TMS320C6474 FFT Problems

I am trying to get the DSPLIB FFTs running on my 6474 processor. I am using CC 3.3 and newest DSPLIB optimized for the C64 architecture. I started with DSP_fft using 16 bit input (using synthetic data) and output. My FFT size is 1024, and I am getting large spikes on the output bins at multiples of 128 no matter how I scale the twiddles, and I never seem to be able to get my signal out cleanly. I left my signal at a low amplitude (~ max value of 5-32, on various experiments) and scaled the twiddles through the entire range of their values. The FFT output is quite noisy, with a lot of noise at DC. I have verified that my memory allocations are correct and get the same results if I pass it real or complex data on the input. I also tried moving up to the DSP_fft32x32 and got the same results.  Any ideas on scaling or using these FFTs? I will have a 14 bit A/D bringing in data, so I need a large dynamic range out of this FFT.