Sonar beamformer with AP7001 running Linux

Go To Last Post
10 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I am interested in implementing a beamformer for Non-Destructive Testing (NDT). The beamformer is planned to have ~100 channels, and could use an AP7001 running Linux.

Normally, I've found that a number of similar applications use DSPs with clocks running at ~1 GHz. Most applications use multiple DSPs.

The first step in any DSP design flow is to write the code for the beamforming algorithm and run it on a host computer. After this is done, the algorithm is then ported to the DSP.

But can the AP700x family handle algorithms involving 3-D recomposition of a reflected acoustic wavefield? Is there any back-of-the-envelope calculations that I can do to determine if the AP700x family can handle these types of recomposition algorithms?

Would it be possible to network three of the AP700x together so that parallel processing can be performed? How would I write code in Linux to take advantage of the parallel processing? Is it possible to link all of the AP700x together on a common bus?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

My first question would be "why AVR32?". Certainly there are more suitable (and faster) DSPs for your application. AVR32 doesn't have an FPU either.

I like cats, too. Let's exchange recipes.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I think your answer is in your post. Develop the code to work on a PC, then port to a processor that can handle your code.

1. Do you need Floating point math?
2. What interface will you have with your sensors.
3. Is there alot of parallel processing that a DSP with 6-8 ALU's could be used for.

I would advise to figure out the problem at hand, then pick the processor to do the job.

I would guess that the AP series of AVR32's would have the speed needed to do a good amount of the work, however, are there hardware features in another processor that would make things even better with a lower clock rate.
Many of the DSP's by TI and Analog, have libraries writing in ASM that do alot of the 3-D matrix stuff. Does the AVR32 have these yet?

MatLab might be a good starting point if you are able to get a copy.

Caleb

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

ninevoltz9: AVR32 is good because of the design tool support and the user community (AVR Freaks). I agree that AVR32 is not the answer in all circumstances, but I'm going to give it a shot with the AP700x.

austca: That is a good idea, and I've been doing much of my coding in Matlab. There are free software libraries available for doing matrix manipulations in Linux.

The best thing to do is to develop on PC and then port to embedded. The best test of the algorithms would really be on the development kit.

Thank you for your replies!

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I doubt very much if you will get a beam-former like that running on anything other than a DSP system. ADI has details of beam-forming using their DSPs.

Leon

Leon Heller G1HSM

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Thanks, Leon. I'll take a look at the ADI chips, still keeping AVR32 in mind. Blackfin with uCLinux perhaps? I found a few app notes:

http://www.analog.com/library/an...

http://electronicdesign.com/Arti...

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I got a Blackfin board which runs uClinux when they were on special offer. It was quite cheap.

Leon

Leon Heller G1HSM

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

One of the STAMP series? Does Blackfin have a well- supported JTAG programmer for your Blackfin board? I'm looking for an inexpensive one which is similar to the "Wiggler", but doesn't require a serial port.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Yes, it's one of the STAMPs. It has the programmer on the board.

Leon

Leon Heller G1HSM

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Thanks, Leon. I'll check out the STAMP, and see if I can take a look at the schematics.