[reginalb]

  Have you disabled any DSPs and applications (lower right corner in the status bar and the Windows control panel) that could cause this?

Have you tried using a different source device than the Sony?

 
I am back at home after traveling, and trying to figure this out. I did nothing but turn off replay gain in Foobar, and now my FR is still crazy, though no spike, it shows a V-shape FR, but we're still talking 30-54dB peaks at each end, which seems a touch excessive. Interesting, while I no longer have a frequency spike at 1K, I'm now showing an insane distortion spike right there. IMD spikes by 80dB. There has to be something really weird going on. 
 
Is there anything special that I need to do when exporting it out of Audacity? I'm just using the default 16-bit WAV. 

 

[HiFiChris]

Hmm, this seems strange.
 
Maybe your recording levels are too low (check the soundcard's settings and the source volume).

Also, RMAA is quite nitpicky and does not like a 0 dB FS signal in some cases. So when you export it in Audacity (do you digitally amplify the signal in Audacity?), be sure to make it a few dB below 0 dB FS.

 

[reginalb]

  Hmm, this seems strange.
 
Maybe your recording levels are too low (check the soundcard's settings and the source volume).

Also, RMAA is quite nitpicky and does not like a 0 dB FS signal in some cases. So when you export it in Audacity (do you digitally amplify the signal in Audacity?), be sure to make it a few dB below 0 dB FS.

 
I'll give that a try. I did mess with levels quite a bit, and ended up with something that was at least more reasonable (though still not great) out of it. I think it might be a case of the input really just being a crappy mic input, I know that he who must not be named, in his RMAA tutorial, indicated that some computers don't have a real line-in and they can cause issues trying to get RMAA working reliably. 

 

[reginalb]

Sony NW-ZX1
 
I picked up a Behringer UCA-222 and I am getting better results. But I have noticed that volume tends to mess with things a bit. And I have some weirdly high noise floor that's messing with my dynamic range measurements. But even frequency response of my ZX1 (no load) seems affected by changes in volume:
 

 
FR Deviations:
 
Max Volume:
40Hz - 15kHz: +0.14, -0.05
20Hz - 20kHz: +0.25, -0.13
 
~75% Volume:
40Hz - 15kHz: +0.09, -0.05
20Hz - 20kHz: +0.14, -0.11
 
Not sure that those differences would be audible, but I found it interesting.
 
Quick question, to use a headphone as a load for the testing, do you just get a 3.5mm y cable, and plug the headphones in to one output, and the line-in cable to the other?
 
EDIT: Loaded results
Sony ZX1 with UM Merlin load (12Ω)

 
Comparisons of the Merlin load and Grado PS1000 load (32Ω with a huge spike at about 80Hz See the impedance measurement here: http://www.innerfidelity.com/images/GradoPS1000.pdf
 

 

[reginalb]

Schiit Fulla
 
Fulla seems to have some issues at and near max volume. Tested twice at 100% volume, and once at around 90-95% on its control dial, and I get really weird FR and tons of distortion, here is the FR
 

 
Same at around 95% volume, but normal at around 75%
 

 
Max Volume:
40Hz - 15kHz: +0.20, -0.35
20Hz - 20kHz: +0.20, -0.73
 
~75% Volume:
40Hz - 15kHz: +0.03, -0.21
20Hz - 20kHz: +0.03, -0.48

 

[reginalb]

Astell&Kern AK300
 
Well, I was definitely expecting reference level performance with no load from the AK300. Maybe Astell's really sound different. Just like HiFiMan players sounded different, or maybe I'm doing something wrong. But this one is almost as good as the Sony, while behaving identically at full and 75 volume, which is definitely a good thing. At 75% volume the $89 Shiit Fulla is actually better, though I doubt that you'd hear the difference. I'm definitely looking forward to seeing what adding a load to these does. 
 

 
Max Volume:
40Hz - 15kHz: +0.07, -0.52
20Hz - 20kHz: +0.07, -1.27
 
~75% Volume:
40Hz - 15kHz: +0.07, -0.52
20Hz - 20kHz: +0.07, -1.28
 
EDIT: Loaded results
A&K AK300 with UM Merlin load (12Ω)

 
 
Comparisons of the Merlin load and Grado PS1000 load (32Ω with a huge spike at about 80Hz See the impedance measurement here: http://www.innerfidelity.com/images/GradoPS1000.pdf
 

 

[bartzky]

You have to be careful setting the volume. The problems that appear in your measurements don't seem to be caused by a bad performing output of the source at that particular volume. It's more likely a problem of either to low input voltage to be measured correctly or to high voltage causing input clipping.

As HiFiChris mentioned before it's important to set the levels correctly. Audacity will help you there. Just set the source's volume to be just under the level Audacity starts to report clipping.

 

[HiFiChris]

 
Quick question, to use a headphone as a load for the testing, do you just get a 3.5mm y cable, and plug the headphones in to one output, and the line-in cable to the other?

 
 
Exactly.
 
 
 
By the way, even if the levels are correctly set, it can be that the other values besides the FR are incorrect and don't represent the actual performance of the tested source. For very accurate and generally reasonable results with all measurements, one would need a professional soundcard/interface that can be quite expen$ive.

 

[reginalb]

You have to be careful setting the volume. The problems that appear in your measurements don't seem to be caused by a bad performing output of the source at that particular volume. It's more likely a problem of either to low input voltage to be measured correctly or to high voltage causing input clipping.

As @HiFiChris mentioned before it's important to set the levels correctly. Audacity will help you there. Just set the source's volume to be just under the level Audacity starts to report clipping.

 
I've been adjusting the input volume in Audacity to the same level (the first two sections of the test track hitting -6 dB in the recording levels leaves you just below clipping later on in the clip). That should, I think, account for that, no? The Astell (and I just did a HiFiMan 602) give me identical numbers at either volume settting. Indicating that they're not causing distortion at max volume (I think).
 

   
 
Exactly.
 
 
 
By the way, even if the levels are correctly set, it can be that the other values besides the FR are incorrect and don't represent the actual performance of the tested source. For very accurate and generally reasonable results with all measurements, one would need a professional soundcard/interface that can be quite expen$ive.

 
Yeah, I don't trust the other results that I'm getting at all. While my FR results have matched others' tests of the same devices, my others don't. I'm getting pretty low dynamic range, for example. Around 10dB below levels that others testing the same devices have gotten. I figure I've got something affecting the noise floor on the input to cause that. But I'm new at this stuff, so I could be way off. All I know is that I've gotten reasonably close to others findings with some of my FR results.

 

[reginalb]

HiFiMan 602 Rockboxed
 
I've wondered for a while if RB would resolve the treble rolloff the old HM players were known for. I'd never seen a rockboxed HM-XXX get tested. So I did it!
 
Nope!
 

 
Taken at low gain
Max Volume:
40Hz - 15kHz: +0.28, -2.49
20Hz - 20kHz: +0.28, -4.74
 
~75% Volume:
40Hz - 15kHz: +0.28, -2.48
20Hz - 20kHz: +0.28, -5.27
 
EDIT: 
Of course, RB does come with a parametric EQ:
 

 
40Hz to 15kHz: +0.17, -0.20

 

[HiFiChris]

   
Yeah, I don't trust the other results that I'm getting at all. [...] Around 10dB below levels that others testing the same devices have gotten.

 
The thing is that one's individual measurements of the other parameters (not including the frequency response) cannot be objectively compared to the ones of other members as long as not the exact same environment is used for measurements.
My soundcard is for example maxing out around -115 dB values (unloaded), so even unloaded measurements among the stuff I have measured are quite useless (with loaded measurements however, the results are more or less halfway usable for "internal use").

 

[bartzky]

-6 dB should be fine (I usually use about -3 to -2 for unloaded measurements). But I would not recommend to use the test track to set the levels. Using a 0 dBFS sine wave is more suitable. You can generate such a tone at audiocheck.net for example. I think RMAA can do this too.

 

[reginalb]

-6 dB should be fine (I usually use about -3 to -2 for unloaded measurements). But I would not recommend to use the test track to set the levels. Using a 0 dBFS sine wave is more suitable. You can generate such a tone at audiocheck.net for example. I think RMAA can do this, too.

 
Cool, thanks for the tip! I'll give it a try with that.

 

[castleofargh]

I have to admit the RMAA pop up to set the level never helped me do something right. I assumed I was doing it wrong until I saw similar reports on the web. I also stopped using RMAA in a direct loop and got used to record in audacity first. again IDK why, but I get more relevant results that way.
 
super friendly tip, try to use an IEM/headphone you don't care about if you don't know what is the actual output you're using in those measurements when doing them loaded. even 1volt can mean a lot for some IEMs.

 

[reginalb]

ZX1 MP3 Stream
 
So, I thought it would be interesting to take the test WAV, convert it to an MP3, upload it to Google Play Music, and stream it back with my ZX1, to see if that would affect things. Still pretty flat, with a steep rolloff north of 19,500Hz:
 

 

 
You can see here that it's well north of anything that matters.