mcasp: how to attenuate audio without signal distortion

Petr Duga

Intellectual 495 points

Other Parts Discussed in Thread: TMS320C6748

Hi,

I work with tms320C6748 lcdk and do audio processing.

I tried to attenuate audio signal on 50% of original volume.

The simple way like:

sample=audioIn;

sample*=0.5

audioOut=sample;

causes the audio distortion.

I thought, that the problem could be in byte sequence, so I modified it in this way:

sample=_swap4(audioIn);

sample*=0.5

audioOut=_swap4(sample);

but it doesn't help.

Do you have any idea how to attenuate audio signal correctly?

Thanks for your answers.

Petr Duga

over 11 years ago

0 Sivaraj Kuppuraj over 11 years ago

TI__Mastermind 35645 points

Hi Petr,

Thanks for your post.

I believe, to control audio attenuation, it should be hardware controlled by choosing the appropriate resistor/series capacitor values at the input of the circuit to cause effect in the output impedance may be chosen little high which would result in low-pass filter effect and this would attenuate the audio at the input side, so that, appropriate attenuation would result in output without signal distortion.

please refer the note mentioned in the below wiki:

http://processors.wiki.ti.com/index.php/SGI_EVM_Hardware_Manual

Also, Please refer the below E2E thread which would give you an idea on digital DAC gain & volume control range to reduce analog gain and proportionately, the SNR too.

http://e2e.ti.com/support/data_converters/audio_converters/f/64/t/171398.aspx#626270

Probably, if you would have post this question to audio converter forum, i hope, it would be better answerable but still, i think, there are challenges to attenuate output audio without loss in signal quality.

Hope it helps!

Thanks & regards,

Sivaraj K

-------------------------------------------------------------------------------------------------------

Please click the Verify Answer button on this post if it answers your question.
--------------------------------------------------------------------------------------------------------

0 RandyP over 11 years ago in reply to Sivaraj Kuppuraj

TI__Guru* 84110 points

Petr,

Perhaps your *= 0.5 operation is converting to/from floating point.

Instead, can you try sample /= 2; to see if the operation occurs differently?

Regards,
RandyP

0 Petr Duga over 11 years ago in reply to RandyP

Intellectual 495 points

Thanks for answers,

the problem was in typecasting. When I overtyped input sample into "short" and output sample into "short" too, the attenuation was started to work. I thought I could use int x,y instead of "float" and could write y direct into txbuffer without overtyping into "short". This works only for direct write back from rxbuffer to txbuffer without processing. There is a working code.

            int i;
            float x,y;

            for(i=0; i<AUDIO_BUF_SIZE/4; i++){
               if(i % 2 == 0){
                   x = (short)*((int *)rxBufPtr[lastFullRxBuf] + i);
                   y = 0.5 * x;
                   *((int *)txBufPtr[lastSentTxBuf] + i) = (short) y;
               }else{
                   x = *((int *)rxBufPtr[lastFullRxBuf] + i);
                   y = x;
                   *((int *)txBufPtr[lastSentTxBuf] + i) = (int) y;
               }
            }

Processors

Processors forum

mcasp: how to attenuate audio without signal distortion