[BC] CD vs LP

[Robert Orban]

I didn't say that dithering improves the dynamic range because there 
is nothing to improve -- dithering is not optional; it is an 
essential part of a system with quantized levels. Without dither, 
such a system is broken because, as you say, it becomes more and more 
nonlinear at low levels.

The correct amount of TPDF dither (the "correct amount" being 
substantially less than -87 dBfs RMS for a 16-bit system) added 
before the quantizer linearizes the system by decorrelating the first 
two statistical moments of the quantization noise from the desired 
signal. (See Lipshitz and Vanderkooy for details.) My measurements 
were made with dither and indicated substantially greater than 87 dB 
range between the largest signal that the system could accommodate 
and the dithered noise floor, both measured with the same 
RMS-responding meter. My THD+N measurement was also made with dither.


At 12:33 AM 9/23/2008, RichardBJohnson at comcast.net wrote:
>Dithering improves RESOLUTION, never DYNAMIC RANGE. Dynamic
>range is the ratio between the maximum number of codes (the maximum
>level) and the minimum --which I even pretend is zero. The documentation
>is absolutely positively wrong and I worked 20 years for the guy, now
>retired, who invented A/D converters and who has over 500 patents on
>them --who drummed this into our heads -- Bernie Gordon.
>
>If you take your test setup and keep reducing the input level by some
>known attenuation step, the output should be reduced by the exact
>same step until you get to residual noise. The residual noise will
>always REDUCE the available dynamic range, never increase it.
>
>It your measurement setup doesn't have a 1:1 correlation between
>the input level and the output level, then you have proved a nonlinear
>system exists. If you have a system in which this nonlinearity exists,
>then you have a quantization problem (missing codes) at low levels.
>This makes the system act like it has a built-in squelch and there
>are papers out there that pretend that this is good. It's like cross-over
>distortion in early SS amplifiers. You keep reducing the level and
>suddenly there is nothing --a great S/N.