How to Measure HDMI Audio Latency: Video Transcript

This video demonstrates measuring HDMI audio latency using the AV Toolkit. First we’ll measure an esports monitor and second an AV receiver. These audio latency measurements can be used to calculate lip sync error if you have also measured the video latency of your display. It is important to note that video latency is slightly different than video input lag: Input lag cannot be used for audio/video synchronization calculations, but video latency can. For an explanation of this difference, you can visit the terminology page of AV

This software toolkit makes use of inexpensive consumer electronics, like this HDMI Audio Extractor, to accurately measure audio latency. It works by using any analog stereo audio input on your computer, such as the microphone port, to record the analog output of the audio extractor and the audio output of the device under test. It then measures the time offset between the two analog audio sources. Of course, this raw time offset doesn’t guarantee any sort of accuracy to the measurements. To correct this and report accurate measurements, the offset between the analog audio output and the HDMI audio output for this audio extractor has been measured using an oscilloscope to produce a profile that is used by this software.

If you don’t have a stereo mic input on your computer, you can use a USB device with a stereo input or a professional audio interface and microphone.

See the video description for a link to other requirements. 

We’ll first measure the headphone output of the monitor and later use a microphone to measure the speaker output of the receiver. A device’s headphone output can be very different than its speaker output, so it is important to measure both. For example, my TV’s speaker output is 28 ms slower than its headphone output.

Let’s get started by downloading the toolkit and extracting it to the desktop.

This toolkit has some other features as well: it is capable of measuring S/PDIF audio latency using a similar technique as shown in this video. It can also measure the time taken to convert HDMI audio to another digital audio format, such as ARC, eARC, or S/PDIF, by using an additional reference audio DAC.

The app presents a cable diagram that we’ll use to set up our devices. The HDMI output of our computer must go into the HDMI Audio Extractor, which is our Dual-Output Reference Device. Its HDMI output must be connected to the device under test.

Next we’ll connect the analog audio cables. Normally white is left and red is right, but some stereo breakout cables are wired backwards, like this one. Also, some audio devices have the left and right input channels reversed, like this Pluggable USB audio adapter. It will be easy to tell which way yours are wired in the upcoming steps when we adjust the volume levels.

The analog output of the audio extractor needs to be connected to the left channel of our stereo breakout and the monitor’s headphone output needs to be connected to the right channel.

The breakout cable can now be connected to the stereo microphone port of the computer.

Our cables match the diagram, so we’re finished with this step.

For your reference, here is a similar cable setup that uses a microphone.

The app shows a reminder to correctly set the EDID switch of the audio extractor. This ensures that all audio formats supported by the device under test will be available to measure.

Next, select the audio output and input devices.

When adjusting volumes, a ticking pattern is played to the output device. I’ll quickly unplug the headphone output of the monitor to demonstrate what this sounds like.

You can verify that your cables are connected correctly by using the peak meters on the left side. By changing the volume of one of your devices, you should see its peak level change without affecting the other channel.

As I turn down the headphone output volume of the monitor, you can see the peak level of the correct channel decreases without affecting the other channel. As this signal becomes quieter, it will eventually be overtaken by crosstalk from the other channel.

By unplugging the HDMI audio extractor’s output, you can see that its peak level changes without affecting the level of the other channel.

It is important that your left and right channels are wired the right way around. Otherwise, results will be incorrect.

Clipping, background noise, or weak signals will never affect the accuracy of final results because of a number of safeguards in the tool. That said, a high input volume can sometimes be problematic and cause measurements to be marked as invalid and discarded. Here is a different computer with a microphone volume level that is too high.

Simply turn off microphone boost and turn down the volume level to correct this.

I would also recommend disabling audio enhancements.

Going back to our previous setup, we see that the peak levels are not too high and the signal quality is OK, so we’re finished adjusting volumes.

Now we’re at the final step before measuring. Select your Dual-Out Reference Device, which is the HDMI audio extractor that we’re using, and the audio formats that you want to measure. I’m going to include one additional sample rate and also decrease the number of measurements to speed up this test.

At this point we could go ahead and start the measurement, but I’m going to fill in these optional test notes that will be automatically saved at the end of the test.

And here is the audio latency of this esports monitor. This is an exceptionally low audio latency, which is needed to match the very low video latency that this monitor has.

In the Detailed Results tab we can see measurements for the other formats we selected.

Lastly, we can open the results folder to find a CSV spreadsheet containing timestamped results that include our test notes. The “Average audio latency” is the value that should be reported and used for calculating lip sync error.

Next, let’s try using a microphone to measure audio latency of the speaker output of this AV receiver.

When using a microphone, it should be positioned as close to the speaker as possible to measure the audio latency that is specific to the receiver. This is a different technique than calibration of a room setup where you might want to include the time it takes sound to travel to a specific listening location.

Similar to last time, the microphone should be connected to the right channel of the breakout cable.

After changing HDMI audio devices like we did here, make sure to click the “Refresh Audio Devices” button.

Since we are now using a microphone, we can expect that the two channels will have very different volume levels and that’s totally fine, so long as neither is too loud. Again, we’ll check to make sure that the cables are connected correctly.

You’ll notice that the number of audio formats supported by this receiver, according to its EDID, are far greater than those supported by previous monitor we tested. I’ll add on a few additional audio formats to this test and update the test notes.

The test will first measure each audio format once. This progress is shown by the bottom progress bar. After all audio formats have been measured once, it will move on to the next pass, according to the number of measurements that were chosen earlier. The overall test progress is shown by the top progress bar.

At any time during the test you can press the stop button to see the results from measurements that have been completed. There’s no need to wait until all the measurements have been completed, other than to increase the accuracy of the final result.

If at any point during the test an audio format fails it will be skipped in the following passes to speed up the test. The number of attempts before skipping an audio format can be set in the Advanced Configuration.

It’s looking like this will take a little while longer because of the time it takes for the receiver to sync to different audio formats, so I’ll fast-forward this video to the results.

Here we can see the audio latency of this receiver. Note that the latency is quite a bit higher for stereo than for 5.1 surround. I have also seen this type of behaviour on a newer Yamaha receiver.

We can also see that the sample rate has an impact on audio latency.

Like last time, the detailed results are saved in a CSV file for your records.

This tool is fully open source under the MIT License. Please see the video description for links to more resources on the topics of audio latency, video latency, and audio video synchronization.

Last updated on May 3rd, 2022.