Fixed-point effects lib? Linux friendly floating-point DSP?

[Amusing Muses]

Hi I'm investigating the Blackfin as a platform for a music DSP
toolkit, unfortunately all the examples for simple effects like
choruses & flangers I can find utilize floating-point
arithmetic--which cripples their performance on the Blackfin because
it has no FPU.

Does anyone know of any open source DSP effects packages that use
fixed-point arithmetic? Could save me a bunch of time rewriting these
effects myself... or should I just use a floating-point DSP? I'm
looking for something that is supported well by open source tools
(gcc, u-boot, Linux, etc) and that is low-cost enough for a high
volume consumer product... The Blackfin meets these requirements
except the lack of floating-point is turning out to be a hassle. Any
suggestions?

Thanks.
--
dupswapdrop -- the music-dsp mailing list and website:
subscription info, FAQ, source code archive, list archive, book reviews, dsp links
http://music.columbia.edu/cmc/music-dsp 
http://music.columbia.edu/mailman/listinfo/music-dsp

[Jerry Evans-4]

> low-cost enough for a high volume consumer


How about the popular, evergreen,  x86 line? floating point, some vector
processing, good support for, and easy access to free tools, operating
systems, examples and test cases. Plus planet-wide suppliers and
technical support.

Google pitches up a UK link to a *retail* 1.5Ghz C7 board for £37 or $58
: http://linitx.com/viewproduct.php?prodid=11878

And it's got HDA, USB 2, SATA, Video, ethernet ... missing a CPU fan.

I bet VIA would have some wickedly competitive prices for this board at
multi K order levels -

Last time I checked a 600Mhz Blackfin evaluation board was $900 direct
from AD.

Spend the money on a proper RT/OS and create great things.

Jerry.
No, not that one, the other one.

Amusing Muses wrote:

> Hi I'm investigating the Blackfin as a platform for a music DSP
> toolkit, unfortunately all the examples for simple effects like
> choruses & flangers I can find utilize floating-point
> arithmetic--which cripples their performance on the Blackfin because
> it has no FPU.
>
> Does anyone know of any open source DSP effects packages that use
> fixed-point arithmetic? Could save me a bunch of time rewriting these
> effects myself... or should I just use a floating-point DSP? I'm
> looking for something that is supported well by open source tools
> (gcc, u-boot, Linux, etc) and that is low-cost enough for a high
> volume consumer product... The Blackfin meets these requirements
> except the lack of floating-point is turning out to be a hassle. Any
> suggestions?
>
> Thanks.
> --
> dupswapdrop -- the music-dsp mailing list and website:
> subscription info, FAQ, source code archive, list archive, book reviews, dsp links
> http://music.columbia.edu/cmc/music-dsp 
> http://music.columbia.edu/mailman/listinfo/music-dsp
>
>  

--
dupswapdrop -- the music-dsp mailing list and website:
subscription info, FAQ, source code archive, list archive, book reviews, dsp links
http://music.columbia.edu/cmc/music-dsp 
http://music.columbia.edu/mailman/listinfo/music-dsp

[Amusing Muses]

I haven't seriously considered the x86 for my embedded application,
intuitively it doesn't seem like people use x86 for embedded DSP.
Though I could try. The algorithms I need run just fine on my 1.5 GHz
Centrino in Linux in just about real-time, and they would probably run
fine on something more lightweight too--maybe I don't need the
specific capabilities of a DSP at all.

A 600MHz BF533 EZKIT goes for about $500 from ADI now and they have
the best Linux & GCC toolchain support around for DSPs it seems.

So as far as the hardware is concerned I'm probably going to try an
ARM w/ FP support from TI, or maybe an x86 like you suggested if
fixed-point is too much of a hurdle.

Thanks.

On Thu, Oct 8, 2009 at 6:30 PM, Jerry Evans <jerry@...> wrote:

>> low-cost enough for a high volume consumer
>
>
> How about the popular, evergreen,  x86 line? floating point, some vector
> processing, good support for, and easy access to free tools, operating
> systems, examples and test cases. Plus planet-wide suppliers and
> technical support.
>
> Google pitches up a UK link to a *retail* 1.5Ghz C7 board for £37 or $58
> : http://linitx.com/viewproduct.php?prodid=11878
>
> And it's got HDA, USB 2, SATA, Video, ethernet ... missing a CPU fan.
>
> I bet VIA would have some wickedly competitive prices for this board at
> multi K order levels -
>
> Last time I checked a 600Mhz Blackfin evaluation board was $900 direct
> from AD.
>
> Spend the money on a proper RT/OS and create great things.
>
> Jerry.
> No, not that one, the other one.
>
> Amusing Muses wrote:
>> Hi I'm investigating the Blackfin as a platform for a music DSP
>> toolkit, unfortunately all the examples for simple effects like
>> choruses & flangers I can find utilize floating-point
>> arithmetic--which cripples their performance on the Blackfin because
>> it has no FPU.
>>
>> Does anyone know of any open source DSP effects packages that use
>> fixed-point arithmetic? Could save me a bunch of time rewriting these
>> effects myself... or should I just use a floating-point DSP? I'm
>> looking for something that is supported well by open source tools
>> (gcc, u-boot, Linux, etc) and that is low-cost enough for a high
>> volume consumer product... The Blackfin meets these requirements
>> except the lack of floating-point is turning out to be a hassle. Any
>> suggestions?
>>
>> Thanks.
>> --
>> dupswapdrop -- the music-dsp mailing list and website:
>> subscription info, FAQ, source code archive, list archive, book reviews, dsp links
>> http://music.columbia.edu/cmc/music-dsp
>> http://music.columbia.edu/mailman/listinfo/music-dsp
>>
>>
>
>
> --
> dupswapdrop -- the music-dsp mailing list and website:
> subscription info, FAQ, source code archive, list archive, book reviews, dsp links
> http://music.columbia.edu/cmc/music-dsp
> http://music.columbia.edu/mailman/listinfo/music-dsp
>

--
dupswapdrop -- the music-dsp mailing list and website:
subscription info, FAQ, source code archive, list archive, book reviews, dsp links
http://music.columbia.edu/cmc/music-dsp 
http://music.columbia.edu/mailman/listinfo/music-dsp

[Richard Dobson]

Amusing Muses wrote:
> I haven't seriously considered the x86 for my embedded application,
> intuitively it doesn't seem like people use x86 for embedded DSP.
> Though I could try. The algorithms I need run just fine on my 1.5 GHz
> Centrino in Linux in just about real-time


"just about"? That describes a very heavy process, if its on a 1.5GHz
machine. even a stereo phase vocoder will not consume much more than 10%
CPU load on such a machine (unless really badly implemented). I
developed my first pvoc VST plugin on a 333MHz machine, and that did
dance around the 50% point, making stereo iffy.


I thought the Blackfin was a 16bit device. That's a bit tight for audio
dsp; a 24bit device would give you far fewer headaches, especially if
you have any interest in full-bandwidth filtering, etc. The issue of
fixed-point v floating point is less central than the raw wordsize
itself. Filter coefficients  (esp for low cutoff frequencies etc) need a
~minimum~ of 24bit precision, as the numbers can get really small. The
classic 24bit devices, such as the Motorola 56K series (now Freescale)
are nominally 24bit devices, but have 56bit accumulators for fixed-point
arithmetic.


Richard Dobson



--
dupswapdrop -- the music-dsp mailing list and website:
subscription info, FAQ, source code archive, list archive, book reviews, dsp links
http://music.columbia.edu/cmc/music-dsp 
http://music.columbia.edu/mailman/listinfo/music-dsp

[Rob Belcham-2]

Just to put that into context - the cheap blackfin are in the order of $5 a
piece. The $900 tag is for the development board, not what it'd cost to
produce a board for yourself.

All depends the application & on what resources you have available i guess.
Rob

--------------------------------------------------
From: "Jerry Evans" <jerry@...>
Sent: Thursday, October 08, 2009 11:30 PM
To: "A discussion list for music-related DSP" <music-dsp@...>
Subject: Re: [music-dsp] Fixed-point effects lib? Linux friendly
floating-point DSP?

>> low-cost enough for a high volume consumer
>
>
> How about the popular, evergreen,  x86 line? floating point, some vector
> processing, good support for, and easy access to free tools, operating
> systems, examples and test cases. Plus planet-wide suppliers and
> technical support.
>
> Google pitches up a UK link to a *retail* 1.5Ghz C7 board for £37 or $58
> : http://linitx.com/viewproduct.php?prodid=11878
>
> And it's got HDA, USB 2, SATA, Video, ethernet ... missing a CPU fan.
>
> I bet VIA would have some wickedly competitive prices for this board at
> multi K order levels -
>
> Last time I checked a 600Mhz Blackfin evaluation board was $900 direct
> from AD.
>
> Spend the money on a proper RT/OS and create great things.
>
> Jerry.
> No, not that one, the other one.
>
> Amusing Muses wrote:
>> Hi I'm investigating the Blackfin as a platform for a music DSP
>> toolkit, unfortunately all the examples for simple effects like
>> choruses & flangers I can find utilize floating-point
>> arithmetic--which cripples their performance on the Blackfin because
>> it has no FPU.
>>
>> Does anyone know of any open source DSP effects packages that use
>> fixed-point arithmetic? Could save me a bunch of time rewriting these
>> effects myself... or should I just use a floating-point DSP? I'm
>> looking for something that is supported well by open source tools
>> (gcc, u-boot, Linux, etc) and that is low-cost enough for a high
>> volume consumer product... The Blackfin meets these requirements
>> except the lack of floating-point is turning out to be a hassle. Any
>> suggestions?
>>
>> Thanks.
>> --
>> dupswapdrop -- the music-dsp mailing list and website:
>> subscription info, FAQ, source code archive, list archive, book reviews,
>> dsp links
>> http://music.columbia.edu/cmc/music-dsp
>> http://music.columbia.edu/mailman/listinfo/music-dsp
>>
>>
>
>
> --
> dupswapdrop -- the music-dsp mailing list and website:
> subscription info, FAQ, source code archive, list archive, book reviews,
> dsp links
> http://music.columbia.edu/cmc/music-dsp
> http://music.columbia.edu/mailman/listinfo/music-dsp
>

--
dupswapdrop -- the music-dsp mailing list and website:
subscription info, FAQ, source code archive, list archive, book reviews, dsp links
http://music.columbia.edu/cmc/music-dsp 
http://music.columbia.edu/mailman/listinfo/music-dsp

[Amusing Muses]

On Thu, Oct 8, 2009 at 7:58 PM, Richard Dobson
<richarddobson@...> wrote:

> Amusing Muses wrote:
>> I haven't seriously considered the x86 for my embedded application,
>> intuitively it doesn't seem like people use x86 for embedded DSP.
>> Though I could try. The algorithms I need run just fine on my 1.5 GHz
>> Centrino in Linux in just about real-time
>
>
> "just about"? That describes a very heavy process, if its on a 1.5GHz
> machine. even a stereo phase vocoder will not consume much more than 10%
> CPU load on such a machine (unless really badly implemented). I
> developed my first pvoc VST plugin on a 333MHz machine, and that did
> dance around the 50% point, making stereo iffy.
>

Whoops, by "just about" I meant that the path introduces a short lag
time, however the pipeline never stalls or anything.

>
> I thought the Blackfin was a 16bit device. That's a bit tight for audio
> dsp; a 24bit device would give you far fewer headaches, especially if
> you have any interest in full-bandwidth filtering, etc. The issue of
> fixed-point v floating point is less central than the raw wordsize
> itself. Filter coefficients  (esp for low cutoff frequencies etc) need a
> ~minimum~ of 24bit precision, as the numbers can get really small. The
> classic 24bit devices, such as the Motorola 56K series (now Freescale)
> are nominally 24bit devices, but have 56bit accumulators for fixed-point
> arithmetic.
>

The Blackfin supports both 16 and 32 bit operations, with 40 bit
accumulators. Thanks for pointing out the importance of the word size,
from a high level my main concern is quality. I figured that a
combination of a decent number of bits per sample, and a high enough
sampling rate would be indicators I could rely on to select components
for use in a professional quality audio application. However it is
good to know that the word size must be at least 24 bits for some
filtering applications.

>
> Richard Dobson
>
>
>
> --
> dupswapdrop -- the music-dsp mailing list and website:
> subscription info, FAQ, source code archive, list archive, book reviews, dsp links
> http://music.columbia.edu/cmc/music-dsp
> http://music.columbia.edu/mailman/listinfo/music-dsp
>

--
dupswapdrop -- the music-dsp mailing list and website:
subscription info, FAQ, source code archive, list archive, book reviews, dsp links
http://music.columbia.edu/cmc/music-dsp 
http://music.columbia.edu/mailman/listinfo/music-dsp

[Florian Schmidt]

On Thursday 08 October 2009, Amusing Muses wrote:

> Hi I'm investigating the Blackfin as a platform for a music DSP
> toolkit, unfortunately all the examples for simple effects like
> choruses & flangers I can find utilize floating-point
> arithmetic--which cripples their performance on the Blackfin because
> it has no FPU.
>
> Does anyone know of any open source DSP effects packages that use
> fixed-point arithmetic? Could save me a bunch of time rewriting these
> effects myself... or should I just use a floating-point DSP? I'm
> looking for something that is supported well by open source tools
> (gcc, u-boot, Linux, etc) and that is low-cost enough for a high
> volume consumer product... The Blackfin meets these requirements
> except the lack of floating-point is turning out to be a hassle. Any
> suggestions?
>
> Thanks.

Hi,

you might also take a look at Intel Atom boards.. They are rather cheap and
can do some number crunching :)

Flo

--
Palimm Palimm!
http://tapas.affenbande.org
--
dupswapdrop -- the music-dsp mailing list and website:
subscription info, FAQ, source code archive, list archive, book reviews, dsp links
http://music.columbia.edu/cmc/music-dsp 
http://music.columbia.edu/mailman/listinfo/music-dsp