[headfinoob]

I wanted to gain some insight into the following phenomenon.
Using REW v5.01 b24 software, calibrated EMM-6 mic, and M-Audio MobilePre MkII interface, I took measurements with the mic orthogonally placed approximately 1cm from the foam covering the driver (touching the imaginary plane connecting the outermost part of the headphone surround.) 
I drove the Beyerdynamic DT990 Pro headphones with the line-level output of the Onkyo SE-90PCI soundcard (blue in the charts.)
I compared the response from the Onkyo SE-90PCI to the below devices with head-phone amplifiers.  Can anyone explain the awful dips from each device?  I did confirm the phenomenon with multiple headphones for each device and repeated each measurement three times (total of 54 measurements).  I am curious to see how a high-end headphone amplifier would compare.
 
FiiO E17:

 
HTC One m7:

 
HTC One m8:

 
 
Samsung Note 2:

 
Samsung Note 3:

 

[headfinoob]

Anybody have any insight to this?
I don't understand why these headphone amps can't drive a 600ohm load without massive dips in the frequency response. They have similar dips with lower impedance headphones as well.

This definitely is a bit unnerving in that I haven't seen this issue discussed before and those frequency response curves are hardly what I would call good sound quality.

 

[mikeaj]

More likely a measurement artifact than anything else. How many runs for each source did you do?
 
I mean, you can also measure the electrical output of the amp while it is driving those headphones.

 

[headfinoob]

More likely a measurement artifact than anything else. How many runs for each source did you do?
 
I mean, you can also measure the electrical output of the amp while it is driving those headphones.

I don't see how it is measurement artifact. 54 measurements were taken covering different headphones with the different amps. Those were taken after a large number of prior measurements were taken when I first discovered the phenomenon.

I may investigate further to see if I can find any more info. The equipment and test setup is very sound, IMO. I also have a calibrated SPL meter that I can use with my calibrated mic.

I'll be getting some Noble Audio CIEMs in in the next few weeks that I can try out. I also have some Sony noise cancelling headphones I can try.

I have indeed taken RMAA measurements of all the above amps and some others. Those dips are only happening when driving headphones.

Something weird is going on. I might make an impedance jig so I can measure the impedance curve myself.

 

[stv014]

Do all of these devices have their own DAC ? Your analysis software may not like the fact that the DAC and the ADC sample rate is not exactly the same, as only the measurements of the output from the sound card itself (which uses the same clock for both the input and the output) lack the frequency response problems. You may need to use different software settings, or resample the recorded files to correct the pitch error.

 

[headfinoob]

  Do all of these devices have their own DAC ? Your analysis software may not like the fact that the DAC and the ADC sample rate is not exactly the same, as only the measurements of the output from the sound card itself (which uses the same clock for both the input and the output) lack the frequency response problems. You may need to use different software settings, or resample the recorded files to correct the pitch error.

 
Yes, all the devices indeed have their own DAC.
 
Thanks for the idea, as any brainstorming is welcome with this perplexing matter.  I would maintain that the input sampling rate and output sampling rate don't matter much, although the only device in measurements that resampled the test file would be the Note 3.
Here are the sampling rates used:
E17:
     REW:  44.1khz test file; played natively at 44.1khz
     RMAA:  44.1khz test file; played natively at 44.1khz
 
HTC One m7:
     REW:  44.1khz test file; played natively at 44.1khz
     RMAA:  44.1khz test file; played natively at 44.1khz
 
HTC One m8: 
     REW:  44.1khz test file; played natively at 44.1khz
     RMAA:  44.1khz test file; played natively at 44.1khz
 
Note 2:
     REW:  44.1khz test file; played natively at 44.1khz
     RMAA:  44.1khz test file; played natively at 44.1khz
 
Note 3:
     REW:  44.1khz test file; played non-natively at 48khz
     RMAA:  44.1khz test file; played non-natively at 48khz
 
 
 
REW software expects a logarithmic frequency sweep.  As long as the measurement waveform is played back within a small period of time after the REW software starts recording, then the waveform is recorded and analyzed.
RMAA software expects a complex waveform containing various elements to be played.  The software displays a timer in which the waveform must be started within the stated timeframe.
In both cases, the sampling rate of the measurement input doesn't relate to the input waveform sampling rate.  D/A in the playback device is out of sync and can be a different sampling rate than the A/D in the measurement device.  Additionally, the measurements in the first post (blue line in each chart) were taken with the same methodology.  Furthermore, each REW sweep consisted of two sweeps that were averaged.  Looking at the RMAA Frequency Response of each device, one can see the full electrical response.  The anomaly is evident only when driving headphones.  So what is going on?  I have no idea!  I'm going to revisit this soon and test various headphones with various amps in various ways...

 
RMAA measurements of each device:

 
RMAA Frequency Response of each device:

 

[stv014]

Originally Posted by headfinoob /img/forum/go_quote.gif
 
Thanks for the idea, as any brainstorming is welcome with this perplexing matter.  I would maintain that the input sampling rate and output sampling rate don't matter much, although the only device in measurements that resampled the test file would be the Note 3.
Here are the sampling rates used:
 
E17:
     REW:  44.1khz test file; played natively at 44.1khz
     RMAA:  44.1khz test file; played natively at 44.1khz

 
What I meant is that the real sample rate of each device is not exactly 44.1 kHz, and the error ranges from tens to hundreds of ppm (portable players tend to be worse). If for example the ADC has an actual sample rate of 44101 Hz, and for the device tested it is 44097 Hz, then REW might be confused by the fact that the sweep is slightly slower and lower pitched than expected. The problem is not that the sample rates are different (like playback at 44.1 kHz and recording at 96 kHz), but that the analysis software does not know about the error. RMAA may tolerate this issue better, as it is often used in configurations where the DAC and ADC are on separate devices.

 
To find out if there is a software problem, you would also need to do the electrical tests with REW. Alternatively, you could upload the actual sweeps recorded, preferably with a tone at a known constant frequency (e.g. 1 kHz) added for calibration purposes, and I will analyze them.