Question

Hi,
  Sorry to bother you with such an uneducated question, well...
I have looked into some live performance audio gear, and it
uses a DSP and a 24bit 192Khz DAC ADC. But each DSP "Box" usually
does only one thing - one effect or compression/limiter or feedback
cancelation etc' - and cost about 130$US.

  So I was thinking to myself why cant the box do all of the above.
Is the DSP too weak? Then I thought to myself whether a PC can do
all of the above, and found out that there are several DirectX
plugins that basically do all the things that the hardware does. So
I wonder what is the difference between using the DSP hardware VS.
Intel PC software. Is there a difference? I read that 10msec latency
is not noticable in realtimes performance, and even 30msec is ok.


Summary:
1. Is the DSP too weak to do more than one functionality at a time?
2. What is the diffence between using a DSP hardware VS. an Intel PC
    and software? (For real time performance with 0-30 msec latency)
3. If you think the PC can handle the audio signal "real time":
    How many such audio signals do you think a PC can handle "real time"?

Hi,
  Sorry to bother you with such an uneducated question, well...
I have looked into some live performance audio gear, and it
uses a DSP and a 24bit 192Khz DAC ADC. But each DSP "Box" usually
does only one thing - one effect or compression/limiter or feedback
cancelation etc' - and cost about 130$US.

  So I was thinking to myself why cant the box do all of the above.
Is the DSP too weak? Then I thought to myself whether a PC can do
all of the above, and found out that there are several DirectX
plugins that basically do all the things that the hardware does. So
I wonder what is the difference between using the DSP hardware VS.
Intel PC software. Is there a difference? I read that 10msec latency
is not noticable in realtimes performance, and even 30msec is ok.


Summary:
1. Is the DSP too weak to do more than one functionality at a time?
2. What is the diffence between using a DSP hardware VS. an Intel PC
    and software? (For real time performance with 0-30 msec latency)
3. If you think the PC can handle the audio signal "real time":
    How many such audio signals do you think a PC can handle "real time"?

I might also point out that there are many boxes that do multiple effects,
e.g. reverb, chorus, flange, echo, EQ, pitch shifting, mixing, compression,
tuning, etc..  The ones targeted for guitarists are usually quite
inexpensive as well.

Also, a PC does generally have significantly more latency than a dedicated
box due to the buffering required.  This may or may not be an important
issue, depending on your application.

Dedicated boxes also have an advantage for fixed installations in that they
don't crash and take much less power and space vs. a PC.

"Mike Rosing" <rosing@neurophys.wisc.edu> wrote in message
news:3F56A367.30900@neurophys.wisc.edu...
> johny wrote:
> > Hi,
> >  Sorry to bother you with such an uneducated question, well...
> > I have looked into some live performance audio gear, and it
> > uses a DSP and a 24bit 192Khz DAC ADC. But each DSP "Box" usually
> > does only one thing - one effect or compression/limiter or feedback
> > cancelation etc' - and cost about 130$US.
> >
> >  So I was thinking to myself why cant the box do all of the above.
> > Is the DSP too weak? Then I thought to myself whether a PC can do
> > all of the above, and found out that there are several DirectX
> > plugins that basically do all the things that the hardware does. So
> > I wonder what is the difference between using the DSP hardware VS.
> > Intel PC software. Is there a difference? I read that 10msec latency
> > is not noticable in realtimes performance, and even 30msec is ok.
> >
> >
> > Summary:
> > 1. Is the DSP too weak to do more than one functionality at a time?
> > 2. What is the diffence between using a DSP hardware VS. an Intel PC
> >    and software? (For real time performance with 0-30 msec latency)
> > 3. If you think the PC can handle the audio signal "real time":
> >    How many such audio signals do you think a PC can handle "real time"?
> >
>
> You want to compare a $130 box to a $1000 box and ask why the $130 box
> is so expensive?  You can reprogram dsp's if you want to, but for a
> commercial application it's a whole lot cheaper (and I think $130 per
> function is pretty reasonable with todays technology) to do one thing
> per box.
>
> If you already have a pc, and you can just "throw it away" for this
> particular application, then certainly you can save some money.  but
> if you want real quality you'll get something that's been designed
> by people whose lives depend on a quality product.
>
> Is this for a hobby?  Then use your pc.  If it's for sound production,
> buy the boxes.
>
> Patience, persistence, truth,
> Dr. mike
>
> --
> Mike Rosing
> www.beastrider.com                   BeastRider, LLC
> SHARC debug tools
>

johny wrote:
> 
> On the cost: That depends on how many audio channels one computer
> can handle. I have asked exactly that question.
> 
> If a 500$ computer does a one box job at most, than you are right - it is
> not economical. But if a 500$ computer does the job of 4 boxes or more,
> than it is economical. I just don't know, maybe one computer can handle
> 20 audio channels, or 40, or 100, I just don't know. Each channel with today
> DAC is 96Khz X 24bit = 288 KByte/sec. That doesn't look  much. A low
> end LAN card does 100Mbit/sec= 12.5MByte/Sec. 43 times a Digital to Analog
> Converter. I googled at a 1.5GHz Pentium 4 Mflops, and found some site
> saying it is about 2000 Mflops. Maybe I ought to find which DSP is in those
> boxes, and compare? (It doesn't say what is inside). And then again there
> are also the latency's...
> 
> How many audio channels can one computer handle?  (with about 10-30msec
> delay)

It depends. 

People are doing this, using a PC for a realtime F/X box, but it's
slightly goofy. By the time you buy enough analog I/O to 
get all the channels you want, dedicated DSP boxes begin 
to make more sense. The dedicated DSP boxes also
cannot be pirated, and plugins can. This means people 
making dedicated boxes have a leg up in terms of cash flow. Some
of the old stalwart plugins makers have abandoned development,
no doubt partially due to piracy. 

The thing you'll find on PC is that being able to control latency is 
very, very iffy, especially driver and O/S latency. 


--
Les Cargill

On the cost: That depends on how many audio channels one computer
can handle. I have asked exactly that question.

If a 500$ computer does a one box job at most, than you are right - it is
not economical. But if a 500$ computer does the job of 4 boxes or more,
than it is economical. I just don't know, maybe one computer can handle
20 audio channels, or 40, or 100, I just don't know. Each channel with today
DAC is 96Khz X 24bit = 288 KByte/sec. That doesn't look  much. A low
end LAN card does 100Mbit/sec= 12.5MByte/Sec. 43 times a Digital to Analog
Converter. I googled at a 1.5GHz Pentium 4 Mflops, and found some site
saying it is about 2000 Mflops. Maybe I ought to find which DSP is in those
boxes, and compare? (It doesn't say what is inside). And then again there
are also the latency's...


How many audio channels can one computer handle?  (with about 10-30msec
delay)

Les Cargill wrote:
> 
> People are doing this, using a PC for a realtime F/X box, but it's
> slightly goofy. By the time you buy enough analog I/O to 
> get all the channels you want, dedicated DSP boxes begin 
> to make more sense. The dedicated DSP boxes also
> cannot be pirated, and plugins can. This means people 
> making dedicated boxes have a leg up in terms of cash flow. Some
> of the old stalwart plugins makers have abandoned development,
> no doubt partially due to piracy. 
> 
> The thing you'll find on PC is that being able to control latency is 
> very, very iffy, especially driver and O/S latency. 
> 

Also note that there are programmable DSP boxes out there, e.g. 
<http://www.soundart-hot.com/english/>, which presumably give you the 
best of both worlds.

Paul

johny wrote:

> Hi,
>   Sorry to bother you with such an uneducated question, well...
> I have looked into some live performance audio gear, and it
> uses a DSP and a 24bit 192Khz DAC ADC. But each DSP "Box" usually
> does only one thing - one effect or compression/limiter or
> feedback cancelation etc' - and cost about 130$US.
> 
>   So I was thinking to myself why cant the box do all of the
>   above.
> Is the DSP too weak? Then I thought to myself whether a PC can do
> all of the above, and found out that there are several DirectX
> plugins that basically do all the things that the hardware does.
> So I wonder what is the difference between using the DSP hardware
> VS. Intel PC software. Is there a difference?

> I read that 10msec
> latency is not noticable in realtimes performance, and even 30msec
> is ok.
> 

If the signals which suffer a delay of 10ms..30ms are mixed together 
with original signals which don't have such a delay, there's a good 
chance that you end up with a mess.
If you "hear" only the delayed result without mixing, no problem.

But there are "different delays".
DSPs which use typical audio ADC/DAC systems (codecs), need 
restauration filters which usually require such a delay.
But this delay is predictable and perfectly constant.

If you process your sounds on a PC, you have the same delay if you 
use the same codecs. But now, your signals will suffer dynamically 
changing delays depending on the implementation.
If you think of Windows or standard Linux as your basis, it will be 
a tough job to program your effects so that they aren't delayed 
because the PC needs time to refresh the desktop or such things.
All this can be done, but keep in mind, that PCs are not designed to 
do such work. 

Processing a lot of channels would bring most PCs to their limit:
only 4 PCI slots for sound cards
or power restrictions if you use USB cards (probably you won't be 
able to handle many more than 4).

So, it's not only a question of calculation power, the interfacing 
might be your biggest issue.

Bernhard
-- 
before sending to the above email-address:
replace deadspam.com by foerstergroup.de

"Paul Russell" <prussell@sonic.net> wrote in message
>
> Also note that there are programmable DSP boxes out there, e.g.
> <http://www.soundart-hot.com/english/>, which presumably give you the
> best of both worlds.
>
> Paul
>



I found that I already have a programable DSP in my computer :P
It is my SoundBlaster Live! Value. See http://kxproject.lugosoft.com/
for details.

"Bernhard Holzmayer" <holzmayer.bernhard@deadspam.com> wrote in message
news:1083636.SaCmMxLLvf@holzmayer.ifr.rt...
> johny wrote:
>
> > Hi,
> >   Sorry to bother you with such an uneducated question, well...
> > I have looked into some live performance audio gear, and it
> > uses a DSP and a 24bit 192Khz DAC ADC. But each DSP "Box" usually
> > does only one thing - one effect or compression/limiter or
> > feedback cancelation etc' - and cost about 130$US.
> >
> >   So I was thinking to myself why cant the box do all of the
> >   above.
> > Is the DSP too weak? Then I thought to myself whether a PC can do
> > all of the above, and found out that there are several DirectX
> > plugins that basically do all the things that the hardware does.
> > So I wonder what is the difference between using the DSP hardware
> > VS. Intel PC software. Is there a difference?
>
> > I read that 10msec
> > latency is not noticable in realtimes performance, and even 30msec
> > is ok.
> >
>
> If the signals which suffer a delay of 10ms..30ms are mixed together
> with original signals which don't have such a delay, there's a good
> chance that you end up with a mess.
> If you "hear" only the delayed result without mixing, no problem.
>
> But there are "different delays".
> DSPs which use typical audio ADC/DAC systems (codecs), need
> restauration filters which usually require such a delay.
> But this delay is predictable and perfectly constant.
>
> If you process your sounds on a PC, you have the same delay if you
> use the same codecs. But now, your signals will suffer dynamically
> changing delays depending on the implementation.

Do you mean that one pluging will produce the output faster then the other?
Or would the same plugin have a varing latency time?

> If you think of Windows or standard Linux as your basis, it will be
> a tough job to program your effects so that they aren't delayed
> because the PC needs time to refresh the desktop or such things.

What if I change the sound processing process to "real time"
priority in the task manager?

> All this can be done, but keep in mind, that PCs are not designed to
> do such work.
>
> Processing a lot of channels would bring most PCs to their limit:
> only 4 PCI slots for sound cards
> or power restrictions if you use USB cards (probably you won't be
> able to handle many more than 4).
>
> So, it's not only a question of calculation power, the interfacing
> might be your biggest issue.
>

I have seen one PCI card that connect to 40 ADC/DAC box. It follows
ASIO2 and has a top latency of 2ms. It sales with a 10 ADC/DAC
box, and cost about 400$. http://www.zzounds.com/item--STODSP2000

Also I have seen that CIRRUS LOGIC started selling a good stereo
ADC/DAC for about 5$ (when you buy 10000 units) on March.
So I guess good interfaces are soon to emarge for a low prices.

> Bernhard
> -- 
> before sending to the above email-address:
> replace deadspam.com by foerstergroup.de

johny wrote:

> 
> "Bernhard Holzmayer" <holzmayer.bernhard@deadspam.com> wrote in
> message news:1083636.SaCmMxLLvf@holzmayer.ifr.rt...
>> johny wrote:
>>
>> > Hi,
>> >   Sorry to bother you with such an uneducated question, well...
>> > I have looked into some live performance audio gear, and it
>> > uses a DSP and a 24bit 192Khz DAC ADC. But each DSP "Box"
>> > usually does only one thing - one effect or compression/limiter
>> > or feedback cancelation etc' - and cost about 130$US.
>> >
>> >   So I was thinking to myself why cant the box do all of the
>> >   above.
>> > Is the DSP too weak? Then I thought to myself whether a PC can
>> > do all of the above, and found out that there are several
>> > DirectX plugins that basically do all the things that the
>> > hardware does. So I wonder what is the difference between using
>> > the DSP hardware VS. Intel PC software. Is there a difference?
>>
>> > I read that 10msec
>> > latency is not noticable in realtimes performance, and even
>> > 30msec is ok.
>> >
>>
>> If the signals which suffer a delay of 10ms..30ms are mixed
>> together with original signals which don't have such a delay,
>> there's a good chance that you end up with a mess.
>> If you "hear" only the delayed result without mixing, no problem.
>>
>> But there are "different delays".
>> DSPs which use typical audio ADC/DAC systems (codecs), need
>> restauration filters which usually require such a delay.
>> But this delay is predictable and perfectly constant.
>>
>> If you process your sounds on a PC, you have the same delay if
>> you use the same codecs. But now, your signals will suffer
>> dynamically changing delays depending on the implementation.
> 
> Do you mean that one pluging will produce the output faster then
> the other? Or would the same plugin have a varing latency time?

The latter. Interrupts could cause a delay (or even loss ) of some 
samples while others pass the system with less delay.
This will probably be no problem as long as the number of channels 
which you process is low. 
But when the number increases, there will be a limitation. Beyond 
that limit, it will be impossible for the system to handle all 
tasks in time.
 
> 
>> If you think of Windows or standard Linux as your basis, it will
>> be a tough job to program your effects so that they aren't
>> delayed because the PC needs time to refresh the desktop or such
>> things.
> 
> What if I change the sound processing process to "real time"
> priority in the task manager?

Smile!
I did this in a certain application which was really critical. 
If the process isn't designed perfectly well with respect to your 
operating system, this can lead to a disaster. Don't think that 
this is easy! And it is certainly not, if you think of just 
cascading plugins or processes which weren't designed exactly for 
that.
Especially Windows is a terribly critical system, because it was 
never designed for real time applications.

> 
>> All this can be done, but keep in mind, that PCs are not designed
>> to do such work.
>>
>> Processing a lot of channels would bring most PCs to their limit:
>> only 4 PCI slots for sound cards
>> or power restrictions if you use USB cards (probably you won't be
>> able to handle many more than 4).
>>
>> So, it's not only a question of calculation power, the
>> interfacing might be your biggest issue.
>>
> 
> I have seen one PCI card that connect to 40 ADC/DAC box. It
> follows ASIO2 and has a top latency of 2ms. It sales with a 10
> ADC/DAC box, and cost about 400$.
> http://www.zzounds.com/item--STODSP2000
Fine, if the specs work for you.
If 40 is your number, why not.
If you think of pluggin 10 such PCI boards in your PC, this will 
certainly fail!
> 
> Also I have seen that CIRRUS LOGIC started selling a good stereo
> ADC/DAC for about 5$ (when you buy 10000 units) on March.
> So I guess good interfaces are soon to emarge for a low prices.

Prices ARE already low. That's not the problem.
You'll even find complete sound cards for 5...10$.
But it's still the question: are you able to connect as many as you 
need?
I don't know how many you need.

I'd write down (or better draw) the scenario, and then find a 
solution for bringing the signals into any system - with respect to 
interfaces and prices.
Then I'd do the same thing for the outputs.
After that, I'd think of what I were going to do with the signals.

After that, and not before, I'd determine, which processing system 
would fit best.

Sometimes it happens to me, that I overestimate a parameter - let's 
say, instead of 4 channels I'd assume that one might wish to handle 
64 channels - and therefore my design gets very complex.
I work and work on it but it's so difficult to solve the details.
In the end, when I recheck every parameter, I end up reducing the 
channel number to 4 again, things getting easier, and the solution 
seems almost trivial. Really, this happens again and again!

Therefore I'd propose that you double or even triple-check, if your 
design ideas are not too big. If you can set a limit to this or 
that parameter, you might end up with a better solution.

Bernhard

-- 
before sending to the above email-address:
replace deadspam.com by foerstergroup.de