Forums

choosing DSP for biosignal processing

Started by Adam Wilson March 8, 2004
Hello,

I am going to be designing a system for processing biomedical signals (from
neural tissue), and I am interested to hear what some people think might be
a good choice for a processor based on some of my criteria. We have a
hardware system now that works great (it uses the Sharc 21065), but we
eventually need to design a more mobile unit. Here are some of my initial
criteria:

1) it will definitely need to be able to support multiple processors,
probably at least 4.
2) It needs to have reasonably low power consumption.
3) A good amount of RAM (at least 64 MB) will be required.
4) It needs to be able to handle multiple data channels (16+)
5) It probably needs to have wireless capability for data transmission to a
computer for anaylsis.

I have worked with the ADSP Sharc 21065 before, and I found it reasonably
easy to use. I'm thinking more of looking at their TigerSHARC line of
processors as well. I am also open to using other brands such as Motorola or
TI if anyone can give me a good reason to. :-)

Thanks for any help you can give me, and let me know if I left out any other
important information, or anything else I should think of.
Thanks!

Adam


Hi Adam,

in your case I would certainly stick with the Sharc family.
Obviously it's quite adequate to the problem itself, and you have 
part of the solution already designed and implemented to that chip.

The SPORT of Sharcs is certainly a good choice if you are able to 
use its sort of daisy chaining to get your 16 channels in. 

If your hardware needs more sophisticated handling, 
integration of a tiny FPGA might be a good idea,
if you use it in combination with a number of 21065s or TigerSharcs.

Did you check the capabilities of the link ports? Might be helpful.
 
Wireless cap. is certainly an issue for an additional component.

Bernhard

Adam Wilson wrote:
> Hello, > > I am going to be designing a system for processing biomedical signals (from > neural tissue), and I am interested to hear what some people think might be > a good choice for a processor based on some of my criteria. We have a > hardware system now that works great (it uses the Sharc 21065), but we > eventually need to design a more mobile unit. Here are some of my initial > criteria: > > 1) it will definitely need to be able to support multiple processors, > probably at least 4. > 2) It needs to have reasonably low power consumption. > 3) A good amount of RAM (at least 64 MB) will be required. > 4) It needs to be able to handle multiple data channels (16+) > 5) It probably needs to have wireless capability for data transmission to a > computer for anaylsis.
You did not specify the resolution of your 16+ signals (bits and sampling rate). That would be my first question when having to specify a processor. Also, how important is stand alone operation (vs. pc hooked via USB or other umbilical)? Regards, Andor
"Andor Bariska" <andor@nospam.net> wrote in message
news:404c8803$1@pfaff2.ethz.ch...
> Adam Wilson wrote: > > Hello, > > > > I am going to be designing a system for processing biomedical signals
(from
> > neural tissue), and I am interested to hear what some people think might
be
> > a good choice for a processor based on some of my criteria. We have a > > hardware system now that works great (it uses the Sharc 21065), but we > > eventually need to design a more mobile unit. Here are some of my
initial
> > criteria: > > > > 1) it will definitely need to be able to support multiple processors, > > probably at least 4. > > 2) It needs to have reasonably low power consumption. > > 3) A good amount of RAM (at least 64 MB) will be required. > > 4) It needs to be able to handle multiple data channels (16+) > > 5) It probably needs to have wireless capability for data transmission
to a
> > computer for anaylsis. > > You did not specify the resolution of your 16+ signals (bits and > sampling rate). That would be my first question when having to specify a > processor. > > Also, how important is stand alone operation (vs. pc hooked via USB or > other umbilical)? > > Regards, > Andor >
Right now, I'd say that the resolution would be 16 bits, and sampled at around 24 kHz. The stand alone operation is pretty important; we have a system now that attaches directly to the PC, which we use to run the DSPs during experiments. The purpose of this system would be to have the same kind of functionality, but in a mobile device that a patient could carry with them. It will need to interface with a computer for programming and data analysis, etc, but it's primarily going to be a mobile device. Thanks for your feedback! Adam
Adam Wilson wrote:
...
>>>1) it will definitely need to be able to support multiple processors, >>>probably at least 4. >>>2) It needs to have reasonably low power consumption. >>>3) A good amount of RAM (at least 64 MB) will be required. >>>4) It needs to be able to handle multiple data channels (16+) >>>5) It probably needs to have wireless capability for data transmission
...
> Right now, I'd say that the resolution would be 16 bits, and sampled at > around 24 kHz.
I would suggest a Blackfin DSP for this kind of purpose. The new dual core 750 MHz version should be able to handle this kind of input data. Or perhaps you'll need a couple of those, depending on what you want to do. They are specifically designed for mobile / low-power situations. Have a look here: http://www.analog.com/processors/processors/blackfin/index.html
> The stand alone operation is pretty important; we have a > system now that attaches directly to the PC, which we use to run the DSPs > during experiments. The purpose of this system would be to have the same > kind of functionality, but in a mobile device that a patient could carry > with them. It will need to interface with a computer for programming and > data analysis, etc, but it's primarily going to be a mobile device.
Some of the high-end Blackfin DSPs are promised (that means they are not yet available) with an Ethernet port. You could use this for umbilical / wireless communication. The Blackfin developement board is also very cheap ($99) to get you started.
> Thanks for your feedback!
Welcome! Regards, Andor