blob: 80a8bd54cbe443e3812f34794ed87cd3fee80562 [file] [log] [blame]
<html devsite>
<head>
<title>Audio Latency Measurements</title>
<meta name="project_path" value="/_project.yaml" />
<meta name="book_path" value="/_book.yaml" />
</head>
<body>
<!--
Copyright 2017 The Android Open Source Project
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
-->
<p><a href="http://en.wikipedia.org/wiki/Latency_%28engineering%29">Latency</a>
is an important system performance metric. While many kinds of
<a href="latency.html">audio latency</a> metrics exist, one useful and
well-understood metric is
<a href="latency_measure.html#measuringRoundTrip">round-trip latency</a>,
defined as the time it takes for an audio signal to enter the input of a mobile
device, be processed by an app running on the application processor, and exit
the output.</p>
<img src="images/round_trip_on_device.png" alt="Round-trip audio latency on
device" id="figure1" />
<p class="img-caption"><strong>Figure 1.</strong> Round-trip audio latency on
device: T<sub>output</sub> - T<sub>input</sub></p>
<p>This page provides round-trip audio latency measurements for select
Nexus/Pixel devices and Android platform versions.</p>
<h2 id="why">Why we measure latency</h2>
<p>Google measures and reports latency so Android application developers have
the data they need to make informed decisions about available latency on actual
devices. By sharing latency numbers for select Nexus and Pixel devices, we hope
to encourage the entire Android community to measure, publish, and reduce
latency on <em>all</em> Android devices. Please join us in our commitment to
reducing audio latency!</p>
<h2 id="app">Application impact on latency</h2>
<p>Signal processing can add the following types of delay to latency:</p>
<ul>
<li><strong>Algorithmic</strong>. This delay is inherent and does not vary with
the CPU. An example is the delay added by a
<a href="http://en.wikipedia.org/wiki/Finite_impulse_response">finite impulse
response</a> (FIR) filter.</li>
<li><strong>Computational</strong>. This delay is related to the number of
required CPU cycles. For example, attenuation of a signal is usually done by a
multiplication operation that takes a varying number of cycles depending on the
CPU.</li>
</ul>
<h2 id="how">How we measure</h2>
<p>We made the measurements listed on this page using the
<a href="loopback.html">Dr. Rick O'Rang audio loopback dongle</a> and an
<a href="latency_measure.html#larsenTest">audio feedback (Larsen effect)
test</a>. Measurements assume the application signal processing adds zero
algorithmic delay and near-zero computational delay.</p>
<p>We measure round-trip latency via the headset connector for several reasons:
</p>
<img src="images/round_trip_via_headset_connector.png" alt="Round-trip latency
via headset connector" id="figure2" />
<p class="img-caption"><strong>Figure 2.</strong> Round-trip latency via headset
connector: T<sub>output</sub> - T<sub>input</sub></p>
<ul>
<li>Important music applications (such as guitar and voice processing) use the
headset connector.</li>
<li>Measuring round-trip latency of the on-device microphone and speaker can be
cumbersome, as it is difficult to keep a feedback loop in open air from entering
uncontrolled oscillation.</li>
<li>On-device transducers are small and sacrifice frequency response to achieve
their small size. To compensate, digital signal processing is applied but
increases algorithmic delay for the on-device path.</li>
</ul>
<p>There are cases where on-device microphone and speaker latencies <em>do</em>
matter, but they are usually for one direction, not round-trip. Techniques for
measuring unidirectional latency are described at
<a href="latency_measure.html#measuringOutput">Measuring Output Latency</a>
and
<a href="latency_measure.html#measuringInput">Measuring Input Latency</a>.</p>
<h2 id="examples">Example measurements</h2>
<p>The measurements listed below are specific to a
<a href="/setup/build-numbers.html">build number</a>. Devices are listed in
approximate order of initial release and by platform version; you can also
<a href="#chart">view latencies in a chart</a>. The test application uses the
Android native audio API based on OpenSL ES.</p>
<table>
<tr>
<th>Model</th>
<th>Platform<br />version</th>
<th>Build<br />number</th>
<th>Sample rate<br />(Hz)</th>
<th>Buffer size<br />(frames)</th>
<th>Buffer size<br />(ms)</th>
<th>Round-trip<br />latency (ms)<br />&plusmn; one buffer</th>
</tr>
<tr>
<td>Nexus One</td>
<td>2.3.6</td>
<td>GRK39F</td>
<td>44100</td>
<td>768</td>
<td>17.4</td>
<td>345</td>
</tr>
<tr>
<td>Nexus S</td>
<td>2.3.6</td>
<td>GRK39F</td>
<td>44100</td>
<td>1024</td>
<td>23.2</td>
<td>260</td>
</tr>
<tr>
<td>Nexus S</td>
<td>4.0.4</td>
<td>IMM76D</td>
<td>44100</td>
<td>1024</td>
<td>23.2</td>
<td>260</td>
</tr>
<tr>
<td>Nexus S</td>
<td>4.1.2</td>
<td>JZO54K</td>
<td>44100</td>
<td>880</td>
<td>20</td>
<td>210</td>
</tr>
<tr>
<td>Galaxy Nexus</td>
<td>4.0.1</td>
<td>ITL41D</td>
<td>44100</td>
<td>976</td>
<td>22.1</td>
<td>270</td>
</tr>
<tr>
<td>Galaxy Nexus</td>
<td>4.3</td>
<td>JWR66Y</td>
<td>44100</td>
<td>144</td>
<td>3.3</td>
<td>130</td>
</tr>
<tr>
<td>Nexus 4</td>
<td>4.2.2</td>
<td>JDQ39E</td>
<td>48000</td>
<td>240</td>
<td>5</td>
<td>195</td>
</tr>
<tr>
<td>Nexus 4</td>
<td>5.1</td>
<td>LMY47O</td>
<td>48000</td>
<td>240</td>
<td>5</td>
<td>58</td>
</tr>
<tr>
<td>Nexus 10</td>
<td>5.0.2</td>
<td>LRX22G</td>
<td>44100</td>
<td>256</td>
<td>5.8</td>
<td>36</td>
</tr>
<tr>
<td>Nexus 10</td>
<td>5.1</td>
<td>LMY47D</td>
<td>44100</td>
<td>256</td>
<td>5.8</td>
<td>35</td>
</tr>
<tr>
<td>Nexus 7<br />2013</td>
<td>4.3</td>
<td>JSR78D</td>
<td>48000</td>
<td>240</td>
<td>5</td>
<td>149</td>
</tr>
<tr>
<td>Nexus 7<br />2013</td>
<td>4.4</td>
<td>KRT16S</td>
<td>48000</td>
<td>240</td>
<td>5</td>
<td>85</td>
</tr>
<tr>
<td>Nexus 7<br />2013</td>
<td>5.0.2</td>
<td>LRX22G</td>
<td>48000</td>
<td>240</td>
<td>5</td>
<td>64</td>
</tr>
<tr>
<td>Nexus 7<br />2013</td>
<td>5.1</td>
<td>LMY47O</td>
<td>48000</td>
<td>240</td>
<td>5</td>
<td>55</td>
</tr>
<tr>
<td>Nexus 7<br />2013</td>
<td>6.0</td>
<td>MRA58K</td>
<td>48000</td>
<td>240</td>
<td>5</td>
<td>55</td>
</tr>
<tr>
<td>Nexus 5</td>
<td>4.4.4</td>
<td>KTU84P</td>
<td>48000</td>
<td>240</td>
<td>5</td>
<td>95</td>
</tr>
<tr>
<td>Nexus 5</td>
<td>5.0.0</td>
<td>LRX21O</td>
<td>48000</td>
<td>240</td>
<td>5</td>
<td>47</td>
</tr>
<tr>
<td>Nexus 5</td>
<td>5.1</td>
<td>LMY47I</td>
<td>48000</td>
<td>240</td>
<td>5</td>
<td>42</td>
</tr>
<tr>
<td>Nexus 5</td>
<td>6.0</td>
<td>MRA58K</td>
<td>48000</td>
<td>192</td>
<td>4</td>
<td>38</td>
</tr>
<tr>
<td>Nexus 9</td>
<td>5.0.0</td>
<td>LRX21L</td>
<td>48000</td>
<td>256</td>
<td>5.3</td>
<td>35</td>
</tr>
<tr>
<td>Nexus 9</td>
<td>5.0.1</td>
<td>LRX22C</td>
<td>48000</td>
<td>256</td>
<td>5.3</td>
<td>38</td>
</tr>
<tr>
<td>Nexus 9</td>
<td>5.1.1</td>
<td>LMY47X</td>
<td>48000</td>
<td>256</td>
<td>5.3</td>
<td>32</td>
</tr>
<tr>
<td>Nexus 9</td>
<td>6.0</td>
<td>MRA58K</td>
<td>48000</td>
<td>128</td>
<td>2.6</td>
<td>15</td>
</tr>
<tr>
<td>Nexus 6</td>
<td>5.0.1</td>
<td>LRX22C</td>
<td>48000</td>
<td>240</td>
<td>5</td>
<td>65</td>
</tr>
<tr>
<td>Nexus 6</td>
<td>5.1</td>
<td>LMY47I</td>
<td>48000</td>
<td>240</td>
<td>5</td>
<td>42</td>
</tr>
<tr>
<td>Nexus 6</td>
<td>6.0</td>
<td>MRA58K</td>
<td>48000</td>
<td>192</td>
<td>4</td>
<td>33</td>
</tr>
<tr>
<td>Nexus 5X</td>
<td>6.0</td>
<td>MDA89E</td>
<td>48000</td>
<td>192</td>
<td>4</td>
<td>18</td>
</tr>
<tr>
<td>Nexus 5X</td>
<td>8.0.0</td>
<td>OPR4.170623.020</td>
<td>48000</td>
<td>192</td>
<td>4</td>
<td>18</td>
</tr>
<tr>
<td>Nexus 5X</td>
<td>8.1.0</td>
<td>OPM2.171019.029.C1</td>
<td>48000</td>
<td>192</td>
<td>4</td>
<td>18</td>
</tr>
<tr>
<td>Nexus 6P</td>
<td>6.0</td>
<td>MDA89D</td>
<td>48000</td>
<td>192</td>
<td>4</td>
<td>18</td>
</tr>
<tr>
<td>Nexus 6P</td>
<td>8.0.0</td>
<td>OPR5.170623.014</td>
<td>48000</td>
<td>192</td>
<td>4</td>
<td>18</td>
</tr>
<tr>
<td>Nexus 6P</td>
<td>8.1.0</td>
<td>OPM5.171019.019</td>
<td>48000</td>
<td>192</td>
<td>4</td>
<td>18</td>
</tr>
<tr>
<td>Pixel</td>
<td>7.1.2</td>
<td>NHG47L</td>
<td>48000</td>
<td>192</td>
<td>4</td>
<td>18</td>
</tr>
<tr>
<td>Pixel</td>
<td>8.0.0</td>
<td>OPR3.170623.013</td>
<td>48000</td>
<td>192</td>
<td>4</td>
<td>18</td>
</tr>
<tr>
<td>Pixel</td>
<td>8.1.0</td>
<td>OPM1.171019.021</td>
<td>48000</td>
<td>192</td>
<td>4</td>
<td>18</td>
</tr>
<tr>
<td>Pixel XL</td>
<td>7.1.2</td>
<td>NHG47L</td>
<td>48000</td>
<td>192</td>
<td>4</td>
<td>18</td>
</tr>
<tr>
<td>Pixel XL</td>
<td>8.0.0</td>
<td>OPR3.170623.013</td>
<td>48000</td>
<td>192</td>
<td>4</td>
<td>18</td>
</tr>
</table>
<a id=chart></a>
<img src="/devices/audio/images/round-trip-latencies.png" />
<p class="img-caption"><strong>Figure 3.</strong> Round trip latencies.</p>
</body>
</html>