Equation 17: en = [An -An-1] tjn ts This analysis is similar to Blesser's approach in examining slew limiting in DAC output stages.12 The frequency domain representation E100%(f) of this error sequence can be obtained by taking the discrete Fourier transform of en: Equation 18:
E100%
(f) =F(en)
where: F is the discrete Fourier operator;
|
signal/jitter sum and difference frequencies; this can be confirmed by simulating a jitter error signal using the model. Fig.24a shows the unjittered spectrum of a 16-bit-quantized, 0dBFS, 10,001Hz audio signal, where the choice of excitation frequency has led to a flat quantization noisefloor even though no dither has been used. Fig.24b shows the same signal, but now corrupted by a lkHz jitter signal of peak amplitude 10ns. As expected, the error components in the jittered spectrum occur as sidebands at 10kHz +- 1kHz at approximately -71dBFS. The An - An-1 factor in Equation 18 makes the magnitude of the error spectrum roughly proportional to the frequency as well as the amplitude of (he audio signal. This is illustrated in the simulated error spectrum of fig.24c for a 0dBFS, 2001Hz sinusoid, again jittered at 1kHz; due to the fivefold drop in excitation frequency, the sidebands are now approximately 14dB lower than the 10kHz example.
|
to previous page Stereophile, March 1996 to next page |