[daltonlanny]

I was just wondering what the general differences in sound between cd players that do upsampling as opposed to regular oversampling?
Which do you prefer and why?
Also, how do the old cd players that did not use either oversampling and upsampling compare sonically to the newer ones that do?
Thanks alot.

 

[i has a can]

the output stage and general design of the cd player is much more important in determining sound quality. Upsampling is seen on higher end designs, while more mundane units have only oversampling dac.

 

[Mr_Sukebe]

From a sound perspective, my experiences with a variety of upsampling sources has been that they tend to sound a little more detailed, have more "air and space" around things. Making up for that, many of them also had a more synthetic feel to the sound.
As already stated, how well it's been implemented is definitely more important than the chip technology in use.

Personally, I now have an upsampling DAC and love the extras it brings, but only because it does seem to have been well implemented.

 

[Herandu]

What exactly is the difference between oversampling and upsampling? And what advantages does upsampling got over oversampling? I fail to see how upsampling can be better than oversampling so anybody care to re-educate me?

 

[DarkAngel]

Before this turns into a long mind numbing discussion of digital signal processing that really means nothing to average listener, the reason designers use upsampling/oversampling is to allow digital filtering to take place at very high frequency ranges that does less damage to audible music range.

Even today some designers use NOS or non oversampling designs, and they can also sound very good if implement properly. So as always actual listening is required to determine which CDP/DAC sounds best

 

[Herandu]

I understand the bit about oversampling using a higher frequency, digital filters, etc. The DAC I use has all of that and sounds better for it. But upsampling seems to be a different beast. Imagine a 44.1KHz signal being upsampled to 88.2Khz. An 8X oversampling DAC would be sampling at 352.8KHz equivalent resolution. So the oversampling DAC should be better than the upsampling DAC.
Am I wrong in my thinking somewhere?

 

[hciman77]

Quote:

Originally Posted by Herandu /img/forum/go_quote.gif I understand the bit about oversampling using a higher frequency, digital filters, etc. The DAC I use has all of that and sounds better for it. But upsampling seems to be a different beast. Imagine a 44.1KHz signal being upsampled to 88.2Khz. An 8X oversampling DAC would be sampling at 352.8KHz equivalent resolution. So the oversampling DAC should be better than the upsampling DAC. Am I wrong in my thinking somewhere?

This has always puzzled me. A CD has 44,100 samples per second. Oversampling that 44,100 means that you should get better error correction . i.e if the CD players misreads a bit once out of 8 times then the seven correct reads should mean that the correct value is output. However you start with 44,100 samples a second on the CD and that is what you get, re-reading the signal 8 times wont get you any extra information, if you divide the 44,100 signal into 8 then 7/8 samples would be all zeros ?

 

[tourmaline]

read this:http://www.audioholics.com/education...-digital-audio

 

[tourmaline]

read this:http://www.audioholics.com/education...-digital-audio

"One thing to note is that many products claim 24-bit word lengths, but yet only process internally at 20 bits."
Probably one of the reasons why some older true 20 bit models still can compete with the "24 bits" models.

 

[Herandu]

Very good link, thanks.
One thing I picked up from there is: "The effects of upsampling are no doubt overstated. By carefully designing the sampler, ADC, digital processing path, and oversampling DAC, the upsampling and asynchronous rate transfer can, in my opinion, be avoided."
So upsampling could be seen as a method to cover up a lot of evil in the design of the DAC, whilst oversampling seems to offer sonic benefits.

 

[webbie64]

Quote:

Originally Posted by DarkAngel /img/forum/go_quote.gif Before this turns into a long mind numbing discussion of digital signal processing that really means nothing to average listener, the reason designers use upsampling/oversampling is to allow digital filtering to take place at very high frequency ranges that does less damage to audible music range. Even today some designers use NOS or non oversampling designs, and they can also sound very good if implement properly. [size=small]So as always actual listening is required to determine which CDP/DAC sounds best[/size]

X2

 

[sejarzo]

Quote:

Originally Posted by Herandu /img/forum/go_quote.gif Very good link, thanks. One thing I picked up from there is: "The effects of upsampling are no doubt overstated. By carefully designing the sampler, ADC, digital processing path, and oversampling DAC, the upsampling and asynchronous rate transfer can, in my opinion, be avoided." So upsampling could be seen as a method to cover up a lot of evil in the design of the DAC, whilst oversampling seems to offer sonic benefits.

Don't jump to that conclusion yet.

Simply Google the string:

upsampling oversampling

and read a variety of information on the topic.

While it seems everyone can agree on the definition of "oversampling" (and most of the current new generation DAC chips are 8x oversamplers by design), there is more than one way to "upsample", and it is indeed a more complex process.

 

[sejarzo]

Quote:

Originally Posted by hciman77 /img/forum/go_quote.gif This has always puzzled me. A CD has 44,100 samples per second. Oversampling that 44,100 means that you should get better error correction . i.e if the CD players misreads a bit once out of 8 times then the seven correct reads should mean that the correct value is output. However you start with 44,100 samples a second on the CD and that is what you get, re-reading the signal 8 times wont get you any extra information, if you divide the 44,100 signal into 8 then 7/8 samples would be all zeros ?

The CD is not being read more often......oversampling occurs once the data is already out of the transport and the error correction process is complete!

The output from the actual D-to-A part of the conversion process doesn't yet "look like" the analog music signal. For an 8x oversampling process, the output is a 352.8 kHz signal, modulated by the actual audio signal.

Picture it this way: Think about a nice, smooth sine wave graph as what the analog audio signal should appear to be. Next, imagine a bar graph superimposed on that same plot, but it has a huge number of bars, each extending up or down from the x-axis to where they meet the smooth sine trace.

That bar graph is more or less what comes out of the D-to-A. Each "bar" represents a conversion of a sample, and then the output drops back to zero and goes back up or down for the next sample.

The filters in a DAC (however they are implemented) then remove that high frequency component and output the smooth signal that is desired.

It really is not about getting more samples to try and be more correct......it's basically about pushing that high frequency signal up so high that it can be removed with less detrimental effect on the audio signal.

 

[hciman77]

Quote:

Originally Posted by tourmaline /img/forum/go_quote.gif read this:http://www.audioholics.com/education...-digital-audio

That explains some benefits of oversampling in terms of filtering, however oversampling at DAC would be different from oversampling at ADC. An oversampled signal at ADC would give you a mathematically smoother digital representation of an analog wave as they show.

But, whatever you do with it beforehand when it is pressed onto a CD it is 16bits rendered at 44,100 so the net sum of information recorded is the same, no more , no less, whatever way it has been processed. That is say you have a 352800 sampled signal - it looks lovely, and smooth but when it is put down on CD that extra resolution has to be compromised i.e lost. It is like downsampling an 3200 x 2400 digital photo to 1200 x 800 , you may not notice the difference but a lot of information is lost by aggregation, i.e zoom it back up to 3200 x 2400 and it will be blocky.

Maybe a 8x oversampled analog signal will be a bit better when down-rendered to 44.1K but it could just as easily be exactly the same as some algorithm has to be used to chuck 7/8 of the information away.

So when we get to the DAC you still have 16 x 44.1 and the net information you can extract is exactly the same. Of course if you had 352800 sampling at ADC pressed as 352800 and converted at DAC at 352800 then you would have something with more information on it.

 

[sejarzo]

Quote:

Originally Posted by hciman77 /img/forum/go_quote.gif That explains some benefits of oversampling in terms of filtering.............

That is the overwhelming benefit, not just "some" of what oversampling achieves.

It's not about getting more info than is actually there, it is in ensuring that the data that is present in the 16/44.1 data stream is converted as accurately as possible to the original analog wave form within the real limits of the filtering process.

When it comes to upsampling, then things are not so clear. Upsampling a 44.1 kHz stream to 96 kHz is not necessarily trivial.