This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

What next with this audio codec project? Thank you!!!

Other Parts Discussed in Thread: TMS320F28335

We are working on an audio codec project with very limited experiences. So far we have implemented the algo in Matlab and ported it as a vs 2005 float-point version. We need to figure out what to do next.

All we know is a fixed-point chip would be adequate for our purpose.

We are planning to

1) convert the float-point into fixed-point C; THEN

2) estimate the computation complexity and storage need; THEN

3) choose the right chip (probably from TI)

4) we don't know

 

The questions are:

a) Does this make sense?

b) We don't know what tools are available for step 2, i.e. estimating the complexity mostly?

c) After we are done selecting chip, what should we do? Contact TI for samples and development kit, and then port the c code?

 

Thank you in advance for any advice!!!

  • Generally when selecting a device the key care-abouts in no particular order are:

    • Performance
    • Price
    • Power
    • Peripherals
    • Size

    What requirements do you have for your project?  For example, since you want to use an audio codec then you'll need 1 serial port to connect to it.  What other peripherals will you need?  Do you have major cost constraints?  For example, would you consider a $17 processor cheap or expensive?  What kind of user interface do you need?  Just a few push-buttons, or fancy 3D graphics?

    Do you have engineers with a lot of experience in fixed-point math?  If not, I suggest you consider a floating point processor like the brand new C6747:

    http://focus.ti.com/docs/prod/folders/print/tms320c6747.html

    It costs under $17 in 100u quantities and supports both fixed-point and floating-point math.  Eliminating "step 1" from your development schedule would save you tremendous amounts of time and money.  In terms of estimating complexity the best thing to do would be to implement the algorithm on an eval board and then measure the number of cycles it takes.  There's hardware inside the core to allow you to easily measure the number of cycles an algorithm takes.

    You can order samples directly from the product page.  This order would be fulfilled by one of TI's distributors.

    Brad

  • Matlab does support the TMS320F28335 floating point device in a more recent release.

    Maybe you could contact them about getting an evaluation copy for your estimation purposes.

    http://www.mathworks.de/matlabcentral/newsreader/view_thread/165968

    You also might want to refer to the below for some benchmarks for simple functions implemented in floating point.

    http://focus.ti.com/docs/toolsw/folders/print/sprc664.html