« Back

Audio quality of bluetooth aptX

Almost six years have passed since the last SE article about bluetooth (BT) audio. It's time to return to the topic following some developments in the field. Even a few years ago it was clear that BT is capable of delivering high quality audio. SE listening tests showed that mandatory audio codec SBC for BT A2DP profile is not bad at all. In “High Quality” mode (@328kbit/s) it is on a par with ATRAC SP (Type-R, @292kbit/s) compression algorithm which is used in Minidisc recorders/players. In “Middle Quality” mode (@229kbit/s) it is roughly comparable with mp3@128kbit/s or aac@96kbit/s. At max possible bitrate 372kbit/s SBC codec is comparable with aac@192kbit/s and most artifacts it produces are beyond human perception. Close positioning of the two codecs – SBC (High Quality) and ATRAC SP (Type-R) – on perceptual quality scale is revealing. Recordings made on Minidisk equipment with ATRAC SP were considered to be of high fidelity by most demanding listeners. It is an example of combination of barely acceptable audio codec with perfectly designed audio circuits. As opposed to ATRAC equipment, early BT A2DP devices, having on board the commensurable audio codec SBC, performed poorly just because their audio circuitry was of simplistic design. In other words BT audio potential was not fully utilized.

At the same time the demand for high quality BT audio was high, because this technology has some clear benefits for consumers:

  • absence of wires, obviously
  • audio quality of such devices are fully determined by receiving side. For example BT headphones can sound exactly the same with different sources of signal – phones/players, tablets, laptops etc., which are just sources of standard digital stream in this scenario.
So, it was clear that sooner or later somebody will hammer all those BT audio drawbacks and will offer a more matured solution for demanding listeners. Finally UK based semiconductor company CSR, which already had experience in developing and manufacturing of BT chips, did this. In particular the company:
  1. developed new BT chips with carefully designed audio circuits
  2. added its propriatery audio codec aptX (acquired with APT Licensing Ltd in 2010) in addition to mandatory SBC codec
  3. implemented more strict control over manufacturers who use CSR BT solutions in their products
As it was shown above the problem of poor BT audio could be solved by taking measures (1) and (3) alone. Improvement in codec efficiency is just an accessory plus. The more so as such improvement is hard to achieve due to tough conditions for any BT audio codec:
  • low complexity of algorithm (to constrain energy consumption)
  • low encoding-decoding delay (for better video and gaming experience)
  • limited bandwidth of A2DP profile (max. 721kbit/s, less in real-life scenarios)
But no matter what was behind the CSR decision to use some other codec in addition to mandatory and pretty acceptable SBC, the choice of aptX looks logical from audio quality perspective. This family of codecs are successfully used in professional audio since 1990s thanks to their low complexity and absence of psychoacoustic masking. This results in low delay and stable/predictive perceptual quality with various audio material. CSR uses in their BT chips aptX codec with the following parameters (specs by CSR):
  • Compression  ratio:  4:1
  • Audio Format: 16-bit, 44.1kHz
  • Data Rates: 352kbps
  • Frequency Response: 10Hz to 22kHz
  • Algorithmic Delay: <1.89ms @ Fs 48KHz
  • Dynamic Range: 16-bit: >92dB
  • THD+N: -68.8dB

We plan to add this codec to SE testing system.

Encoding and decoding of AptX streams for this research turned out to be not easy task, as the software codec is not publicly available. We contacted CSR plс several times with no success. Multiple combinations of USB dongles/BT software stacks/notebooks/smartphones were tried to produce digital AptX records without any result. As SoundExpert is a low-to-non-budget research project and can't afford buying devices for testing, a few guys, interested in the research, donated around $30 to buy hardware BT receiver with digital outputs. The delivered device had so badly designed digital interfaces that bit-perfect recording was not possible (for details see "Audio quality of Bluetooth music receiver TS-BTAD01").

If somebody knows a reliable procedure for bit-perfect recording of AptX stream using ordinary consumer equipment (Mac/PC/phone), please, contact us. Also you can download and encode/decode SE test vectors on your side and send them back to us. Encoded/decoded Sample#1 would be enough for adding AptX codec to SE listening tests, more lengthy Sample#2 will allow some objective measurements and comparison with SBC/ADPCM codecs. Thanks in advance.

Hope somebody will find this "message in a bottle" )) The link to this request is also sent to Qualcomm, current owner of AptX BT codecs.

 

To be continued ...

 

Comments
Trackback URL :

Serge Smirnoff
The first $5 are mine.
Please feel free to report your donations in comments.
Posted on 12/8/15 7:57 PM.
Serge Smirnoff
Thank you, Pierre for donation.
Posted on 12/9/15 10:11 AM.
Serge Smirnoff
Thanks to Alexey who donated $20 and Chinese manufacturers who offer now cheaper analogues of required device the long story of testing aptX has moved forward. This aptX-enabled BT receiver with s/pdif output - http://www.dx.com/p/lx-ts-btad01-bluetooth-v4-0-fiber-coaxial-audio-hi-fi-lossle­ss-output-speaker-receiver-black-246256 - will be delivered within two weeks.
Posted on 10/24/16 11:15 PM.