[benjamincburns]

>You've been taking the wrong classes :P

I soooo agree with this.

"Twiddle factor? What the..."

You might scratch the surface of DSP by taking a Matlab-centric course, but if you'd like to do it as a career you're far better off to go in and implement some of the more fundamental algorithms you're using.

Implementing FFT or DCT on actual hardware (which supports a multiply/accumulate instruction over a true DSP memory architecture) will give you the tools that you need when it comes time to troubleshoot (read: use) some vendor's shitty proprietary DSP library.

That said, learn the Matlab stuff too. It's endlessly useful for prototyping, and it's pretty much industry standard in this use. Just don't take it for granted.

[aylons]

For prototyping, I go with SciPy. So I can deploy at least a demonstration without the need to reprogram the algorithm in a new language.

Specially useful when the algorithm is actually a tool which can be used several times, changing the parameters. I put it on a web server and share it wit my colleagues.

[choochootrain]

how do you recommend getting into this sort of stuff? i have some basic experience with microcontrollers and i'm interested in some dsp hardware projects but i'm not sure where to start as the classes offered do not touch hardware..

[stinos]

the only real way is to get hardware. If you have some experience already, you shouldn't need a course: I can't find the current prices immdeiately, but a couple of years ago something like 250$ got you a fully functional development board (including some analog I/O) + corresponding IDE + signal processing example code of a true DSP like the TI C6000 series. Example of a current version would be http://www.ti.com/tool/tmdxevm6670