[Bluez-devel] Motorola HT820 audio fidelity

[Bluez-devel] Motorola HT820 audio fidelity

[Fredrik Tolf]

Hi all!

I just got a pair of Motorola HT820 bluetooth headphones, which support
A2DP and work with a2play. However, the audio fidelity could definitely
be a lot better. I'm no audiophile, but it sounds to me as though the
higher frequency bands ("higher" as in, I'd guestimate, ~10 kHz and
above) are seriously distorted (I can't put a better word on it than
that they are completely transformed into "chirping"). All the lower
bands are completely unaffected as far as I can hear, though.

A friend told me that they work perfectly fine with his BT-enabled cell
phone. I haven't tried myself, though, since my phone doesn't support
A2DP.

I don't know A2DP very well -- could this be the result of bad
transmission? I'd guess not, since when I move away, the audio tends to
break up rather than lose quality.

I found this little tidbit on Wikipedia:

 Most bluetooth stacks implement the SCMS-T copyright protection. In
 these cases it is not possible to connect the A2DP headphones for high
 quality audio. E.g. the Motorola HT820 can be used for high quality
 audio only with certain versions of the Toshiba bluetooth stack.
 (From <http://en.wikipedia.org/wiki/Bluetooth>)

I don't know if that is related to this, but I thought I should at least
mention it.

Does anyone know what could be at fault?

Fredrik Tolf




-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel

Re: [Bluez-devel] Motorola HT820 audio fidelity

[Brad Midgley]

Fredrik

> I just got a pair of Motorola HT820 bluetooth headphones, which support
> A2DP and work with a2play. However, the audio fidelity could definitely
> be a lot better. I'm no audiophile, but it sounds to me as though the
> higher frequency bands ("higher" as in, I'd guestimate, ~10 kHz and
> above) are seriously distorted (I can't put a better word on it than
> that they are completely transformed into "chirping"). All the lower
> bands are completely unaffected as far as I can hear, though.

I noticed some quality loss at high frequencies, such as high notes on a
violin, possibly when using fixed point to build libsbc. It is subtle.
On the other hand, if you hear an unexpected "pop" using fixed point,
it's definitely an overflow error in the encoder. There's still a
problem in there I didn't hunt down.

Steven Singer suggested we to track the error "K" after an
encode-decode. K is derived from the root-mean-square error. We could
use this to rule out a misbehaving codec.

It doesn't help that we're typically decoding mp3 and reencoding in sbc.
Loss is really compounded. Try to test with something in a lossless
format or with a high-quality first-generation mp3 or ogg.

Brad


-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel

Re: [Bluez-devel] Motorola HT820 audio fidelity

[Fredrik Tolf]

On Thu, 2006-04-27 at 21:19 -0600, Brad Midgley wrote:
> It doesn't help that we're typically decoding mp3 and reencoding in sbc.
> Loss is really compounded.

Indeed, that seems to be the problem. I was always under the
(uninformed) impression that SBC was lossless, but I just tried
re-encoding an MP3 to SBC and playing that SBC with sbcenc and sbcdec,
and the exact same audio artifacts are showing themselves when I play
back to my normal speaker system as through the headphones.

So, at least it's not a hardware or BT stack problem. Is this an
inherent problem in SBC, or is there a way to tune the encoder to
compensate? Is it possibly so that a better encoder could be written
altogether?

Fredrik Tolf




-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel

Re: [Bluez-devel] Motorola HT820 audio fidelity

[Brad Midgley]

Fredrik

>>It doesn't help that we're typically decoding mp3 and reencoding in sbc.
>>Loss is really compounded.
> 
> 
> Indeed, that seems to be the problem. I was always under the
> (uninformed) impression that SBC was lossless, but I just tried
> re-encoding an MP3 to SBC and playing that SBC with sbcenc and sbcdec,
> and the exact same audio artifacts are showing themselves when I play
> back to my normal speaker system as through the headphones.

You haven't ruled out our implementation of sbc however. We could have
problems still. Also, there are also some defaults for encoding that
although set fairly high are not at the max iirc.

There is/was a reference SBC encoder that is a windows executable
somewhere at bluetooth.org. Testing with it would let us rule out (or
implicate) our implementation.

> So, at least it's not a hardware or BT stack problem. Is this an
> inherent problem in SBC, or is there a way to tune the encoder to
> compensate? Is it possibly so that a better encoder could be written
> altogether?

The spec allows us to negotiate a different codec with the headset. SBC
is the only required codec. fwiw, I've never seen a headset that accepts
mp3 or others. I should modify a2play to flag these so they would be
easier to spot.

Brad


-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel

Re: [Bluez-devel] Motorola HT820 audio fidelity

[Mayank Batra]

[-- Attachment #1: Type: text/plain, Size: 2201 bytes --]

Brad,

Have a look at the following headset (Sonorix):

http://www.assisttechnology.co.uk/documents/Sonorix_Bluetooth_Headset-Pamphlet.pdf

It supports both SBC and MP3 codecs.
And it is BT1.2 compliant.

Mayank


On 5/1/06, Brad Midgley <bmidgley@xmission.com> wrote:
>
> Fredrik
>
> >>It doesn't help that we're typically decoding mp3 and reencoding in sbc.
> >>Loss is really compounded.
> >
> >
> > Indeed, that seems to be the problem. I was always under the
> > (uninformed) impression that SBC was lossless, but I just tried
> > re-encoding an MP3 to SBC and playing that SBC with sbcenc and sbcdec,
> > and the exact same audio artifacts are showing themselves when I play
> > back to my normal speaker system as through the headphones.
>
> You haven't ruled out our implementation of sbc however. We could have
> problems still. Also, there are also some defaults for encoding that
> although set fairly high are not at the max iirc.
>
> There is/was a reference SBC encoder that is a windows executable
> somewhere at bluetooth.org. Testing with it would let us rule out (or
> implicate) our implementation.
>
> > So, at least it's not a hardware or BT stack problem. Is this an
> > inherent problem in SBC, or is there a way to tune the encoder to
> > compensate? Is it possibly so that a better encoder could be written
> > altogether?
>
> The spec allows us to negotiate a different codec with the headset. SBC
> is the only required codec. fwiw, I've never seen a headset that accepts
> mp3 or others. I should modify a2play to flag these so they would be
> easier to spot.
>
> Brad
>
>
> -------------------------------------------------------
> Using Tomcat but need to do more? Need to support web services, security?
> Get stuff done quickly with pre-integrated technology to make your job
> easier
> Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
> http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
> _______________________________________________
> Bluez-devel mailing list
> Bluez-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/bluez-devel
>

[-- Attachment #2: Type: text/html, Size: 3034 bytes --]

Re: [Bluez-devel] Motorola HT820 audio fidelity

[Brad Midgley]

Mayank

> http://www.assisttechnology.co.uk/documents/Sonorix_Bluetooth_Headset-Pamphlet.pdf

Nice find.

It brings up an interesting challenge. When using this headset, the
driver should have access to the encoded audio so it can choose whether
it's appropriate to decode the stream and then encode it in sbc or if it
can pass the stream as-is to the set. I'll bet a gst bin could be
constructed that would configure its elements accordingly for the input
type and the headset caps.

iirc, a2play assumes the first codec reported from the set is sbc. we
should test it with the sonorix if somebody here has it to make sure we
handle multiple codec capabilities properly.

A bit of good news. It takes a couple of tries to get gxine to talk to
my a2dp headset via the newly working plugin, but once it works the
audio is very close to being in sync with the video and it is pretty
solid for the duration of playing some video.

Brad


-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel

Re: [Bluez-devel] Motorola HT820 audio fidelity

[Mayank Batra]

[-- Attachment #1: Type: text/plain, Size: 2252 bytes --]

Brad,

The simplest approach would be to analyse all SEPs from the sink and if we
find an MP3 SEP, we can very well send the mp3 we receive from mpg123,
otherwise use the SBC codec.

Actually somewhere in the command line arguments we can take an input from
the user which codec does he prefer.

So if he writes something like:
a2play -c mp3
(Where -c is for the codec)
This should first try searching mp3 on the sink.
If not found, it should try SBC (which is of course mandatory and should be
found)

Whereas if he writes something like:
a2play -c sbc

This will straight away connect on SBC SEP.


Mayank


On 5/2/06, Brad Midgley <bmidgley@xmission.com> wrote:
>
> Mayank
>
> >
> http://www.assisttechnology.co.uk/documents/Sonorix_Bluetooth_Headset-Pamphlet.pdf
>
> Nice find.
>
> It brings up an interesting challenge. When using this headset, the
> driver should have access to the encoded audio so it can choose whether
> it's appropriate to decode the stream and then encode it in sbc or if it
> can pass the stream as-is to the set. I'll bet a gst bin could be
> constructed that would configure its elements accordingly for the input
> type and the headset caps.
>
> iirc, a2play assumes the first codec reported from the set is sbc. we
> should test it with the sonorix if somebody here has it to make sure we
> handle multiple codec capabilities properly.
>
> A bit of good news. It takes a couple of tries to get gxine to talk to
> my a2dp headset via the newly working plugin, but once it works the
> audio is very close to being in sync with the video and it is pretty
> solid for the duration of playing some video.
>
> Brad
>
>
> -------------------------------------------------------
> Using Tomcat but need to do more? Need to support web services, security?
> Get stuff done quickly with pre-integrated technology to make your job
> easier
> Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
> http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
> _______________________________________________
> Bluez-devel mailing list
> Bluez-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/bluez-devel
>

[-- Attachment #2: Type: text/html, Size: 3116 bytes --]

Re: [Bluez-devel] Motorola HT820 audio fidelity

[Shawn Rutledge]

On 5/1/06, Brad Midgley <bmidgley@xmission.com> wrote:
> It brings up an interesting challenge. When using this headset, the
> driver should have access to the encoded audio so it can choose whether
> it's appropriate to decode the stream and then encode it in sbc or if it
> can pass the stream as-is to the set. I'll bet a gst bin could be
> constructed that would configure its elements accordingly for the input
> type and the headset caps.

What about the plan to support A2DP via ALSA?

Maybe there needs to be a new kind of sound device (/dev/sound/play or
something) which for typical sound hardware, does the decoding, and
for A2DP can either decode or send the packets across, depending on
whether it can negotiate a compatible decoder on the device.  Having
it work only with one library/toolkit like gstreamer is kindof
limiting.  And some day, I bet there will be A2DP support for even
more codecs.  FLAC would be awesome (but you'd need EDR I suppose).=20
So say if the device supports SBC, MP3 and FLAC, you could send MP3
directly, and encode everything else as FLAC.  A player program would
only need to open the /dev file and write MP3/Ogg/whatever packets to
the device, after stripping off any metadata.

The ALSA A2DP driver was supposed to be in userspace anyway, right?


-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel

Re: [Bluez-devel] Motorola HT820 audio fidelity

[Brad Midgley]

Shawn

> Maybe there needs to be a new kind of sound device (/dev/sound/play or
> something) which for typical sound hardware, does the decoding, and
> for A2DP can either decode or send the packets across, depending on
> whether it can negotiate a compatible decoder on the device.  Having
> it work only with one library/toolkit like gstreamer is kindof
> limiting. 

alsa doesn't transparently provide an api like this. it's either
something on top of alsa or approximated by writing a plugin for each
type of codec. alsa plugins are only supposed to know about some very
simple encoding audio schemes though.

ironic that you are making the case for gstreamer at the same time you
say it's limiting. :)

> The ALSA A2DP driver was supposed to be in userspace anyway, right?

We have the alsa plugin mostly working now. We haven't abandoned that
idea. alsa plugins are relatively simple in what they expose to the client.

i plan to keep improving the alsa plugin and get into gstreamer to get
some of the more advanced stuff. i have a little picture illustrating
how we tear btsco into two projects that go into these two camps.

http://bluetooth-alsa.sourceforge.net/wavez.png

Brad


-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel

Re: [Bluez-devel] Motorola HT820 audio fidelity

[Fredrik Tolf]

On Mon, 2006-05-01 at 08:48 -0600, Brad Midgley wrote:
> Fredrik
> 
> >>It doesn't help that we're typically decoding mp3 and reencoding in sbc.
> >>Loss is really compounded.
> > 
> > 
> > Indeed, that seems to be the problem. I was always under the
> > (uninformed) impression that SBC was lossless, but I just tried
> > re-encoding an MP3 to SBC and playing that SBC with sbcenc and sbcdec,
> > and the exact same audio artifacts are showing themselves when I play
> > back to my normal speaker system as through the headphones.
> 
> You haven't ruled out our implementation of sbc however. We could have
> problems still. Also, there are also some defaults for encoding that
> although set fairly high are not at the max iirc.
> 
> There is/was a reference SBC encoder that is a windows executable
> somewhere at bluetooth.org. Testing with it would let us rule out (or
> implicate) our implementation.

I have Windows in a VMware exactly for things like these, so I could
test that. However, I could find no such thing on bluetooth.org. In
fact, googling for "sbc site:bluetooth.org" only yielded two
insignificant results, so I'm beginning to wonder if I'm doing something
wrong (whatever that'd be...). Any pointers?

> The spec allows us to negotiate a different codec with the headset. SBC
> is the only required codec. fwiw, I've never seen a headset that accepts
> mp3 or others. I should modify a2play to flag these so they would be
> easier to spot.

Unfortunately, that seems to be no relief in this case. a2play gets only
one acp_seid_info from the headset, so I guess that means that it only
supports SBC, correct?

Fredrik Tolf




-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel

Re: [Bluez-devel] Motorola HT820 audio fidelity

[Brad Midgley]

Fredrik

> I have Windows in a VMware exactly for things like these, so I could
> test that. However, I could find no such thing on bluetooth.org. In
> fact, googling for "sbc site:bluetooth.org" only yielded two
> insignificant results, so I'm beginning to wonder if I'm doing something
> wrong (whatever that'd be...). Any pointers?

You can get it but you probably need to be logged into a (free) account:

https://www.bluetooth.org/foundry/adopters/document/A2DP_Bitstream/en/1/A2DP_Bitstream.zip

we can also compare with our floating point implementation which we have
more faith in.

> Unfortunately, that seems to be no relief in this case. a2play gets only
> one acp_seid_info from the headset, so I guess that means that it only
> supports SBC, correct?

It wouldn't be hard to add mp3 support. We might have to know just a bit
of mp3 smarts like how to break up a stream into packets appropriately.
I don't think we'd ever want to encode mp3 on the fly so it's not as
complex as sbc support.

Brad


-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel

Re: [Bluez-devel] Motorola HT820 audio fidelity

[Fredrik Tolf]

On Tue, 2006-05-02 at 10:33 -0600, Brad Midgley wrote:
> Fredrik
> 
> > I have Windows in a VMware exactly for things like these, so I could
> > test that. However, I could find no such thing on bluetooth.org. In
> > fact, googling for "sbc site:bluetooth.org" only yielded two
> > insignificant results, so I'm beginning to wonder if I'm doing something
> > wrong (whatever that'd be...). Any pointers?
> 
> You can get it but you probably need to be logged into a (free) account:
> 
> https://www.bluetooth.org/foundry/adopters/document/A2DP_Bitstream/en/1/A2DP_Bitstream.zip

I could get it without an account. I find it rather weird that google
didn't find it, but oh well.

Anyway, I tried it, and it was an *enormous* difference. It is true that
the file encoded by the encoder from bluetooth.org is almost twice as
large as that encoded by sbcenc as well, but nonetheless, the difference
in audio quality was huge.

> we can also compare with our floating point implementation which we have
> more faith in.

Is this in the CVS tree? I can't seem to find any other encoder than
sbcenc. (Speaking of which, what's the difference between the "sbc" and
"sbc32" directories?)

Fredrik Tolf




-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel

Re: [Bluez-devel] Motorola HT820 audio fidelity

[Brad Midgley]

Fredrik

>>> I have Windows in a VMware exactly for things like these, so I could
>>> test that. However, I could find no such thing on bluetooth.org. In
>>> fact, googling for "sbc site:bluetooth.org" only yielded two
>>> insignificant results, so I'm beginning to wonder if I'm doing something
>>> wrong (whatever that'd be...). Any pointers?
>> You can get it but you probably need to be logged into a (free) account:
>>
>> https://www.bluetooth.org/foundry/adopters/document/A2DP_Bitstream/en/1/A2DP_Bitstream.zip
> 
> I could get it without an account. I find it rather weird that google
> didn't find it, but oh well.

google won't necessarily "deep link" into sites like bluetooth.org

> Anyway, I tried it, and it was an *enormous* difference. It is true that
> the file encoded by the encoder from bluetooth.org is almost twice as
> large as that encoded by sbcenc as well, but nonetheless, the difference
> in audio quality was huge.

twice as large? could be they're using 4 subbands.

maybe our defaults are too lossy. after calling sbc_init, you can change
sbc.subbands (default 8), sbc.blocks (default 16), sbc.bitpool (default
32). subbands can be 4,8. blocks can be 4,8,12,16. i need to check the
spec on bitpool ranges.

>> we can also compare with our floating point implementation which we have
>> more faith in.
> 
> Is this in the CVS tree? I can't seem to find any other encoder than
> sbcenc. (Speaking of which, what's the difference between the "sbc" and
> "sbc32" directories?)

a lot of this will change, but for now, "sbc" has floating, 64-bit
fixed, and 32-bit fixed and chooses one at compile time. you'll get
floating point implementation if you configure without --enable-fixed.
(heh, maybe that's what you're getting even now)

sbc32 was the precursor to separating out the library and has only the
32-bit implementation. it probably isn't even being built.

a2play "sort of" has a flag for choosing encoder settings (-t for
thrifty) but it is pretty rough.

brad


-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel

Re: [Bluez-devel] Motorola HT820 audio fidelity

[Sergey Krivov]

Hi,

> maybe our defaults are too lossy. after calling
> sbc_init, you can change
> sbc.subbands (default 8), sbc.blocks (default 16),
> sbc.bitpool (default
> 32). subbands can be 4,8. blocks can be 4,8,12,16. i
> need to check the
> spec on bitpool ranges.

I tried to change bitpool to 53 (that is reccomended
value in a2dp specs for 8 subbands and 16 blocks) in
pcm_a2dp. i did not hear any difference. but maybe my
HP headphones are not good. The traffic did increase
by some amount.

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 


-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel

Re: [Bluez-devel] Motorola HT820 audio fidelity

[Fredrik Tolf]

On Wed, 2006-05-03 at 23:31 -0600, Brad Midgley wrote:
> > Anyway, I tried it, and it was an *enormous* difference. It is true that
> > the file encoded by the encoder from bluetooth.org is almost twice as
> > large as that encoded by sbcenc as well, but nonetheless, the difference
> > in audio quality was huge.
> 
> twice as large? could be they're using 4 subbands.

I don't entirely remember, but I really think it was using 8 subbands
(when checking with sbcinfo).

> maybe our defaults are too lossy. after calling sbc_init, you can change
> sbc.subbands (default 8), sbc.blocks (default 16), sbc.bitpool (default
> 32). subbands can be 4,8. blocks can be 4,8,12,16. i need to check the
> spec on bitpool ranges.

Except the bitpool value, doesn't that mean that it's already using the
best possible values? Or do lower values of sbc.subbands or sbc.blocks
increase quality?

> (heh, maybe that's what you're getting even now)

Indeed. :)

> a2play "sort of" has a flag for choosing encoder settings (-t for
> thrifty) but it is pretty rough.

Again, wouldn't that make the audio quality worse (given how the thrifty
value only decreases the given values)?

Fredrik Tolf




-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel

Re: [Bluez-devel] Motorola HT820 audio fidelity

[Brad Midgley]

Fredrik

> Except the bitpool value, doesn't that mean that it's already using the
> best possible values? Or do lower values of sbc.subbands or sbc.blocks
> increase quality?

iirc, changing subbands to 4 will increase the traffic and it looks
computationally simpler. try it out with your problematic audio. also it
isn't clear what allocation method does in terms of quality.

this is interesting at http://www.otolith.com/otolith/olt/sbc.html

"Most SBC encoders use a structure like this. First, a time-frequency
mapping (a filter bank, or FFT, or something else) decomposes the input
signal into subbands. The psychoacoustic model looks at these subbands
as well as the original signal, and determines masking thresholds using
psychoacoustic information."

We are just trying to follow the spec for our codec, so anything
optional like analyzing the psychoacoustic model might as well be rocket
science.

brad



-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel