Blind test: 6 DACs compared
Jan 29, 2013 at 5:25 AM Post #31 of 176
Quote:
Quote:
Did you hook headphones in the chain when testing? I recall previous discussion about the sound being different was with people who were plugging headphones into them.


That's a fair point: no, I didn't hook up headphones while recording, for two reasons:
1) I wanted to test the DACs more than the attached headphone amps
2) Headphones would have altered the recordings to a certain extent (frequency response due to high impedance, stereo crosstalk, and possibly some level of distortion), but different headphones would have had different effects, so nothing close to a common ground could have been met.
Quote:
That will be totally different to plugging them into a EMU0404 or whatever which will have a very high input impedance, leading them to very likely sounding pretty much identical IME. 
smile.gif


Ha! That didn't take long. Are we claiming that all DACs sound the same now, and that it's the headphone amp that makes all the difference? That is not what I've been told again and again.

 
Well, people listen to DAPs with headphones.  Hard to say how they are going to sound as sources unless they are using radically different digital filters though depending on how they output a signal there might be subtle differences. No two pieces of electronics that are different will perform exactly the same, but it's ultimately going to be up to people to decide individually if any difference is worth caring about, either empirically or subjectively. 
 
There's a lot of technical discussion that can be had about it. For example, a slow-roll-off filter versus a brick wall filter might be audible to people who can hear high frequencies with certain music, but in a DAP where most people are likely to listen to MP3s where everything about 16kHz has been cut off already the filter likely wouldn't make any difference. Even then, we'd have to consider what music we're listening to, especially on Head-Fi where there are plenty of people who fill their DAPs with lossless files and use better than average IEMs or headphones and value hearing exactly how many millimetres the drummer hit the cymbal from the edge whether or not they can hear the nostril hairs of the artist swaying when they breath and so-on and so-forth so they will care about the tiny differences to the point of overkill sometimes.
 
I think it is interesting and valuable to do listening experiments, but for the sake of fun and learning rather than trying to prove a point. It's supposed to be a fun hobby after all. 
smile.gif

 
Jan 29, 2013 at 5:26 AM Post #32 of 176
i should also say you can't actually call this test a DAC comparison, since most of your tests consist of a DAC+amp combination.


Except that the amps weren't loaded and acted pretty much like line outs. When you run RMAA, you want the amps to be loaded. The opposite is true here.

And there's no other way of testing the DACs in those devices.
 
Jan 29, 2013 at 7:02 AM Post #35 of 176
here is an example of how I use my 'resample' and 'sinetest' utilities to correct the pitch, delay, and amplitude of a loopback recording from RealTek ALC887 onboard audio.


Interesting, though some people will certainly complain that artificially modifying any sample would invalidate the test.
Just to check, would you care to PM me and tell me which sample you think is the original, and which is the EMU loopback? Not in this thread please, unless you're guessing only by listening to the samples.
 
Jan 29, 2013 at 7:59 AM Post #38 of 176
Just that they are obviously technically different. They are bit tracking captures in wavelab pro. Could be the d2a or simply the a2d fault but those 2 files show a good amount of difference. It basically shows that d2a to a2d is clearly not perfect in this particular case. How imperfect it needs to be for audiblity etc is still up for grabs but it certainly suggests that all conversions are not identical.
 
Those are samples from your 1st 2 files supplied in the zipped download. Right and left from each, respectively.
 
Jan 29, 2013 at 8:28 AM Post #41 of 176
Quote:
Interesting, though some people will certainly complain that artificially modifying any sample would invalidate the test.

 
That is true, but if they do not like my resampler, they are free to use another (and preferably post their files so that I can check that an audible difference was not introduced at the fault of the other software), as all the source files are made available, and the processing is entirely documented and repeatable. However, software sample rate conversion and delay, if properly implemented, is more transparent than pretty much any analog converter, let alone onboard audio. Additionally, converting the original sample to 96/24 format with an idealized "virtual" DAC, and performing the recording and comparison in 96/24 is good practice in my opinion, because it minimizes the effects of both the recording ADC, and the test subjects' playback DAC, and it also preserves ultrasonic content in the 22.05-48 kHz range (like imaging from the DAC under test), for those who believe they can hear it.
 
Jan 29, 2013 at 9:22 AM Post #43 of 176
Note also that one of the files has inverted phase.


DOH! Indeed! One of my devices plays audio with inverted phase, very strange. I just repeated the recording process to make sure that this wasn't a mistake on my part.

Now, to upload a corrected sample, or not? One reason not to, is that a user would normally play back music with that device, without manually correcting it. So, if it's audible (which I doubt), then so be it, I guess? Fair game?
 
Jan 29, 2013 at 9:29 AM Post #44 of 176
Quote:
Now, to upload a corrected sample, or not? One reason not to, is that a user would normally play back music with that device, without manually correcting it. So, if it's audible (which I doubt), then so be it, I guess? Fair game?

 
I did not refer to the inverted phase as a fault in the test, but rather as an obvious source of a major visual difference (that is, for the short time it usually takes to find out).
 
Jan 29, 2013 at 10:48 AM Post #45 of 176
Quote:
 
The most likely explanation is that the files are simply not very accurately synchronized, and that causes the (visual) sample differences, even if they in fact sound the same. Note also that one of the files has inverted phase. Differences in very high or low frequency content can also be more visible than audible.
Try the same comparison on my files (see above), those are synchronized to a fraction of a sample
normal_smile%20.gif

Those are Skamp's original files. Nothing is inverted relative to it's compare. They are synchronized to a about 0.1 milliseconend. The lower one being that much behind. Unloaded amp, line stages should be extremely linear and is there anything on that recording above 15k or below 50hz anyways? Again, whether there's enough there to be sonically different is something only our ears can tell us so it's not an opinion on the overall significance but I took a quick look at out of curiosity and thought I would share it.
 

Users who are viewing this thread

Back
Top