Need help for DSP interfacing (newbie)

Hi,

I am collaborating with my friend to create a custom-made amplifier; we
are going to add a programmable effects feature to it. Even if my
friend is enough experienced with "normal" processors, we know very
little about the DSP world (most theorical knowledge, not practical).

I have a few questions whose answers should allow us to finish our
design stage:

1) Do modern DSPs have ADC and DAC integrated? I suppose *yes*
2) Do modern DSPs have an internal MCU?
3) What protocols are usually available to link the DSP to my main CPU?
4) What do you suggest me to read about interfacing a DSP to a "normal"
CPU?

We have already made two amplifiers without DSPs so what we need now is
some knowledge to "fill the gap", even if I am sure we can't really get
it without experience.

Thank you
--
  Daniele C.

Daniele C. wrote:
> Hi,
> 
> I am collaborating with my friend to create a custom-made amplifier; we
> are going to add a programmable effects feature to it. Even if my
> friend is enough experienced with "normal" processors, we know very
> little about the DSP world (most theorical knowledge, not practical).
> 
> I have a few questions whose answers should allow us to finish our
> design stage:
> 
> 1) Do modern DSPs have ADC and DAC integrated? I suppose *yes*
> 2) Do modern DSPs have an internal MCU?
> 3) What protocols are usually available to link the DSP to my main CPU?
> 4) What do you suggest me to read about interfacing a DSP to a "normal"
> CPU?
> 
> We have already made two amplifiers without DSPs so what we need now is
> some knowledge to "fill the gap", even if I am sure we can't really get
> it without experience.

I'd like to suggest, if it doesn't discourage you, that you consult here 
not only about how to implement an architecture, but also about how to 
choose one. I deduce from your question about D/A and A/D that the 
amplifiers you built are analog, but from your questions about 
interfacing a typical DSP to another processor that there is a computer 
involved somehow. Can you explain?

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

Daniele C. wrote:

> Hi,
> 
> I am collaborating with my friend to create a custom-made amplifier; we
> are going to add a programmable effects feature to it

What effects exactly are you trying to implement?


> Even if my
> friend is enough experienced with "normal" processors, we know very
> little about the DSP world (most theorical knowledge, not practical).
> 
> I have a few questions whose answers should allow us to finish our
> design stage:
> 
> 1) Do modern DSPs have ADC and DAC integrated? I suppose *yes*
> 2) Do modern DSPs have an internal MCU?
> 3) What protocols are usually available to link the DSP to my main CPU?
> 4) What do you suggest me to read about interfacing a DSP to a "normal"
> CPU?

Judging on your extremely naive questions, I can suggest you one of the 
two options:

* Buying a DSP board such as eZDSP 55xx or BlackFin EZ Kit lite, or All 
Clark's SHARC board.

* Using all-in-one DAP such as TI TAS3004.


> We have already made two amplifiers without DSPs so what we need now is
> some knowledge to "fill the gap", even if I am sure we can't really get
> it without experience.

I see. You have one advantage compared to the others: your experience 
grows at every minute :)


Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com

Daniele C. wrote:

> Hi,
>
> I am collaborating with my friend to create a custom-made amplifier; we
> are going to add a programmable effects feature to it. Even if my
> friend is enough experienced with "normal" processors, we know very
> little about the DSP world (most theorical knowledge, not practical).
>
> I have a few questions whose answers should allow us to finish our
> design stage:


>
> 1) Do modern DSPs have ADC and DAC integrated? I suppose *yes*

On occasion. Mostly, interestingly, no. [not surprising for a number of
reasons I won't go into now]

> 2) Do modern DSPs have an internal MCU?
By definition they have a processor that might be called an MCU. Do you
mean 'do they have a separate MCU'? in which case the answer is
'rarely'.

> 3) What protocols are usually available to link the DSP to my main CPU?
The standard ones are almost always available; SPI, I2C, UARTS. Shared
memory (via BRQ, BGNT) may or may not be available. What do you need to
connect and at what bandwidth?

> 4) What do you suggest me to read about interfacing a DSP to a "normal"
> CPU?

A DSP is an ordinary CPU with a different instruction set. The devices
reflect this, of course, in the IO pins available. As there is no
'standard' DSP CPU just as there is no 'standard' CPU it would help if
you were more forthcoming on details :)

As noted above the amount of data you need to transfer or share is of
paramount inportance.


>
> We have already made two amplifiers without DSPs so what we need now is
> some knowledge to "fill the gap", even if I am sure we can't really get
> it without experience.

This shows you have some wisdom :)

Cheers

PeteS


> 
> Thank you
> --
>   Daniele C.

Yes, the amplifiers are analog. Sorry for my *bad* questions...that's
probably because I miss experience. The fact is that the other guy is
the electronic engineer but cannot speak english while I am just the
programmer...so even if I basically understand how things work I can
rarely find the correct term to describe them to other people (and with
no success at all, when other people have experience as you have).

There will be a microcomputer with a listening software that will catch
(via IrDA) amplifier commands and programs that will use the DSP
processor to make effects. This is mostly a sperimental project for
ourselves.

Thanks.

--
  Daniele

Jerry Avins ha scritto:

> Daniele C. wrote:
> > Hi,
> >
> > I am collaborating with my friend to create a custom-made amplifier; we
> > are going to add a programmable effects feature to it. Even if my
> > friend is enough experienced with "normal" processors, we know very
> > little about the DSP world (most theorical knowledge, not practical).
> >
> > I have a few questions whose answers should allow us to finish our
> > design stage:
> >
> > 1) Do modern DSPs have ADC and DAC integrated? I suppose *yes*
> > 2) Do modern DSPs have an internal MCU?
> > 3) What protocols are usually available to link the DSP to my main CPU?
> > 4) What do you suggest me to read about interfacing a DSP to a "normal"
> > CPU?
> >
> > We have already made two amplifiers without DSPs so what we need now is
> > some knowledge to "fill the gap", even if I am sure we can't really get
> > it without experience.
>
> I'd like to suggest, if it doesn't discourage you, that you consult here
> not only about how to implement an architecture, but also about how to
> choose one. I deduce from your question about D/A and A/D that the
> amplifiers you built are analog, but from your questions about
> interfacing a typical DSP to another processor that there is a computer
> involved somehow. Can you explain?
>
> Jerry
> --
> Engineering is the art of making what you want from things you can get.
> =AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=
=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=
=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF

PeteS ha scritto:

> Daniele C. wrote:
>
> > Hi,
> >
> > I am collaborating with my friend to create a custom-made amplifier; we
> > are going to add a programmable effects feature to it. Even if my
> > friend is enough experienced with "normal" processors, we know very
> > little about the DSP world (most theorical knowledge, not practical).
> >
> > I have a few questions whose answers should allow us to finish our
> > design stage:
>
>
> >
> > 1) Do modern DSPs have ADC and DAC integrated? I suppose *yes*
>
> On occasion. Mostly, interestingly, no. [not surprising for a number of
> reasons I won't go into now]

Ok, thanks. Will further investigate.

>
> > 2) Do modern DSPs have an internal MCU?
> By definition they have a processor that might be called an MCU. Do you
> mean 'do they have a separate MCU'? in which case the answer is
> 'rarely'.
>

Yes I meant a /common/ MCU whose task would have been to /drive/ the
DSP, but as the below answer clarifies there's no need for such an
unit.

> > 3) What protocols are usually available to link the DSP to my main CPU?
> The standard ones are almost always available; SPI, I2C, UARTS. Shared
> memory (via BRQ, BGNT) may or may not be available. What do you need to
> connect and at what bandwidth?
>
Uhm...I cannot yet answer because we have not choosen the main CPU; I
think we will use I2C because other peripherals already use that.

> > 4) What do you suggest me to read about interfacing a DSP to a "normal"
> > CPU?
>
> A DSP is an ordinary CPU with a different instruction set. The devices
> reflect this, of course, in the IO pins available. As there is no
> 'standard' DSP CPU just as there is no 'standard' CPU it would help if
> you were more forthcoming on details :)

Well...I will have to read something about having the DSP run a program
loaded by the CPU. As you would probably say, Google's my friend :)

>
> As noted above the amount of data you need to transfer or share is of
> paramount inportance.
>

We will handle 2 channels (left+right) at the same time, they will be
converted to digital and then converted back to analog after DSP
processing. The amount of data will depend from the sampling rate we
are going to use (correct?).

>
> >
> > We have already made two amplifiers without DSPs so what we need now is
> > some knowledge to "fill the gap", even if I am sure we can't really get
> > it without experience.
>
> This shows you have some wisdom :)

Yeees, a drop of water in the ocean!

>
> Cheers
>
> PeteS

Many thanks!
--
  Daniele

> 
> 
> > 
> > Thank you
> > --
> >   Daniele C.

Vladimir Vassilevsky ha scritto:

> Daniele C. wrote:
>
> > Hi,
> >
> > I am collaborating with my friend to create a custom-made amplifier; we
> > are going to add a programmable effects feature to it
>
> What effects exactly are you trying to implement?
>

Real-time (not really, but with a delay of 0.5 s ca. I suppose) DSP
effects, the program would be transferred via IrDA to the DSP (passing
through the MCU first).

>
> > Even if my
> > friend is enough experienced with "normal" processors, we know very
> > little about the DSP world (most theorical knowledge, not practical).
> >
> > I have a few questions whose answers should allow us to finish our
> > design stage:
> >
> > 1) Do modern DSPs have ADC and DAC integrated? I suppose *yes*
> > 2) Do modern DSPs have an internal MCU?
> > 3) What protocols are usually available to link the DSP to my main CPU?
> > 4) What do you suggest me to read about interfacing a DSP to a "normal"
> > CPU?
>
> Judging on your extremely naive questions, I can suggest you one of the
> two options:
>
> * Buying a DSP board such as eZDSP 55xx or BlackFin EZ Kit lite, or All
> Clark's SHARC board.
>
Hey, that would steal us all the fun! :)

> * Using all-in-one DAP such as TI TAS3004.
>

Uhm...I have already heard that, I can't remember the (technical)
reason why we chose not to use it.

>
> > We have already made two amplifiers without DSPs so what we need now is
> > some knowledge to "fill the gap", even if I am sure we can't really get
> > it without experience.
>
> I see. You have one advantage compared to the others: your experience
> grows at every minute :)
>

Yes, that's why we started this experimental project :)

>
> Vladimir Vassilevsky
>
> DSP and Mixed Signal Design Consultant
> 
> http://www.abvolt.com

Thank you!!!
--
  Daniele

Daniele C. wrote:
> Yes, the amplifiers are analog. Sorry for my *bad* questions...that's
> probably because I miss experience. The fact is that the other guy is
> the electronic engineer but cannot speak english while I am just the
> programmer...so even if I basically understand how things work I can
> rarely find the correct term to describe them to other people (and with
> no success at all, when other people have experience as you have).
> 
> There will be a microcomputer with a listening software that will catch
> (via IrDA) amplifier commands and programs that will use the DSP
> processor to make effects. This is mostly a sperimental project for
> ourselves.

I appreciate your difficulty. Let me assure that we aren't too proud 
help you in your own terms, but we need to know what you want to do. 
Also, in order to communicate with us, you will need to learn a little 
of our language.

As a programmer, you undoubtedly know what a flow chart is. For other 
systems, we call those block diagrams, and we use fewer stylized symbols 
and maybe more words. I suggest that you think through the system you 
want to build well enough so that you can draw a block diagram to 
represent it. Then we can get down to the business of filling in the 
digital blocks.

DSP chips have special hardware to expedite the execution of common 
signal processing algorithms, and they tend to use less power than 
general-purpose CPUs of the same clock speed. There is nothing they can 
do that other processors can't if speed and power are ignored. Consider 
the software synthesizers that run on desktop computers: one could be 
proud to build something as good. If a desktop computer will be part of 
the system you envision, it's likely that you'll need nothing more. 
(Your sound card has the A/D and D/A you need.)

There are several freeware sound manipulation programs available for 
download. I don't use any enough to be able to compare, so I'll leave 
recommendations to others. It will help them to know what you want to 
accomplish.

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

Jerry Avins ha scritto:

> Daniele C. wrote:
> > Yes, the amplifiers are analog. Sorry for my *bad* questions...that's
> > probably because I miss experience. The fact is that the other guy is
> > the electronic engineer but cannot speak english while I am just the
> > programmer...so even if I basically understand how things work I can
> > rarely find the correct term to describe them to other people (and with
> > no success at all, when other people have experience as you have).
> >
> > There will be a microcomputer with a listening software that will catch
> > (via IrDA) amplifier commands and programs that will use the DSP
> > processor to make effects. This is mostly a sperimental project for
> > ourselves.
>
> I appreciate your difficulty. Let me assure that we aren't too proud
> help you in your own terms, but we need to know what you want to do.
> Also, in order to communicate with us, you will need to learn a little
> of our language.
>
> As a programmer, you undoubtedly know what a flow chart is. For other
> systems, we call those block diagrams, and we use fewer stylized symbols
> and maybe more words. I suggest that you think through the system you
> want to build well enough so that you can draw a block diagram to
> represent it. Then we can get down to the business of filling in the
> digital blocks.

The engineer is already on that...unfortunately I cannot post an URL to
the work-in-progress diagram for now.

>
> DSP chips have special hardware to expedite the execution of common
> signal processing algorithms, and they tend to use less power than
> general-purpose CPUs of the same clock speed. There is nothing they can
> do that other processors can't if speed and power are ignored.

Their biggest difference is (for me) the parallel execution feature;
when programming them (I haven't yet) I will consider the circular
buffers too.

> Consider
> the software synthesizers that run on desktop computers: one could be
> proud to build something as good. If a desktop computer will be part of
> the system you envision, it's likely that you'll need nothing more.
> (Your sound card has the A/D and D/A you need.)
>

Uhm...there will be no desktop computer in this system; this system
will be a "normal" amplifier with an infrared remote control. The same
infrared port will be used to interface with a  PC and transfer small
programs into the device. Once finished the program transfer, those
little programs will exist on a memory chip inside the amplifier and
ready to run on user's selection. Actually, each program will generate
a DSP effect.

> There are several freeware sound manipulation programs available for
> download. I don't use any enough to be able to compare, so I'll leave
> recommendations to others. It will help them to know what you want to
> accomplish.

I have used (professionally) SoundForge in the past, and I think it is
still one of the best (now it has been bought by Sony if I remember
correctly); this project is more a training exercise rather than a
professional attempt to create something. For example, we should have
had to use switching-mode power supply if we were going to make
something better...

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

Thank you Jerry! You have been very kind and explanatory, and your
patience is great too! I know how bad sound technical terms written by
me :)

--
  Daniele

Daniele C. wrote:
> Jerry Avins ha scritto:

   ...

>>  I suggest that you think through the system you
>> want to build well enough so that you can draw a block diagram to
>> represent it. Then we can get down to the business of filling in the
>> digital blocks.
> 
> The engineer is already on that...unfortunately I cannot post an URL to
> the work-in-progress diagram for now.
> 
>> DSP chips have special hardware to expedite the execution of common
>> signal processing algorithms, and they tend to use less power than
>> general-purpose CPUs of the same clock speed. There is nothing they can
>> do that other processors can't if speed and power are ignored.
> 
> Their biggest difference is (for me) the parallel execution feature;
> when programming them (I haven't yet) I will consider the circular
> buffers too.

Parallel execution need not be a conceptual hurdle. Consider the MAC 
instruction. It fetches two values from locations specified by pointers 
in registers and then increments those registers. It also multiplies 
them and adds the product to an accumulator. Whether those actions are 
performed sequentially or in a single machine cycle is merely a matter 
of how fast is happens and how many instructions are needed to specify it.

>> Consider
>> the software synthesizers that run on desktop computers: one could be
>> proud to build something as good. If a desktop computer will be part of
>> the system you envision, it's likely that you'll need nothing more.
>> (Your sound card has the A/D and D/A you need.)
>>
> 
> Uhm...there will be no desktop computer in this system; this system
> will be a "normal" amplifier with an infrared remote control. The same
> infrared port will be used to interface with a  PC and transfer small
> programs into the device. Once finished the program transfer, those
> little programs will exist on a memory chip inside the amplifier and
> ready to run on user's selection. Actually, each program will generate
> a DSP effect.

If the programs are not computed during a performance, why not have the 
programs all permanently resident where they will be used. Memory is 
cheap. People throw away the little 16-Mbyte cards that come in Digital 
cameras to make room for Gbyte cards. Don't think too small.

   ...

> Thank you Jerry! You have been very kind and explanatory, and your
> patience is great too! I know how bad sound technical terms written by
> me :)

You're certainly welcome. Technical terms facilitate communication. We 
don't seem to suffer for their lack. :-)

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