GUITAR EFFECT DSP CHIP

Hi all

I want to design a guitar effect with some DSP chip.
Which DSP chip might be good for this ?

Thanks
EC

On 18/08/2010 08:53, RealInfo wrote:
> Hi all
>
> I want to design a guitar effect with some DSP chip.
> Which DSP chip might be good for this ?
>
> Thanks
> EC

This is a question with NO information on which to base an answer. It 
sounds extremely lazy. "String" and "length" come to mind. The regulars 
on this list will come down on you like a ton of bricks. They may even 
consider you a troll. What research have you done so far? Have you read 
the comp.dsp FAQ?

Anyway, as I am feeling generous, here is one link for you to look at:

http://line6.com/tcddk

Richard Dobson

RealInfo wrote:

> Hi all
> 
> I want to design a guitar effect with some DSP chip.
> Which DSP chip might be good for this ?

Drop this idea. The nature of the question suggests that you'd be better 
do something else.

VLV

On Aug 18, 10:19=A0am, Vladimir Vassilevsky <nos...@nowhere.com> wrote:
> RealInfo wrote:
> > Hi all
>
> > I want to design a guitar effect with some DSP chip.
> > Which DSP chip might be good for this ?
>
> Drop this idea. The nature of the question suggests that you'd be better
> do something else.

i dunno, i think Richard had a good answer to the question.

i oughta get one of those Line6 pedals.

r b-j

>On 18/08/2010 08:53, RealInfo wrote:
>> Hi all
>>
>> I want to design a guitar effect with some DSP chip.
>> Which DSP chip might be good for this ?
>>
>> Thanks
>> EC
>
>This is a question with NO information on which to base an answer. It 
>sounds extremely lazy. "String" and "length" come to mind. The regulars 
>on this list will come down on you like a ton of bricks. They may even 
>consider you a troll. What research have you done so far? Have you read 
>the comp.dsp FAQ?
>
>Anyway, as I am feeling generous, here is one link for you to look at:
>
>http://line6.com/tcddk
>
>Richard Dobson
>

FWIW, I looked seriously at developing for this platform but ultimately
decided against it. The support is awful to nonexistent, and on the forum
one representative claimed the pedals were discontinued before another
stepped in an corrected him. In short, the writing may be on the wall.

My biggest gripe was the cost of the pedal unit. The deck, which should be
the least expensive part, is $70! The swappable, programmable module is
$35. So you have to convince someone to pay $105 + whatever you want to
charge for your program. They're obviously marketing this hoping most
people will have the deck (false) and simply will want to purchase your
module. And they offer no developer discount/bulk pricing.

Bryan

>>On 18/08/2010 08:53, RealInfo wrote:
>>> Hi all
>>>
>>> I want to design a guitar effect with some DSP chip.
>>> Which DSP chip might be good for this ?
>>>
>>> Thanks
>>> EC
>>
>>This is a question with NO information on which to base an answer. It 
>>sounds extremely lazy. "String" and "length" come to mind. The regulars 
>>on this list will come down on you like a ton of bricks. They may even 
>>consider you a troll. What research have you done so far? Have you read 
>>the comp.dsp FAQ?
>>
>>Anyway, as I am feeling generous, here is one link for you to look at:
>>
>>http://line6.com/tcddk
>>
>>Richard Dobson
>>
>
>FWIW, I looked seriously at developing for this platform but ultimately
>decided against it. The support is awful to nonexistent, and on the forum
>one representative claimed the pedals were discontinued before another
>stepped in an corrected him. In short, the writing may be on the wall.
>
>My biggest gripe was the cost of the pedal unit. The deck, which should
be
>the least expensive part, is $70! The swappable, programmable module is
>$35. So you have to convince someone to pay $105 + whatever you want to
>charge for your program. They're obviously marketing this hoping most
>people will have the deck (false) and simply will want to purchase your
>module. And they offer no developer discount/bulk pricing.
>
>Bryan

It might not be a route to low cost volume production, but it seems like a
pretty good way to get a few well made units out to people to try your
algorithms in the real world.

Steve

On Aug 18, 12:00=A0pm, "Bryan52803" <bryan.paul@n_o_s_p_a_m.gmail.com>
wrote:
> >On 18/08/2010 08:53, RealInfo wrote:
> >> Hi all
>
> >> I want to design a guitar effect with some DSP chip.
> >> Which DSP chip might be good for this ?
>
> >> Thanks
> >> EC
>
> >This is a question with NO information on which to base an answer. It
> >sounds extremely lazy. "String" and "length" come to mind. The regulars
> >on this list will come down on you like a ton of bricks. They may even
> >consider you a troll. What research have you done so far? Have you read
> >the comp.dsp FAQ?
>
> >Anyway, as I am feeling generous, here is one link for you to look at:
>
> >http://line6.com/tcddk
>
> >Richard Dobson
>
> FWIW, I looked seriously at developing for this platform but ultimately
> decided against it. The support is awful to nonexistent, and on the forum
> one representative claimed the pedals were discontinued before another
> stepped in an corrected him. In short, the writing may be on the wall.
>
> My biggest gripe was the cost of the pedal unit. The deck, which should b=
e
> the least expensive part, is $70! The swappable, programmable module is
> $35. So you have to convince someone to pay $105 + whatever you want to
> charge for your program. They're obviously marketing this hoping most
> people will have the deck (false) and simply will want to purchase your
> module. And they offer no developer discount/bulk pricing.

it also seems to me that since there is no difference between the
programmable modules other than the bits that live in, what i guess
would be, flash memory, it seems to me that a cost effect multieffects
pedal would be one with a switch of some sort.  so you turn a knob
rather than swap hardware to change your pedal from one function to
another.

the Freescale thing is a concern.  essentially Motorola simply ditched
every development effort they had going with the 56K and later
generations more than a decade ago.  so to milk out whatever they
could from what was left of their ca. 1995 technology, they spun off
Freescale.  but nothing new will be done with it.  you have,
essentially, a 56300 core competing against whatever TI and ADI
continue to crank out.  this does sorta sadden me, because back in the
90s, i *was* a real Mot partisan.  now there is really nothing left
there to be a partisan about.  oh well.

i wonder what L6 plans to use for their future pods and pedals?  i
heard that Digi is planning on changing their TDM gear from 56K to
some TI chip.  it wouldn't be the decision i would make about it, but
some change is inevitable.

r b-j

On 08/18/2010 05:22 AM, Richard Dobson wrote:
> On 18/08/2010 08:53, RealInfo wrote:
>> Hi all
>>
>> I want to design a guitar effect with some DSP chip.
>> Which DSP chip might be good for this ?
>>
>> Thanks
>> EC
>
> This is a question with NO information on which to base an answer. It
> sounds extremely lazy. "String" and "length" come to mind. The regulars
> on this list will come down on you like a ton of bricks. They may even
> consider you a troll. What research have you done so far? Have you read
> the comp.dsp FAQ?
>
> Anyway, as I am feeling generous, here is one link for you to look at:
>
> http://line6.com/tcddk

Oh, some of us will be nice.

-- 

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" was written for you.
See details at http://www.wescottdesign.com/actfes/actfes.html

On 08/18/2010 12:53 AM, RealInfo wrote:
> Hi all
>
> I want to design a guitar effect with some DSP chip.
> Which DSP chip might be good for this ?

42.  Now that you know the answer, would you like to have your question 
explained?

You can do _some_ guitar effect with _any_ DSP chip -- or even a cheap 
old little microprocessor, or a logic gate, or a single transistor 
(think Fuzzbox).

But any _specific_ guitar effect is going to take some specific amount 
of processing power and audio fidelity, and some of those are going to 
demand more from the DSP than others.

At a guess I would say that for a one-off you want a chip that can do 24 
or 32-bit math (note that a fast '16 bit' DSP will do 32 bit math just 
fine), you want the chip to be able to do it fast enough to keep up with 
a sampling rate no less than 40kHz and probably well in excess of that, 
and you want the chip to have 16 bit audio going in and out, at the very 
least*.

For a production pedal you'd want to look at the needs of a specific 
algorithm or set thereof and choose the least expensive set of hardware 
that'll let them run in whatever combinations you want to allow.

For prototyping a floating-point DSP will make your algorithmic 
development much easier -- but if you go into production it'll either 
cost big bundles of $$ in production costs, or it'll cost engineering 
time to port the algorithms to fixed-point.

It may be a good idea to record some licks cleanly off of the guitar 
into a PC, then use something like Scilab (or Matlab or Octave) to 
experiment with different effects algorithms.  This can't be done in 
real time, but these tools make the algorithm development very easy; 
once you know that you like it then you can start estimating what it'll 
take in terms of processor resources in for-real hardware.

Hopefully this'll spawn enough questions that you can get your thoughts 
narrowed down.

* Depending on the effect -- for a fuzzbox, some would argue that you 
just need 1 :-).  Audiophiles may argue for more, but I think that for 
just one instrument going in, and that one not having a terribly huge 
dynamic range, that you could get away with 16.  But if you can afford 
24, run with it.

-- 

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" was written for you.
See details at http://www.wescottdesign.com/actfes/actfes.html

Tim Wescott <tim@seemywebsite.com> wrote in 
news:GaKdnUW5xaSJgvHRnZ2dnUVZ_jSdnZ2d@web-ster.com:

> On 08/18/2010 12:53 AM, RealInfo wrote:
>> Hi all
>>
>> I want to design a guitar effect with some DSP chip.
>> Which DSP chip might be good for this ?
> 
> 42.  Now that you know the answer, would you like to have your question 
> explained?
> 
> You can do _some_ guitar effect with _any_ DSP chip -- or even a cheap 
> old little microprocessor, or a logic gate, or a single transistor 
> (think Fuzzbox).
> 
> But any _specific_ guitar effect is going to take some specific amount 
> of processing power and audio fidelity, and some of those are going to 
> demand more from the DSP than others.
> 
> At a guess I would say that for a one-off you want a chip that can do 24 
> or 32-bit math (note that a fast '16 bit' DSP will do 32 bit math just 
> fine), you want the chip to be able to do it fast enough to keep up with 
> a sampling rate no less than 40kHz and probably well in excess of that, 
> and you want the chip to have 16 bit audio going in and out, at the very 
> least*.

I think you probably want to sample internally at very high rates for some 
algorithms to reduce harmonic aliasing. Of course, this does not 
necessarily say anything about the sample rate of the converters.

Relatively good ADCs with 100dB+ S/N are not expensive. SOA converters are 
in the 20 to 21 bit range. Lots of converters are labeled "24 bit". 

The really good converters draw lots of current. Not so good if you want to 
run on batteries. 


> 
> For a production pedal you'd want to look at the needs of a specific 
> algorithm or set thereof and choose the least expensive set of hardware 
> that'll let them run in whatever combinations you want to allow.
> 
> For prototyping a floating-point DSP will make your algorithmic 
> development much easier -- but if you go into production it'll either 
> cost big bundles of $$ in production costs, or it'll cost engineering 
> time to port the algorithms to fixed-point.

Most people spend too much time pricing out components without factoring in 
dev costs, realistic product estimates, time to market, risk, etc.

If you are trying to eat Line 6's lunch, you are looking at high volumes 
before you make money. My guess, this is very unlikely given the original 
question in the first place.

If you are looking at a niche market, then cost is probably less of an 
issue and dev cost will dominant regardless.

I have had the privilege of competing for available inventory with Line 6 
for floating point DSPs (SHARCs). Floating point DSPs are much easier to 
code with and they are not necessarily expensive. 

You also have to look at the supporting cast of parts when looking at a 
DSP. You will need memory, core supplies, etc. 



> 
> It may be a good idea to record some licks cleanly off of the guitar 
> into a PC, then use something like Scilab (or Matlab or Octave) to 
> experiment with different effects algorithms.  This can't be done in 
> real time, but these tools make the algorithm development very easy; 
> once you know that you like it then you can start estimating what it'll 
> take in terms of processor resources in for-real hardware.
> 
> Hopefully this'll spawn enough questions that you can get your thoughts 
> narrowed down.
> 
> * Depending on the effect -- for a fuzzbox, some would argue that you 
> just need 1 :-).  Audiophiles may argue for more, but I think that for 
> just one instrument going in, and that one not having a terribly huge 
> dynamic range, that you could get away with 16.  But if you can afford 
> 24, run with it.
> 

Stomp boxes are fun to think about, A project like this might be a good 
place to learn some actual DSP skils. I doubt it makes much business sense 
as a product opportunity. The market is crowded and apparently, you do not 
have specific skills already that would set you apart in this area.

With this assumption, go with a SHARC if you want real hardware solution or 
maybe play around in a PC. (Disclosure: I build SHARC boards so this 
comment is somewhat self-serving)

56K has no future. It hasn't been king in audio for at least 10 years.

The TI parts are also candidates. They have their own partisans. They are 
not as popular in audio (non-telephony) space as ADI, but have good DSP 
market share overall.

Regardless, I would avoid fixed point DSPs for this kind of application. It 
is much, much easier to code most of the algorithms in floating point. 

Modern DSPs will have MIPs to burn so most anything will work.

Al Clark
www.danvillesignal.com