[DACLadder]

  How so?
 
Wouldn't the interpolation filter be set at 20kHz all the same?

I would think one purpose of oversampling and correctly interpolating is to increase bandwidth of the system.  This would relax requirements for the anti-aliasing filter and avoid the need for drastic brick wall filtering.  So instead of steep cutoff you have a extended response above the original Nyquist frequency with a gradual filter slope depending upon amount of oversampling.   I am no expert though on digital filtering.   

 

[aive]

I would think one purpose of oversampling and correctly interpolating is to increase bandwidth of the system.  This would relax requirements for the anti-aliasing filter and avoid the need for drastic brick wall filtering.  So instead of steep cutoff you have a extended response above the original Nyquist frequency with a gradual filter slope depending upon amount of oversampling.   I am no expert though on digital filtering.   

You're correct.

 

[Clemmaster]

 
I would think one purpose of oversampling and correctly interpolating is to increase bandwidth of the system.  This would relax requirements for the anti-aliasing filter and avoid the need for drastic brick wall filtering.  So instead of steep cutoff you have a extended response above the original Nyquist frequency with a gradual filter slope depending upon amount of oversampling.   I am no expert though on digital filtering.   

 
That's indeed the purpose of oversampling.
You still need to filter out anything above the original nyquist frequency (22.05 or 24) during the interpolation process.
 
I think the problem at 0dB is an overshoot issue (the interpolation filter creates intermediate values >0dB, which produce this wrapping effect).
Resonessence reduces the amplitude by 10dB on their Concero when using "up-sampling" (which is equivalent to oversampling) to prevent the overshoot issue.

 

[Kingwa]

An user send me email about the issue.
I ask him send me the test waves and I test by a Samsung notebook line out, it is same issue so I am consider this is the waves issue.
Try to test on the computer line out to confirm the waves please.

 

[aive]

An user send me email about the issue.
I ask him send me the test waves and I test by a Samsung notebook line out, it is same issue so I am consider this is the waves issue.
Try to confirm the waves please.

Hey kingwa, DACLadder used his own square wave files and got same looking waveforms....

 

[Clemmaster]

Hey kingwa, DACLadder used his own square wave files and got same looking waveforms....

Technically, a square wave does not meet the nyquist requitements of redbook (or whatever FS you choose), at any frequency. It obviously gets worse at higher frequencies.
While you can generate a very nice looking square wave digitally, you will never be able to get a nice result from a DAC.

Only ladder NOS DAC would reproduce something that looks like a square wave at its output (before analog filtering).

 

[aive]

Technically, a square wave does not meet the nyquist requitements of redbook (or whatever FS you choose), at any frequency. It obviously gets worse at higher frequencies.
While you can generate a very nice looking square wave digitally, you will never be able to get a nice result from a DAC.

Only ladder NOS DAC would reproduce something that looks like a square wave at its output (before analog filtering).

Yep I understand that a square wave has an infinite sequence of harmonics in theory. But what we seem to see if considerable distortion/filtering that is unexpected (to me at least) particularly at lower frequencies.

How do you correlate the waveform with performance? I don't know. I also don't have an alternative dac in my home to measure for comparison.

Simply, I just thought I'd see a cleaner square wave at 1 kHz than what I'm measuring. Thoughts?

Edit: I also had received a comment from Kevin G. stating that the 10 kHz square wave seemed heavily distorted and abnormal. So I'm presuming that an issue does exist. But as I said I have no comparison point of what a good DACs square wave output should look like at various frequencies.

Edit2: Oh and the obvious thing that I forgot to mention is that Kingwa seems to be agreeing that this is an issue via email and on here.

 

[Clemmaster]

Yep I understand that a square wave has an infinite sequence of harmonics in theory. But what we seem to see if considerable distortion/filtering that is unexpected (to me at least) particularly at lower frequencies.

How do you correlate the waveform with performance? I don't know. I also don't have an alternative dac in my home to measure for comparison.

Simply, I just thought I'd see a cleaner square wave at 1 kHz than what I'm measuring. Thoughts?

Edit: I also had received a comment from Kevin G. stating that the 10 kHz square wave seemed heavily distorted and abnormal. So I'm presuming that an issue does exist. But as I said I have no comparison point of what a good DACs square wave output should look like at various frequencies.

The frequency is not the issue here. The full scale square wave creates distortion in the reconstruction stage after upsampling (interpolation). When the filter computes the missing samples, it can produces values that are greater than the original ones (overshoot). This is a problem if the starting sample is already at maximum level (0dB). Since the filter cannot create values greater than max, it will wrap around and produce a negative value instead.
Then the filter will uses this erroneous data to compute the next value so the error is not singular but actually looks like the part of the sinewave that's been wrapped to negative frequencies.

 

[DACLadder]

My apologies to everyone.   The web site where I downloaded the squarewave files is performing prefiltering based upon the selected sampling rate.  Therefore the waveforms do not look like squarewaves.   And the M7 is faithfully reproducing data in the files.   Here's a snapshot of the 10kHz 0dB opened in an editor and it looks exactly like the output of the M7.   Same with the others as well.  Sorry to mislead...
 

 

[mowglycdb]

So that means there's no issue?
 
( the HDMI kit may be on it's way to my home, *getting ready to solder in the DI-2014* )

 

[DACLadder]

OK I redid the test using files I generated with Sonic Foundry tools.   The Master 7 10kHz 0dB 48k-16-bit squarewave response looks very good!!!   The generated file is not oversampled so unevenness of the squarewave period is due to some steps having two samples while others have three.  Again the Master 7 looks great when presented with the correct files.  Sorry for the misleading earlier post.
 

 
 
 

 

[Articnoise]

Okay probably a stupid question, but why do you guys use and are interested in the square wave response at 10 kHz 48 kHz 16-bit and not for example 1 kHz 44 kHz 24-bit? Is it saying anything special about the filter in the DAC or something?

 

[DACLadder]

 
Okay probably a stupid question, but why do you guys use and are interested in the square wave response at 10 kHz 48 kHz 16-bit and not for example 1 kHz 44 kHz 24-bit? Is it saying anything special about the filter in the DAC or something?

I wouldn't read too much into it.  Without further complicating the issue and gracefully redacting my original post I just used the same sampling rate today (48kHz) as were the files I downloaded.  It would be almost the same with 44.1kHz sampling but slight worse.  The number of bits (16 vs 24) is not going to make that much difference with a square wave.  There are only two samples involved (highest step and lowest step) with formation of the waveform.   Originally 1khz square wave seemed normal versus 10kHz and the focus on that frequency.  
 
If I could rewind to a few days ago I would have asked "aive" to first verify the playback files with an editor.  Also to use as high a sampling rate as possible to generate the playback test signal.  96 or 192Khz sampling would improve the 10khz cycle-to-cycle period versus the lower 48kHz rate and hopefully the square wave risetime as well. 

 

[Clemmaster]

 
My apologies to everyone.   The web site where I downloaded the squarewave files is performing prefiltering based upon the selected sampling rate.  Therefore the waveforms do not look like squarewaves.   And the M7 is faithfully reproducing data in the files.   Here's a snapshot of the 10kHz 0dB opened in an editor and it looks exactly like the output of the M7.   Same with the others as well.  Sorry to mislead...
 

 
Now that looks like what I get in Audacity after LP filtering @ 20kHz.
 
Good to know the M7 does not interpolate, just over-samples 

 

[aive]

 
OK I redid the test using files I generated with Sonic Foundry tools.   The Master 7 10kHz 0dB 48k-16-bit squarewave response looks very good!!!   The generated file is not oversampled so unevenness of the squarewave period is due to some steps having two samples while others have three.  Again the Master 7 looks great when presented with the correct files.  Sorry for the misleading earlier post.
 

 

 
Yah that looks very similar to my 10 kHz waveform. I wasn't expecting that to be 'normal' - but it's understandable given the filter cut-off and relatively low sample rate. Learn something every day.
 
I created my waveforms in Matlab and then checked them by reading them in Matlab and plotting the data...
 
So DACLadder, just to confirm - the 1 kHz square waves we're seeing, they're good/acceptable even with the ripple we're seeing?