Fix RTSP clock pacing at 20 fps

screen-rtsp.ps1

@@ -12,7 +12,10 @@ param(
     [int]$ChannelCount = 2,
 
     [ValidateRange(1, 240)]
-    [int]$FrameRate = 30,
+    [int]$FrameRate = 20,
+
+    [ValidateRange(1, 10)]
+    [int]$GopSeconds = 2,
 
     [ValidatePattern("^[0-9]+[kKmM]$")]
     [string]$VideoBitrate = "8000k",
@@ -536,9 +539,9 @@ namespace ScreenRtsp.Audio
                 ThrowIfFailed(client.Start(), "IAudioClient.Start");
 
                 byte[] buffer = new byte[1024 * 1024];
-                int silenceFrames = Math.Max(1, (int)(format.SamplesPerSec / 100));
-                int silenceBytes = checked(silenceFrames * format.BlockAlign);
-                System.Diagnostics.Stopwatch packetClock =
+                int maximumSilenceFrames = Math.Max(1, (int)(format.SamplesPerSec / 20));
+                long framesWritten = 0;
+                System.Diagnostics.Stopwatch streamClock =
                     System.Diagnostics.Stopwatch.StartNew();
                 while (pipe.IsConnected)
                 {
@@ -548,17 +551,23 @@ namespace ScreenRtsp.Audio
 
                     if (packetFrames == 0)
                     {
-                        // Some endpoints stop producing loopback packets while
-                        // completely silent or power-saving. Keep FFmpeg and
-                        // RTSP clocks moving with one 10 ms silent PCM block.
-                        if (packetClock.ElapsedMilliseconds >= 10)
+                        // Sleep(2) can wake at roughly 15.6 ms on Windows. A
+                        // fixed 10 ms block per wake therefore makes the audio
+                        // clock run slow. Pace silence from elapsed wall time.
+                        long targetFrames = (long)(
+                            streamClock.Elapsed.TotalSeconds * format.SamplesPerSec);
+                        long missingFrames = targetFrames - framesWritten;
+                        if (missingFrames > 0)
                         {
+                            int silenceFrames = (int)Math.Min(
+                                missingFrames, maximumSilenceFrames);
+                            int silenceBytes = checked(silenceFrames * format.BlockAlign);
                             if (buffer.Length < silenceBytes) buffer = new byte[silenceBytes];
                             Array.Clear(buffer, 0, silenceBytes);
                             pipe.Write(buffer, 0, silenceBytes);
-                            packetClock.Restart();
+                            framesWritten += silenceFrames;
                         }
-                        Thread.Sleep(2);
+                        Thread.Sleep(1);
                         continue;
                     }
 
@@ -582,7 +591,7 @@ namespace ScreenRtsp.Audio
                         try
                         {
                             pipe.Write(buffer, 0, bytes);
-                            packetClock.Restart();
+                            framesWritten += frames;
                         }
                         finally
                         {
@@ -710,7 +719,7 @@ function Assert-IniSettings {
     param($Configuration)
     $allowed = @{
         General = @("ChannelCount", "StableTimeSec", "PollIntervalSec")
-        Video = @("FrameRate", "Bitrate", "EncoderPriority")
+        Video = @("FrameRate", "GopSeconds", "Bitrate", "EncoderPriority")
         Audio = @("Enabled", "MicrophoneEnabled", "Bitrate", "SignalThreshold")
         Network = @("NicDescription", "HealthTarget")
         RTSP = @(
@@ -779,6 +788,10 @@ if ($ConfigWasLoaded) {
     if ($setting.Found -and -not $ExplicitParameters.ContainsKey("FrameRate")) {
         $FrameRate = [int]$setting.Value
     }
+    $setting = Get-IniSetting $ini "Video" "GopSeconds"
+    if ($setting.Found -and -not $ExplicitParameters.ContainsKey("GopSeconds")) {
+        $GopSeconds = [int]$setting.Value
+    }
     $setting = Get-IniSetting $ini "Video" "Bitrate"
     if ($setting.Found -and -not $ExplicitParameters.ContainsKey("VideoBitrate")) {
         $VideoBitrate = [string]$setting.Value
@@ -888,6 +901,7 @@ Assert-IntegerRange "General.ChannelCount" $ChannelCount 1 16
 Assert-IntegerRange "General.StableTimeSec" $StableTimeSec 1 60
 Assert-IntegerRange "General.PollIntervalSec" $PollIntervalSec 1 60
 Assert-IntegerRange "Video.FrameRate" $FrameRate 1 240
+Assert-IntegerRange "Video.GopSeconds" $GopSeconds 1 10
 Assert-IntegerRange "RTSP.Port" $RtspPort 1 65535
 Assert-IntegerRange "RTSP.HealthCheckIntervalSec" $RtspHealthCheckIntervalSec 1 3600
 Assert-IntegerRange "RTSP.HealthTimeoutSec" $RtspHealthTimeoutSec 1 60
@@ -902,6 +916,7 @@ if ($AudioBitrate -notmatch '^[0-9]+[kKmM]$') {
 if ($AudioSignalThreshold -lt 0.0 -or $AudioSignalThreshold -gt 1.0) {
     throw "Audio.SignalThreshold 값은 0.0 이상 1.0 이하여야 합니다: $AudioSignalThreshold"
 }
+$GopSize = $FrameRate * $GopSeconds
 
 if (-not [string]::IsNullOrWhiteSpace($EncoderPriorityCsv)) {
     $EncoderPriority = @($EncoderPriorityCsv -split ',' | ForEach-Object { $_.Trim() })
@@ -1059,6 +1074,7 @@ function Get-HiddenWorkerArgumentList {
         "-EncoderPriorityCsv", ($EncoderPriority -join ","),
         "-ChannelCount", [string]$ChannelCount,
         "-FrameRate", [string]$FrameRate,
+        "-GopSeconds", [string]$GopSeconds,
         "-VideoBitrate", $VideoBitrate,
         "-AudioBitrate", $AudioBitrate,
         "-AudioEnabled", [string]$AudioEnabled,
@@ -1153,6 +1169,8 @@ if ($ShowConfiguration) {
         StableTimeSec = $StableTimeSec
         PollIntervalSec = $PollIntervalSec
         FrameRate = $FrameRate
+        GopSeconds = $GopSeconds
+        GopSize = $GopSize
         VideoBitrate = $VideoBitrate
         EncoderPriority = ($EncoderPriority -join " > ")
         AudioEnabled = $AudioEnabled
@@ -1662,7 +1680,7 @@ function Get-EncoderDefinition {
                 Vendor = "AMD"; Name = "h264_amf"
                 Arguments = @(
                     "-usage","lowlatency","-quality","speed","-rc","cbr","-latency","1",
-                    "-forced_idr","1","-header_spacing",[string]$FrameRate
+                    "-forced_idr","1","-header_spacing",[string]$GopSize
                 )
             }
         }
@@ -1747,8 +1765,10 @@ function Test-HardwareEncoder {
     $testDirectory = New-DependencyTempDirectory
     $bitstreamPath = Join-Path $testDirectory ("{0}.h264" -f $Definition.Name)
     $testFrameRate = [Math]::Max(1, [int]$FrameRate)
-    $frameCount = $testFrameRate * [Math]::Max(3, $TestSeconds)
-    $expectedIdrFrames = [Math]::Max(3, [int][Math]::Floor(($frameCount - 1) / $testFrameRate) + 1)
+    $testGopSize = [Math]::Max(1, [int]$GopSize)
+    $effectiveTestSeconds = [Math]::Max([Math]::Max(3, $TestSeconds), (($GopSeconds * 2) + 1))
+    $frameCount = $testFrameRate * $effectiveTestSeconds
+    $expectedIdrFrames = [Math]::Max(3, [int][Math]::Floor(($frameCount - 1) / $testGopSize) + 1)
     $arguments = @(
         "-hide_banner", "-loglevel", "verbose", "-nostdin",
         "-f", "lavfi", "-i", "testsrc2=size=1280x720:rate=$testFrameRate",
@@ -1756,8 +1776,8 @@ function Test-HardwareEncoder {
         "-c:v", $Definition.Name
     ) + $Definition.Arguments + @(
         "-b:v", "4000k", "-maxrate", "4000k", "-bufsize", "4000k",
-        "-g", [string]$testFrameRate, "-bf", "0",
-        "-force_key_frames", "expr:gte(t,n_forced*1)",
+        "-g", [string]$testGopSize, "-bf", "0",
+        "-force_key_frames", "expr:gte(t,n_forced*$GopSeconds)",
         "-f", "h264", "-y", $bitstreamPath
     )
     try {
@@ -1799,7 +1819,7 @@ function Test-HardwareEncoder {
             Encoder = $Definition.Name
             Passed = $passed
             ExitCode = $result.ExitCode
-            TestSeconds = [Math]::Max(3, $TestSeconds)
+            TestSeconds = $effectiveTestSeconds
             Frames = if ($metrics) { $metrics.DecodedFrames } else { 0 }
             IdrFrames = if ($metrics) { $metrics.IdrFrameCount } else { 0 }
             NalIdrCount = if ($metrics) { $metrics.NalIdrCount } else { 0 }
@@ -2147,7 +2167,7 @@ function Get-EnvironmentFingerprint {
     }
     else { "" }
     $source = (Get-GpuSignature) + "|" + $ffmpegHash + "|" + $mediaHash + "|" + $scriptHash + "|" +
-        ($EncoderPriority -join ",") + "|$FrameRate|$VideoBitrate"
+        ($EncoderPriority -join ",") + "|$FrameRate|$GopSeconds|$VideoBitrate"
     $sha = [System.Security.Cryptography.SHA256]::Create()
     try {
         $bytes = [System.Text.Encoding]::UTF8.GetBytes($source)
@@ -2948,7 +2968,14 @@ function Get-VideoFilter {
     $filters = @("hwdownload", "format=bgra")
     $rotationFilter = Get-RotationFilter $Screen.Orientation
     if ($rotationFilter) { $filters += $rotationFilter.Split(',') }
-    $filters += @("setpts=N/(${FrameRate}*TB)", "format=nv12")
+    # ddagrab emits one probe frame before slower inputs (notably DirectShow)
+    # finish opening. Drop that stale frame, then preserve DDA elapsed-time PTS.
+    # Do not synthesize missed frames: doing so creates an encode backlog and
+    # makes media time lag wall time on an overloaded GPU.
+    $filters += @(
+        "trim=start_frame=1", "setpts=PTS-STARTPTS", "settb=AVTB",
+        "format=nv12"
+    )
     return ($filters -join ",")
 }
 
@@ -2966,37 +2993,38 @@ function New-FFmpegArguments {
         "-hide_banner", "-loglevel", "warning", "-nostdin"
     )
     $filterParts = @()
-
-    for ($index = 0; $index -lt $SelectedScreens.Count; $index++) {
-        $screen = $SelectedScreens[$index]
-        $ddaInput = "ddagrab=output_idx=$($screen.OutputIndex):draw_mouse=1:framerate=${FrameRate}:dup_frames=1"
-        $arguments += @("-f", "lavfi", "-i", $ddaInput)
-        $filterParts += "[$index`:v]$(Get-VideoFilter $screen)[v$index]"
-    }
+    $videoInputOffset = 0
 
     if ($AudioEnabled -eq 1) {
-        $playbackInputIndex = $SelectedScreens.Count
+        # Open slower audio devices first. ddagrab creates a probe frame as
+        # soon as its input is opened; opening it before DirectShow gives the
+        # first two DDA frames timestamps that are several seconds apart.
+        $playbackInputIndex = 0
         $arguments += @(
             "-thread_queue_size", "1024", "-f", $PlaybackEndpoint.SampleFormat,
             "-ar", $PlaybackEndpoint.SampleRate, "-ac", $PlaybackEndpoint.Channels,
             "-i", $AudioPipePath
         )
-        $filterParts += "[$playbackInputIndex`:a]aresample=48000:async=1:first_pts=0," +
+        $filterParts += "[$playbackInputIndex`:a]asetpts=PTS-STARTPTS," +
+            "aresample=48000:async=1:first_pts=0," +
             "aformat=sample_fmts=fltp:channel_layouts=stereo[sys]"
 
         if ($MicrophoneEnabled -eq 1) {
             $microphoneInputIndex = $playbackInputIndex + 1
             $arguments += @(
-                "-thread_queue_size", "1024", "-f", "dshow",
+                "-thread_queue_size", "1024", "-rtbufsize", "32M", "-f", "dshow",
                 "-i", "audio=$MicrophoneInputName"
             )
-            $filterParts += "[$microphoneInputIndex`:a]aresample=48000:async=1:first_pts=0," +
+            $filterParts += "[$microphoneInputIndex`:a]asetpts=PTS-STARTPTS," +
+                "aresample=48000:async=1:first_pts=0," +
                 "aformat=sample_fmts=fltp:channel_layouts=stereo[mic]"
             $filterParts += "[sys][mic]amix=inputs=2:duration=longest:" +
                 "dropout_transition=2:normalize=1[mixed]"
+            $videoInputOffset = 2
         }
         else {
             $filterParts += "[sys]anull[mixed]"
+            $videoInputOffset = 1
         }
 
         if ($SelectedScreens.Count -eq 1) {
@@ -3008,6 +3036,14 @@ function New-FFmpegArguments {
         }
     }
 
+    for ($index = 0; $index -lt $SelectedScreens.Count; $index++) {
+        $screen = $SelectedScreens[$index]
+        $videoInputIndex = $videoInputOffset + $index
+        $ddaInput = "ddagrab=output_idx=$($screen.OutputIndex):draw_mouse=1:framerate=${FrameRate}:dup_frames=1"
+        $arguments += @("-f", "lavfi", "-i", $ddaInput)
+        $filterParts += "[$videoInputIndex`:v]$(Get-VideoFilter $screen)[v$index]"
+    }
+
     $arguments += @("-filter_complex", ($filterParts -join ";"))
 
     for ($index = 0; $index -lt $SelectedScreens.Count; $index++) {
@@ -3018,8 +3054,10 @@ function New-FFmpegArguments {
         $arguments += @($Encoder.Arguments)
         $arguments += @(
             "-b:v", $VideoBitrate, "-maxrate", $VideoBitrate, "-bufsize", $VideoBitrate,
-            "-g", $FrameRate, "-bf", "0", "-pix_fmt", "nv12",
-            "-force_key_frames", "expr:gte(t,n_forced*1)"
+            "-g", $GopSize, "-bf", "0", "-pix_fmt", "nv12",
+            "-force_key_frames", "expr:gte(t,n_forced*$GopSeconds)",
+            "-enc_time_base", "filter",
+            "-fps_mode", "passthrough"
         )
         if ($AudioEnabled -eq 1) {
             $arguments += @(
@@ -3027,7 +3065,8 @@ function New-FFmpegArguments {
             )
         }
         $arguments += @(
-            "-f", "rtsp", "-rtsp_transport", "tcp", "$RTSPPublishBase/monitor$channel"
+            "-f", "rtsp", "-max_interleave_delta", "100000", "-flush_packets", "1",
+            "-rtsp_transport", "tcp", "$RTSPPublishBase/monitor$channel"
         )
     }
 
@@ -3652,6 +3691,13 @@ function Invoke-SimulationTests {
                 $qsv -contains "-forced_idr" -and $qsv -contains "-idr_interval")
         }
 
+        Invoke-SimulationCase "무음 WASAPI loopback은 경과시간 기준으로 PCM 시계 유지" {
+            return ($AudioInteropSource -match 'streamClock\.Elapsed\.TotalSeconds' -and
+                $AudioInteropSource -match 'targetFrames\s*-\s*framesWritten' -and
+                $AudioInteropSource -match 'framesWritten\s*\+=\s*silenceFrames' -and
+                $AudioInteropSource -notmatch 'packetClock\.Restart')
+        }
+
         Invoke-SimulationCase "자동 실행은 숨김 worker와 config.ini만 전달" {
             $arguments = Get-StartupArgumentList
             return ($arguments -match '-WindowStyle\s+Hidden' -and
@@ -3715,6 +3761,7 @@ function Invoke-SimulationTests {
             $filterIndex = [Array]::IndexOf($arguments, "-filter_complex")
             $filter = if ($filterIndex -ge 0) { $arguments[$filterIndex + 1] } else { "" }
             return ($filter -match 'amix=inputs=2' -and $filter -match 'asplit=2' -and
+                @([regex]::Matches($filter, 'asetpts=PTS-STARTPTS')).Count -eq 2 -and
                 $arguments -contains 'dshow' -and
                 $arguments -contains '-force_key_frames' -and
                 @($arguments | Where-Object { $_ -eq 'h264_nvenc' }).Count -eq 2 -and
@@ -3722,6 +3769,32 @@ function Invoke-SimulationTests {
                 @($arguments | Where-Object { $_ -eq "$RTSPPublishBase/monitor2" }).Count -eq 1)
         }
 
+        Invoke-SimulationCase "DDA와 인코더가 20fps/GOP40 및 원본 PTS를 일관되게 사용" {
+            $script:AudioEnabled = 1
+            $script:MicrophoneEnabled = 1
+            $arguments = @(New-FFmpegArguments `
+                -SelectedScreens @($screenPrimary,$screenSecond) `
+                -PlaybackEndpoint $defaultAudio -MicrophoneEndpoint $defaultMicrophone `
+                -MicrophoneInputName $defaultMicrophone.Name `
+                -Encoder (Get-EncoderDefinition "Intel") -AudioPipePath "\\.\pipe\simulation")
+            $filterIndex = [Array]::IndexOf($arguments, "-filter_complex")
+            $filter = if ($filterIndex -ge 0) { $arguments[$filterIndex + 1] } else { "" }
+            $joined = $arguments -join " "
+            $pipePosition = $joined.IndexOf('\\.\pipe\simulation')
+            $ddaPosition = $joined.IndexOf('ddagrab=')
+            return ($FrameRate -eq 20 -and $GopSize -eq 40 -and
+                $pipePosition -ge 0 -and $ddaPosition -gt $pipePosition -and
+                $joined -match 'framerate=20' -and
+                $filter -match 'trim=start_frame=1' -and
+                $filter -match 'setpts=PTS-STARTPTS' -and $filter -match 'settb=AVTB' -and
+                $filter -notmatch 'setpts=N/' -and
+                @($arguments | Where-Object { $_ -eq "40" }).Count -ge 2 -and
+                @($arguments | Where-Object { $_ -eq "filter" }).Count -eq 2 -and
+                @($arguments | Where-Object { $_ -eq "passthrough" }).Count -eq 2 -and
+                @($arguments | Where-Object { $_ -eq "100000" }).Count -eq 2 -and
+                $joined -match 'n_forced\*2')
+        }
+
         Invoke-SimulationCase "마이크 OFF이면 시스템 재생음만 두 채널에 분배" {
             $script:AudioEnabled = 1
             $script:MicrophoneEnabled = 0
@@ -3966,7 +4039,7 @@ try {
         "${RtspHealthCheckIntervalSec}초 주기 / ${RtspHealthTimeoutSec}초 제한 / " +
         "${RtspHealthFailureThreshold}회 연속 실패 시 재시작"
     )
-    Write-Log "영상 설정: ${FrameRate}fps / 채널당 $VideoBitrate"
+    Write-Log "영상 설정: ${FrameRate}fps / GOP ${GopSize}프레임(${GopSeconds}초) / DDA PTS 유지 / 채널당 $VideoBitrate"
 
     $lastSystemSignature = ""
     $healthySince = $null