Discrete Fourier Transform Stationary Noise Spectral Density

[Brainchild]

I am using a digital audio recorder (Fostex MR-8HD) which uses 24-bit 192kHz AD chips from Cirrus Logic and records to 16-bit 44.1kHz PCM files. I recorded the ambient noise of the device with all levels at minimum (shorted hot and cold pins of a balanced input for test). The image below is the spectral analysis of a normalized file of a two-minute recording.

**broken link removed**

I am completely unfamiliar with all of these issues, but have been seeking for the past two days to understand the cause of the periodic cyclical pattern in the noise. The closest I have been able to figure out is that this may be the result of phase modulation caused by either the AD converter itself or the capacitor layout in the device, but that is the closest I have come.

Any help is much appreciated as this issue has been haunting me day and night all weekend.

 

[FvM]

Unfortunately no amplitude scale is given. I understand, that you mean that the pattern is cyclic in frequency, is this correct?

 

[Brainchild]

Re: Discrete Fourier Transform Stationary Noise Spectral Den

FvM, thank you for your response. Yes, what I mean is that the pattern is cyclic in frequency.


FFT analysis of the non-normalized file using the program Audacity, the amplitudes are given as follows (with Hann windowing):

2048 bins: troughs -126dBFS; peaks -124.
4096 bins: troughs -129dBFS; peaks -127.
8192 bins: troughs -132dBFS; peaks -130.

 

[Brainchild]

Re: Discrete Fourier Transform Stationary Noise Spectral Den

Well as far as I have been able to determine what I am seeing is passband ripple resulting from signal convolution in the windowing process of the digital filter.