[Armaegis]

We were waiting for you.

se

 
Can't let you have all the fun Steve

Waitasec, how did I get lulled into soundscience?!?
 

   
Currawong and I basically said that. ^_^
 
(Although experiments for the sake of experimentation are kinda cool nonetheless.)

 
Assuming that the line drivers from the balanced outputs are basically referencing around "ground" without actually using it... it might somewhat better to try making a single ended adapter by dropping the negative legs and pulling over ground from somewhere else.
 
Fun fact: a lot of balanced prosumer gear at the entry level doesn't even have symmetric output. The cold pin is often simply zero (it shouldn't be "ground", but sometimes they cheat and it is). How this interacts with "balanced" amps is a crapshoot.

 

[stv014]

 
I know, but likewise a 30 Hz square wave measurement will be very sensitive as well.
 
I did just do a quick comparison of most of the noise portion to try to avoid this (1:33 to 2:01 on balanced, 1:34 to 2:02 on unbalanced) and the peaks are eliminated but the higher unbalanced sub-bass still exists. I'm not making any claims of what caused it, external noise during the test or the effects of a balanced signal or differences in equipment like impedance, but in any of these cases I think it matches up with the 30 Hz square wave results.

 
On my graphs above, which show the actual frequency response extracted from the sine sweeps, it can be clearly seen that it is the balanced signal that has rolled off sub-bass response.

 

[stv014]

  Focusrite claims <10Ω on the output impedance. I am not sure if it's the line balanced output or the headphone output impedance, but I will note this in the next few measurements that I make.
 
Worthy of note is that these are not hi-fi devices, but rather are purpose-built to be reference sound. It is their design to match sound as close to what it should sound like; as such I am currently assuming that the drive circuitry should reflect that.

 
Well, the point of "high fidelity" is originally reference sound (minimal changes to the signal) as well. In any case, "pro audio" hardware is not inherently more transparent or neutral than competently designed "consumer" devices. Even the HD audio codecs on PC motherboards have flat frequency response and low distortion (but often higher than ideal noise floor) on the line output once any unwanted software processing is disabled. The main differences are in features, connectivity (e.g. having balanced inputs, microphone pre-amplifier with 48V phantom power, etc.), and bundled software.

 
However, bad headphone outputs can easily be non-transparent, often because of having too high output impedance. Even 10 ohms can be too much for example for balanced armature IEMs, but for your particular headphones, which have relatively high and not very reactive impedance, it is not a major issue.
 
On the other hand, if the balanced output on the Focusrite is intended to be used as a line output only (which I suspect it is), then using it to drive headphones is definitely a potential source of problems, such as the obvious bass roll-off that can be seen on the frequency response. Line outputs are designed for high impedance (>1000 ohms) loads, and may have high serial resistance, and/or use small output capacitors.
 
If you want to test with the Focusrite only, then it would indeed be a better idea to create a balanced output from the two channels of the headphone output by inverting the signal on one of the channels. Even then, the doubled output impedance will change the frequency response slightly, but not as much as a typical line output.
 
In the end, whatever results you get, they are unlikely to show differences that are inherent (and audible) to balanced vs. unbalanced outputs. With normal dynamic headphones, either can be audibly transparent without much difficulty if well implemented. When a significant difference does exist, it is very likely because of some specific deficiencies of the particular device(s) being tested.

 

[SilverEars]

In regards to capacitive output, low impedance such as iems will suffer greater bass roll-off than something of much greater resistance value because the load reacting to the cap in the path creating high pass response which the cutt-off(roff-off) depends on the load value.

 

[stv014]

In regards to capacitive output, low impedance such as iems will suffer greater bass roll-off than something of much greater resistance value because the load reacting to the cap in the path creating high pass response which the cutt-off(roff-off) depends on the load value.

 
Yes, that is why if it was designed to be used as a line output, it will have no significant roll-off in the audio band with the intended kiloohm range loads.

 

[prodo123]

I'm back!
 
True to my word, I have tested the "pseudo-balanced" drive setup. This was done with the right channel only. Balanced output was achieved by duplicating the right channel in the previous test audio track then inverting it. For unbalanced output, I did a mono feed to right channel only.
 
Recording setup has changed as well. This time I went with my CAD GXL2200 condenser mic, with the can secured to it by rubber band. Graphs were mostly done using raw data extracted from WAV to CSV, then plotted on R. Frequency response graph was done using Audacity's built-in analyzer. Although Focusrite does not specify the output impedance of the headphone out, electrically the balanced drive should have double the impedance of the unbalanced line.
 
The results show that the main difference is in output volume. This is clearly seen in the frequency response graph:
 


Accounting for the volume difference yields this graph:


 
 
Ignoring the bass response below 20 Hz as data collection error, the frequency response between balanced and unbalanced remain nominal. 
 
 
Other waveforms follow this pattern of volume difference. The general shape of each one remains unchanged, showing little difference between balanced and unbalanced in terms of audio quality. This is best seen in the sine graphs, where other than magnitude the two graphs are identical.
 
Square 300 Hz:

 
Square 30 Hz:


Sine 1000 Hz:

 
Saw 100 Hz:

 
Ping:

 
After this testing, I arrived at the conclusion that balanced drive offers no tangible benefits other than complete elimination of crosstalk between left and right channels. Crosstalk, as shown before, is a measurable and audible phenomenon that do muddle up sound quality. However, other than that, I do not see the reason for going balanced.
 
Data can be downloaded here: https://mega.co.nz/#!2p4zlSYI!tAo62kDWxV3JB3wOBVoL42Ceu9jEQuxCGXhHkO3Oo-Q

 

[Armaegis]

Except... now all you've done is show that the amplifiers are simply working as intended because you're just measuring frequency response.

 

[prodo123]

  Except... now all you've done is show that the amplifiers are simply working as intended because you're just measuring frequency response.

Well, yes and no. What I've done is see what happens when everything is kept constant except for the way the cans are driven.
 
The waveform graphs show the behavior of the diaphragm. If a "balanced" signal shows any difference from the unbalanced signal, then there should be a change in how the diaphragm moves (e.g. overshoot due to voltage spike, rattle due to instability in the upper registers) and as a result the waveforms should have a different shape and pattern. Only the sine sweep and pink noise measure overall frequency response.
 
Any change in diaphragm behavior should manifest into a frequency response change as well if significant enough. However, neither the waveform nor frequency response show any change between unbalanced and balanced.
 
The amps and DACs have been kept constant to make the balanced drive the only variable in the measurement because anything otherwise would make the experiment pointless.

 

[castleofargh]

maybe check crosstalk if there is an important benefit. I would expect 2 separated  amp section to be able to keep each side well separated even on high current(small loads). not sure it's of any use when SE amp can already offer pretty great stereo values. but that's one of the possible use I could think of(outside of the voltage swing of course).

 

[stv014]

Originally Posted by prodo123 /img/forum/go_quote.gif
 
After this testing, I arrived at the conclusion that balanced drive offers no tangible benefits other than complete elimination of crosstalk between left and right channels.
Crosstalk, as shown before, is a measurable and audible phenomenon that do muddle up sound quality. However, other than that, I do not see the reason for going balanced.

 
Actually, the level of crosstalk needs to be fairly high to be audible under normal conditions (i.e. not something like playing digital silence on one channel and then only listening to that channel at high volume). A competently made unbalanced headphone amplifier can easily make it low enough that it remains inaudible, especially since the only place where crosstalk performance is really limited is the TRS connector. Using an 1/8" connector, the O2 achieves better than 60 dB channel separation driving a 16 ohms load. I do not know the exact threshold of audibility, but from the results of various ABX tests in the past, I would expect 60 dB to be good enough, and only really worry about 40 dB or less (which is basically still just ~1% narrower sound stage).

 

[prodo123]

Actually, the level of crosstalk needs to be fairly high to be audible under normal conditions (i.e. not something like playing digital silence on one channel and then only listening to that channel at high volume). A competently made unbalanced headphone amplifier can easily make it low enough that it remains inaudible, especially since the only place where crosstalk performance is really limited is the TRS connector. Using an 1/8" connector, the O2 achieves better than 60 dB channel separation driving a 16 ohms load. I do not know the exact threshold of audibility, but from the results of various ABX tests in the past, I would expect 60 dB to be good enough, and only really worry about 40 dB or less (which is basically still just ~1% narrower sound stage).

Again, you're just talking numbers. I actually measured crosstalk with mics. No matter what the specifications and the theoretical say, the fact remains that crosstalk, at least with this audio interface, is an audible and measurable phenomenon, as noted in post 3. With regards to why it occurs when it shouldn't is out of my scope of knowledge.

My ears are subjective but my mics aren't.

 

[stv014]

Again, you're just talking numbers. I actually measured crosstalk with mics. No matter what the specifications and the theoretical say, the fact remains that crosstalk, at least with this audio interface, is an audible and measurable phenomenon, as noted in post 3. With regards to why it occurs when it shouldn't is out of my scope of knowledge.

My ears are subjective but my mics aren't.

 
How much crosstalk did you measure exactly, and how did you find out if it is audible, and that any audible difference is indeed the result of crosstalk ? Your uploaded recordings seem to be mono, and I did not find anything in the samples that could be used for calculating crosstalk, but maybe I missed something. Also, I checked post #1 and post #3, and found nothing concrete regarding crosstalk, only some theory, and the subjective analysis, which does not tell much if it was sighted listening and/or did not exclude other factors like level and frequency response differences.

 

[prodo123]

   
How much crosstalk did you measure exactly, and how did you find out if it is audible, and that any audible difference is indeed the result of crosstalk ? Your uploaded recordings seem to be mono, and I did not find anything in the samples that could be used for calculating crosstalk, but maybe I missed something. Also, I checked post #1 and post #3, and found nothing concrete regarding crosstalk, only some theory, and the subjective analysis, which does not tell much if it was sighted listening and/or did not exclude other factors like level and frequency response differences.

Here's a GIF of the graphs.
 


The red circles show the area of issue. It is clear that these are not caused by an echo since each channel gets them when the other channel plays the pluck.
 

 
This shows that channel isolation eliminates the issue. The mic and cable positions were identical, recording method and all that the same.
 
As for measuring, my DAW measures -34 dB peak crosstalk for each channel while the noise floor is -43 dB. If I did my math right, that is 42.4 dB of crosstalk.
 
Not to mention you can hear this in the recording with just your ears. Try downloading it and playing back each channel individually.

 

[stv014]

  Here's a GIF of the graphs.
 

 
It looks more like a problem with your setup (hardware or software) than normal crosstalk, which is a linear time invariant effect. That very low frequency noise was not there in the original sample, it does not appear consistently, and it can even be seen on the right channel well before the last pluck begins on the left channel. It also explains why your frequency response graphs show a ~20 dB boost at 2 Hz for the unbalanced sample. Actual crosstalk of the plucks themselves is barely visible on these pictures, and it is not obviously better for the balanced sample, where it can be seen too on the right channel when a pluck begins on the left channel.
 

Originally Posted by prodo123 /img/forum/go_quote.gif
 
As for measuring, my DAW measures -34 dB peak crosstalk for each channel while the noise floor is -43 dB. If I did my math right, that is 42.4 dB of crosstalk.
 
Not to mention you can hear this in the recording with just your ears. Try downloading it and playing back each channel individually.

 
How did you measure the crosstalk exactly with a microphone, and made sure it is not acoustic crosstalk ? You could also have posted the stereo samples.
 
The recordings may or may not sound different (did you compare them level matched with ABX software, by the way ?), but even if they do, it does not tell much, since there are clearly still unexplained issues with your recording setup.

 

[Ike1985]

I don't know anything about any of this, but form what I can tell, investing much moolah in a balanced cable for my iems would be a complete waste.