EmbeddedRelated.com
Forums

Help.....DSP on ARM9 or ARM7!!!!

Started by jagdish.dsp January 30, 2006
Dear Friends,

In my design I need to develope a DSP algorithm on one DSP processor and
then to interface same with ARM which is handling all the peripherals.

I am thinking if my DSP algorithm can be implemented on ARM or not...so
need you help. Right now I am using evaluation board of Texas DSK, so
taking the input from microphone is given to CODEC(A to D conversion)
whose sampling freq is 32KHz.

I have to implement 8192 point FFT on real time sound data....
And also want to implement Low pass FIR filter on Real time sound data
before doing FFT. So can this be implemented on ARM 9 platform?????? 
while pareallaly I want to interface one modem, IR receiver,key board, and
external memory with ARM. If possible please let me know which will be the
best ARM9 based chip to implement the same. 

So is it possible to implement all this on ARM7 or ARM9 platform so with
one processor only I will have my solution.
What chip should I use for A/D conversion with good sampling of sound.

Please help me.......
Thanks and Regards,

Jagdish.





jagdish.dsp wrote:
> Dear Friends, > > In my design I need to develope a DSP algorithm on one DSP processor and > then to interface same with ARM which is handling all the peripherals. > > I am thinking if my DSP algorithm can be implemented on ARM or not...so > need you help. Right now I am using evaluation board of Texas DSK, so > taking the input from microphone is given to CODEC(A to D conversion) > whose sampling freq is 32KHz. > > I have to implement 8192 point FFT on real time sound data....
You need 16 bits or 2x12bits A2D. I don't think there is 13 bits A2D.
> And also want to implement Low pass FIR filter on Real time sound data > before doing FFT. So can this be implemented on ARM 9 platform??????
You need over 1GHz ARM, preferrable with muti-media instructions.
> while pareallaly I want to interface one modem, IR receiver,key board, and > external memory with ARM. If possible please let me know which will be the > best ARM9 based chip to implement the same. > > So is it possible to implement all this on ARM7 or ARM9 platform so with > one processor only I will have my solution.
What's wrong with DSP?
> What chip should I use for A/D conversion with good sampling of sound. > > Please help me....... > Thanks and Regards, > > Jagdish.
I agree as well, a 8192 point FFT in realtime is going to need a lot of
horsepower. It is probably best to leave it to the DSP and have the
ARM7/9 control the DSP.

"jagdish.dsp" <jagdish.dsp@gmail.com> wrote in message 
news:uuSdnROkSq5X0UPeRVn-uQ@giganews.com...
> Dear Friends, > > In my design I need to develope a DSP algorithm on one DSP processor and > then to interface same with ARM which is handling all the peripherals. > > I am thinking if my DSP algorithm can be implemented on ARM or not...so > need you help. Right now I am using evaluation board of Texas DSK, so > taking the input from microphone is given to CODEC(A to D conversion) > whose sampling freq is 32KHz. > > I have to implement 8192 point FFT on real time sound data....
An ARM7 takes 2Mhz to perform a 8192 complex Q15 FFT. ARM9E is twice as fast.
> And also want to implement Low pass FIR filter on Real time sound data > before doing FFT. So can this be implemented on ARM 9 platform??????
Your requirements add up to around 10Mhz on an ARM7. Unless power consumption is key, no need for a DSP. Wilco
Wilco Dijkstra wrote:
> "jagdish.dsp" <jagdish.dsp@gmail.com> wrote in message > news:uuSdnROkSq5X0UPeRVn-uQ@giganews.com... > > Dear Friends, > > > > In my design I need to develope a DSP algorithm on one DSP processor and > > then to interface same with ARM which is handling all the peripherals. > > > > I am thinking if my DSP algorithm can be implemented on ARM or not...so > > need you help. Right now I am using evaluation board of Texas DSK, so > > taking the input from microphone is given to CODEC(A to D conversion) > > whose sampling freq is 32KHz. > > > > I have to implement 8192 point FFT on real time sound data.... > > An ARM7 takes 2Mhz to perform a 8192 complex Q15 FFT. ARM9E is twice as > fast.
That's for doing one per second, which is unlikely for any reasonable algorithm.
> > > And also want to implement Low pass FIR filter on Real time sound data > > before doing FFT. So can this be implemented on ARM 9 platform?????? > > Your requirements add up to around 10Mhz on an ARM7. Unless power > consumption is key, no need for a DSP.
DSPs are designed exactly for these things.
> > Wilco
"Wilco Dijkstra" <Wilco_dot_Dijkstra@ntlworld.com> wrote in message
news:omwDf.18838$mf2.1624@newsfe6-win.ntli.net...
> > "jagdish.dsp" <jagdish.dsp@gmail.com> wrote in message > news:uuSdnROkSq5X0UPeRVn-uQ@giganews.com... > > Dear Friends, > > > > In my design I need to develope a DSP algorithm on one DSP processor and > > then to interface same with ARM which is handling all the peripherals. > > > > I am thinking if my DSP algorithm can be implemented on ARM or not...so > > need you help. Right now I am using evaluation board of Texas DSK, so > > taking the input from microphone is given to CODEC(A to D conversion) > > whose sampling freq is 32KHz. > > > > I have to implement 8192 point FFT on real time sound data.... > > An ARM7 takes 2Mhz to perform a 8192 complex Q15 FFT. ARM9E is twice as > fast. > > > And also want to implement Low pass FIR filter on Real time sound data > > before doing FFT. So can this be implemented on ARM 9 platform?????? > > Your requirements add up to around 10Mhz on an ARM7. Unless power > consumption is key, no need for a DSP. > > Wilco
In support of Wilco Dijkstra argument: The FFT is O( n log2(n) ). As a quick estimate, at fs = 32KHz that's around 400.000 butterflies per second if no overlap is needed (please corect me if I'm wrong). That would not be a big problem to do on the ARM then. The OP does not detail the FIR requirements, but with some luck the system may allow the FIR to simply be a scaling of the FFT bins - in which case the FIR comes almost for free. DJ --
"linnix" <me@linnix.info-for.us> wrote in message 
news:1138664199.493845.11670@g43g2000cwa.googlegroups.com...
> > Wilco Dijkstra wrote: >> "jagdish.dsp" <jagdish.dsp@gmail.com> wrote in message >> news:uuSdnROkSq5X0UPeRVn-uQ@giganews.com...
>> > I have to implement 8192 point FFT on real time sound data.... >> >> An ARM7 takes 2Mhz to perform a 8192 complex Q15 FFT. ARM9E is twice as >> fast. > > That's for doing one per second, which is unlikely for any reasonable > algorithm.
Without more detailed information it is reasonable to assume only one FFT would be needed for each set of 8192 samples, ie. only 3.9 needed per second. ARM7 goes up to 120Mhz, so it can handle around 50 complex 8192-point FFTs per second. In another post you wrote:
>You need over 1GHz ARM, preferrable with muti-media instructions.
The fastest ARM (Cortex-A8) is about as fast as the TMS320C64x (TI's fastest DSP). Exactly which audio codec needs _that_ much DSP power? Wilco
Hello,

I strongly disagree with (some of) the other posters:

- That processing won't need THAT much computational power as someone
mentioned (1GHz+ ARM9).
- Either case, DSPs are made for that sort of thing.

But, what if you need control processing, besides number crunching?
Here, a much more compact and flexible solution is desired.

I strongly recommend you to have a look at Analog Devices' Blackfin
family of processors. They have both computational (800+ MMACs) and
control capabilities, as well as nice set of peripherals: SDRAM
controller, USB, Ehternet, CAN, UARTs, Timers, RTC, core voltage
regulators, PPI, etc.

What you mention is exactly the kind of job that Blackfin was designed
for!

Regards,

JaaC

jaac wrote:
> > I strongly disagree with (some of) the other posters: > > - That processing won't need THAT much computational power as > someone mentioned (1GHz+ ARM9). > - Either case, DSPs are made for that sort of thing. > > But, what if you need control processing, besides number crunching? > Here, a much more compact and flexible solution is desired. > > I strongly recommend you to have a look at Analog Devices' Blackfin > family of processors. They have both computational (800+ MMACs) and > control capabilities, as well as nice set of peripherals: SDRAM > controller, USB, Ehternet, CAN, UARTs, Timers, RTC, core voltage > regulators, PPI, etc. > > What you mention is exactly the kind of job that Blackfin was > designed for!
Please include adequate context in your posts. This can be done even from the broken google usenet interface. If you want to post a followup via groups.google.com, don't use the broken "Reply" link at the bottom of the article. Click on "show options" at the top of the article, then click on the "Reply" at the bottom of the article headers. More details at: <http://cfaj.freeshell.org/google/> Also see <http://www.safalra.com/special/googlegroupsreply/> -- "The power of the Executive to cast a man into prison without formulating any charge known to the law, and particularly to deny him the judgement of his peers, is in the highest degree odious and is the foundation of all totalitarian government whether Nazi or Communist." -- W. Churchill, Nov 21, 1943
"CBFalconer" <cbfalconer@yahoo.com> wrote in message 
news:43DFD395.AE7D79C0@yahoo.com...
> jaac wrote: >> >> I strongly disagree with (some of) the other posters: >> >> - That processing won't need THAT much computational power as >> someone mentioned (1GHz+ ARM9). >> - Either case, DSPs are made for that sort of thing. >> >> But, what if you need control processing, besides number crunching? >> Here, a much more compact and flexible solution is desired. >> >> I strongly recommend you to have a look at Analog Devices' Blackfin >> family of processors. They have both computational (800+ MMACs) and >> control capabilities, as well as nice set of peripherals: SDRAM >> controller, USB, Ehternet, CAN, UARTs, Timers, RTC, core voltage >> regulators, PPI, etc. >> >> What you mention is exactly the kind of job that Blackfin was >> designed for! > > Please include adequate context in your posts. This can be done > even from the broken google usenet interface. > > If you want to post a followup via groups.google.com, don't use the > broken "Reply" link at the bottom of the article. Click on "show > options" at the top of the article, then click on the "Reply" at > the bottom of the article headers. > > More details at: <http://cfaj.freeshell.org/google/> > Also see <http://www.safalra.com/special/googlegroupsreply/>
Hello, Thanks for the reminder. I am the same poster (jaac). Getting used to google, I forgot that option. Will keep it in my mind for later posts. Regards, JaaC