First I like to say I don't believe this is an EVGA Nu Audio issue in any way (I've measured this on a few different machines).  I think this is related to Windows 10 itself and the speaker output path.  The speaker output path will not handle a full scale digital audio signal.  First I have made a test WAV file with Matlab that is a full scale digital WAV file with a 997Hz sine wave.  I have also built a Matlab script that analyzes WAV files for distortion components.  I following pictures are outputs from my analyzer Matlab script.  I am using Audacity and its ability to capture audio in parallel with the device being tested.  I play my precision test file with VLC and direct the output of VLC to the device under test and Audacity records an effective digital copy of the signal being tested.  I tested the speaker output on my Dell workstation with an EVGA Nu Audio 712-P1-AN01-KR installed and on an Intel D54250WYKH NUC with Realtek Audio (I also tested the Intel Display HDMI Audio output).  I have all sound inputs/outputs set to 44.1kHz sample rate and 24 bit depth.  Here are my results:
 
This is the test file analyzed.  It is 997seconds of a 997Hz sinewave at full scale (numerically the peaks of the sinewave are +8388607 and -8388608 and no DC component).  Keep an eye on the fundamental signal "Power (dBFS)" which is 0dB at 0.997kHz.

 
This signal is about the limit for 24bit audio.
 
The Intel Display Audio HDMI output has no issues playing the full scale signal.  Note the distortion components are an exact copy of the original signal.  There are only tiny spectral variations due to different sampling positions in the WAV file.

Digitally perfect at the HDMI output of the Intel NUC.
 
Now lets plug in the headphone jack and enable the Realtek Speaker out.  The following signal is the same signal as played and recorded above.
 
 
The Speaker is clipped at -0.12dB from full scale.  Inducing the harmonic distortion.
 
(to be continued in the next post)
 

Could you please reupload images to some image hosting? EVGA forum says files not found.

 
Trying using Imgur to upload images. I use it.
 
https://imgur.com/ 

Yea, its all fine now. Except the fact of distortion itself.

Macaroni323
Not sure why this forum does not support pictures very well.

Minor IT issue.  Should be fixed now.

Macaroni323
Thanks Lee,
I see no attached images anymore but the images included in the post are there.  Is that the way the forum should operate normally?

I still see the images attached, and also see them when logged out.  Checked on FF, Chrome, and IE.
 

Lee,
 
I have had a look at the Audio Precision website.  Not sure they have any demo software that I could test with.  I'm nervous the guys at Audio Note are missing this.  It is happening and I can recreate it on many machines.  Last night I tested Windows 10 on VMware on my Mac and it does exactly the same thing.  I would love to send the guys one of my Matlab generated test files so they could take a look at it.  All the software that I've tested and generated sine waves at maximum full scale amplitude with will generate this issue when played with several different applications and the Nu Audio card outputs the distorted audio (similar to what I've measured and posted).  Again the Nu Audio card is not generating this distortion.  I may have a driver or something in my machine that is causing this but I find it hard to believe that every machine I test does this.
 
1. Could it just be a file generation issue?  For example, a two's complement coding issue?
 
I have used my analyzer, built in a Matlab script, as well as few other apps to test my file over and over.  Again, 1st and second plots in the post are (1st plot) my test file analyzed with my Matlab script directly, and (2nd plot) my test file played on VLC (volume locked at 100%) playing through Intel Display Audio HDMI and captured by Audacity WASAPI (Intel Display Audio loopback).  There is NO distortion generated in this case.  The audio is a digital replication of my test file.  The distortion is on the Speaker Output only.
 
2. Are the FFT plots full of even harmonics, meaning asymmetrical clipping?
 
In the first post the 3rd plot shows the Realtek Speaker out.  This is identical to the the 4th plot (2nd post) the Nu Audio Speaker out.  Both of these plots are predominantly odd order harmonics indicating symmetrical clipping.  My text output on the analyzer determines some statistics on the captured distorted data in the WAV file and shows the signal has minimal DC offset positive (sub sample of test file captured) yet the maximum amplitude for the positive and negative peaks are slightly offset negative.
 
Spectral Analysis ----------------------------------------------------------------------------------------------------
Filename: D:\Users\Rick\Documents\MATLAB\Test_Tones\VLC_AudacityWASAPISpeakers(NuAudioLoopback)_997Hz_FS_44.1k_24b.wav
   Title:
  Artist:
 Comment:
Uncompressed audio, 2 channels, 24 bit, 2691030 samples @ 44100 samples/sec, for 61.0210884 seconds

    Max Native Pos Pk (8388607 max) = +8268231         +8268231         
        Max Pos Amplitude Peak (dB) = -0.12554494      -0.12554494      
             Max Pos Normalized Mag = +0.98565006      +0.98565006      
---------> DC Offset Percentage (%) = +0.00000106      +0.00000106      
--------------> DC Offset (integer) = +0               +0               
             Max Neg Normalized Mag = -0.98565435      -0.98565435      
        Max Neg Amplitude Peak (dB) = -0.12550711      -0.12550711      
   Max Native Pos Pk (-8388608 max) = -8268268         -8268268         

           Total RMS Amplitude (dB) = -3.13462991      -3.13462991      
     Total RMS Normalized Magnitude = 0.697057298      0.697057298      
          Total Int Loudness (LUFS) = -3.13567282      -3.13567282      

Elapsed time is 17.703251 seconds.
 
Here is the text output for the statistics on the source test file (spectrum in plot 1 of 1st post).
 
Spectral Analysis ----------------------------------------------------------------------------------------------------
Filename: D:\Users\Rick\Documents\MATLAB\Test_Tones\Sine_1ch_997s_997Hz@FS_44.1k_24b_int.wav
   Title: Sine_1ch_997s_997Hz@FS_44.1k_24b_int
  Artist: Rick
 Comment: 1 Tone (Generate_Test_Tones12)
Uncompressed audio, 1 channel, 24 bit, 43967700 samples @ 44100 samples/sec, for 997.0000000 seconds

    Max Native Pos Pk (8388607 max) = +8388607         
        Max Pos Amplitude Peak (dB) = +0.00000000      
             Max Pos Normalized Mag = +1.00000000      
---------> DC Offset Percentage (%) = +0.00000001      
--------------> DC Offset (integer) = +0               
             Max Neg Normalized Mag = -1.00000000      
        Max Neg Amplitude Peak (dB) = +0.00000000      
   Max Native Pos Pk (-8388608 max) = -8388608         

           Total RMS Amplitude (dB) = -3.01029996      
     Total RMS Normalized Magnitude = 0.707106739      
          Total Int Loudness (LUFS) = -3.01134556      

Elapsed time is 15.565559 seconds.
 
 

Sent you a PM.