POLL Do You Have Any Plans to Develop with Bluetooth LE Audio?

I was reading a whathifi.com article a while back.

It's not a great article, the author bases his conclusion that LE Audio is not good, on two main things:

(1) he believes a higher bitrate is needed for better audio, but then immediately concedes that high quality audio can potentially come about with lower bitrate through lossy compression, which is stating the obvious!

(2) he believes that Fraunhofer are not to be trusted since MP3 was not originally great - again very unfair, given that MP3 changed the music world, enabling solid-state players and eventually music streaming services.

Crazy they would write such an opinion before they have even listened to LE Audio!

Plus, music aside, the relatively low latency (in reality increased somewhat due to practical implementations) may provide better voice quality for hearing boosters and wireless headsets/speakerphones and so on. The ultra-low-power is exciting too.

Anyway, although I currently have no plan to develop with LE Audio, I'm excited to see what products and development kits appear, and get to try them out for myself. It took actually listening to an MP3 to realize how amazing that tech was, back in the 1990's.

That is a very interesting article, but it seems to be one written by an audiophile of the "lossless or nothing" philosophy.

There is a truth to the "garbage-in, garbage-out" issue - it is impossible to reconstruct the original signal perfectly once you start throwing out a lot of the bits as lossless compression exploits redundancy in the data which quickly is exhausted for many signals by the time you get to a compression ratio of 2:1. Beyond that, lossy compression techniques are all about transforming the signal into more efficient representations that allow some detail to be thrown away that humans would not easily hear. The more that is thrown away, the more compression artifacts will be audible.

For someone of the "lossless or nothing" persuasion, MP3 was a devastating development, as it heralded the beginning of a lossy-compression era of music where audio is distributed with lossy compression. Peer-to-peer sharing, streaming services are just some examples - the purist generally detests these sources because they are not "transparent" and alter the sound (although, frequently, not in ways the majority would notice except in critical listening codec tests - things that I have previously participated in).

As much as MP3 was revolutionary in terms of enabling higher compression ratios than MPEG-1 Layer 2 audio with a claim of half the bitrate, the reality was that crtiical listeners did not find the audio quite as good as the claim. If you have a chance, you can wind back the clock and try one of the first-available MP3 encoders from Fraunhofer - L3enc. Even 128kbit/s sounds like swimming depending on the source audio. There was also the Xing MP3 encoder which was fast, but also not great for quality.

The main reason is the computational limitations of reasonable encoding times on the hardware of the era (e.g. 486 DX2/66MHz) and the limitations in the psychoacoustic model used. The reason MP3 lived so long was that the bitstream was well defined and remained constant, while quality improved over time as encoders improved their psychoacoustic models and free encoders (e.g. LAME) pushed the boundaries of what MP3 were capable of. By the late 2000's, most encoders (including Fraunhofer's later MP3Enc, then FastEnc) were neck-and-neck, having squeezed out almost all the compression scheme was capable of delivering.

Nowadays, MP3 is a bit of a legacy codec, with MP4-AAC/HE-AAC/xHE-AAC based codecs are more often used for distribution, delivering better quality than MP3 at a claimed "half" the bitrate. It's no surprise that Fraunhofer is still involved - they're pioneers in the field. But issues with patents and licensing are still with us, so some services have moved towards open-source codecs such as Opus.

I think the article bases itself on a premise that LC3 was made to compete with audiophile-targeted codecs (e.g. LDAC) when it was not. Of course, LC3 solves a real need - better audio quality and performance under constrained bitrate applications of a BLE channel. Newer codecs generally have better compression techniques, more accurate psychoacoustic models which deliver better quality for the same bitrate. Higher-bitrate codecs (aptX, LDAC, AAC etc) which claim to offer better quality may not scale down to low-bitrates gracefully and would not work for BLE. Issues of latency and packet-loss concealment are also important. Compatibility and licensing are issues with some of these alternative codecs, which is why baseline standardised codecs like SBC/LC3 are perhaps more important.

- Gough

There's no dispute that MP3, and other codecs, had poor quality compared to CD back in the 90's.

It was a game-changer that enabled solid-state music players and streaming services, and improvements took time. They had to start somewhere. You cannot drag users away from spotify, they prefer that rather than being given a music CD.

I was involved in listening tests for compression schemes (not MP3) about the same time, and artifacts were very noticeable on some music. But you had to listen to it to learn the applications for it. I started work on a car MP3 player back then! A perfect application for it.

The author's opinion is removed from what people actually listen to in real life. There's no major consumer market for anything beyond uncompressed 16-bit CD audio either, even though 24-bit is better, so there is still a grey line with uncompressed audio, at which point it would be acceptable for so-called audiophiles. People still listen to FM radio in their cars, and the FM quality has reduced over the years. There are plenty of applications where a lower bitrate, or lower latency or lower power, win out.

I've spent many years working with voice over IP andf conferencing systems (I developed H.323 and SIP based systems) where another important factor is what happens when there is packet loss. Also, how simple or complex is the codec? Since it directly relates to cost! in terms of hardware and software resources for conf-calls. 

I agree that the author's conclusions seem to be based on some idea that the codec is intended to provide him with audiophile grade audio : ) however the article, even though he recognises that there just might be uses for audio beyond audiophile, suggests that Bluetooth audio sounds like <expletive> - when millions of users use Bluetooth (albeit not LE) with existing compressed codecs every day with their Beats or Bose headphones or whatever - I certainly do. I don't think it is practical listening on wired speakers except on infrequent occasions, otherwise you're annoying others with your music choice, unless you live in a detached home. And the author has not listened to it, and claims to have read the spec, but we all know that's not really possible beyond a high level at best, unless you're coding for it : ) 

Well, I think we can tell the article isn't written by the most-knowledgeable with this particular snippet -

"But there has also been Bluetooth LE (Low Energy) alongside, which is sometimes used as a secondary Bluetooth connection made by headphones or audio gear for either control or for low-bit-rate voice transmission. Just occasionally we’ve had headphones that have accidentally jumped to LE for their audio transmission, which is easy to spot, because it sounds like ***."

As far as I can tell, this is technically wrong - they seem to confuse hands-free profile (HFP) and headset profile (HSP) with LE Audio which hadn't existed and wouldn't have been supported by most devices on the market even today. Of course, those profiles are most often implementing mono, 8kHz sample-rate for call audio, hence they sound terrible compared to A2DP.

Human physiology is definitely a limiting factor - I think the obsession with high-res is unfounded. Given the range of human hearing, there are very few circumstances where CD quality would not be able to convey everything we can perceive. But a lot of audiophile opinions are based on placebo ... because somehow the effort and money one puts into improving their system convinces them that they have achieved something.

That being said, in the early days of A2DP, SBC did sound horrible. I've had headsets and Bluetooth stacks cause issues - my first A2DP headset probably had such a low bitpool value that the bitrate caused notable artifacting and its internal integrated DAC was very noisy. It probably wasn't helped that I used Bluesoleil. After upgrading to a USB dongle that was supported by Widcomm, the audio quality improved a little but then under adverse signal conditions, the sample rate would wander resulting in audio playback speed variations rather than just gracefully leaving a "gap". It wasn't until I graduated to CSR Harmony and several generations of headsets later that A2DP in its full bitpool 53 configuration that it became more difficult to distinguish SBC from the source. There is still a little harshness in the treble, but it is not objectionable not like my first experience.

Technology progresses over time ... and perhaps convenience is a greater draw to many consumers than the minuscule difference in quality, especially when it doesn't affect enjoyment. But as one ages and as their hearing declines, I suspect the differences will be even less likely to be perceived.

- Gough

Speaking of which, I entirely forgot that I published this a few years back -

https://goughlui.com/2014/02/04/tech-flashback-mp3-encoding-through-the-years/

A look at how MP3 encoders evolved over the years. The L3enc and Blade results are quite audibly interesting.

I suspect LE Audio is much more refined, out of the gate, given the high standards set for it.

- Gough

I think I went through a similar MP3 encoder experimentation phase as you, having used many of the paid and free encoders! : )

I didn't purchase much music beyond CD format back then, and using MP3 conversions in the car etc.

I think I went through a similar MP3 encoder experimentation phase as you, having used many of the paid and free encoders! : )

I didn't purchase much music beyond CD format back then, and using MP3 conversions in the car etc.

The DAC matters too. The one you sent me for the SDR project, turned out to be an excellent (for me) device.
Noticeably better than what's coming out of my PCs and laptops (and my hearing, at 55, is not golden )

This device has reviews online that aren't favourable. For me, it's the best sound card in the house.

Agreed - the DAC, headphones, background noise, your own hearing, even how familiar you are with the music  ... they all have an influence.

I'm personally quite partial to using the Audio-Technica ATH-M50x at home, paired with an Asus Xonar Essence STX (from ages ago which still works well with its integrated headphone amp). Before that, I was using a Creative X-Fi XtremeMusic PCI ... quite decent for the price and not a bad card at all.

But that being said, over the years, the integrated solutions have been playing catch-up, and the latest ones aren't like the AC'97 or early HD Audio implementations where the background noise was easily discernable along with intermodulation distortions. Later products, whether by advancement of the chip design itself or motherboard design to route analog signals more carefully and decouple power - I suspect both - have achieved sound quality which is adequate for most.

- Gough

Audio-Technica stuff is awesome. I have a lower-cost Beyerdynamic DT 770 PRO and the soft cushion comfort was the driving factor more than the sound beyond a certain point. I have a pair of on-ear Bose headphones (foldable, mainly for travel) but they get uncomfortable and I cannot wear them longer than a few hours : (