- G.722: decode as 16 kHz audio, keep the 8 kHz RTP clock separate

G722Codec::info() reported an 8 kHz sample rate for its 16 kHz output (640
bytes of PCM per 20 ms frame, read as 40 ms at 8 kHz). AudioReceiver therefore
fetched packets at half the rate they arrived, and the high-water trim
discarded about 44% of every G.722 stream: 109 of 250 packets per 5 s, with the
buffer at 360 ms instead of 100 ms.

The 8 kHz figure is right for one thing: RFC 3551 keeps G.722's RTP clock at
8000. Codec::Info already had mTimestampUnit for that. G.722 now reports
mSamplerate 16000 and mTimestampUnit 1/8000, and Codec::rtpClockRate() gives
the clock rate. The uses of samplerate() that mean RTP time now use it:
- the jitter statistics in AudioReceiver::add;
- the timestamp-gap-to-milliseconds conversion in AudioReceiver;
- AudioStream's transmit timestamp unit and increment (numerically unchanged
  for every codec, including G.722).
PCM uses stay on samplerate(), so AudioStream now also feeds the G.722
encoder 16 kHz audio instead of 8 kHz. The factory still advertises G722/8000
in SDP, as it must; only its backwards comment is fixed.

AudioReceiver::infoFor() also applied a "G729A silence" rule to payload type
9, which is G.722; G.729 is 18. It now applies to PT 18. Other fixed-frame
payloads are timed proportionally, as AudioReceiver::add already does, so a
10 ms G.722 packet (half a 160-byte frame) counts as 10 ms, not 20.

Measured in vq-core with 16 replayed calls (NISQA speech; G.722 from ffmpeg
and pjmedia, G.711 controls of the same speech; 3 runs each, all identical):
- G.722 packet drops: 44% -> 0; buffer 100 ms like G.711; no underruns;
- Sevana MOS of every G.722 call now equals its G.711 control, normal and loud
  (with the previous code three calls read 4.45 against 4.25);
- DeadAir-01 0.68-0.89 -> 0.01-0.27 (G.711: 0.00-0.14); SilentCall within 0.01
  of G.711 for the ffmpeg-encoded calls (was 0.15-0.17 higher);
- AmpClipping on loud calls 0.004/0.001, as G.711; the 0.02-0.03 seen after
  the codec swap alone came from the dropped packets;
- G.711 results, packet counts, loss, jitter and audio durations unchanged.
The transmit path (AudioStream) builds but was not exercised; vq-core only
receives.

Needs the vq_net change that computes jitter with rtpClockRate().

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

src/engine/media/MT_AudioCodec.cpp

@@ -1132,13 +1132,15 @@ G722Codec::~G722Codec()
 
 Codec::Info G722Codec::info()
 {
-    // ToDo: double check the G722 calls - remember RFC has bug about samplerate
+    // G.722 decodes to 16 kHz (640 bytes of PCM per 20 ms frame), but RFC 3551 keeps its RTP
+    // clock at 8 kHz for historical reasons: the timestamp unit is 1/8000, not 1/samplerate.
     return {.mName = G722_MIME_NAME,
-            .mSamplerate = 8000,
+            .mSamplerate = 16000,
             .mChannels = 1,
             .mPcmLength = 640,
             .mFrameTime = 20,
-            .mRtpLength = 160};
+            .mRtpLength = 160,
+            .mTimestampUnit = 1.0f / 8000};
 }
 
 Codec::EncodeResult G722Codec::encode(std::span<const uint8_t> input, std::span<uint8_t> output)
@@ -1186,7 +1188,8 @@ const char* G722Codec::G722Factory::name()
 
 int G722Codec::G722Factory::samplerate()
 {
-    // Although G722 uses 16000 as rate for timestamping RTP frames - in fact it is 8KHz codec. So return 8KHz here.
+    // The rate SDP advertises (rtpmap G722/8000) is the RTP clock, 8 kHz by RFC 3551, although the
+    // audio is 16 kHz; see G722Codec::info().
     return 8000;
 }
 

src/engine/media/MT_AudioReceiver.cpp

@@ -526,7 +526,7 @@ Codec* AudioReceiver::add(const std::shared_ptr<jrtplib::RTPPacket>& p)
                 time_length = lround(double(payloadLength) / codec->rtpLength() * codec->frameTime());
 
             if (codec)
-                samplerate = codec->samplerate();
+                samplerate = codec->rtpClockRate(); // RTP timestamp units, not PCM rate (G.722)
         }
 
         // Process jitter anyway - can we decode payload or not
@@ -724,7 +724,7 @@ AudioReceiver::DecodeResult AudioReceiver::decodePacketTo(Audio::DataWindow& out
     if (mLastPacketTimestamp && mLastPacketTimeLength && mCodec)
     {
         int units = rtp.GetTimestamp() - *mLastPacketTimestamp;
-        int milliseconds = units / (mCodec->samplerate() / 1000);
+        int milliseconds = units / (mCodec->rtpClockRate() / 1000);
         if (milliseconds > mLastPacketTimeLength)
         {
             auto silenceLength = std::chrono::milliseconds(milliseconds - mLastPacketTimeLength);
@@ -1157,11 +1157,16 @@ AudioReceiver::MediaInfo AudioReceiver::infoFor(jrtplib::RTPPacket& p)
 
     if (codec->rtpLength() != 0)
     {
-        int frameCount = static_cast<int>(p.GetPayloadLength() / codec->rtpLength());
+        // A G.729 (PT 18) SID frame is shorter than a voice frame but stands for one; other
-        if (p.GetPayloadType() == 9 /*G729A silence*/ && p.GetPayloadLength() % codec->rtpLength())
+        // codecs scale with the payload (G.722 at 10 ms ptime carries half a 160-byte frame).
-            frameCount++;
+        if (p.GetPayloadType() == 18 && p.GetPayloadLength() % codec->rtpLength())
-
+        {
-        packetTime = std::chrono::milliseconds(frameCount * codec->frameTime());
+            int frameCount = static_cast<int>(p.GetPayloadLength() / codec->rtpLength()) + 1;
+            packetTime = std::chrono::milliseconds(frameCount * codec->frameTime());
+        }
+        else
+            packetTime = std::chrono::milliseconds(
+                lround(double(p.GetPayloadLength()) / codec->rtpLength() * codec->frameTime()));
     }
     else if (typeid(*codec) == typeid(OpusCodec))
     {

src/engine/media/MT_AudioStream.cpp

@@ -114,7 +114,7 @@ void AudioStream::setTransmittingCodec(Codec::Factory& factory, int payloadType)
     mTransmittingCodec = factory.create();
     mTransmittingPayloadType = payloadType;
     if (mRtpSession.IsActive())
-        mRtpSession.SetTimestampUnit(1.0 / mTransmittingCodec->samplerate());
+        mRtpSession.SetTimestampUnit(mTransmittingCodec->timestampUnit());
 }
 
 PCodec AudioStream::transmittingCodec()
@@ -241,7 +241,7 @@ void AudioStream::addData(const void* buffer, int bytes)
                 ICELogMedia(<< "Sending RTP packet pt = " << mTransmittingPayloadType
                             << ", plength = " << (int)mEncodedAudio.size() << " to ");
                 mRtpSession.SendPacketEx(mEncodedAudio.data(), mEncodedAudio.size(), mTransmittingPayloadType, false,
-                                         packetTime * codec->samplerate() / 1000, 0, nullptr, 0);
+                                         packetTime * codec->rtpClockRate() / 1000, 0, nullptr, 0);
                 mEncodedAudio.clear();
                 encodedTime = 0;
             }

src/engine/media/MT_Codec.h

@@ -66,6 +66,10 @@ public:
     int          frameTime() { return info().mFrameTime; }
     std::string  name() { return info().mName; }
     float timestampUnit() { return info().mTimestampUnit == 0.0f ? 1.0f / info().mSamplerate : info().mTimestampUnit; }
+    // RTP clock rate in Hz, the unit of RTP timestamps. Equals samplerate() except for codecs
+    // whose RTP clock differs from the decoded audio: G.722 decodes to 16 kHz but its RTP clock
+    // is 8 kHz (RFC 3551). Use it for timestamp arithmetic and jitter, samplerate() for PCM.
+    int   rtpClockRate() { return static_cast<int>(1.0f / timestampUnit() + 0.5f); }
 
     Audio::Format getAudioFormat() { return Audio::Format(this->info().mSamplerate, this->info().mChannels); }