by all hardware.
So, the DLS Level 1 specification is relatively rigid. It defines a preset routing
arrangement that is simple enough to be supportable by existing hardware. Fortunately, it
is a musically logical and useful configuration.
Importantly, the specification maps the routing on a flexible system, so it can grow into
Level 2 and beyond.
It is important to understand that this is purely a symbolic system that can be mapped on
many different hardware designs. Do not think of it as the recipe for a specific method
for building synthesis architectures. It should be flexible enough to provide a common
language between hardware implementations.
*/
#ifndef __SYNTH_H__
#define __SYNTH_H__
#include "clist.h"
#include "dmdls.h"
#define MIDI_NOTEOFF 0x80
#define MIDI_NOTEON 0x90
#define MIDI_PTOUCH 0xA0
#define MIDI_CCHANGE 0xB0
#define MIDI_PCHANGE 0xC0
#define MIDI_MTOUCH 0xD0
#define MIDI_PBEND 0xE0
#define MIDI_SYSX 0xF0
#define MIDI_MTC 0xF1
#define MIDI_SONGPP 0xF2
#define MIDI_SONGS 0xF3
#define MIDI_EOX 0xF7
#define MIDI_CLOCK 0xF8
#define MIDI_START 0xFA
#define MIDI_CONTINUE 0xFB
#define MIDI_STOP 0xFC
#define MIDI_SENSE 0xFE
// controller numbers
#define CC_BANKSELECTH 0x00
#define CC_BANKSELECTL 0x20
#define CC_MODWHEEL 0x01
#define CC_VOLUME 0x07
#define CC_PAN 0x0A
#define CC_EXPRESSION 0x0B
#define CC_SUSTAIN 0x40
#define CC_ALLSOUNDSOFF 0x78
#define CC_RESETALL 0x79
#define CC_ALLNOTESOFF 0x7B
#define CC_MONOMODE 0x7E
#define CC_POLYMODE 0x7F
// rpn controllers
#define CC_DATAENTRYMSB 0x06
#define CC_DATAENTRYLSB 0x26
#define CC_NRPN_LSB 0x62
#define CC_NRPN_MSB 0x63
#define CC_RPN_LSB 0x64
#define CC_RPN_MSB 0x65
// registered parameter numbers
#define RPN_PITCHBEND 0x00
#define RPN_FINETUNE 0x01
#define RPN_COARSETUNE 0x02
/* Sample format and Sample playback flags are organized
together because together they determine which
mix loop to use.
*/
#define SFORMAT_16 1 // Sixteen bit sample.
#define SFORMAT_8 2 // Eight bit sample.
#define SPLAY_MMX 0x10 // Use MMX processor (16 bit only).
#define SPLAY_STEREO 0x40 // Stereo output.
/* For internal representation, volume is stored in Volume Cents,
where each increment represents 1/100 of a dB.
Pitch is stored in Pitch Cents, where each increment
represents 1/100 of a semitone.
*/
typedef long PREL; // Pitch cents, for relative pitch.
typedef short PRELS; // Pitch cents, in storage form.
typedef long VREL; // Volume cents, for relative volume.
typedef short VRELS; // Volume cents, in storage form.
typedef long TREL; // Time cents, for relative time
typedef short TRELS; // Time Cents, in storage form.
typedef LONGLONG STIME; // Time value, in samples.
typedef long MTIME; // Time value, in milliseconds.
typedef long PFRACT; // Pitch increment, where upper 20 bits are
// the index and the lower 12 are the fractional
// component.
typedef long VFRACT; // Volume, where lower 12 bits are the fraction.
typedef long TCENT;
typedef short SPERCENT;
#define MIN_VOLUME -9600 // Below 96 db down is considered off.
#define PERCEIVED_MIN_VOLUME -8000 // But, we cheat.
#define SAMPLE_RATE_22 22050 // 22 kHz is the standard rate.
#define SAMPLE_RATE_44 44100 // 44 kHz is the high quality rate.
#define SAMPLE_RATE_11 11025 // 11 kHz should not be allowed!
#define STEREO_ON 1
#define STEREO_OFF 0
#define FORCEBOUNDS(data,min,max) {if (data < min) data = min; else if (data > max) data = max;}
#define FORCEUPPERBOUNDS(data, max) {if (data > max) data = max;}
class Collection;
class CControlLogic;
class CInstManager;
class CSynth;
/*****************************************************************************
* class CSourceLFO
*****************************************************************************
* CSourceLFO is the file format definition of the LFO in an
* instrument. This is used to represent an LFO as part of
* a specific articulation set within an instrument that
* has been loaded from disk. Once the instrument is chosen
* to play a note, this is also copied into the CVoice
* object.
*/
class CSourceLFO
{
public:
CSourceLFO();
void Init(DWORD dwSampleRate);
void SetSampleRate(long lDirection);
void Verify(); // Verifies that the data is valid.
PFRACT m_pfFrequency; // Frequency, in increments through the sine table.
STIME m_stDelay; // How long to delay in sample units.
VRELS m_vrMWVolumeScale; // Scaling of volume LFO by Mod Wheel.
PRELS m_prMWPitchScale; // Scaling of pitch LFO by Mod Wheel.
VRELS m_vrVolumeScale; // Scaling of straight volume signal from LFO.
PRELS m_prPitchScale; // Scaling of straight pitch signal from LFO.
};
/*****************************************************************************
* class CSourceEG
*****************************************************************************
* CSourceEG is the file format definition of an Envelope
* generator in an instrument.
*/
class CSourceEG
{
public:
CSourceEG();
void SetSampleRate(long lDirection);
void Init();
void Verify(); // Verifies valid data.
STIME m_stAttack; // Attack rate.
STIME m_stDecay; // Decay rate.
STIME m_stRelease; // Release rate.
TRELS m_trVelAttackScale; // Scaling of attack by note velocity.
TRELS m_trKeyDecayScale; // Scaling of decay by note value.
SPERCENT m_pcSustain; // Sustain level.
short m_sScale; // Scaling of entire signal.
};
/*****************************************************************************
* class CSourceArticulation
*****************************************************************************
* CSourceArticulation is the file format definition of
* a complete articulation set: the LFO and two
* envelope generators.
*
* Since several regions within one Instrument can
* share one articulation, a counter is used to keep
* track of the usage.
*/
class CSourceArticulation
{
public:
CSourceArticulation();
HRESULT Download( DMUS_DOWNLOADINFO * pInfo,
void * pvOffsetTable[], DWORD dwIndex,
DWORD dwSampleRate, BOOL fNewFormat);
void Init(DWORD dwSampleRate);
void Verify(); // Verifies valid data.
void AddRef();
void Release();
void SetSampleRate(DWORD dwSampleRate);
CSourceEG m_PitchEG; // Pitch envelope.
CSourceEG m_VolumeEG; // Volume envelope.
CSourceLFO m_LFO; // Low frequency oscillator.
DWORD m_dwSampleRate;
WORD m_wUsageCount; // Keeps track of how many times in use.
short m_sDefaultPan; // default pan (for drums)
};
/*****************************************************************************
* class CWave
*****************************************************************************
* Since multiple regions may reference
* the same Wave, a reference count is maintained to
* keep track of how many regions are using the sample.
*/
class CWave : public CListItem
{
public:
CWave();
~CWave();
void Verify(); // Verifies that the data is valid.
void Release(); // Remove reference.
void AddRef(); // Add reference.
void PlayOn(); // Increment play count.
void PlayOff(); // Decrement play count.
BOOL IsPlaying(); // Is currently playing?
CWave * GetNext() {return(CWave *)CListItem::GetNext();};
DWORD m_dwSampleLength; // Length of sample.
DWORD m_dwSampleRate;
HRESULT ( CALLBACK *m_lpFreeHandle)(HANDLE,HANDLE);
HANDLE m_hUserData; // Used to notify app when wave released.
short * m_pnWave;
DWORD m_dwID; // ID for matching wave with regions.
WORD m_wUsageCount; // Keeps track of how many times in use.
WORD m_wPlayCount; // Wave is currently being played.
BYTE m_bSampleType;
DMUS_DOWNLOADINFO * m_pWaveMem;
};
/*****************************************************************************
* class CWavePool
*****************************************************************************
* implements a list of wave objects.
*/
class CWavePool : public CList
{
public:
CWave * GetHead() {return (CWave *)CList::GetHead();};
CWave * GetItem(DWORD dwID) {return (CWave *)CList::GetItem((LONG)dwID);};
CWave * RemoveHead() {return (CWave *)CList::RemoveHead();};
};
/*****************************************************************************
* class CSourceSample
*****************************************************************************
* The CSourceSample class describes one sample in an
* instrument. The sample is referenced by a CSourceRegion
* structure.
*/
class CSourceSample
{
public:
CSourceSample();
~CSourceSample();
BOOL CopyFromWave();
void Verify(); // Verifies that the data is valid.
CWave * m_pWave; // Wave in pool.
DWORD m_dwLoopStart; // Index of start of loop.
DWORD m_dwLoopEnd; // Index of end of loop.
DWORD m_dwSampleLength; // Length of sample.
DWORD m_dwSampleRate; // Sample rate of recording.
PRELS m_prFineTune; // Fine tune to correct pitch.
DWORD m_dwID; // Wave pool id.
BYTE m_bSampleType; // 16 or 8.
BYTE m_bOneShot; // Is this a one shot sample?
BYTE m_bMIDIRootKey; // MIDI note number for sample.
};
/*****************************************************************************
* class CSourceRegion
*****************************************************************************
* The CSourceRegion class defines a region within an instrument.
* The sample is managed with a pointer instead of an embedded
* sample. This allows multiple regions to use the same sample.
*
* Each region also has an associated articulation. For drums, there
* is a one to one matching. For melodic instruments, all regions
* share the same articulation. So, to manage this, each region
* points to the articulation.
*/
class CSourceRegion : public CListItem
{
public:
CSourceRegion();
~CSourceRegion();
CSourceRegion *GetNext()
{return(CSourceRegion *)CListItem::GetNext();};
void Verify(); // Verifies that the data is valid.
void SetSampleRate(DWORD dwSampleRate);
HRESULT Download( DMUS_DOWNLOADINFO * pInfo, void * pvOffsetTable[],
DWORD *pdwRegionIX, DWORD dwSampleRate, BOOL fNewFormat);
CSourceSample m_Sample; // Sample structure.
CSourceArticulation * m_pArticulation; // Pointer to associated articulation.
VRELS m_vrAttenuation; // Volume change to apply to sample.
PRELS m_prTuning; // Pitch shift to apply to sample.
BYTE m_bAllowOverlap; // Allow overlapping of note.
BYTE m_bKeyHigh; // Upper note value for region.
BYTE m_bKeyLow; // Lower note value.
BYTE m_bGroup; // Logical group (for drums.)
};
/*****************************************************************************
* class CSourceRegionList
*****************************************************************************
* implements a list of CSourceRegion objects.
*/
class CSourceRegionList : public CList
{
public:
CSourceRegion *GetHead() {return (CSourceRegion *)CList::GetHead();};
CSourceRegion *RemoveHead() {return (CSourceRegion *)CList::RemoveHead();};
};
/*****************************************************************************
* class CInstrument
*****************************************************************************
* The CInstrument class is really the file format definition
* of an instrument.
*
* The CInstrument can be either a Drum or a Melodic instrument.
* If a drum, it has up to 128 pairings of articulations and
* regions. If melodic, all regions share the same articulation.
* ScanForRegion is called by ControlLogic to get the region
* that corresponds to a note.
*/
class CInstrument : public CListItem
{
public:
CInstrument();
~CInstrument();
void Init(DWORD dwSampleRate);
void Verify(); // Verifies that the data is valid.
CInstrument * GetInstrument(DWORD dwProgram,DWORD dwAccept);
CInstrument * GetNext()
{return(CInstrument *)CListItem::GetNext();};
void SetSampleRate(DWORD dwSampleRate);
CSourceRegion * ScanForRegion(DWORD dwNoteValue);
HRESULT LoadRegions( BYTE *p, BYTE *pEnd, DWORD dwSampleRate);
HRESULT Load( BYTE *p, BYTE *pEnd, DWORD dwSampleRate);
CSourceRegionList m_RegionList; // Linked list of regions.
DWORD m_dwProgram; // Which program change it represents.
};
/*****************************************************************************
* class CInstrumentList
*****************************************************************************
* Implements a list of CInstrument objects.
*/
class CInstrumentList : public CList
{
public:
CInstrument * GetHead() {return (CInstrument *)CList::GetHead();};
CInstrument * RemoveHead() {return (CInstrument *)CList::RemoveHead();};
};
#define WAVE_HASH_SIZE 31 // Keep waves in a hash table of linked lists to speed access.
#define INSTRUMENT_HASH_SIZE 31 // Same with instruments.
/*****************************************************************************
* class CInstManager
*****************************************************************************
* Manages the instruments, including sample rates, downloads, and waves.
* Utilizes a hash scheme for quick location of waves and instruments
* (they can become numerous).
*/
class CInstManager
{
public:
CInstManager();
~CInstManager();
CInstrument * GetInstrument(DWORD dwPatch,DWORD dwKey);
void Verify(); // Verifies that the data is valid.
void SetSampleRate(DWORD dwSampleRate);
HRESULT Download(LPHANDLE phDownload,
void * pvData,
LPBOOL pbFree);
HRESULT Unload(HANDLE hDownload,
HRESULT ( CALLBACK *lpFreeHandle)(HANDLE,HANDLE),
HANDLE hUserData);
private:
HRESULT DownloadInstrument(LPHANDLE phDownload,
DMUS_DOWNLOADINFO *pInfo,
void *pvOffsetTable[],
void *pvData,
BOOL fNewFormat);
HRESULT DownloadWave(LPHANDLE phDownload,
DMUS_DOWNLOADINFO *pInfo,
void *pvOffsetTable[],
void *pvData);
CInstrumentList m_InstrumentList[INSTRUMENT_HASH_SIZE];
CWavePool m_WavePool[WAVE_HASH_SIZE];
CWavePool m_FreeWavePool; // Track waves still in use, but unloaded.
DWORD m_dwSampleRate; // Sample rate requested by app.
public:
CRITICAL_SECTION m_CriticalSection; // Critical section to manage access.
BOOL m_fCSInitialized;
};
/*****************************************************************************
* class CMIDIData
*****************************************************************************
* Represents a single MIDI event.
*/
class CMIDIData : public CListItem
{
public:
CMIDIData();
CMIDIData * GetNext() {return (CMIDIData *)CListItem::GetNext();};
STIME m_stTime; // Time this event was recorded.
long m_lData; // Data stored in event.
};
/*****************************************************************************
* class CMIDIDataList
*****************************************************************************
* Implements a list of CMIDIData objects.
*/
class CMIDIDataList : public CList
{
public:
CMIDIData *GetHead() {return (CMIDIData *)CList::GetHead();};
CMIDIData *RemoveHead() {return (CMIDIData *)CList::RemoveHead();};
};
#define MAX_MIDI_EVENTS 1000
/*****************************************************************************
* class CMIDIRecorder
*****************************************************************************
* CMIDIRecorder is used to keep track of a time
* slice of MIDI continuous controller events.
* This is subclassed by the PitchBend, Volume,
* Expression, and ModWheel Recorder classes, so
* each of them may reliably manage MIDI events
* coming in.
*
* CMIDIRecorder uses a linked list of CMIDIData
* structures to keep track of the changes within
* the time slice.
*
* Allocation and freeing of the CMIDIData events
* is kept fast and efficient because they are
* always pulled from the static pool m_pFreeList,
* which is really a list of events pulled directly
* from the static array m_sEventBuffer. This is
* safe because we can make the assumption that
* the maximum MIDI rate is 1000 events per second.
* Since we are managing time slices of roughly
* 1/16 of a second, a buffer of 100 events would
* be overkill.
*
* Although CMIDIRecorder is subclassed to several
* different event types, they all share the one
* static free list.
*/
class CMIDIRecorder
{
public:
CMIDIRecorder();
~CMIDIRecorder(); // Be sure to clear local list.
void Init(); // Inits the free list.
static void InitTables();
BOOL FlushMIDI(STIME stTime); // Clear after time stamp.
BOOL ClearMIDI(STIME stTime); // Clear up to time stamp.
BOOL RecordMIDI(STIME stTime, long lData); // MIDI input goes here.
long GetData(STIME stTime); // Gets data at time.
static VREL VelocityToVolume(WORD nVelocity);
static CMIDIDataList * m_sFreeList; // Free list of events.
static ULONG sm_cRefCnt; // ref count
protected:
void GrabSpinLock();
void ReleaseSpinLock();
CMIDIDataList m_EventList; // This recorder's list.
STIME m_stCurrentTime; // Time for current value.
long m_lCurrentData; // Current value.
private:
KSPIN_LOCK m_SpinLock;
KIRQL m_OldIrql;
};
/*****************************************************************************
* class CNote
*****************************************************************************
* Represents a single note. This can also represent fakes notes that
* represent special MIDI commands (Master Volume, etc).
*/
class CNote
{
public:
STIME m_stTime;
BYTE m_bPart;
BYTE m_bKey;
BYTE m_bVelocity;
};
/*****************************************************************************
* Fake note values held in CNoteIn's queue to indicate changes in the
* sustain pedal and "all notes off".
*
* This is a grab bag for synchronous events that should be queued in time,
* not simply done as soon as received.
*
* By putting them in the note queue, we ensure they are evaluated in the
* exact same order as the notes themselves.
*****************************************************************************/
const BYTE NOTE_PROGRAMCHANGE = 0xF1;
const BYTE NOTE_CC_BANKSELECTH = 0xF2;
const BYTE NOTE_CC_BANKSELECTL = 0xF3;
const BYTE NOTE_CC_POLYMODE = 0xF4;
const BYTE NOTE_CC_MONOMODE = 0xF5;
const BYTE NOTE_CC_RPN_MSB = 0xF6;
const BYTE NOTE_CC_RPN_LSB = 0xF7;
const BYTE NOTE_CC_NRPN = 0xF8;
const BYTE NOTE_CC_DATAENTRYLSB = 0xF9;
const BYTE NOTE_CC_DATAENTRYMSB = 0xFA;
const BYTE NOTE_ASSIGNRECEIVE = 0xFB;
const BYTE NOTE_MASTERVOLUME = 0xFC;
const BYTE NOTE_SOUNDSOFF = 0xFD;
const BYTE NOTE_SUSTAIN = 0xFE;
const BYTE NOTE_ALLOFF = 0xFF;
/*****************************************************************************
* class CNoteIn
*****************************************************************************
* Implements a note receptor. This is used by CControlLogic to queue notes.
*/
class CNoteIn : public CMIDIRecorder
{
public:
void FlushMIDI(STIME stTime);
void FlushPart(STIME stTime, BYTE bChannel);
BOOL RecordNote(STIME stTime, CNote * pNote);
BOOL RecordEvent(STIME stTime, DWORD dwPart, DWORD dwCommand, BYTE bData);
BOOL GetNote(STIME stTime, CNote * pNote); // Gets the next note.
};
/*****************************************************************************
* class CModWheelIn
*****************************************************************************
* CModWheelIn handles one channel of Mod Wheel
* input. As such, it is not embedded in the CVoice
* class, rather it is in the Channel class.
* CModWheelIn's task is simple: keep track of MIDI
* Mod Wheel events, each tagged with millisecond
* time and value, and return the value for a specific
* time request.
*
* CModWheelIn inherits almost all of its functionality
* from the CMIDIRecorder Class.
* CModWheelIn receives MIDI mod wheel events through
* the RecordMIDI() command, which stores the
* time and value of the event.
*
* CModWheelIn is called by CVoiceLFO to get the
* current values for the mod wheel to set the amount
* of LFO modulation for pitch and volume.
*/
class CModWheelIn : public CMIDIRecorder
{
public:
DWORD GetModulation(STIME stTime); // Gets the current Mod Wheel value.
};
/*****************************************************************************
* class CPitchBendIn
*****************************************************************************
* CPitchBendIn handles one channel of Pitch Bend
* input. Like the Mod Wheel module, it inherits
* its abilities from the CMIDIRecorder class.
*
* It has one additional routine, GetPitch(),
* which returns the current pitch bend value.
*/
class CPitchBendIn : public CMIDIRecorder
{
public:
CPitchBendIn();
PREL GetPitch(STIME stTime); // Gets the current pitch in pitch cents.
// current pitch bend range. Note that this is not timestamped!
PREL m_prRange;
};
/*****************************************************************************
* class CVolumeIn
*****************************************************************************
* CVolumeIn handles one channel of Volume
* input. It inherits its abilities from
* the CMIDIRecorder class.
*
* It has one additional routine, GetVolume(),
* which returns the volume in decibels at the
* specified time.
*/
class CVolumeIn : public CMIDIRecorder
{
public:
CVolumeIn();
VREL GetVolume(STIME stTime); // Gets the current volume in db cents.
};
/*****************************************************************************
* class CExpressionIn
*****************************************************************************
* CExpressionIn handles one channel of Expression
* input. It inherits its abilities from
* the CMIDIRecorder class.
*
* It has one additional routine, GetVolume(),
* which returns the volume in decibels at the
* specified time.
*/
class CExpressionIn : public CMIDIRecorder
{
public:
CExpressionIn();
VREL GetVolume(STIME stTime); // Gets the current volume in db cents.
};
/*****************************************************************************
* class CPanIn
*****************************************************************************
* CPanIn handles one channel of Volume
* input. It inherits its abilities from
* the CMIDIRecorder class.
*
* It has one additional routine, GetPan(),
* which returns the pan position (MIDI value)
* at the specified time.
*/
class CPanIn : public CMIDIRecorder
{
public:
CPanIn();
long GetPan(STIME stTime); // Gets the current pan.
};
/*****************************************************************************
* class CVoiceLFO
*****************************************************************************
* The CVoiceLFO class is used to track the behavior
* of an LFO within a voice. The LFO is hard wired to
* output both volume and pitch values, through separate
* calls to GetVolume and GetPitch.
*
* It also manages mixing Mod Wheel control of pitch and
* volume LFO output. It tracks the scaling of Mod Wheel
* for each of these in m_nMWVolumeScale and m_nMWPitchScale.
* It calls the Mod Wheel module to get the current values
* if the respective scalings are greater than 0.
*
* All of the preset values for the LFO are carried in
* the m_CSource field, which is a replica of the file
* CSourceLFO structure. This is initialized with the
* StartVoice call.
*/
class CVoiceLFO
{
public:
CVoiceLFO();
STIME StartVoice(CSourceLFO *pSource,
STIME stStartTime,CModWheelIn * pModWheelIn);
VREL GetVolume( STIME stTime, STIME *pstTime); // Returns volume cents.
PREL GetPitch( STIME stTime, STIME *pstTime); // Returns pitch cents.
private:
long GetLevel( STIME stTime, STIME *pstTime);
CSourceLFO m_Source; // All of the preset information.
STIME m_stStartTime; // Time the voice started playing.
CModWheelIn *m_pModWheelIn; // Pointer to Mod Wheel for this channel.
STIME m_stRepeatTime; // Repeat time for LFO.
};
/*****************************************************************************
* class CVoiceEG
*****************************************************************************
* The CVoiceEG class is used to track the behavior of
* an Envelope Generator within a voice. There are two
* EG's, one for pitch and one for volume. However, they
* behave identically.
*
* All of the preset values for the EG are carried in
* the m_Source field, which is a replica of the file
* CSourceEG structure. This is initialized with the
* StartVoice call.
*/
class CVoiceEG
{
public:
CVoiceEG();
STIME StartVoice(CSourceEG *pSource, STIME stStartTime,
WORD nKey, WORD nVelocity);
void StopVoice(STIME stTime);
void QuickStopVoice(STIME stTime, DWORD dwSampleRate);
VREL GetVolume(STIME stTime, STIME *pstTime); // Returns volume cents.
PREL GetPitch(STIME stTime, STIME *pstTime); // Returns pitch cents.
BOOL InAttack(STIME stTime); // is voice still in attack?
BOOL InRelease(STIME stTime); // is voice in release?
private:
long GetLevel(STIME stTime, STIME *pstTime, BOOL fVolume);
CSourceEG m_Source; // Preset values for envelope, copied from file.
STIME m_stStartTime; // Time note turned on
STIME m_stStopTime; // Time note turned off
};
/*****************************************************************************
* class CDigitalAudio
*****************************************************************************
* The CDigitalAudio class is used to track the playback
* of a sample within a voice.
*
* It manages the loop points, the pointer to the sample.
* and the base pitch and base volume, which it initially sets
* when called via StartVoice().
*
* Pitch is stored in a fixed point format, where the leftmost
* 20 bits define the sample increment and the right 12 bits
* define the factional increment within the sample. This
* format is also used to track the position in the sample.
* Mix is a critical routine. It is called by the CVoice to blend
* the instrument into the data buffer. It is handed relative change
* values for pitch and volume (semitone cents and decibel
* cents.) These it converts into three linear values:
* Left volume, Right volume, and Pitch.
*
* It then compares these new values with the values that existed
* for the previous slice and divides by the number of samples to
* determine an incremental change at the sample rate.
* Then, in the critical mix loop, these are added to the
* volume and pitch indices to give a smooth linear slope to the
* change in volume and pitch.
*/
#define MAX_SAMPLE 4095
#define MIN_SAMPLE (-4096)
#define MAXDB 0
#define MINDB -100
#define TEST_WRITE_SIZE 3000
#define TEST_SOURCE_SIZE 44100
class CDigitalAudio
{
public:
CDigitalAudio();
STIME StartVoice(CSynth *pSynth,
CSourceSample *pSample,
VREL vrBaseLVolume, VREL vrBaseRVolume,
PREL prBasePitch, long lKey);
BOOL Mix(short *pBuffer,DWORD dwLength,
VREL dwVolumeL, VREL dwVolumeR, PREL dwPitch,
DWORD dwStereo);
void ClearVoice();
static PFRACT PRELToPFRACT(PREL prPitch); // Pitch cents to pitch.
private:
DWORD Mix8(short * pBuffer, DWORD dwLength,DWORD dwDeltaPeriod,
VFRACT vfDeltaLVolume, VFRACT vfDeltaRVolume,
PFRACT pfDeltaPitch,
PFRACT pfSampleLength, PFRACT pfLoopLength);
DWORD MixMono8(short * pBuffer, DWORD dwLength,DWORD dwDeltaPeriod,
VFRACT vfDeltaVolume,
PFRACT pfDeltaPitch,
PFRACT pfSampleLength, PFRACT pfLoopLength);
DWORD Mix16(short * pBuffer, DWORD dwLength, DWORD dwDeltaPeriod,
VFRACT vfDeltaLVolume, VFRACT vfDeltaRVolume,
PFRACT pfDeltaPitch,
PFRACT pfSampleLength, PFRACT pfLoopLength);
DWORD MixMono16(short * pBuffer, DWORD dwLength,DWORD dwDeltaPeriod,
VFRACT vfDeltaVolume,
PFRACT pfDeltaPitch,
PFRACT pfSampleLength, PFRACT pfLoopLength);
DWORD _cdecl Mix8X(short * pBuffer, DWORD dwLength, DWORD dwDeltaPeriod,
VFRACT vfDeltaLVolume, VFRACT vfDeltaRVolume,
PFRACT pfDeltaPitch,
PFRACT pfSampleLength, PFRACT pfLoopLength);
DWORD _cdecl Mix16X(short * pBuffer, DWORD dwLength, DWORD dwDeltaPeriod,
VFRACT vfDeltaLVolume, VFRACT vfDeltaRVolume,
PFRACT pfDeltaPitch,
PFRACT pfSampleLength, PFRACT pfLoopLength);
DWORD MixMono16X(short * pBuffer, DWORD dwLength,DWORD dwDeltaPeriod,
VFRACT vfDeltaVolume,
PFRACT pfDeltaPitch,
PFRACT pfSampleLength, PFRACT pfLoopLength);
DWORD MixMono8X(short * pBuffer, DWORD dwLength,DWORD dwDeltaPeriod,
VFRACT vfDeltaVolume,
PFRACT pfDeltaPitch,
PFRACT pfSampleLength, PFRACT pfLoopLength);
void BeforeBigSampleMix();
void AfterBigSampleMix();
static VFRACT VRELToVFRACT(VREL vrVolume); // dB to absolute.
CSourceSample m_Source; // Preset values for sample.
CSynth * m_pSynth; // For access to sample rate, etc.
BOOL m_fMMXEnabled;
short * m_pnWave; // Private pointer to wave.
VREL m_vrBaseLVolume; // Overall left volume.
VREL m_vrBaseRVolume; // Overall left volume.
PFRACT m_pfBasePitch; // Overall pitch.
VFRACT m_vfLastLVolume; // The last left volume value.
VFRACT m_vfLastRVolume; // The last right volume value.
PFRACT m_pfLastPitch; // The last pitch value.
VREL m_vrLastLVolume; // The last left volume value, in VREL.
VREL m_vrLastRVolume; // Same for right.
PREL m_prLastPitch; // Same for pitch, in PREL.
PFRACT m_pfLastSample; // The last sample position.
PFRACT m_pfLoopStart; // Start of loop.
PFRACT m_pfLoopEnd; // End of loop.
PFRACT m_pfSampleLength; // Length of sample buffer.
BOOL m_fElGrande; // Indicates larger than 1m wave.
ULONGLONG m_ullLastSample; // Used to track > 1m wave.
ULONGLONG m_ullLoopStart; // Used to track > 1m wave.
ULONGLONG m_ullLoopEnd; // Used to track > 1m wave.
ULONGLONG m_ullSampleLength; // Used to track > 1m wave.
DWORD m_dwAddressUpper; // Temp storage for upper bits of address.
};
/*****************************************************************************
* class CVoice
*****************************************************************************
* The CVoice class pulls together everything needed to perform
* one voice. It has the envelopes, lfo, and sample embedded
* within it.
*
* StartVoice() initializes a voice structure for playback. The
* CSourceRegion structure carries the region and sample as well
* as a pointer to the articulation, which is used to set up
* the various articulation modules. It also carries pointers to
* all the MIDI modulation inputs and the values for the note key
* and channel which are used by the parent ControlLogic object
* to match incoming note off events with the right voice.
*/
class CVoice : public CListItem
{
public:
CVoice();
CVoice * GetNext()
{return (CVoice *)CListItem::GetNext();};
BOOL StartVoice(CSynth *pControl,
CSourceRegion *pRegion, STIME stStartTime,
CModWheelIn * pModWheelIn,
CPitchBendIn * pPitchBendIn,
CExpressionIn * pExpressionIn,
CVolumeIn * pVolumeIn,
CPanIn * pPanIn,
WORD nKey,WORD nVelocity,
VREL vrVolume, // Added for GS
PREL prPitch); // Added for GS
void StopVoice(STIME stTime); // Called on note off event.
void QuickStopVoice(STIME stTime); // Called to get quick release.
void SpeedRelease(); // Force an already off envelope to release quickly.
void ClearVoice(); // Release use of sample.
PREL GetNewPitch(STIME stTime); // Return current pitch value
void GetNewVolume(STIME stTime, VREL& vrVolume, VREL &vrVolumeR);
// Return current volume value
DWORD Mix(short *pBuffer,DWORD dwLength,STIME stStart,STIME stEnd);
private:
CVoiceLFO m_LFO; // LFO.
CVoiceEG m_PitchEG; // Pitch Envelope.
CVoiceEG m_VolumeEG; // Volume Envelope.
CDigitalAudio m_DigitalAudio; // The Digital Audio Engine structure.
CPitchBendIn * m_pPitchBendIn; // Pitch bend source.
CExpressionIn * m_pExpressionIn; // Expression source.
CVolumeIn * m_pVolumeIn; // Volume source, if allowed to vary
CPanIn * m_pPanIn; // Pan source, if allowed to vary
CSynth * m_pSynth; // To access sample rate, etc.
STIME m_stMixTime; // Next time we need a mix.
long m_lDefaultPan; // Default pan
STIME m_stLastMix; // Last sample position mixed.
public:
STIME m_stStartTime; // Time the sound starts.
STIME m_stStopTime; // Time the sound stops.
BOOL m_fInUse; // This is currently in use.
BOOL m_fNoteOn; // Note is considered on.
BOOL m_fTag; // Used to track note stealing.
VREL m_vrVolume; // Volume, used for voice stealing...
BOOL m_fSustainOn; // Sus pedal kept note on after off event.
WORD m_nPart; // Part that is playing this (channel).
WORD m_nKey; // Note played.
BOOL m_fAllowOverlap; // Allow overlapped note.
DWORD m_dwGroup; // Group this voice is playing now
DWORD m_dwProgram; // Bank and Patch choice.
DWORD m_dwPriority; // Priority.
CControlLogic * m_pControl; // Which control group is playing voice.
};
/*****************************************************************************
* class CVoiceList
*****************************************************************************
* Implements a list of CVoice objects.
*/
class CVoiceList : public CList
{
public:
CVoice * GetHead() {return (CVoice *) CList::GetHead();};
CVoice * RemoveHead() {return (CVoice *) CList::RemoveHead();};
CVoice * GetItem(LONG lIndex) {return (CVoice *) CList::GetItem(lIndex);};
};
/*****************************************************************************
* struct PerfStats
*****************************************************************************
* Contains statistics on the synthesizer, updated continuously.
*/
typedef struct PerfStats
{
DWORD dwTotalTime;
DWORD dwTotalSamples;
DWORD dwNotesLost;
DWORD dwVoices;
DWORD dwCPU;
DWORD dwMaxAmplitude;
} PerfStats;
#define MIX_BUFFER_LEN 500 // Set the sample buffer size to 500 mils
#define MAX_NUM_VOICES 32
#define NUM_EXTRA_VOICES 8 // Extra voices for when we overload.
/*****************************************************************************
* class CControlLogic
*****************************************************************************
* CControlLogic object, implementing the control logic for the DLS
* instrument. This handles MIDI events, plus selection of instrument, sample
* and articulation.
*
*
* Essentially, ControlLogic is the big Kahuna that manages
* the whole system. It parses incoming MIDI events
* by channel and event type. And, it manages the mixing
* of voices into the buffer.
*
* MIDI Input:
*
* The most important events are the note on and
* off events. When a note on event comes in,
* ControlLogic searches for an available voice.
* ControlLogic matches the channel and finds the
* instrument on that channel. It then call the instrument's
* ScanForRegion() command which finds the region
* that matches the note. At this point, it can copy
* the region and associated articulation into the
* voice, using the StartVoice command.
*
* When it receives the sustain pedal command,
* it artificially sets all notes on the channel on
* until a sustain off arrives. To keep track of notes
* that have been shut off while the sustain was on
* it uses an array of 128 shorts, with each bit position
* representing a channel. When the sustain releases,
* it scans through the array and creates a note off for
* each bit that was set.
*
* Additional continuous controller events are managed
* by the CModWheelIn, CPitchBendIn, etc., MIDI input recording
* modules.
*
* Mixing:
*
* Control Logic is also called to mix the instruments into a buffer
* at regular intervals. The buffer is provided by the calling sound
* driver. Each voice is called to mix its sample into the buffer.
*/
class CControlLogic
{
public:
CControlLogic();
~CControlLogic();
HRESULT Init(CInstManager *pInstruments, CSynth *pSynth);
void Flush(STIME stTime); // Clears all events after time.
BOOL RecordMIDI(STIME stTime,BYTE bStatus, BYTE bData1, BYTE bData2);
HRESULT RecordSysEx(DWORD dwSysExLength,BYTE *pSysExData, STIME stTime);
void QueueNotes(STIME stEndTime);
void ClearMIDI(STIME stEndTime);
void SetGainAdjust(VREL vrGainAdjust);
HRESULT SetChannelPriority(DWORD dwChannel,DWORD dwPriority);
HRESULT GetChannelPriority(DWORD dwChannel,LPDWORD pdwPriority);
CSynth * m_pSynth;
private:
void GMReset();
CInstManager * m_pInstruments;
CNoteIn m_Notes; // All Note ons and offs.
CModWheelIn m_ModWheel[16]; // Sixteen channels of Mod Wheel.
CPitchBendIn m_PitchBend[16]; // Sixteen channels of Pitch Bend.
CVolumeIn m_Volume[16]; // Sixteen channels of Volume.
CExpressionIn m_Expression[16]; // Sixteen channels of Expression.
CPanIn m_Pan[16]; // Sixteen channels of Pan.
BOOL m_fSustain[16]; // Sustain on / off.
short m_nCurrentRPN[16]; // RPN number.
BYTE m_bBankH[16]; // Bank selects for instrument.
BYTE m_bBankL[16];
DWORD m_dwProgram[16]; // Instrument choice.
BOOL m_fEmpty; // Indicates empty lists, no need to flush.
VREL m_vrGainAdjust; // Final stage gain adjust
DWORD m_dwPriority[16]; // Priorities for each channel.
BOOL m_fXGActive; // Is XG Active?
BOOL m_fGSActive; // Is GS enabled?
WORD m_nData[16]; // Used to track RPN reading.
VREL m_vrMasterVolume; // Master Volume.
PREL m_prFineTune[16]; // Fine tune for each channel.
PREL m_prScaleTune[16][12];// Alternate scale for each channel.
PREL m_prCoarseTune[16]; // Coarse tune.
BYTE m_bPartToChannel[16]; // Channel to Part converter.
BYTE m_bDrums[16]; // Melodic or which drum?
BOOL m_fMono[16]; // Mono mode?
public:
CRITICAL_SECTION m_CriticalSection; // Critical section to manage access.
BOOL m_fCSInitialized;
};
#endif // __SYNTH_H__