text stringlengths 1 112 |
|---|
its first transmission th e transport delay can be less than a millisecond but if one or more |
retransmissions have been scheduled it will take a few more milliseconds before the last |
possible transmission would be received at the synchronisation reference point the |
earliest that can occur with bluetooth le audio is around ms from the point where the |
first sample for that audio frame was taken and is the fixed point in time at which every |
acceptor can start to decode their received packets the synchronisation reference point is |
where the presentation delay starts within this period the acceptor needs to decode the |
lc packet which takes a few milliseconds fo r the lc decoder and apply the packet loss |
concealment if that is required which takes another few milliseconds although its not |
needed for most packets the time to run the algorithm has to be allocated for the occasions |
where it is required if a ny other audio processing needs to be done such as algorithms for |
noise cancellation or speech enhancement they need to be completed before the end of the |
section quality of service qos |
presentation delay which is where each acceptor renders the reconstituted audio stream |
because t he basic audio profile requires that e very acceptor must support a value of ms |
for presentation delay they need to support around ms of buffering to hold the decoded |
audio data before the rendering point in practice the value of presentation delay may be |
lower or greater manufacturers of receiving devices may support a range from as low as |
ms up to several hundred milliseconds |
if everything is optimised t he quickest this whole process can happen for a ms lc |
packet is just over ms using a ms frame makes little difference as the shorter frame |
needs a longer look ahead delay so the saving is only around ms |
a ms delay is equivalent to the time it takes sound to travel m our hearing has evolved |
to cope with th is level of delay if we hear an original sound and an echo ms later |
the brain processes it without any difficulty that means that we can use bluetooth le |
audio streams for earbuds and hearing aids which pick up the ambient sound as well as a |
bluetooth stream with out the wearer being distracted by any echo effects however the |
example above involved a lot of optimisations it assumes that transmission can start as |
soon as the encoding is complete and that all retransmissions happen within a quarter of a |
frame which is only valid for small packets and the lower sampling frequencies in most |
real applications other factors come into play which brings us to the quality of service or |
qos |
quality of service qos |
quality of service is a term applied to the received audio signal and encompasses latency the |
perceived sound of the decoded audio and the incidence of any audio artefacts such as pops |
crackles and gaps all of these features have trade offs with each other the latency |
example described above is a highly idealised one where packets arrive when theyre |
expected in practice they dont the human body is very efficient at absorbing bluetooth |
signals so if someone is wearing earbuds but has their phone in the back pocket of their |
jeans the signal between the phone and the earbuds may be attenuated by up to db if |
youre in a room that may not matter as the earbud will probably pick up reflected signals |
from the walls or ceiling but if youre outsid e where you dont have those reflecting |
surfaces far more packets will be lost |
in the previous chapter we saw that t he design of isochronous channel s adds robustness by |
using retransmissions pretransmissions burst numbers flush timeouts and frequenc y |
hopping if we apply enough of these features we have a very high confidence that almost |
every packet will get through and the small number that dont arrive intact can be filled in |
using plc however as we apply these techniques latency starts to increase as does power |
consumption spreading retransmissions across more than a single isochronous interval |
adds an extra frame time to the latency for each additional isochronous interval putting |
more retransmissions within a single frame limits the n umber of different streams which can |
be accommodated and pushes up the power both for the transmitter and also for the |
receiver which needs to stay active to look for consecutive transmission slots |
chapter lc latency and qos |
these robustness features are separate from the codec settings but the codec settings also |
have an effect on robustness h igher quality encoding with khz sampling will produce |
larger packets which will be more susceptible to interference similarly if you encode |
multi ple audio streams into a single packet ie the channel allocation is greater than that |
sdu will contain multiple codec frames resulting in larger packets with the same problem |
given the large number of possible codec and robustness configurations that are allowed |
bap has def ined sets of standard combinations aimed at the two major use cases low |
latency and high reliability which can be found in tables and of bap they cover |
both ms and ms frame intervals |
low latency is interpreted as settings which will allow all of the retransmissions to fit within |
a single isochronous interval for sampling rates of khz to khz the larger packets for |
khz extend into two isochronous intervals going up to four isochr onous intervals for |
khz high reliability qos configurations prioritise retransmission over latency allowing |
retransmissions to be spread across six or more isochronous intervals for broadcast and ten |
or more for unicast |
table shows the maximum number of isochronous interval s allowed for each sampling |
frequency derived from these tables the reason that the low latency khz sampled |
setting needs two isochronous interval s is to allow a sufficient number of retransmissions |
with the larger packets as only a limited number fit into a single isochronous interval for |
low sampling frequencies the smaller packets mean that more retransmissions can be fitted |
into each isochronous interval how many retransmissions are allocated is ultimately down |
to the scheduler in the controller |
table the m aximum number of isochronous intervals allowed for bap qos settings |
in the latency example of figure we saw that using lc with a ms frame and a |
significant degree of optimisation gave an overall latency of just over ms that u sed a low latency |
high reliability |
unicast broadcast |
sampling frequency ms ms ms ms ms ms |
khz |
khz |
khz |
khz |
khz |
khz kbps |
khz kbps |
the khz figures differ because they are defined for framed pdus all other sampling |
rates are unframed |
section quality of service qos |
presentation delay of just over ms i f low latency is important to an application |
implement ers need to be careful about their choice of qos and codec configuration as it can |
result in the latency increasing significantly table shows the actual overall latency values |
which are likely to achieved these have been calculated using a value of ms for the lc |
to sampling and encod e a ms fr ame |
low latency high reliability |
unicast and broadcast unicast broadcast |
sampling |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.