Forums

GSM AMR on blackfin.

Started by Akshay Mishra August 1, 2006
I am working on a GSM AMR codec. I have optimized the codec for real
time operation on a Blackfin DSP on the 533 evaluation board. I capture
the input and play it out. I have put the system in two configurations:

(a) No codec, plain loopback of the captured sound on the output
speaker. I get good sound.
(b) When GSM AMR is involved in the operation, the output has the input
but it is subdued heavily in noise. The noise is akin to radio static
but very pronounced.

The codec has been tested/verified with the standard test vectors and
found correct.

What can be reason for the same ?


Best Regards,
Akshay

What about the levels? Do you use the full input range of the AD converter?

Best regards,

Andre


Akshay Mishra wrote:
> I am working on a GSM AMR codec. I have optimized the codec for real > time operation on a Blackfin DSP on the 533 evaluation board. I capture > the input and play it out. I have put the system in two configurations: > > (a) No codec, plain loopback of the captured sound on the output > speaker. I get good sound. > (b) When GSM AMR is involved in the operation, the output has the input > but it is subdued heavily in noise. The noise is akin to radio static > but very pronounced. > > The codec has been tested/verified with the standard test vectors and > found correct. > > What can be reason for the same ? > > > Best Regards, > Akshay >
-- For emailing me, please exchange abuse by a . l o d w i g
Yes. I used the full input range.

Thanks,
Akshay
Andre wrote:
> What about the levels? Do you use the full input range of the AD converter? > > Best regards, > > Andre > > > Akshay Mishra wrote: > > I am working on a GSM AMR codec. I have optimized the codec for real > > time operation on a Blackfin DSP on the 533 evaluation board. I capture > > the input and play it out. I have put the system in two configurations: > > > > (a) No codec, plain loopback of the captured sound on the output > > speaker. I get good sound. > > (b) When GSM AMR is involved in the operation, the output has the input > > but it is subdued heavily in noise. The noise is akin to radio static > > but very pronounced. > > > > The codec has been tested/verified with the standard test vectors and > > found correct. > > > > What can be reason for the same ? > > > > > > Best Regards, > > Akshay > > > > > -- > > For emailing me, please exchange abuse by > a . l o d w i g
Akshay Mishra wrote:
> Yes. I used the full input range. > > Thanks, > Akshay > Andre wrote: >> What about the levels? Do you use the full input range of the AD converter? >> >> Best regards, >> >> Andre >> >> >> Akshay Mishra wrote: >>> I am working on a GSM AMR codec. I have optimized the codec for real >>> time operation on a Blackfin DSP on the 533 evaluation board. I capture >>> the input and play it out. I have put the system in two configurations: >>> >>> (a) No codec, plain loopback of the captured sound on the output >>> speaker. I get good sound. >>> (b) When GSM AMR is involved in the operation, the output has the input >>> but it is subdued heavily in noise. The noise is akin to radio static >>> but very pronounced. >>> >>> The codec has been tested/verified with the standard test vectors and >>> found correct. >>> >>> What can be reason for the same ? >>> >>> >>> Best Regards, >>> Akshay
Do you overload it? Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Akshay Mishra wrote:
> Yes. I used the full input range. > > Thanks, > Akshay > Andre wrote: > > What about the levels? Do you use the full input range of the AD converter? > > > > Best regards, > > > > Andre > > > > > > Akshay Mishra wrote: > > > I am working on a GSM AMR codec. I have optimized the codec for real > > > time operation on a Blackfin DSP on the 533 evaluation board. I capture > > > the input and play it out. I have put the system in two configurations: > > > > > > (a) No codec, plain loopback of the captured sound on the output > > > speaker. I get good sound. > > > (b) When GSM AMR is involved in the operation, the output has the input > > > but it is subdued heavily in noise. The noise is akin to radio static > > > but very pronounced. > > > > > > The codec has been tested/verified with the standard test vectors and > > > found correct. > > > > > > What can be reason for the same ? > > > > > > > > > Best Regards, > > > Akshay > > > > > > > > > -- > > > > For emailing me, please exchange abuse by > > a . l o d w i g
The problem you state is identical to the result one will get if the data are misaligned. I had precisely this problem a little while ago when an object purporting to provide PCM style long frame sync actually provided short frame sync. Makes sense if you think about it. Misaligned data will blow away your sign bit, which is replaced with some random value. The original signal will now exist on an (unknown) 'DC' offset that is much larger than the signal, with the added feature that the 'DC' offset is in fact jumping positive and negative in a random fashion, giving large signal excursions. That's a definition of noise :) Cheers PeteS