Scope graphing rewrite: speech/music merged, allows 8- or 16-bit, auto-adjust level; bug #1064

git-svn-id: svn://svn.code.sf.net/p/sc2/code/trunk@3181 8092fc87-c524-0410-9efc-e669fe64eaf9

sc2/ChangeLog

@@ -1,4 +1,6 @@
 Changes towards version 0.7:
+- Allow any sound data format to be graphed by comm oscilloscope; also
+  auto-adjust the scope for different gain levels (bug #1064) - Alex
 - Game settings Quit menu now delegates to F10 quit (bug #462) - Alex
 - Do not pause the game in places where not relevant (bug #984) - Alex
 - Fixed crashes and potential weirdness when loading savegames from

sc2/src/sc2code/libs/sound/audiocore.c

@@ -237,3 +237,31 @@ audio_BufferData (audio_Object bufobj, uint32 format, void* data,
 	audiodrv.BufferData (bufobj, audiodrv.EnumLookup[format], data, size,
 			freq);
 }
+
+bool
+audio_GetFormatInfo (uint32 format, int *channels, int *sample_size)
+{
+	switch (format)
+	{
+	case audio_FORMAT_MONO8:
+		*channels = 1;
+		*sample_size = sizeof (uint8);
+		return true;
+
+	case audio_FORMAT_STEREO8:
+		*channels = 2;
+		*sample_size = sizeof (uint8);
+		return true;
+	
+	case audio_FORMAT_MONO16:
+		*channels = 1;
+		*sample_size = sizeof (sint16);
+		return true;
+
+	case audio_FORMAT_STEREO16:
+		*channels = 2;
+		*sample_size = sizeof (sint16);
+		return true;
+	}
+	return false;
+}

sc2/src/sc2code/libs/sound/audiocore.h

@@ -164,4 +164,6 @@ void audio_GetBufferi (audio_Object bufobj, audio_BufferProp pname,
 void audio_BufferData (audio_Object bufobj, uint32 format, void* data,
 		uint32 size, uint32 freq);
 
+bool audio_GetFormatInfo (uint32 format, int *channels, int *sample_size);
+
 #endif /* _AUDIOCORE_H */

sc2/src/sc2code/libs/sound/sound.h

@@ -75,11 +75,11 @@ typedef struct tfb_soundsource
 	sint32 start_time;
 	void *positional_object;
 
-	// for oscilloscope
+	// Cyclic waveform buffer for oscilloscope
 	void *sbuffer; 
-	uint32 sbuf_start;
+	uint32 sbuf_start;        // cyclic buffer tail (confusing, eh?)
 	uint32 sbuf_size;
-	uint32 sbuf_offset;
+	uint32 sbuf_offset;       // cyclic buffer head
 	uint32 sbuf_lasttime;
 	// keep track for paused tracks
 	uint32 pause_time;

sc2/src/sc2code/libs/sound/stream.c

@@ -449,3 +449,175 @@ StreamDecoderTaskFunc (void *data)
 	return 0;
 }
 
+inline sint32
+readSoundSample (void *ptr, int sample_size)
+{
+	if (sample_size == sizeof (uint8))
+		return (*(uint8*)ptr - 128) << 8;
+	else
+		return *(sint16*)ptr;
+}
+
+// Graphs the current sound data for the oscilloscope.
+// Includes a rudimentary automatic gain control (AGC) to properly graph
+// the streams at different gain levels (based on running average).
+// We use AGC because different pieces of music and speech can easily be
+// at very different gain levels, because the game is moddable.
+int
+GraphForegroundStream (uint8 *data, sint32 width, sint32 height)
+{
+	int source_num;
+	TFB_SoundSource *source;
+	TFB_SoundDecoder *decoder;
+	int channels;
+	int sample_size;
+	int full_sample;
+	int step;
+	long played_time;
+	long delta;
+	uint8 *sbuffer;
+	unsigned long pos;
+	int scale;
+	sint32 i;
+	// AGC variables
+#define DEF_PAGE_MAX    28000
+#define AGC_PAGE_COUNT  16
+	static int page_sum = DEF_PAGE_MAX * AGC_PAGE_COUNT;
+	static int pages[AGC_PAGE_COUNT] =
+	{
+		DEF_PAGE_MAX, DEF_PAGE_MAX, DEF_PAGE_MAX, DEF_PAGE_MAX,
+		DEF_PAGE_MAX, DEF_PAGE_MAX, DEF_PAGE_MAX, DEF_PAGE_MAX,
+		DEF_PAGE_MAX, DEF_PAGE_MAX, DEF_PAGE_MAX, DEF_PAGE_MAX,
+		DEF_PAGE_MAX, DEF_PAGE_MAX, DEF_PAGE_MAX, DEF_PAGE_MAX,
+	};
+	static int page_head;
+#define AGC_FRAME_COUNT  8
+	static int frame_sum;
+	static int frames;
+	static int avg_amp = DEF_PAGE_MAX; // running amplitude (sort of) average
+	int target_amp;
+	int max_a;
+#define VAD_MIN_ENERGY  100
+	long energy;
+
+
+	if (speechVolumeScale != 0.0f)
+	{	// Use speech waveform when speech is enabled
+		source_num = SPEECH_SOURCE;
+		// Step is picked experimentally. Using step of 1 sample at 11025Hz,
+		// because human speech is mostly in the low frequencies, and it looks
+		// better this way.
+		step = 1;
+
+	}
+	else if (musicVolumeScale != 0.0f)
+	{	// Use music waveform when speech is disabled
+		source_num = MUSIC_SOURCE;
+		// Step is picked experimentally. Using step of 4 samples at 11025Hz.
+		// It looks better this way.
+		step = 4;
+	}
+	else
+	{
+		return 0;
+	}
+
+	source = &soundSource[source_num];
+	LockMutex (source->stream_mutex);
+	if (!PlayingStream (source_num) || !source->sample
+			|| !source->sample->decoder || !source->sbuffer
+			|| source->sbuf_size == 0)
+	{	// We don't have data to return, oh well.
+		UnlockMutex (source->stream_mutex);
+		return 0;
+	}
+	decoder = source->sample->decoder;
+
+	assert (audio_GetFormatInfo (decoder->format, &channels, &sample_size));
+	full_sample = channels * sample_size;
+
+	// See how far into the buffer we should be now
+	played_time = GetTimeCounter () - source->sbuf_lasttime;
+	delta = played_time * decoder->frequency * full_sample / ONE_SECOND;
+	// align delta to sample start
+	delta = delta & ~(full_sample - 1);
+
+	if (delta < 0)
+	{
+		log_add (log_Debug, "GraphForegroundStream(): something is messed"
+				" with timing, delta %ld", delta);
+		delta = 0;
+	}
+	else if (delta > (long)source->sbuf_size)
+	{	// Stream decoder task has just had a heart attack, not much we can do
+		delta = 0;
+	}
+
+	// Step is in 11025 Hz units, so we need to adjust to source frequency
+	step = decoder->frequency * step / 11025;
+	if (step == 0)
+		step = 1;
+	step *= full_sample;
+
+	sbuffer = source->sbuffer;
+	pos = source->sbuf_offset + delta;
+
+	// We are not basing the scaling factor on signal energy, because we
+	// want it to *look* pretty instead of sounding nice and even
+	target_amp = (height >> 1) >> 1;
+	scale = avg_amp / target_amp;
+
+	max_a = 0;
+	energy = 0;
+	for (i = 0; i < width; ++i, pos += step)
+	{
+		sint32 s;
+		int t;
+
+		pos %= source->sbuf_size;
+
+		s = readSoundSample (sbuffer + pos, sample_size);
+		if (channels > 1)
+			s += readSoundSample (sbuffer + pos + sample_size, sample_size);
+
+		energy += (s * s) / 0x10000;
+		t = abs(s);
+		if (t > max_a)
+			max_a = t;
+
+		s = (s / scale) + (height >> 1);
+		if (s < 0)
+			s = 0;
+		else if (s > height - 1)
+			s = height - 1;
+		
+		data[i] = s;
+	}
+	energy /= width;
+
+	// Very basic VAD. We don't want to count speech pauses in the average
+	if (energy > VAD_MIN_ENERGY)
+	{
+		// Record the maximum amplitude (sort of)
+		frame_sum += max_a;
+		++frames;
+		if (frames == 8)
+		{	// Got a full page
+			frame_sum /= AGC_FRAME_COUNT;
+			// Record the page
+			page_sum -= pages[page_head];
+			page_sum += frame_sum;
+			pages[page_head] = frame_sum;
+			page_head = (page_head + 1) % AGC_PAGE_COUNT;
+
+			frame_sum = 0;
+			frames = 0;
+
+			avg_amp = page_sum / AGC_PAGE_COUNT;
+		}
+	}
+
+	UnlockMutex (source->stream_mutex);
+	return 1;
+}
+

sc2/src/sc2code/libs/sound/stream.h

@@ -30,4 +30,6 @@ TFB_SoundTag* FindTaggedBuffer (TFB_SoundSample* sample, audio_Object buffer);
 void TFB_ClearBufferTag (TFB_SoundTag* ptag);
 void TFB_TagBuffer (TFB_SoundSample* sample, audio_Object buffer, void* data);
 
+int GraphForegroundStream (uint8 *data, sint32 width, sint32 height);
+
 #endif

sc2/src/sc2code/libs/sound/trackplayer.c

@@ -19,9 +19,7 @@
 #include "libs/sound/trackint.h"
 #include "libs/log.h"
 #include "comm.h"
-#include "sis.h"
 #include "options.h"
-#include <assert.h>
 #include <ctype.h>
 #include <stdlib.h>
 
@@ -821,179 +819,6 @@ FastForward_Page ()
 	return TRUE;
 }
 
-// processes sound data to oscilloscope
-int
-GetSoundData (void *data) 
-{
-	// XXX: These two variants are begging to be merged
-	if (speechVolumeScale != 0.0f)
-	{
-		// speech is enabled
-
-		LockMutex (soundSource[SPEECH_SOURCE].stream_mutex);
-		if (soundSource[SPEECH_SOURCE].sample && soundSource[SPEECH_SOURCE].sample->decoder && 
-			PlayingStream (SPEECH_SOURCE) && soundSource[SPEECH_SOURCE].sbuffer && 
-			soundSource[SPEECH_SOURCE].sbuf_size > 0)
-		{
-			float played_time = (GetTimeCounter () - soundSource[SPEECH_SOURCE].sbuf_lasttime) / 
-				(float)ONE_SECOND;
-			long delta = (int) (played_time * (float)soundSource[SPEECH_SOURCE].
-				sample->decoder->frequency * 2.0f);
-			unsigned long pos;
-			int i;
-			int step;
-			UBYTE *scopedata = (UBYTE *) data;
-			UBYTE *sbuffer = soundSource[SPEECH_SOURCE].sbuffer;
-
-			assert (soundSource[SPEECH_SOURCE].sample->decoder->frequency >= 11025);
-			assert (soundSource[SPEECH_SOURCE].sample->decoder->format == audio_FORMAT_MONO16);
-
-			// Using step of 1 sample at 11025Hz, because the human speech
-			// is mostly in the low frequencies
-			step = soundSource[SPEECH_SOURCE].sample->decoder->frequency * 2 / 11025;
-			step = (step + 1) & ~1;
-
-			if (delta < 0)
-			{
-				log_add (log_Debug, "GetSoundData(): something's messed"
-						" with timing, delta %ld", delta);
-				delta = 0;
-			}
-			else if (delta > (int)(soundSource[SPEECH_SOURCE].sbuf_size * 2))
-			{
-#if 0
-				log_add (log_Debug, "GetSoundData(): something's messed"
-						" with timing, delta %d", delta);
-#endif
-				delta = 0;
-			}
-#if 0
-			log_add (log_Debug, "played_data %d total_decoded %d delta %d",
-					played_data, soundSource[SPEECH_SOURCE].total_decoded,
-					delta);
-#endif
-			pos = soundSource[SPEECH_SOURCE].sbuf_offset + delta;
-			if (pos % 2 == 1)
-				pos++;
-
-			// pos is an unsigned data type; this assertion cannot fail!
-			// assert (pos >= 0);
-
-			for (i = 0; i < RADAR_WIDTH - 2; ++i)
-			{
-				SDWORD s;
-
-				for (;;)
-				{
-					if (pos >= soundSource[SPEECH_SOURCE].sbuf_size)
-						pos = pos - soundSource[SPEECH_SOURCE].sbuf_size;
-					else
-						break;
-				}
-
-				s = *(SWORD*) (&sbuffer[pos]);
-				s = (s / 1360) + (RADAR_HEIGHT >> 1);
-				if (s < 1)
-					s = 1;
-				else if (s > RADAR_HEIGHT - 2)
-					s = RADAR_HEIGHT - 2;
-				scopedata[i] = (UBYTE) s;
-
-				pos += step;
-			}
-
-			UnlockMutex (soundSource[SPEECH_SOURCE].stream_mutex);
-			return 1;
-		}
-		UnlockMutex (soundSource[SPEECH_SOURCE].stream_mutex);
-	}
-	else if (musicVolumeScale != 0.0f)
-	{
-		// speech is disabled but music is not so process it instead
-
-		LockMutex (soundSource[MUSIC_SOURCE].stream_mutex);
-		if (soundSource[MUSIC_SOURCE].sample && soundSource[MUSIC_SOURCE].sample->decoder && 
-			PlayingStream (MUSIC_SOURCE) && soundSource[MUSIC_SOURCE].sbuffer && 
-			soundSource[MUSIC_SOURCE].sbuf_size > 0)
-		{
-			float played_time = (GetTimeCounter () - soundSource[MUSIC_SOURCE].sbuf_lasttime) / 
-				(float)ONE_SECOND;
-			int delta = (int) (played_time * (float)soundSource[MUSIC_SOURCE].
-				sample->decoder->frequency * 4.0f);
-			unsigned long pos;
-			int i, step;
-			UBYTE *scopedata = (UBYTE *) data;
-			UBYTE *sbuffer = soundSource[MUSIC_SOURCE].sbuffer;
-
-			assert (soundSource[MUSIC_SOURCE].sample->decoder->frequency >= 11025);
-			assert (soundSource[MUSIC_SOURCE].sample->decoder->format == audio_FORMAT_STEREO16);
-
-			// Using step of 4 samples at 11025Hz for the music
-			step = soundSource[MUSIC_SOURCE].sample->decoder->frequency / 11025 * 16;
-			if (step % 2 == 1)
-				step++;
-
-			if (delta < 0)
-			{
-#if 0				
-				log_add (log_Debug, "GetSoundData(): something's messed"
-						" with timing, delta %d", delta);
-#endif
-				delta = 0;
-			}
-			else if (delta > (int)(soundSource[MUSIC_SOURCE].sbuf_size * 2))
-			{
-#if 0				
-				log_add (log_Debug, "GetSoundData(): something's messed"
-						" with timing, delta %d", delta);
-#endif
-				delta = 0;
-			}
-#if 0			
-			log_add (log_Debug, "played_data %d total_decoded %d delta %d",
-					played_data, soundSource[MUSIC_SOURCE].total_decoded,
-					delta);
-#endif
-			pos = soundSource[MUSIC_SOURCE].sbuf_offset + delta;
-			if (pos % 2 == 1)
-				pos++;
-
-			// pos is an unsigned type; this assertion cannot fail!
-			// assert (pos >= 0);
-
-			for (i = 0; i < RADAR_WIDTH - 2; ++i)
-			{
-				SDWORD s;
-
-				for (;;)
-				{
-					if (pos >= soundSource[MUSIC_SOURCE].sbuf_size)
-						pos = pos - soundSource[MUSIC_SOURCE].sbuf_size;
-					else
-						break;
-				}
-
-				s = (*(SWORD*)(&sbuffer[pos])) + (*(SWORD*)(&sbuffer[pos + 2]));
-				
-				s = (s / 1800) + (RADAR_HEIGHT >> 1);
-				if (s < 1)
-					s = 1;
-				else if (s > RADAR_HEIGHT - 2)
-					s = RADAR_HEIGHT - 2;
-				scopedata[i] = (UBYTE) s;
-
-				pos += step;
-			}
-
-			UnlockMutex (soundSource[MUSIC_SOURCE].stream_mutex);
-			return 1;
-		}
-		UnlockMutex (soundSource[MUSIC_SOURCE].stream_mutex);
-	}
-
-	return 0;
-}
-
 // tells current position of streaming speech
 int
 GetSoundInfo (int max_len)

sc2/src/sc2code/libs/sound/trackplayer.h

@@ -37,7 +37,6 @@ void FastReverse_Page(void);
 void StopTrack(void);
 void SpliceTrack(UNICODE *filespec, UNICODE *textspec, UNICODE *TimeStamp, TFB_TrackCB cb);
 void SpliceMultiTrack (UNICODE *TrackNames[], UNICODE *TrackText);
-int GetSoundData (void *data);
 int GetSoundInfo (int max_len);
 
 #endif

sc2/src/sc2code/oscill.c

@@ -21,6 +21,7 @@
 #include "sis.h"
 #include "libs/graphics/gfx_common.h"
 #include "libs/graphics/drawable.h"
+#include "libs/sound/sound.h"
 #include "libs/sound/trackplayer.h"
 
 
@@ -92,15 +93,15 @@ Oscilloscope (DWORD grab_data)
 		return;
 
 	TFB_DrawImage_Image (scope_bg, 0, 0, 0, NULL, scope_surf);
-	if (GetSoundData (scope_data)) 
+	if (GraphForegroundStream (scope_data, RADAR_WIDTH - 2, RADAR_HEIGHT - 2)) 
 	{
 		int i, r, g, b;		
 		TFB_DrawCanvas_GetPixel (scope_bg->NormalImg,
 				scope_bg->extent.width / 2, scope_bg->extent.height / 2,
 				&r, &g, &b);
 		for (i = 0; i < RADAR_WIDTH - 3; ++i)
-			TFB_DrawImage_Line (i + 1, scope_data[i], i + 2,
+			TFB_DrawImage_Line (i + 1, scope_data[i] + 1, i + 2,
-					scope_data[i + 1], r, g, b, scope_surf);
+					scope_data[i + 1] + 1, r, g, b, scope_surf);
 	}
 	TFB_DrawImage_Image (scope_surf, 0, 0, 0, NULL, scope_frame->image);