Why does dithering sacrifice "effective sampling frequency"? You're just adding extremely small amounts of white noise to reduce quantization distortion or (in more advanced cases) noise with a power spectrum that puts the power of the noise mainly in parts of the audio spectrum that humans hear poorly.
This is not that (or it's an extreme special case of that); I'm talking about the thing the grandparent link is suggesting, representing amplitudes below 1 by having the sample sometimes be 0 and sometimes be 1. If you represent a waveform that should be 0.5 0.5 0.5 0.5 0 0 0 0 -0.5 -0.5 -0.5 -0.5 0 0 0 0 by doing 1 0 1 0 0 0 0 0 -1 0 -1 0 0 0 0 0, then yes you've increased your effective bit depth by 1, but you've halved your effective sample frequency.
