Dweller is confusing "upsampling" with "resolution enhancement". Upsampling means that we resample the input stream (i.e. 44.1 khz CD) at a higher sampling rate, retaining the original 16 bit word length.
If the higher sample rate is a non-integral multiple of 44.1 (i.e. 96 or 192 khz) one must calculate totally new sample values for each point in the new data stream. If the higher rate is an integral multiple 044.1 (i.e. 88.2, 176.4 etc.) then the original samples become a direct part of the new output stream. This is standard oversampling. Again we retain the original 16 bit word length.
Resolution enchancement is an attempt to extrapolate extra data beyond the original 16 bits in the CD data stream. There are 3 ways which may be used together. First, the digital reconstruction filter will generate extra bits as part of its calculation process. These may be used to extend the signal. Second the designer may add some form of dither (a small amount of digital noise) to create extra bits beyond 16. BTW, note that any didgital audio systems requires a certain amountof dither to achieve good linearity at small signal levels. Third, the designer may use a software algorithm to guess at the extra bits by analyzing the original 16 bit data stream.
Many manufacturers really mean upsampling + resolution enhancement when they advertise upsampling. They are independent processes, however.
If the higher sample rate is a non-integral multiple of 44.1 (i.e. 96 or 192 khz) one must calculate totally new sample values for each point in the new data stream. If the higher rate is an integral multiple 044.1 (i.e. 88.2, 176.4 etc.) then the original samples become a direct part of the new output stream. This is standard oversampling. Again we retain the original 16 bit word length.
Resolution enchancement is an attempt to extrapolate extra data beyond the original 16 bits in the CD data stream. There are 3 ways which may be used together. First, the digital reconstruction filter will generate extra bits as part of its calculation process. These may be used to extend the signal. Second the designer may add some form of dither (a small amount of digital noise) to create extra bits beyond 16. BTW, note that any didgital audio systems requires a certain amountof dither to achieve good linearity at small signal levels. Third, the designer may use a software algorithm to guess at the extra bits by analyzing the original 16 bit data stream.
Many manufacturers really mean upsampling + resolution enhancement when they advertise upsampling. They are independent processes, however.