[nichifanlema]

I've seen that some of the higher priced DACs have Oversampling or Upsampling function.

What's the difference between the two?

I'm no engineer, so I'd appreciate if you'd explain in layman's terms..

 

[Garbz]

Oversampling takes the sampling rate and multiplies it by a factor of 2^n. so 2, 4, or 8 usually. The purpose is to aid in digital filtering only, to allow a different filter to be used so the garbage produced from this digital filter is now further from the audio band. This means softer and less agressive analogue filters can be used on the output of the DAC and IMHO it produces better sound. Although there are plenty of people on this board who for whatever reason grasp onto their antiquated designs for fear of death lol.

Upsampling on the other hand takes any sampling rate and produces any other sampling rate. The most common numbers are 96khz or 192khz, neither of which is a multiple of 44.1khz sound that CDplayers produce. This is a lossy process. The original signal is destroyed in the process by mathematical approximation of what values the new samples are supposed to be.

By contrast Oversampling only adds values with approximation, the originals are unchanged. Oversampling tends to sound very similar in most implimentations, whereas Upsampling can sound very good (Simple Rabit Code resampler), or bad (Foobar's PHSS resampler).

 

[Carl]

Quote:

Originally Posted by Garbz Oversampling takes the sampling rate and multiplies it by a factor of 2^n. so 2, 4, or 8 usually. The purpose is to aid in digital filtering only, to allow a different filter to be used so the garbage produced from this digital filter is now further from the audio band. This means softer and less agressive analogue filters can be used on the output of the DAC and IMHO it produces better sound. Although there are plenty of people on this board who for whatever reason grasp onto their antiquated designs for fear of death lol. Upsampling on the other hand takes any sampling rate and produces any other sampling rate. The most common numbers are 96khz or 192khz, neither of which is a multiple of 44.1khz sound that CDplayers produce. This is a lossy process. The original signal is destroyed in the process by mathematical approximation of what values the new samples are supposed to be. By contrast Oversampling only adds values with approximation, the originals are unchanged. Oversampling tends to sound very similar in most implimentations, whereas Upsampling can sound very good (Simple Rabit Code resampler), or bad (Foobar's PHSS resampler).

That's synchronious and asynchronious upsampling.

Oversampling means increasing the sampling rate a DAC works at to something over that of the Nyquest frequency (ie, twice the highest frequency) for whatever reason.

Upsampling simply means increasing the sampling rate of a signal anywhere, for whatever reason at all. All oversampling is upsampling.

 

[Scrith]

The important thing to remember is that the original data on a regular CD is 16-bit and sampled at 44.1K. Anything else is going to be some engineer's idea of what might sound good based on a small sample of the original data. The other thing to keep in mind is that higher sample rates require even more precise timing, meaning that a small amount of jitter becomes more significant, because each sample's duration is much shorter.

 

[nichifanlema]

So, if the DAC is Non-Oversampling, then it will just receive the digital data as it is, convert it to analog, and send it out on analog output?


And Oversampling is one specific type of Upsampling,
(Oversampling) ⊂ (Upsampling) is this correct?


And so DAC with Oversampling function multiplies original digital data at 2x 4x 8x, etc etc, so the data will be the multiples of 44.1khz?
Or is that Oversampling function will just x2 the original signal before it converts to analog data?

and DAC with Upsampling function just amplifies original signal to ex) 96khz or 192khz ?

 

[alex_cs]

turning the signal from digital to analog will also be an engineers approximation

some approximations just sound better than others.

 

[Herandu]

The 1st consumer end CD players were 12 bits. They were also very good at masking master tape noise. If you played say a Philips CD-100 12 bit machine against a Marantz CDM-73 16 bit 4X oversampling), you could hear more "noise" from the source material. As a matter of fact, I used for many years an old 12 bit player to do my recordings onto cassette. They sounded far nicer in the car when played back compared to the 16 bit unit.
Whether oversampling is good or not is one issue. The end result is however mainly determined by the analogue output stage after the DAC. If that audio stage is too reactiive in nature, the whole oversampling effort just gets wasted.

 

[Garbz]

Quote:

Originally Posted by Carl That's synchronious and asynchronious upsampling. Oversampling means increasing the sampling rate a DAC works at to something over that of the Nyquest frequency (ie, twice the highest frequency) for whatever reason. Upsampling simply means increasing the sampling rate of a signal anywhere, for whatever reason at all. All oversampling is upsampling.

Umm that's what i said, except that oversampling is always done in 2^n values. This has nothing to do with asynchronous or synchronous. That is related to the incomming and outgoing clock rates.

 

[ezkcdude]

It would be easier if people started using the more general terms resampling or sample rate conversion instead of upsampling. My DAC takes 192 kHz and resamples it to 96 kHz. Strictly speaking, that is not upsampling, but most people refer to my design as an upsampling DAC. To be more general, we should call it a resampling DAC. Well, that's my two cents.

 

[Carl]

Quote:

Originally Posted by Garbz Umm that's what i said, except that oversampling is always done in 2^n values. This has nothing to do with asynchronous or synchronous. That is related to the incomming and outgoing clock rates.

The FIR filters in oversampling DAC chips in practice have 2^n values, but there is nothing inherant to oversampling DACs that means they have to. If a DAC manufacturer wanted to implement a really complex asynchronious oversampling scheme using IIR filters into a DAC chip they most certainlly could, and it'd still be an oversampling DAC, as they sampling rate would still be at a rate above that of the Nyquest frequency.