I am trying to use the KISS FFT library found on www.sourceforge.net with the TMS320F2812 DSP from Texas Instruments. After a bit of work, I have been able to compile the code and even run a couple of small-scale FFT calculations, but am having trouble with larger FFT sizes. I have a program set up to trigger the ADC periodically and commence the FFT calculation after enough samples have been collected. The interrupt routine that runs whenever an ADC conversion completes causes a light on the ezDSP board I'm using to toggle on and off, to indicate that the processor is working. When I run FFTs with small (less than 100) sample sizes, everything works fine. However, when I use sample sizes above this, the light no longer blinks and the FFT calculation never seems to finish. I have been looking through the source code for KISS FFT, but haven't been able to find anything. Does anyone have any suggestions? -Brad
Using KISS FFT with Ti's TMS320F2812 DSP
Started by ●October 11, 2004
Reply by ●October 11, 20042004-10-11
By default the watchdog will be enabled at startup. When computing larger FFTs you probably don't have enough time to complete the processing before the watchdog resets the board. Brad "BradK" <b_kartchner@yahoo.com> wrote in message news:743175b4.0410111602.76069aa1@posting.google.com...>I am trying to use the KISS FFT library found on www.sourceforge.net > with the TMS320F2812 DSP from Texas Instruments. After a bit of work, > I have been able to compile the code and even run a couple of > small-scale FFT calculations, but am having trouble with larger FFT > sizes. > > I have a program set up to trigger the ADC periodically and commence > the FFT calculation after enough samples have been collected. The > interrupt routine that runs whenever an ADC conversion completes > causes a light on the ezDSP board I'm using to toggle on and off, to > indicate that the processor is working. > > When I run FFTs with small (less than 100) sample sizes, everything > works fine. However, when I use sample sizes above this, the light no > longer blinks and the FFT calculation never seems to finish. > > I have been looking through the source code for KISS FFT, but haven't > been able to find anything. > > Does anyone have any suggestions? > > -Brad
Reply by ●October 13, 20042004-10-13
Thanks for the suggestion. I hadn't thought about the watchdog, but it turns out that is was disabled anyway. I discovered that the problem was that I didn't have enough memory on or off the chip to store the FFT samples and results (If I had only checked the value returned by the far_malloc function...). So, I'll be adding more memory. Thanks for the reply! -Brad "Brad Griffis" <bradgriffis@hotmail.com> wrote in message news:<mkGad.8749$Rf1.8204@newssvr19.news.prodigy.com>...> By default the watchdog will be enabled at startup. When computing larger > FFTs you probably don't have enough time to complete the processing before > the watchdog resets the board. > > Brad > > "BradK" <b_kartchner@yahoo.com> wrote in message > news:743175b4.0410111602.76069aa1@posting.google.com... > >I am trying to use the KISS FFT library found on www.sourceforge.net > > with the TMS320F2812 DSP from Texas Instruments. After a bit of work, > > I have been able to compile the code and even run a couple of > > small-scale FFT calculations, but am having trouble with larger FFT > > sizes. > > > > I have a program set up to trigger the ADC periodically and commence > > the FFT calculation after enough samples have been collected. The > > interrupt routine that runs whenever an ADC conversion completes > > causes a light on the ezDSP board I'm using to toggle on and off, to > > indicate that the processor is working. > > > > When I run FFTs with small (less than 100) sample sizes, everything > > works fine. However, when I use sample sizes above this, the light no > > longer blinks and the FFT calculation never seems to finish. > > > > I have been looking through the source code for KISS FFT, but haven't > > been able to find anything. > > > > Does anyone have any suggestions? > > > > -Brad