Another controversial digital thread

[Gordon Gray]

Haha. I can't wait till I hear you doing some fancy things with that k-01. So how much better is it than the Cary??? You had the 306 right?

The jury is discussing the verdict at this time.

I ended up breaking the solder connect on one of my reference IC's a week ago but should have back today.

The K01 is definitely "more refined" than the Cary but I think I'm at that point of somewhat diminishing returns on the $ / performance scale. I believe that the Cary has held its own in many ways. The question is if the additional performance justifies the price differential.

GG

 

[bonzo]

You will be able to improve the performance with the 10Mhz clock and experiementing with the upsampling, so it will be a while before you can actually make a fair comparison

 

[Gordon Gray]

Not getting the clock for the K01 at this time. K01 has 3 upsampling options. As does the Cary 306.

Will certainly be able to compare the units "straight up" against each other.

Sorry but I have no idea what you are talking about.

 

[RCHeliGuy]

Quick description.

CD contains 44.1 thousand samples per second.
Then they came out with oversampling CD players 2x, 4x, 8x, etc.. These would interpolate between samples to create a "smoother" signal.
Now you can download native 192 thousand samples per second from HDTracks.

Past that you can continue to interpolate between samples and have the DAC try to fill in the blanks between smaller and smaller time intervals.



So if you start with 192,000 samples per second 2x would be 384,000, and 4x would be 768,000, and 8x would be 1,536,000 or 1.5 MHz.

If I understand him correctly he's saying that you can get up to 10,000,000 samples per second. (PLEASE CORRECT ME if I misunderstood you )

So compared to a lowly CD at 44.1kHz 10MHz would be oversampling by 226,757 times.

Obviously the differences have much more impact in the 2x, 4x, 8x range. By the time you are at several thousand times that there are obviously diminishing returns.

I would also expect that at some point the analog signal would blend together pretty seamlessly. I won't suggest where that point may be.

 

[amey01]

I would also expect that at some point the analog signal would blend together pretty seamlessly. I won't suggest where that point may be.

Not quite. For disclosure, there is oversampling in recording, and oversampling in reconstruction. I’m talking about reconstruction as that is what we are discussing here – playback hardware.

As music is always sinewave, you only ever need two samples to perfectly reconstruct the analogue waveform (Nyquist). That is, a sample to locate the positive peak, and a sample to locate the negative peak.

It is the job of the DAC to do the interpolation, and two samples is all that is sufficient to reconstruct the analogue waveform perfectly, and any additional samples are wasted and are not “filling in the gaps” in any way.

So I’ll make that clear: At sub-Nyquist frequencies, the analogue signal always blends together seamlessly. It is the job of the DAC to do this.

Oversampling is not used to make this waveform any smoother (it can’t be because the source is all you’ve got). The signal that comes out of the DAC is already pure, smooth analogue without gaps. That is what the "A" stands for in DAC. Nor can it make it more accurate - oversampling can't "generate" or create accuracy that was never there in the first place.

What it can do however, is keep [I use that word specifically] the analogue output more accurate. Oversampling is used for a variety of unrelated reasons, such as to push the conversion artefacts (like quantisation noise) out to a higher frequency so that the reconstruction filters can be more easily / cheaply implemented, or implemented with better quality and results.

If anyone wants to discuss the benefits of oversampling / upsampling in recording, it's probably best to start a new thread.

 

[RCHeliGuy]

I understand that 44.1 is still low enough to allow for some ringing in the 20kHz range, which most of us have lost the ability to hear.

Is that the type of artifact you are referring to?

 

[bonzo]

Not getting the clock for the K01 at this time. K01 has 3 upsampling options. As does the Cary 306.

Will certainly be able to compare the units "straight up" against each other.

Sorry but I have no idea what you are talking about.

Oh ok I didn't know the Cary had upsampling options. Well I thought you would be getting the clock, and then who knows, maybe the difference would have been more?

 

[Bernard]

Not quite. For disclosure, there is oversampling in recording, and oversampling in reconstruction. I’m talking about reconstruction as that is what we are discussing here – playback hardware.

As music is always sinewave, you only ever need two samples to perfectly reconstruct the analogue waveform (Nyquist). That is, a sample to locate the positive peak, and a sample to locate the negative peak.

And what happens if a peak happens to fall between two quantization levels?

It is the job of the DAC to do the interpolation, and two samples is all that is sufficient to reconstruct the analogue waveform perfectly, and any additional samples are wasted and are not “filling in the gaps” in any way.

So I’ll make that clear: At sub-Nyquist frequencies, the analogue signal always blends together seamlessly. It is the job of the DAC to do this.

Oversampling is not used to make this waveform any smoother (it can’t be because the source is all you’ve got). The signal that comes out of the DAC is already pure, smooth analogue without gaps. That is what the "A" stands for in DAC. Nor can it make it more accurate - oversampling can't "generate" or create accuracy that was never there in the first place.

What it can do however, is keep [I use that word specifically] the analogue output more accurate. Oversampling is used for a variety of unrelated reasons, such as to push the conversion artefacts (like quantisation noise) out to a higher frequency so that the reconstruction filters can be more easily / cheaply implemented, or implemented with better quality and results.

If anyone wants to discuss the benefits of oversampling / upsampling in recording, it's probably best to start a new thread.

So what you're saying is that the ADC/DAC process produces as output an analog signal that is exactly the same as the input signal so long as the Nyquist criterion is satisfied, even if the peaks fall between quantization levels, and if your sample is not taken at the exact time the peak occurs?

 

[InterMechanico]

Given a sample rate of 44.1 khz, I would assume that a signal is likely sampled multiple times on any given peak.

 

[amey01]

I understand that 44.1 is still low enough to allow for some ringing in the 20kHz range, which most of us have lost the ability to hear.

Is that the type of artifact you are referring to?

Pre-ringing, post-ringing, phase shifts, alteration of linearity, etc. are all artifacts of filtering.

These artifacts can be splayed right into the audible range - even if they are caused by frequencies we can't hear, so they pose a major impact on ultimate sound quality.

I said very specifically I wasn't discussing the benefits of oversampling from a recording perspective.

Ringing is an artifact of the digital or analogue filter. 44.1 kHz sample rate requires a brick-wall filter at 22.05 kHz at the recording source, and yes - this sort of filter may induce ringing.

That is exactly why oversampling is necessary.

But it does NOT affect the ability of the DAC to reconstruct the waveform -- perfectly.

 

[amey01]

And what happens if a peak happens to fall between two quantization levels?

So what you're saying is that the ADC/DAC process produces as output an analog signal that is exactly the same as the input signal so long as the Nyquist criterion is satisfied, even if the peaks fall between quantization levels, and if your sample is not taken at the exact time the peak occurs?

Again Bernard - I said specifically that I was not discussing this from a recording perspective.

If the peak occurs between samples, it is the responsibility of the ADC to cater for this.

This is where upsampling on the ADC side is beneficial. But we were very specifically not discussing that. The upsampling will enable the ADC to capture everything before downsampling to 16/44.1 (or whatever the target rate is.

But the above scenario is a case for the benefits of high resolution recording/playback, and has nothing to do with the reconstruction elements of 16/44.1.

I never said 16/44.1 is perfect, and that the world doesn't need high-res audio. But please:

1. There are no gaps in the sound
2. The 16/44.1 sound cannot me "made more accurate" or "made smoother" by upsampling or "filling in the gaps" as alluded to in previous posts.

We've discussed before. See picture here for comparison of one sample versus multiple samples.

http://www.martinloganowners.com/fo...-Vinyl-junkies&p=119897&viewfull=1#post119897

 

[timm]

Ok. So maybe something not as interesting as the creation of a sine wave...... I downloaded the springsteen album in 96/24 from hd tracks as I said. I compared to cd and to compare is very difficult given we are comparing my denon 5900 dac to the squeezebox touch. Comparing the two I find the download clearer but maybe a little hotter as well. So to say one was better or worse is difficult as they sound different. Both have good attributes. The denon seems a bit more 'fleshed out' but the sbt seems clearer. Only wat to really compare is via an outboard dac which I do not have. So take that for what it's worth. In my setup the download was not a failure nor was it a stunning revelation.

 

[amey01]

Ok. So maybe something not as interesting as the creation of a sine wave...... I downloaded the springsteen album in 96/24 from hd tracks as I said. I compared to cd and to compare is very difficult given we are comparing my denon 5900 dac to the squeezebox touch. Comparing the two I find the download clearer but maybe a little hotter as well. So to say one was better or worse is difficult as they sound different. Both have good attributes. The denon seems a bit more 'fleshed out' but the sbt seems clearer. Only wat to really compare is via an outboard dac which I do not have. So take that for what it's worth. In my setup the download was not a failure nor was it a stunning revelation.

Why would you compare with two separate DACs? The SBT will output analogue for both 16/44 and 24/96, so at the very least you can compare apples with apples.

 

[timm]

If I'm outputting analog on both is it not the DACs I am comparing?? I don't have an outboard dac. I'm going straight to a Cary slp98p - all analog. So I can't go digital out of either as of now.