mpeg1 audio psychoacustic model in fixed point

Hi,

We are implementing an mpeg1 audio encoder in a fixed point dsp but we have problems in migrating the psychoacustic model to fixed point (all the log al pow instructions).

Can anybody help us?.

Regards,

Juan Start a New Thread

Hello,
You can implement floating point operations in a fixed point DSP. In a fixed point DSP, you need to be careful to avoid loosing precision in those operations.

Regards,
Pablo

Hi,
>
>We are implementing an mpeg1 audio encoder in a fixed point dsp but we have problems in migrating the psychoacustic model to fixed point (all the log al pow instructions).
>
>Can anybody help us?.
>
>Regards,
>
>Juan
> Start a New Thread

Juan,

This is pretty difficult task.

The computations are not predictable in the implementations and you
just cannot fix one format for the psycho acoustic model.

One way to address the same if to use a block fixed point format in
which you can have common exponent component for a block of data which
can be dynamic.

Regards,

Tony

________________________________

From: a... [mailto:a...] On
Behalf Of p...@hotmail.com
Sent: Sunday, September 30, 2007 4:47 AM
To: a...
Subject: [audiodsp] Re: mpeg1 audio psychoacustic model in fixed point

Hello,
You can implement floating point operations in a fixed point DSP. In a
fixed point DSP, you need to be careful to avoid loosing precision in
those operations.

Regards,
Pablo

Hi,
>
>We are implementing an mpeg1 audio encoder in a fixed point dsp but we
have problems in migrating the psychoacustic model to fixed point (all
the log al pow instructions).
>
>Can anybody help us?.
>
>Regards,
>
>Juan
> Start a New Thread