Fix audio clock drift and adaptive jitter buffering

dotnet/src/VoiceCat.Audio/AdaptivePcmBuffer.cs

@@ -0,0 +1,97 @@
+namespace VoiceCat.Audio;
+
+// SPSC PCM handoff with a small occupancy-controlled sample-rate correction. Producers and
+// consumers remain non-waiting; the correction keeps independent hardware and managed clocks
+// from periodically reaching the hard underflow/overflow edges of a conventional ring.
+public sealed class AdaptivePcmBuffer
+{
+    private const int SampleRate = 48_000;
+    private const double MaximumCorrection = 0.005;
+    private readonly short[] samples;
+    private readonly int channels, frameMask;
+    private int readFrame, writtenFrame, producer, targetFrames;
+    private double phase;
+    private bool primed;
+
+    public AdaptivePcmBuffer(int channels, int bufferMilliseconds = 40, int capacityFrames = 16_384)
+    {
+        if (channels is not (1 or 2)) throw new ArgumentOutOfRangeException(nameof(channels));
+        if (capacityFrames < 2 || (capacityFrames & (capacityFrames - 1)) != 0) throw new ArgumentOutOfRangeException(nameof(capacityFrames));
+        this.channels = channels; samples = new short[checked(capacityFrames * channels)]; frameMask = capacityFrames - 1;
+        BufferMilliseconds = bufferMilliseconds;
+    }
+
+    public int Channels => channels;
+    public int CountFrames => unchecked(Volatile.Read(ref writtenFrame) - Volatile.Read(ref readFrame));
+    public int BufferMilliseconds
+    {
+        get => Volatile.Read(ref targetFrames) * 1000 / SampleRate;
+        set
+        {
+            if (value is not (20 or 40 or 60)) throw new ArgumentOutOfRangeException(nameof(value));
+            Volatile.Write(ref targetFrames, value * SampleRate / 1000);
+        }
+    }
+
+    public bool TryWrite(ReadOnlySpan<short> source)
+    {
+        if (source.Length == 0 || source.Length % channels != 0 || Interlocked.CompareExchange(ref producer, 1, 0) != 0) return false;
+        try
+        {
+            int frames = source.Length / channels, index = writtenFrame;
+            if (frames > frameMask + 1 - unchecked(index - Volatile.Read(ref readFrame))) return false;
+            for (int frame = 0; frame < frames; frame++)
+            {
+                int target = ((index + frame) & frameMask) * channels;
+                for (int channel = 0; channel < channels; channel++) samples[target + channel] = source[frame * channels + channel];
+            }
+            Volatile.Write(ref writtenFrame, unchecked(index + frames)); return true;
+        }
+        finally { Volatile.Write(ref producer, 0); }
+    }
+
+    // Returns interleaved samples written. A zero return means the caller should treat the
+    // already-cleared destination as silence. Once primed, short scheduling stalls re-prime
+    // instead of repeatedly clicking at the ring edge.
+    public int Read(Span<short> destination)
+    {
+        if (destination.Length % channels != 0) throw new ArgumentException("PCM must contain complete interleaved frames.", nameof(destination));
+        int requestedFrames = destination.Length / channels;
+        if (requestedFrames == 0) return 0;
+        int available = CountFrames, target = Volatile.Read(ref targetFrames);
+        if (!primed)
+        {
+            if (available < target) { destination.Clear(); return 0; }
+            primed = true; phase = 0;
+        }
+        if (available <= 0) { primed = false; phase = 0; destination.Clear(); return 0; }
+
+        double correction = Math.Clamp((available - target) / (SampleRate * 2.0), -MaximumCorrection, MaximumCorrection);
+        double step = 1.0 + correction;
+        int produced = 0, read = readFrame;
+        for (int frame = 0; frame < requestedFrames; frame++)
+        {
+            int baseOffset = (read & frameMask) * channels;
+            int nextOffset = ((read + 1) & frameMask) * channels;
+            int remaining = unchecked(Volatile.Read(ref writtenFrame) - read);
+            if (remaining <= 0) break;
+            double fraction = phase;
+            for (int channel = 0; channel < channels; channel++)
+            {
+                int first = samples[baseOffset + channel];
+                int second = remaining > 1 ? samples[nextOffset + channel] : first;
+                destination[produced++] = (short)Math.Clamp((int)Math.Round(first + (second - first) * fraction), short.MinValue, short.MaxValue);
+            }
+            phase += step;
+            int advance = (int)phase;
+            if (advance > remaining) advance = remaining;
+            read = unchecked(read + advance); phase -= advance;
+        }
+        Volatile.Write(ref readFrame, read);
+        if (produced < destination.Length)
+        {
+            destination[produced..].Clear(); primed = false; phase = 0;
+        }
+        return produced;
+    }
+}

dotnet/src/VoiceCat.Audio/AudioDeviceBackend.cs

@@ -3,7 +3,11 @@ namespace VoiceCat.Audio;
 public sealed record AudioDeviceInfo(string Id, string Name, bool IsDefault);
 public delegate void CapturePcmHandler(ReadOnlySpan<short> pcm, int channels);
 public interface IAudioCapture : IDisposable { }
-public interface IAudioPlayback : IDisposable { void Write(ReadOnlySpan<short> stereoPcm); }
+public interface IAudioPlayback : IDisposable
+{
+    int BufferMilliseconds { get; set; }
+    void Write(ReadOnlySpan<short> stereoPcm);
+}
 public interface IAudioDeviceBackend
 {
     IReadOnlyList<AudioDeviceInfo> Enumerate(bool input);

dotnet/src/VoiceCat.Audio/AudioEngine.cs

@@ -23,6 +23,17 @@ public sealed class AudioEngine : IDisposable
     public volatile float InputGain = 1, OutputGain = 1, VadThreshold = 0.02f;
     public volatile bool InputNoiseReduction, MicMuted, Deafened, PushToTalk;
     public volatile AudioInputMode InputMode = AudioInputMode.VoiceActivation;
+    private int deviceBufferMilliseconds = 40;
+    public int DeviceBufferMilliseconds
+    {
+        get => Volatile.Read(ref deviceBufferMilliseconds);
+        set
+        {
+            if (value is not (20 or 40 or 60)) throw new ArgumentOutOfRangeException(nameof(value));
+            Volatile.Write(ref deviceBufferMilliseconds, value);
+            foreach (LocalStream stream in Volatile.Read(ref routes).Local) stream.BufferMilliseconds = value;
+        }
+    }
     public Exception? Failure { get; private set; }
 
     public AudioEngine(EncodedVoiceSender sender, bool startWorker = true)
@@ -42,7 +53,7 @@ public sealed class AudioEngine : IDisposable
         lock (gate)
         {
             ObjectDisposedException.ThrowIf(disposed != 0, this);
-            var stream = new LocalStream(info, captureChannels);
+            var stream = new LocalStream(info, captureChannels, DeviceBufferMilliseconds);
             Routes previous = routes;
             var locals = previous.Local.Where(s => s.Info.StreamId != info.StreamId).Append(stream).ToArray();
             Publish(locals, previous.Remote);
@@ -58,7 +69,7 @@ public sealed class AudioEngine : IDisposable
         lock (gate)
         {
             var current = routes.Local.FirstOrDefault(s => s.Info.StreamId == streamId) ?? throw new ArgumentException("Stream not found.");
-            if (current.CaptureChannels != channels) Publish(routes.Local.Select(s => s == current ? new LocalStream(s.Info, channels) : s).ToArray(), routes.Remote);
+            if (current.CaptureChannels != channels) Publish(routes.Local.Select(s => s == current ? new LocalStream(s.Info, channels, DeviceBufferMilliseconds) : s).ToArray(), routes.Remote);
         }
     }
 

dotnet/src/VoiceCat.Audio/LocalStream.cs

@@ -15,7 +15,7 @@ internal sealed class LocalStream : IDisposable
 {
     internal readonly StreamInfo Info;
     internal readonly int CaptureChannels;
-    internal readonly PcmRing Input = new(16384);
+    internal readonly AdaptivePcmBuffer Input;
     internal volatile float Level;
     internal volatile bool Talking;
     internal volatile float Gain = 1;
@@ -27,6 +27,7 @@ internal sealed class LocalStream : IDisposable
     private readonly short[] converted = new short[16384];
     private int feeding;
     private int buffered;
+    private int starvedSamples;
     private uint timestamp;
     private bool wasTransmitting, marker;
 
@@ -45,9 +46,12 @@ internal sealed class LocalStream : IDisposable
         finally { Volatile.Write(ref feeding, 0); }
     }
 
-    internal LocalStream(StreamInfo stream, int captureChannels)
+    internal int BufferMilliseconds { get => Input.BufferMilliseconds; set => Input.BufferMilliseconds = value; }
+
+    internal LocalStream(StreamInfo stream, int captureChannels, int bufferMilliseconds = 40)
     {
         Info = stream.Clone(); CaptureChannels = captureChannels;
+        Input = new(captureChannels, bufferMilliseconds);
         encoder = new(new()
         {
             Channels = stream.Audio.Mode == ChannelMode.ModeStereo ? 2 : 1,
@@ -64,10 +68,14 @@ internal sealed class LocalStream : IDisposable
     internal void Process(AudioEngine engine, EncodedVoiceSender sender)
     {
         var input = capture.AsSpan(0, 960 * CaptureChannels);
-        if (Input.Count < input.Length) { Level = 0; Talking = false; buffered = 0; wasTransmitting = false; return; }
-        while (Input.Count > input.Length * 6) Input.Read(input);
+        if (Input.Read(input) != input.Length)
+        {
+            Level = 0; Talking = false; starvedSamples += 960;
+            if (starvedSamples >= 9600) { buffered = 0; wasTransmitting = false; }
+            return;
+        }
+        starvedSamples = 0;
         if (buffered == 0) timestamp = engine.SampleClock;
-        Input.Read(input);
         bool mic = Info.Kind == StreamKind.StreamMic;
         if (mic && engine.InputNoiseReduction)
         {

dotnet/src/VoiceCat.Audio/ReceiveStream.cs

@@ -20,28 +20,36 @@ internal sealed class ReceiveStream : IDisposable
     private readonly byte[][] packets = Enumerable.Range(0, 64).Select(_ => new byte[1275]).ToArray();
     private readonly VoiceFrameHeader[] headers = new VoiceFrameHeader[64];
     private readonly int[] lengths = new int[64];
+    private readonly long[] arrivals = new long[64];
     private int read, written;
-    private readonly byte[][] jitter = Enumerable.Range(0, 6).Select(_ => new byte[1275]).ToArray();
-    private readonly uint[] timestamps = new uint[6];
-    private readonly int[] sizes = new int[6];
+    private readonly byte[][] jitter;
+    private readonly uint[] timestamps;
+    private readonly int[] sizes;
     private int count, available, offset, missing, waiting;
     private uint expected;
     private bool started, hasTimestamp;
     private bool hasMarker;
     private uint lastMarker;
     private readonly int channels, frameSamples, maximumDepth;
+    private long lastArrival;
+    private uint lastArrivalTimestamp;
+    private double jitterSamples;
+    private readonly TimeProvider clock;
     internal int Depth => count;
+    internal int TargetDepthSamples => TargetSamples();
     internal int ConcealedFrames { get; private set; }
     internal int DredFrames { get; private set; }
     internal int FecFrames { get; private set; }
 
-    internal ReceiveStream(uint userId, StreamInfo info)
+    internal ReceiveStream(uint userId, StreamInfo info, TimeProvider? clock = null)
     {
         if (info.Audio.FrameMs is not (5 or 10 or 20 or 40 or 60) || !Enum.IsDefined(info.Audio.Mode)) throw new ArgumentException("Unsupported remote audio configuration.", nameof(info));
-        UserId = userId; Info = info.Clone();
+        UserId = userId; Info = info.Clone(); this.clock = clock ?? TimeProvider.System;
         channels = info.Audio.Mode == ChannelMode.ModeStereo ? 2 : 1;
         frameSamples = checked((int)info.Audio.FrameMs * 48);
-        maximumDepth = Math.Clamp(120 / (int)info.Audio.FrameMs, 2, 6);
+        maximumDepth = Math.Clamp(500 / (int)info.Audio.FrameMs + 2, 8, 104);
+        jitter = Enumerable.Range(0, maximumDepth).Select(_ => new byte[1275]).ToArray();
+        timestamps = new uint[maximumDepth]; sizes = new int[maximumDepth];
         decoder = new(48000, channels);
         try { left = new(); } catch { decoder.Dispose(); throw; }
         try { right = new(); } catch { left.Dispose(); decoder.Dispose(); throw; }
@@ -53,7 +61,7 @@ internal sealed class ReceiveStream : IDisposable
     {
         int index = written;
         if (payload.Length is < 1 or > 1275 || unchecked(index - Volatile.Read(ref read)) >= 64) return false;
-        int slot = index & 63; payload.CopyTo(packets[slot]); headers[slot] = header; lengths[slot] = payload.Length;
+        int slot = index & 63; payload.CopyTo(packets[slot]); headers[slot] = header; lengths[slot] = payload.Length; arrivals[slot] = clock.GetTimestamp();
         Volatile.Write(ref written, unchecked(index + 1)); return true;
     }
 
@@ -67,8 +75,8 @@ internal sealed class ReceiveStream : IDisposable
             if ((headers[source].Flags & VoiceFrameFlags.Marker) != 0 && (!hasMarker || unchecked((int)(timestamp - lastMarker)) > 0))
             {
                 hasMarker = true; lastMarker = timestamp;
-                sizes.AsSpan().Clear(); count = available = offset = missing = waiting = 0;
-                expected = timestamp; started = false; delta = 0;
+                DropBefore(timestamp); available = offset = missing = waiting = 0;
+                expected = timestamp; started = false; delta = 0; lastArrival = 0; jitterSamples = 0;
             }
             bool duplicate = false;
             for (int i = 0; i < sizes.Length; i++) if (sizes[i] != 0 && timestamps[i] == timestamp) duplicate = true;
@@ -79,14 +87,61 @@ internal sealed class ReceiveStream : IDisposable
                     int oldest = Oldest(); sizes[oldest] = 0; count--;
                 }
                 int target = Array.IndexOf(sizes, 0);
-                timestamps[target] = timestamp; sizes[target] = lengths[source]; packets[source].AsSpan(0, lengths[source]).CopyTo(jitter[target]); count++;
-                int oldestRemaining = Oldest();
-                if (count >= maximumDepth && unchecked((int)(timestamps[oldestRemaining] - expected)) > 0) expected = timestamps[oldestRemaining];
+                timestamps[target] = timestamp; sizes[target] = lengths[source];
+                packets[source].AsSpan(0, lengths[source]).CopyTo(jitter[target]); count++;
+                ObserveArrival(timestamp, arrivals[source]);
             }
             Volatile.Write(ref read, unchecked(read + 1));
         }
     }
 
+    private void ObserveArrival(uint timestamp, long arrival)
+    {
+        if (lastArrival != 0)
+        {
+            int timestampDelta = unchecked((int)(timestamp - lastArrivalTimestamp));
+            if (timestampDelta <= 0) return;
+            if (timestampDelta > 0 && timestampDelta <= frameSamples * 10)
+            {
+                double arrivalDelta = clock.GetElapsedTime(lastArrival, arrival).TotalSeconds * 48_000;
+                double deviation = Math.Abs(arrivalDelta - timestampDelta);
+                jitterSamples += (deviation - jitterSamples) / 16.0;
+            }
+        }
+        lastArrival = arrival; lastArrivalTimestamp = timestamp;
+    }
+
+    private int TargetSamples()
+    {
+        int recovery = Info.Audio.Dred || Info.Audio.Fec ? frameSamples : 0;
+        int variation = checked((int)Math.Ceiling(4 * jitterSamples / frameSamples)) * frameSamples;
+        return Math.Min(5760, recovery + variation);
+    }
+
+    private int Newest()
+    {
+        int newest = -1;
+        for (int i = 0; i < sizes.Length; i++) if (sizes[i] != 0 && (newest < 0 || unchecked((int)(timestamps[i] - timestamps[newest])) > 0)) newest = i;
+        return newest;
+    }
+
+    private void DropBefore(uint timestamp)
+    {
+        for (int i = 0; i < sizes.Length; i++)
+            if (sizes[i] != 0 && unchecked((int)(timestamps[i] - timestamp)) < 0) { sizes[i] = 0; count--; }
+    }
+
+    private void CatchUp()
+    {
+        if (!started || available != 0 || count == 0) return;
+        int newest = Newest(); int target = TargetSamples();
+        int lead = unchecked((int)(timestamps[newest] - expected));
+        if (lead <= target + frameSamples) return;
+        int keepBehind = target / frameSamples * frameSamples;
+        expected = unchecked(timestamps[newest] - (uint)keepBehind);
+        DropBefore(expected); missing = 0;
+    }
+
     private int Oldest()
     {
         int oldest = -1;
@@ -108,11 +163,12 @@ internal sealed class ReceiveStream : IDisposable
         else
         {
             int next = Oldest(); missing++;
-            if (next >= 0 && unchecked((int)(timestamps[next] - expected)) == frameSamples)
+            int recoveryOffset = next < 0 ? 0 : unchecked((int)(timestamps[next] - expected));
+            if (next >= 0 && recoveryOffset > 0 && recoveryOffset % frameSamples == 0)
             {
                 var packet = jitter[next].AsSpan(0, sizes[next]);
-                if (dred?.TryRecover(decoder, packet, pcm, frameSamples) == true) { decoded = true; DredFrames++; }
-                else if (Info.Audio.Fec && decoder.TryDecode(packet, pcm, frameSamples, out int recovered, true) && recovered == frameSamples) { decoded = true; FecFrames++; }
+                if (dred?.TryRecover(decoder, packet, pcm, frameSamples, recoveryOffset) == true) { decoded = true; DredFrames++; }
+                else if (recoveryOffset == frameSamples && Info.Audio.Fec && decoder.TryDecode(packet, pcm, frameSamples, out int recovered, true) && recovered == frameSamples) { decoded = true; FecFrames++; }
             }
         }
         int maximumConcealment = Math.Max(1, 200 / (int)Info.Audio.FrameMs);
@@ -127,11 +183,16 @@ internal sealed class ReceiveStream : IDisposable
     internal void Mix(Span<int> output, bool deafened, PcmStreamHandler? sink)
     {
         Drain();
+        CatchUp();
         if (!started)
         {
             waiting++;
-            if (!hasTimestamp || count < Math.Min(3, maximumDepth) && waiting < 3) return;
-            expected = timestamps[Oldest()]; started = true;
+            if (!hasTimestamp || count == 0) return;
+            int oldest = Oldest(), newest = Newest();
+            if (unchecked((int)(timestamps[newest] - timestamps[oldest])) < TargetSamples()) return;
+            int keepBehind = TargetSamples() / frameSamples * frameSamples;
+            expected = unchecked(timestamps[newest] - (uint)keepBehind);
+            DropBefore(expected); started = true; waiting = 0;
         }
         int copied = 0;
         while (copied < 960)

dotnet/tests/VoiceCat.Tests/AudioEngineTests.cs

@@ -7,6 +7,84 @@ namespace VoiceCat.Tests;
 
 public class AudioEngineTests
 {
+    [Theory]
+    [InlineData(-1000)]
+    [InlineData(1000)]
+    public void AdaptivePcmBufferAbsorbsIndependentClockDrift(int partsPerMillion)
+    {
+        var buffer = new AdaptivePcmBuffer(1, 40); short[] input = new short[962], output = new short[960];
+        input.AsSpan().Fill(1234); Assert.True(buffer.TryWrite(input.AsSpan(0, 960)));
+        Assert.True(buffer.TryWrite(input.AsSpan(0, 960)));
+        double produced = 0;
+        for (int cycle = 0; cycle < 10_000; cycle++)
+        {
+            produced += 960 * (1 + partsPerMillion / 1_000_000.0);
+            int frames = (int)produced; produced -= frames;
+            Assert.True(buffer.TryWrite(input.AsSpan(0, frames)));
+            Assert.Equal(output.Length, buffer.Read(output));
+        }
+        Assert.InRange(buffer.CountFrames, 480, 3840);
+        long before = GC.GetAllocatedBytesForCurrentThread();
+        for (int i = 0; i < 100; i++) { buffer.TryWrite(input.AsSpan(0, 960)); buffer.Read(output); }
+        Assert.Equal(0, GC.GetAllocatedBytesForCurrentThread() - before);
+    }
+
+    [Fact]
+    public void OneCaptureMissDoesNotRestartTalkspurtButSustainedStarvationDoes()
+    {
+        var sent = new List<(uint Timestamp, VoiceFrameFlags Flags)>();
+        using var engine = new AudioEngine((_, timestamp, _, flags) => { sent.Add((timestamp, flags)); return true; }, false)
+        { InputMode = AudioInputMode.AlwaysOn, DeviceBufferMilliseconds = 20 };
+        engine.AddLocalStream(Stream()); short[] tone = Tone();
+        engine.FeedPcm(1, tone, 1); engine.ProcessCycle();
+        engine.ProcessCycle();
+        engine.FeedPcm(1, tone, 1); engine.ProcessCycle();
+        Assert.Equal(2, sent.Count); Assert.True((sent[0].Flags & VoiceFrameFlags.Marker) != 0); Assert.Equal(VoiceFrameFlags.None, sent[1].Flags & VoiceFrameFlags.Marker);
+        for (int i = 0; i < 10; i++) engine.ProcessCycle();
+        engine.FeedPcm(1, tone, 1); engine.ProcessCycle();
+        Assert.True((sent[^1].Flags & VoiceFrameFlags.Marker) != 0);
+    }
+
+    [Theory]
+    [InlineData(5)] [InlineData(10)] [InlineData(20)] [InlineData(40)] [InlineData(60)]
+    public void RecoveryLookaheadTracksChannelFrameDuration(int frameMilliseconds)
+    {
+        using var stream = new ReceiveStream(2, Stream(frameMilliseconds));
+        Assert.Equal(frameMilliseconds * 48, stream.TargetDepthSamples);
+    }
+
+    [Fact]
+    public void JitterTargetAdaptsInSampleTimeAndRemainsCapped()
+    {
+        var clock = new ManualAudioClock(); StreamInfo info = Stream(); info.Audio.Fec = false;
+        using var stream = new ReceiveStream(2, info, clock); using var encoder = new OpusEncoder(new() { Bitrate = 32000 });
+        byte[] packet = new byte[1275]; int length = encoder.Encode(Tone(), packet);
+        for (uint i = 0; i < 40; i++)
+        {
+            clock.Advance(i % 2 == 0 ? 5 : 35);
+            stream.Enqueue(new(MediaFrameType.Voice, 0, 0, 42, i, i * 960), packet.AsSpan(0, length));
+        }
+        int[] output = new int[1920]; stream.Mix(output, false, null);
+        Assert.InRange(stream.TargetDepthSamples, 960, 5760);
+    }
+
+    [Fact]
+    public void DredUsesTimestampOffsetForConsecutiveMissingShortFrames()
+    {
+        using var probe = new OpusEncoder();
+        if (!probe.SupportsDeepRedundancy) return;
+        StreamInfo info = Stream(10, dred: true); using var stream = new ReceiveStream(2, info);
+        using var encoder = new OpusEncoder(new() { FrameDurationMilliseconds = 10, DeepRedundancy = true, ExpectedPacketLossPercent = 30, Bitrate = 64000 });
+        byte[] packet = new byte[1275]; short[] tone = new short[480]; int[] output = new int[1920];
+        for (uint i = 0; i < 50; i++)
+        {
+            CodecTests.FillTone(tone, 480, 1, 48000, (int)i); int length = encoder.Encode(tone, packet);
+            if (i is not (25 or 26)) stream.Enqueue(new(MediaFrameType.Voice, 0, 0, 42, i, i * 480), packet.AsSpan(0, length));
+            if (i % 2 == 1) { output.AsSpan().Clear(); stream.Mix(output, false, null); }
+        }
+        Assert.True(stream.DredFrames >= 2);
+    }
+
     [Theory]
     [InlineData(true)]
     [InlineData(false)]
@@ -114,4 +192,12 @@ public class AudioEngineTests
         for (int i = 0; i < 15; i++) { send.FeedPcm(1, stereo, 2); send.ProcessCycle(); receive.ProcessCycle(); }
         Assert.True(energy > 100000);
     }
+
+    private sealed class ManualAudioClock : TimeProvider
+    {
+        private long milliseconds;
+        public override long TimestampFrequency => 1000;
+        public override long GetTimestamp() => milliseconds;
+        internal void Advance(int value) => milliseconds += value;
+    }
 }