Audio processing -- fixed or floating point?

Are you breaking your 6th order filter into 3 second order sections (in 
series/cascade)?  If not, that is the first step.  If you are already doing 
that, as an alternative to throwing more bits at it, you could also try a 
different filter topology.  I find that the "4-multiply normalized 
lattice-ladder" structure is much better in terms of stability.

-- 
Jon Harris
SPAM blocker in place:
Remove 99 (but leave 7) to reply

"Roman Rumian" <usun_torumian@agh.edu.pl> wrote in message 
news:do6tf9$vfc$1@news.agh.edu.pl...
> Hi Vladimir,
>
> Vladimir Vassilevsky napisa&#4294967295;(a):
> (...)
>> Technically speaking, the single precision IEEE-754 float accuracy may not be 
>> sufficient for the certain aspects of audio processing. The 32-bit int may 
>> produce better results.
>
> thats interesting for me. I have problems with IIR filter stability (6th order 
> Butterworth highpass at 20 Hz and 48ksps sampling) on Motorola DSP56300 
> (24-bit fixed point). Plan to change platform to SHARC Melody, but it seems 
> that should still use fixed point representation ? !
>
> Best regards
>
> Roman Rumian 

"john" <johns@xetron.com> schrieb im Newsbeitrag 
news:1134909175.895541.297360@g49g2000cwa.googlegroups.com...

> Not being an audio person, I was wondering what some of the experts
> here would say about that assertion. Is it really always better to use
> floating point for audio processing such as SRC, mixing, eq, etc?

I'd say that all current DAW (digitial-audio-workstation) applications for 
PCs or Macs
do in fact already use floats instead of integer.

One reason is the fact, that float-performance on all PC CPUs in the last 
couple of
years was way ahead of integer. And obviously you will get quite a bit more
precision this way.

One of the posters mentioned "denormals" as a performance-problem. This was
in fact a common thing which became aware with the introduction of P4 CPUs
(if I remember correctly).  All of the DAW-software-manufacturers and
vendors of software-effects & instruments have since then done quite a lot 
of
work that those denormals won't cause performance-issues any more...

Some effect-manufacturers have even started to use 64bit floats for internal
effect-processsing which is supposed to achieve even higher audio-quality.

But there a lot of people in that field seem quite esotheric about how 
something
has to sound ;-) lamenting on how everything was better in those old analog 
times ;-)

Tobias 

Tobias Erichsen wrote:

   ...

> I'd say that all current DAW (digitial-audio-workstation) applications for 
> PCs or Macs do in fact already use floats instead of integer.
> 
> One reason is the fact, that float-performance on all PC CPUs in the last 
> couple of years was way ahead of integer.

> And obviously you will get quite a bit more precision this way.

What makes that obvious?

   ...

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

"Jerry Avins" <jya@ieee.org> schrieb im Newsbeitrag 
news:b62dnc6euJjbsDXeRVn-vA@rcn.net...

> What makes that obvious?

Well - the original poster was referring to 16bit integer applications and 
32bit floating-point
apps... so how would that not be obvious?

Tobias 

Tobias Erichsen wrote:
> "Jerry Avins" <jya@ieee.org> schrieb im Newsbeitrag 
> news:b62dnc6euJjbsDXeRVn-vA@rcn.net...
> 
> 
>>What makes that obvious?
> 
> 
> Well - the original poster was referring to 16bit integer applications and 
> 32bit floating-point
> apps... so how would that not be obvious?

The difference is the bit count, not whether the binary point floats. 
That wasn't obvious to me from what you wrote.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;