revid: matching up audio packet sizes, chunk sizes and rates throughout revid pipeline

cmd/revid-cli/main.go

@@ -30,14 +30,17 @@ package main
 
 import (
 	"flag"
+	"math"
 	"os"
 	"runtime/pprof"
 	"strconv"
 	"strings"
 	"time"
 
+	"bitbucket.org/ausocean/av/codec/adpcm"
 	"bitbucket.org/ausocean/av/container/mts"
 	"bitbucket.org/ausocean/av/container/mts/meta"
+	"bitbucket.org/ausocean/av/container/mts/pes"
 	"bitbucket.org/ausocean/av/revid"
 	"bitbucket.org/ausocean/iot/pi/netsender"
 	"bitbucket.org/ausocean/iot/pi/sds"
@@ -61,8 +64,6 @@ const (
 	pkg                 = "revid-cli:"
 	defaultLogVerbosity = logger.Info
 	defaultSleepTime    = 60 // Seconds
-	sampleSize          = 2     // Bytes
-	chunkSize           = 16000 // Bytes
 )
 
 // canProfile is set to false with revid-cli is built with "-tags profile".
@@ -193,13 +194,6 @@ func handleFlags() revid.Config {
 		log.Log(logger.Error, pkg+"bad input argument")
 	}
 
-	switch *inputPtr {
-	case "Audio":
-		cfg.WriteRate = float64(*sampleRatePtr*sampleSize) / float64(chunkSize)
-	default:
-		cfg.WriteRate = float64(*frameRatePtr)
-	}
-
 	switch *inputCodecPtr {
 	case "H264":
 		cfg.InputCodec = revid.H264
@@ -212,6 +206,26 @@ func handleFlags() revid.Config {
 		log.Log(logger.Error, pkg+"bad input codec argument")
 	}
 
+	switch *inputPtr {
+	case "Audio":
+		byteDepth := *bitDepthPtr / 8
+		PCMRate := *sampleRatePtr * byteDepth * *channelsPtr * *recPeriodPtr
+		var byteRate int
+		switch cfg.InputCodec {
+		case revid.PCM:
+			byteRate = PCMRate
+		case revid.ADPCM:
+			byteRate = adpcm.BytesOutput(PCMRate)
+		}
+		if byteRate < pes.MaxPesSize {
+			cfg.WriteRate = 1
+		} else {
+			cfg.WriteRate = uint(math.Ceil(float64(byteRate) / float64(pes.MaxPesSize)))
+		}
+	default:
+		cfg.WriteRate = *frameRatePtr
+	}
+
 	if len(outputs) == 0 {
 		cfg.Outputs = make([]uint8, 1)
 	}

codec/adpcm/adpcm.go

@@ -57,6 +57,13 @@ type decoder struct {
 	step  int16
 }
 
+// BytesOutput will return the number of adpcm bytes that will be generated for the given pcm data
+func BytesOutput(pcm int) int {
+	// for X pcm bytes, 2 bytes are left uncompressed, the rest is compressed by a factor of 4
+	// and a start index and padding byte are added.
+	return (pcm-2)/4 + 2 + 1 + 1
+}
+
 // PcmBS is the size of the blocks that an encoder uses.
 // 'encodeBlock' will encode PcmBS bytes at a time and the output will be AdpcmBS bytes long.
 const PcmBS = 1010

codec/lex/lex.go

@@ -262,14 +262,14 @@ func PCM(dst io.Writer, src io.Reader, delay time.Duration) error {
 		tick = ticker.C
 	}
 
-	r := bufio.NewReader(src)
 	for {
 		buf := make([]byte, 0, audioChunkSize)
-		_, err := r.Read(buf)
+		_, err := src.Read(buf)
 		if err != nil {
 			return err
 		}
 		<-tick
+		fmt.Printf("LEXED AUDIO: %v\n", buf[:64])
 		_, err = dst.Write(buf)
 		if err != nil {
 			return err

container/mts/encoder.go

@@ -249,9 +249,7 @@ func (e *Encoder) Write(data []byte) (int, error) {
 			pkt.PCR = e.pcr()
 			pusi = false
 		}
-		bytes := pkt.Bytes(e.tsSpace[:PacketSize])
+		_, err := e.dst.Write(pkt.Bytes(e.tsSpace[:PacketSize]))
-		fmt.Printf("Packet: %v", bytes)
-		_, err := e.dst.Write(bytes)
 		if err != nil {
 			return len(data), err
 		}

container/mts/pes/pes.go

@@ -26,7 +26,7 @@ LICENSE
 
 package pes
 
-const MaxPesSize = 64 * 1 << 10
+const MaxPesSize = 64 * 1 << 10 // 65536
 
 /*
 The below data struct encapsulates the fields of an PES packet. Below is

revid/config.go

@@ -55,7 +55,7 @@ type Config struct {
 	IntraRefreshPeriod uint
 	RtpAddress         string
 	SendRetry          bool
-	WriteRate          float64 // How many times a second revid encoders will be written to.
+	WriteRate          uint // How many times a second revid encoders will be written to.
 
 	// Video
 	Height        uint