Add overhead to audio bwe min, max.

webrtc/media/engine/webrtcvoiceengine.cc

 /*
  *  Copyright (c) 2004 The WebRTC project authors. All Rights Reserved.
  *
  *  Use of this source code is governed by a BSD-style license
  *  that can be found in the LICENSE file in the root of the source
  *  tree. An additional intellectual property rights grant can be found
  *  in the file PATENTS.  All contributing project authors may
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
@@ skipping 1402 matching lines @@
     RTC_DCHECK(worker_thread_checker_.CalledOnValidThread());
     if (stream_) {
       call_->DestroyAudioSendStream(stream_);
       stream_ = nullptr;
     }
     RTC_DCHECK(!stream_);
     if (webrtc::field_trial::FindFullName("WebRTC-Audio-SendSideBwe") ==
         "Enabled") {
       // TODO(mflodman): Keep testing this and set proper values.
       // Note: This is an early experiment currently only supported by Opus.
-      config_.min_bitrate_bps = kOpusMinBitrateBps;
-      config_.max_bitrate_bps = kOpusBitrateFbBps;
+      if (webrtc::field_trial::FindFullName(
+              "WebRTC-SendSideBwe-WithOverhead") == "Enabled") {
+        // OverheadPerPacket(58) = Ipv4(20B) + UDP(8B) + SRTP(10B) + RTP(12)
+        //     + RTP_ext(8);
+        // kMaxOverheadBps = OverheadPerPacket * 8 * 1000ms / frame_length(20ms)

[minyue-webrtc, 2016/11/24 10:42:10]

use 10ms I think due to
https://cs.chromium.org/chromium/src/third_party/webrtc/media/engine/webrtcvo...

[michaelt, 2016/11/24 10:51:24]

I think we should not us 10ms frames since they are not used in ANA. Setting
here 10ms as min frame length would lead to a higher max encoding rate which
would be a waste of bandwidth. 

[minyue-webrtc, 2016/11/28 15:27:06]

It does not sound convincing to me. Are we only defining a lower/upper bounds
for the overall bitrate here. Or are you saying that the limit is a function of
frame length?

+        // kMinOverheadBps = OverheadPerPacket * 8 * 1000ms / frame_length(60ms)
+        constexpr int kMaxOverheadBps = 23200;
+        constexpr int kMinOverheadBps = 7733;

[stefan-webrtc, 2016/11/24 10:48:27]

This will be hard to maintain. can't we compute them instead based on
https://cs.chromium.org/chromium/src/third_party/webrtc/media/engine/webrtcvo...

[minyue-webrtc, 2016/11/28 15:27:06]

It looks like that the problem is that the min frame length in
webrtcvoiceengine.cc is one thing and the min frame length that can ever happen
in audio network adaptation is another.

But since we are defining lower/upper bound now, it is supposed to work if go
choose global min and max (Opus min and max sounds good to me, too), unless we
made error somewhere.

[stefan-webrtc, 2016/12/22 09:06:25]

I still don't really follow. How are they different? I think that must be made
clear here, otherwise it will be hard for others to change this code.

Can the adaptive configuration override the frame lengths configured here:
https://cs.chromium.org/chromium/src/third_party/webrtc/media/engine/webrtcvo...
?

Seems like we should be restricted by what frame lengths are negotiated?

+        config_.min_bitrate_bps = kOpusMinBitrateBps + kMinOverheadBps;
+        config_.max_bitrate_bps = kOpusBitrateFbBps + kMaxOverheadBps;
+      } else {
+        config_.min_bitrate_bps = kOpusMinBitrateBps;
+        config_.max_bitrate_bps = kOpusBitrateFbBps;
+      }
     }
     stream_ = call_->CreateAudioSendStream(config_);
     RTC_CHECK(stream_);
     UpdateSendState();
   }
 
   rtc::ThreadChecker worker_thread_checker_;
   rtc::RaceChecker audio_capture_race_checker_;
   webrtc::AudioTransport* const voe_audio_transport_ = nullptr;
   webrtc::Call* call_ = nullptr;
@@ skipping 1203 matching lines @@
   RTC_DCHECK(worker_thread_checker_.CalledOnValidThread());
   const auto it = send_streams_.find(ssrc);
   if (it != send_streams_.end()) {
     return it->second->channel();
   }
   return -1;
 }
 }  // namespace cricket
 
 #endif  // HAVE_WEBRTC_VOICE