Components of Bluetooth audio delay/latency?

[NTL2009]

TL;DR Summary

What is responsible for Bluetooth audio latency?

I often watch instructional YouTube videos on my tablet with Bluetooth headphones. The delay/latency in the audio often isn't so noticeable, but when I see someone using a hammer, or sandpaper, the delay between the visual of the hammer hit or sandpaper swipe and the audible "bam" or "swoosh" is jarring. Which always reminds me of those film clapperboard out-takes, which were used to sync video-audio.

When I try to search on this subject, I get a bunch of generic info about making sure drivers are up to date, you have a strong signal, etc. But I'm not talking about out-of-the-ordinary delay, I'm talking about the inherent, unavoidable delays in the process.

In practice, the common protocols seem to have a latency of ~ 150 msec, which is very noticeable under the conditions I mentioned.

I'm somewhat familiar with GSM transmission, and I assume it's fairly similar. I think the inherent delays there come from:

A) Sample the audio for 1/50th of a second (20 ms).

B) Only after that 20 msec is captured, can the the data be compressed and put into a "packet". I'm not sure how long this takes, but I'm pretty sure it can't begin until the entire 20 msec sample is collected, 20 msec plus processing time.

C) The packet is transmitted. Time will depend on available time slice (and maybe bandwidth?). Though BT may be different, I'm pretty sure GSM is set up such that the entire packet fits into one time slice, so bandwidth is not a limiting factor, by design. Or maybe I should say, the bandwidth delay is fixed by design?

D) The packet is received. Again, I'm pretty sure the entire packet must be received before the de-compression can begin. And only after it is fully decompressed, can the audio be output. And, I'd imagine some amount of buffering is needed to assure the packets can be seamlessly connected (and maybe to request a resend of data for error detection/correction?).

So if that is reasonable analogous to how BT works, does anyone know how long of a sample BT takes for audio? And would that be the biggest factor? I believe that BT uses the SBC codec. https://en.wikipedia.org/wiki/SBC_(codec)

TIA

 

[anorlunda]

I see the problem too, but more severely. I don't think it has anything to do with bluetooth. I do believe that the video and audio streams are sent separately over the Internet.

I have low bandwidth. Sometimes, I lose lip-synch up to 30 seconds. Sometimes, the closed caption texts lag behind the video by 5-10 minutes. I can fix it temporarily by using fast-forward or rewind by only a few seconds. That flushes the caches and forces a restart with perfect synch.

The worst is recent. On Netflix, everything that I watch seems to stop the audio 30 seconds before the video ends. I miss the last 30 seconds of audio. If I rewind to 60 seconds before the end and replay, audio and video work OK. That leads me to conclude that video/audio/text streams are independent and can lose synch.

 

[Wrichik Basu]

One interesting reason for delay in Bluetooth headphones is using 2.4 GHz WiFi. Bluetooth also uses the 2.4 GHz band, which probably causes interference. I have seen that the Bluetooth delays are negligible if I am using a 5 GHz WiFi.

 

[NTL2009]

I see the problem too, but more severely. I don't think it has anything to do with bluetooth. I do believe that the video and audio streams are sent separately over the Internet. ...

That can certainly be an issue, but I'm really trying to look into the inherent latency of BT, outside of any external slow-downs.

Hmmm, I tried replicating this on my computer. Look at this video at 10:30, he strikes a mark on two chisels (the 1st seems to have an edit, but the 2nd is clear).



On my computer, the strike and the sound are in pretty good sync over BT. On my tablet, a very noted delay over BT (same headphones - Boltune BT-H001)). And when I listen/watch over the tablet speakers, it is in sync, so it's not a lag on youtube. So I guess my tablet is adding a BT delay. If I get ambitious, I may try to measure it later. I could make an audio recording in Audacity, watch and clap when I see it, and then measure the time between my clap and the sound. I'll try to find a good repeating video later, something with ~ 1 second between hits, so I can sync up and clap in time.

 

[jedishrfu]

More likely the delay is due to processing the audio stream. We had done a project at work with students where we had a stereo mike setup as a headset someone could wear. We used a raspberry pi to beamform the signal simply by adding the left and right channels together and adjusting the volume and then sending it to the headset speakers.

You could notice the delay get input signals, add and send and it would give an echo effect from the original signal.

We were trying to make a hearing assist for a hearing-impaired person giving a lecture. By wearing the directional headset they could look in the direction of an audience member questioner and hear the question better.

Shortly after we completed the project BOSE introduced the Hearphones which did the exact same thing only with less of a delay likely due to using a more analog approach over our raspberry pi digital approach.

https://www.bose.com/en_us/support/products/bose_wellness_support/hearphones.html

BOSE developed the product in anticipation of the OTC Hearing Aide law that was in the works and is now law. Yay, cheaper hearing aides maybe.

https://www.nidcd.nih.gov/health/over-counter-hearing-aids

 

[f95toli]

It also depends quite a bit on which Bluetooth version you are using and which CODECs are supported.
Some quick searching on the Bluetooth Alliance website led to this Blog

https://www.telink-semi.com/low-latency-headsets-tws-earbuds/

This bit might be relevant

Bluetooth has been a cornerstone of the wireless audio market for years, and Bluetooth LE Audio in particular is built to offer ultra-low latency for audio functions. In fact, the theoretical latency with LE Audio can be as low as just 20 ms. The actual latency is expected to be between 30-40 ms when we take into consideration re-transmission and other processing delays. For reference, a range of 20 ms to 40 ms is considered optimal when it comes to online games.

For music production people typically want a latency of about 1.5 ms at most (about 128 samples); anything more becomes very noticeable if you are e.g. playing an instrument and listening back using headphones (this is something anyone who has ever tried using even a simple DAW will be familiar with).
This is also why wireless systems for music are typically based on proprietary formats, not Bluetooth.