I am designing a monitoring system in which the sound of an industrial motor should be analyzed to tell the condition of the motor. For this application I need to use FFT. So I chose DSP as the capable processor. Now I need help in selecting a cost effective DSP. Any kind of help would be appreciated.

My questions;

1. How do I decide how many points will be needed for FFT?

2. Floating point or fixed point?

3. Number of bits?

4. I need to serve 10 channels

5. ADC/DAC separately or embedded in DSP? which would be a cost effective choice?

Thanks for your time.

Hi,

The answers to your questions really depend on the specifics of the application. Here are some general answers to get you started:

1. This really depends on what you are using the FFT for. Larger sizes lead to better performance in terms of resolution and accuracy. Smaller sizes reduce computation and storage requirements.

2. This is usually a tradeoff between cost/power consumption and accuracy/dynamic range. You may find a high-level discussion about this point here: http://www.analog.com/en/embedded-processing-dsp/content/Fixed-Point_vs_Floating-Point_DSP/fca.html

3. Depends on your application and on accuracy/resolution requirements.

4 & 5. An integrated option is usually more cost-effective. An example of such option is the BF506F product, which has a Blackfin fixed-point processor, a 12-channel 12-bit ADC, and an integrated FLASH.

Hope this helps.

Regards,

Wassim