Name		Name	Last commit message	Last commit date
parent directory ..
AdpfWrapper.cpp		AdpfWrapper.cpp
AdpfWrapper.h		AdpfWrapper.h
AudioSourceCaller.cpp		AudioSourceCaller.cpp
AudioSourceCaller.h		AudioSourceCaller.h
AudioStream.cpp		AudioStream.cpp
AudioStreamBuilder.cpp		AudioStreamBuilder.cpp
DataConversionFlowGraph.cpp		DataConversionFlowGraph.cpp
DataConversionFlowGraph.h		DataConversionFlowGraph.h
FilterAudioStream.cpp		FilterAudioStream.cpp
FilterAudioStream.h		FilterAudioStream.h
FixedBlockAdapter.cpp		FixedBlockAdapter.cpp
FixedBlockAdapter.h		FixedBlockAdapter.h
FixedBlockReader.cpp		FixedBlockReader.cpp
FixedBlockReader.h		FixedBlockReader.h
FixedBlockWriter.cpp		FixedBlockWriter.cpp
FixedBlockWriter.h		FixedBlockWriter.h
LatencyTuner.cpp		LatencyTuner.cpp
MonotonicCounter.h		MonotonicCounter.h
OboeDebug.h		OboeDebug.h
OboeExtensions.cpp		OboeExtensions.cpp
QuirksManager.cpp		QuirksManager.cpp
QuirksManager.h		QuirksManager.h
README.md		README.md
SourceFloatCaller.cpp		SourceFloatCaller.cpp
SourceFloatCaller.h		SourceFloatCaller.h
SourceI16Caller.cpp		SourceI16Caller.cpp
SourceI16Caller.h		SourceI16Caller.h
SourceI24Caller.cpp		SourceI24Caller.cpp
SourceI24Caller.h		SourceI24Caller.h
SourceI32Caller.cpp		SourceI32Caller.cpp
SourceI32Caller.h		SourceI32Caller.h
StabilizedCallback.cpp		StabilizedCallback.cpp
Trace.cpp		Trace.cpp
Trace.h		Trace.h
Utilities.cpp		Utilities.cpp
Version.cpp		Version.cpp

README.md

Notes on Implementation

Latency from Resampling

There are two components of the latency. The resampler itself, and a buffer that is used to adapt the block sizes.

The resampler is an FIR running at the target sample rate. So its latency is the number of taps. From MultiChannelResampler.cpp, numTaps is

Fastest: 2 Low: 4 Medium: 8 High: 16 Best: 32

For output, the device sampling rate is used, which is typically 48000.For input, the app sampling rate is used.

There is a block size adapter that collects odd sized blocks into larger blocks of the correct size.

The adapter contains one burst of frames, from getFramesPerBurst(). But if the app specifies a particular size using setFramesPerCallback() then that size will be used. Here is some pseudo-code to calculate the latency.

latencyMillis = 0
targetRate = isOutput ? deviceRate : applicationRate
// Add latency from FIR
latencyMillis += numTaps * 1000.0 / targetRate
// Add latency from block size adaptation
adapterSize = (callbackSize > 0) ? callbackSize : burstSize
if (isOutput && isCallbackUsed) latencyMillis += adapterSize * 1000.0 / deviceRate
else if (isInput && isCallbackUsed) latencyMillis += adapterSize * 1000.0 / applicationRate
else if (isInput && !isCallbackUsed) latencyMillis += adapterSize * 1000.0 / deviceRate

oboe/src/common at 1.11.0 · google/oboe · GitHub

Expand file tree

Breadcrumbs

common

Directory actions

More options

Directory actions

More options

Latest commit

History

Breadcrumbs

common

Folders and files

parent directory

README.md

Notes on Implementation

Latency from Resampling