Can you share your circuit diagram, a picture of how you build it (PCB, breadboard?), and scope shots (wave form before and after the filter, and the FFT result)? And maybe your project as well.
Please let me know...
1) Did you observe side band spurious around the 25kHz peak signal? In other words, are there 20kHz, 22kHz, 24kHz, 25kHz, 26kHz, 28kHz, 30kHz spectra?
2) Or did you observe lower frequency spectra like 1kHz, 3kHz, 5kHz?
3) What is the sampling rate of the FFT and the number of samples used for the FFT?
If 1) is true, it is assumed that the PWM output amplitude is modulated by a 1kHz square wave. The possible noise source is power supply. Are there any switching devices using a power supply for the PSoC's VDDIO?
If 2) is true, it is assumed that the ground level is affected by a noise source. Please connect the ground line of the FFT probe to a ground pattern close to the PSoC.