[prodo123]

Update: See post #3 for download link to graphs and recordings.
 
 
**UPDATE 2: post #36 has even better data.**
There is no real difference in sound quality between balanced and unbalanced other than the elimination of crosstalk.
 
 
Introduction
 
I work as an audio engineer on the side. Having worked with numerous mics, I am very familiar with the concept of balanced signals. When I first heard of the concept of "balanced headphones" I thus immediately called it snake oil, as there is no return ground in its implementation. This is what we call a quasi-balanced line:
 


Sourced from Rane.com
 
What most "balanced headphones" do is snip the ground/shield from a true balanced cable, resulting in a cable as described in the diagram above as #17/18, except the RCA jack is connected to the headphone earpiece. The ramification of this implementation is a mere doubling of the voltage applied to the speakers, which leads to the claim of "double the slew rate." Most claims concerning this effect claims better bass response since the speakers are driven in push/pull.
 
This "slew rate" myth is somewhat unfounded, since headphones are already driven in push/pull in an unbalanced line. The signal itself is alternating voltages to create sound, and the amount of time it takes to form the charge in the coil necessary to move it is beyond what humans can perceive: ~3µs, or 3x10^-6 seconds. In practice, both sides of the coil are charged instantaneously by an unbalanced signal, and is electrically identical to a quasi-balanced signal being sent at half of the unbalanced signal's voltage.
 
Another claim being made for balanced headphones stem from the fact that balanced microphones are able to take advantage of circuitry to reject common-mode noise. This has more credibility in terms of theory. In a properly grounded but unbalanced system, any voltage applied to the ground is drawn to the ground and thus any interference in the signal line remains. In a closed system such as a PMP, however, the ground is simply a common return path to the battery, but this common ground is shared by all the components in the device, causing noise and crosstalk.
 
On a balanced headphone line, however, interference creates a voltage difference on both sides of the speaker. As long as the voltage induced by interference is equal on both the hot and cold lines of a balanced connection there should be no noticeable degradation in the signal. This is similar to the way that phantom power does not affect a mic signal. By isolating left and right channels and eliminating crosstalk, this noise rejection should theoretically provide some quality boosts. Just how much of a difference it makes depends on how long the cables are, though.
 
This brings another issue addressed by balanced headphones: crosstalk. This is the claim that in my opinion makes most sense, as not all of the power sent to the solenoid inside the headphone is not converted to mechanical energy. By separating left ground/cold from right ground/cold one eliminates any chance that the left channel might leak into the right and vice versa.
 
(If you find any errors in any of these, please let me know as I've only briefly thought about this)
 
Methodology
 
The project involves a recabled Sennseiher HD 25-1 II, using 26 AWG 19-strand SPC cable provided by navships, gold-plated connectors for the headphone, gold-plated 3.5mm TRRS plug, gold-plated 3.5mm TRRS-to-TRS adapter and gold-plated 3.5mm TRRS to dual 6.25mm TRS balanced breakout cable.
 
TRRS was chosen for the 4-conductor and slim profile. The conductors are configured so that left and right alternate:
T - Left +/Hot
R - Right +/Hot
R - Left -/Cold
S - Right -/Cold
 

 
The headphone was braided in a 4-strand Litz braid such that left hot/cold and right hot/cold form twisted pairs respectively. This twisted pair configuration was extended into dual TRS using the custom adapter.
 

 
 
The unbalanced 3.5mm TRS adapter simply shorts the left cold and right cold to form a common ground. As such, this only uses a 3-wire braid.
 

 
 
The audio source is a Focusrite Scarlett 2i2, which has balanced monitor outputs in the rear and unbalanced headphone output in the front, running at 24/96. All audio samples were 16/44.1 lossless to ensure maximum quality. Output volumes were subjectively matched using the Monitor and Headphone volume controls.
 

 
 
The choice of audio source may cause some issues with some readers, since it is not an "audiophile" source, but it is of note that it provides both unbalanced and balanced output. This is the main reason I chose this: it eliminates any variation caused by different audio sources.
 
Music files tested were:
 
Sweet Dreams (Are Made of This) by Eurythmics
Bach's Unaccompanied Cello Suite No.1, BWV 1007: Sarabande, performed by Yo-Yo Ma
Lateralus by Tool
Babe I'm Gonna Leave You by Led Zeppelin
 
Subjective Analysis
 
Sweet Dreams (Are Made of This) by Eurythmics
 
The intro to Sweet Dreams has heavy pan effects and quick bass beats that should bring out the main advantages of balanced lines. Unsurprisingly, the lack of crosstalk cleans up the presentation immensely and gives the impression of a wider soundstage. Bass notes are cleaner and smoother, but the response is identical between unbalanced and balanced. Note separation is also better.
 
Bach's Unaccompanied Cello Suite No.1, BWV 1007: Sarabande, performed by Yo-Yo Ma
 
This piece is much quieter than the others; as such, it tests for fine control of the diaphragm. Instead what I got from the balanced mode was, again, wider soundstage and better note separation. Notes sounded slightly smoother, but not by much. Unbalanced mode gave a harsher quality of sound.
 
Lateralus by Tool
 
This song is laden with intricate drumming that really benefits from a balanced headphone signal. Wider soundstage and cleaner notes are again present, but what really stood out was the impact of the drum beats. The introduction and the heavy guitar riffs also have much more impact with the balanced mode. However, bass response was the same.
 
Babe I'm Gonna Leave You by Led Zeppelin
 
The beginning measures of this song has weird panning from right to left which is made extremely apparent by balanced mode. Separation of left and right does some good for note response, but not much as the channels carry very different sounds already. What I did notice with balanced mode here was the ability to pick out minute details that are harder to distinguish with an unbalanced connection. For example, the small sample of "I can hear it calling me" at 1:40 is much, much cleaner with balanced mode.
 
 
 
Conclusion
 
Better Soundstage and Note Separation
 
I attribute this to the channel isolation that arises as a consequence of balanced mode. Without crosstalk, the left and right channels do not interfere with each other. This interference may also be caused by a slight mismatch of clocks between the channels, which again electrical isolation solves. The result is a cleaner sound which "widens" the soundstage and improves note response.
 
Detailed, Smoother Presentation, but Identical Bass Response
 
I still think this is a consequence of channel isolation, but it may also be some proof of this "slew rate" myth. The testing largely disproved the claimed improved bass response, but the amount of detail and smooth sound that results from balanced mode leads me to believe that this claim of "slew rate" should be investigated further and with better equipment than my ears.
 
Noise Rejection?
 
Short length cables are not privy to noise as much as longer ones. Thus it was natural that I experienced no problems with noise on both balanced and unbalanced cables. Mind you, this is in a place with over 300 wireless devices within 50 feet of me; if there should be any noise, this is the place.
 
Whether It's Worth It
 
I really wish it was. The difference in audio quality is absolutely there. They are tangible enough that I always prefer the balanced sound over the unbalanced, and this is coming from someone who completely did not believe in this. But the amount of trouble I had to go through to make this setup took months and months to source the right parts. And even then you are left with a headphone that cannot play on regular devices without an adapter.
 
But for the hardcore audiophile, balanced mode undoubtedly provides better sound than unbalanced. This may be because of the inherent channel isolation that comes with balanced mode, or it might also be the "slew rate" that people talk about; I do not know for sure and am not able to test for such at the moment. If you're willing to make the compromises that follow such a setup, go for it. The difference was so significant that it unwarranted my consideration of going for further ABX testing. Your experiences may vary.
 
-prodo123

 

[esldude]

The difference was so significant that it unwarranted my consideration of going for further ABX testing. 
 
Where have we have read this before.......oh like hundreds of claims that turn out not to be true.
 
Then the description of the sound being smoother, better, wider etc. etc. etc. etc. etc.   Like virtually every snake oil product made, and nearly all subjective evaluations
 
Now you may be hearing a real difference, but those two things just leap out as cliches of subjective listening evaluation. 
 
When differences are really so large as to need no testing to be sure, measurements are easily able to discern what the differences are caused by.    I also am skeptical of your level matching by ear.  Better than nothing, but it would be trivially easy to measure the two connectors and match exact volume.  Only .2 db can cause one to sound better than the other in quality when it is the only difference.  You won't notice a difference until it is beyond 1 db.  So that would be my first piece of advice.  Match levels carefully.
 
Otherwise, if they still sound different to you, then measure FR.  Most differences are from frequency response changes. 
 
If by crosstalk you are referring to L/R channel separation, 15 db is enough for max imaging with headphones.  For good measure 25 db and above is more than enough.  Not likely common grounds are causing that level of crosstalk. 
 
So your methodology of evaluating is a text book example of how to bias and fool yourself.  Not meant as a personal accusation or attack.  Just an explanation that your situation screams expectation bias.  Your results sound exactly like such a situation would cause.  So if you wish to pursue it further I would suggest level matching with a multi-meter and whatever measurements of FR you can manage if they still sound different to you.

 

[prodo123]

  The difference was so significant that it unwarranted my consideration of going for further ABX testing. 
 
Where have we have read this before.......oh like hundreds of claims that turn out not to be true.
 
Then the description of the sound being smoother, better, wider etc. etc. etc. etc. etc.   Like virtually every snake oil product made, and nearly all subjective evaluations
 
Now you may be hearing a real difference, but those two things just leap out as cliches of subjective listening evaluation. 
 
When differences are really so large as to need no testing to be sure, measurements are easily able to discern what the differences are caused by.    I also am skeptical of your level matching by ear.  Better than nothing, but it would be trivially easy to measure the two connectors and match exact volume.  Only .2 db can cause one to sound better than the other in quality when it is the only difference.  You won't notice a difference until it is beyond 1 db.  So that would be my first piece of advice.  Match levels carefully.
 
Otherwise, if they still sound different to you, then measure FR.  Most differences are from frequency response changes. 
 
If by crosstalk you are referring to L/R channel separation, 15 db is enough for max imaging with headphones.  For good measure 25 db and above is more than enough.  Not likely common grounds are causing that level of crosstalk. 
 
So your methodology of evaluating is a text book example of how to bias and fool yourself.  Not meant as a personal accusation or attack.  Just an explanation that your situation screams expectation bias.  Your results sound exactly like such a situation would cause.  So if you wish to pursue it further I would suggest level matching with a multi-meter and whatever measurements of FR you can manage if they still sound different to you.

Here's data backing my findings. I tried to record as best as I could, using two Samson C02s in XY configuration at point blank; the mics did not move between recordings. The samples played were in order: Square 300Hz no alias, Square 30Hz no alias, Saw 100Hz, Sine 1000Hz, Logarithmic sine sweep from 0-25kHz, pink noise (separate samples for each channel), plucks on left and right with right channel 0.3 seconds behind the left. The original sample as well as the recordings are available inside the attached ZIP. These were then analyzed using Audacity.
 
Note the instability in the pluck samples. They are only present in the unbalanced mode. The 30Hz square sample has drastically different behavior for balanced mode. Also note the higher peaks in the square waveforms, higher arch on the sawtooth, and the identical frequency response overall. What this tells me is that the differences are small enough to be averaged out in frequency response analysis, but minute behaviors in the diaphragm are still there, and are quite clear in certain situations.
 
Note that although this is not as controlled as I would have liked it to be (mic frequency response unaccounted for, no anechoic chamber) the conditions have been kept as identical as in my control.
 
Thanks for prompting me to give actual data. I did not realize how significant the difference would look like on graph.
 
Link to data: https://mega.co.nz/#!684E3TIQ!p-7ASLCB5iVe0OxGGmrOkaZD7p-PqFM0A4clSx-7Gtg

 

[Music Alchemist]

But the amount of trouble I had to go through to make this setup took months and months to source the right parts. And even then you are left with a headphone that cannot play on regular devices without an adapter.

 
Cool experiment. Just wanted to add that balanced headphones (as well as any type of adapters) are rather common and easy to come by.
 
Also, certain amplifiers are designed to sound better with balanced connections. This is not due to the fact that it's balanced in itself, but merely the design.

 

[prodo123]

   
Cool experiment. Just wanted to add that balanced headphones (as well as any type of adapters) are rather common and easy to come by.
 
Also, certain amplifiers are designed to sound better with balanced connections. This is not due to the fact that it's balanced in itself, but merely the design.

I wanted to make the adapters myself with the SPC wire. The trouble was getting reliable parts from China, especially the TRRS female connector; shipping time exceeded 3 weeks. Then there's the issue of Litz braiding in proper twisted pairs...but at that point I think I just went overboard
 
And the amplifier part I do note and had little control over, unfortunately. If someone could do a better controlled experiment I would definitely look into that.

 

[Head Injury]

  Note the instability in the pluck samples. They are only present in the unbalanced mode. The 30Hz square sample has drastically different behavior for balanced mode. Also note the higher peaks in the square waveforms, higher arch on the sawtooth, and the identical frequency response overall. What this tells me is that the differences are small enough to be averaged out in frequency response analysis, but minute behaviors in the diaphragm are still there, and are quite clear in certain situations.

Flipping quickly between the images, there do appear to be audible frequency response changes. Most importantly, the balanced graph is about 1 dB higher overall. You can see this easily with the peak near 1 kHz since it appears in both graphs. Balanced tops out at -59 dB, unbalanced at -60. BTW, logarithmic graphs are more useful IMO, because linear ones like these are dominated by high frequencies and the bass is hard to pick out.
 
What I'm not seeing are the drastic differences you see in the square waves. They're virtually identical, differences are much smaller than ones I've seen in manufacturer variations between two of the same headphone. The saw wave shows more difference, but even the differences between channels are much larger in the graphs you provided.
 
Am I missing something?

 

[prodo123]

  Flipping quickly between the images, there do appear to be audible frequency response changes. Most importantly, the balanced graph is about 1 dB higher overall. You can see this easily with the peak near 1 kHz since it appears in both graphs. Balanced tops out at -59 dB, unbalanced at -60. BTW, logarithmic graphs are more useful IMO, because linear ones like these are dominated by high frequencies and the bass is hard to pick out.
 
What I'm not seeing are the drastic differences you see in the square waves. They're virtually identical, differences are much smaller than ones I've seen in manufacturer variations between two of the same headphone. The saw wave shows more difference, but even the differences between channels are much larger in the graphs you provided.
 
Am I missing something?

The 30Hz square graph dips more after the peak on the balanced samples than on the unbalanced samples. The balanced sawtooth samples have a higher arch than the unbalanced ones. 
Quick overlay does a better job than my words.
 

 

 

[mikeaj]

Sorry, only have time for a quick skim through, so this response will be shallow, but you use the headphone output compared to the balanced line output of the same device?
 
I'm pretty sure you'll find the output impedance to be different for the two outputs (and probably not the balanced being the same or double the other), or is that the point? Any difference in behavior between the outputs is not necessarily due to the state of being balanced or unbalanced. It's probably different drive circuitry. What are the measured impedances, anyway?

 

[Head Injury]

Okay, thanks. I was paying more attention to the 300 Hz wave which shows almost no difference.
 
Using a 4 second clip of the noise section of the left recording (1:34 to 1:38) I'm seeing a pretty big difference in the bass which might explain the square wave responses. Top is balanced and bottom is unbalanced.
 

 

 
Like I mentioned, balanced is about 1 dB higher from 1 kHz on, and the 80 Hz peak is about 1 dB higher as well. However, down at 30 Hz the balanced is at about -56.5 dB and the unbalanced is about -55.5 dB, a 1 dB difference in the opposite direction. If volume matched, balanced would be 2 dB lower at 30 Hz. Notches and peaks are larger on the balanced system, it's back up to +1 at 40 Hz but down to -1 again at 50 Hz for example, relative to unbalanced. It rolls off much faster, then swoops back up at very low frequencies, probably electrical noise or something, but may explain the square wave's dip then faster climb back to 0.
 
I'd post the spectrum exports for more exact numbers, but they're huge.
 
A very imperfect test but I tried. Thoughts?

 

[esldude]

I downloaded the files and took a look.  There is something like a 1 db difference in level with balanced being louder.  There is a difference in FFT's mostly below 50 hz with balanced having a bit higher level of noise below 50 hz.  I think that is likely environmental noise.  Low frequencies are tough to stop.  Otherwise as far as I can tell these look pretty much identical.  They look the same in a spectrum view as well as far as I can tell. 
 
When I FFT the sweep and flip from one to the other there is a bit of move from level difference and changes below 50 hz.  The general response otherwise looks the same.  Ditto on the noise section.  Ditto on square and sawtooth. 
 
I think the overlays you posted fit with the level and low frequency differences.  The slightly louder one will dip lower and arch slightly higher from being louder in a view like that.  The amount of difference fits as the balanced one is a touch louder.  Try amping the unbalanced by 1 db in software and do the overlay again if want to see that.  But I don't see much difference between these. 
 
Nice job doing the measurement and posting. 

 

[esldude]

To those looking at the low end differences.  Having done room measurements for Room correction this is likely just a change in background noise.  A truck a block away pulling up a hill in one instance and not being there in the other will do that.  Those low frequencies travel a long way.  I have done impulse response tests that way and seen the range below 50 hz swing wildly just a few minutes apart. 

 

[Head Injury]

  To those looking at the low end differences.  Having done room measurements for Room correction this is likely just a change in background noise.  A truck a block away pulling up a hill in one instance and not being there in the other will do that.  Those low frequencies travel a long way.  I have done impulse response tests that way and seen the range below 50 hz swing wildly just a few minutes apart. 

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.
 
These actually make the difference clearer (balanced on top again):
 

 

 
Bonus dancing spectrum just for fun (shake dat bass):

 

[Music Alchemist]

  I wanted to make the adapters myself with the SPC wire. The trouble was getting reliable parts from China, especially the TRRS female connector; shipping time exceeded 3 weeks. Then there's the issue of Litz braiding in proper twisted pairs...but at that point I think I just went overboard
 
And the amplifier part I do note and had little control over, unfortunately. If someone could do a better controlled experiment I would definitely look into that.

 
Ah, yes, DIY is always trickier. It's just that the way you worded it seemed to imply that balanced connections were difficult to get in general.
 
Any experiment would really be between an amp's balanced design versus its single-ended design; for example, if you look at the Schiit Audio Ragnarok threads, there is discussion about how its new technology only works its magic with balanced connections, and if you are only using SE, there are better alternatives.

 

[esldude]

 
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.
 
 

Hey love that dancing spectrum.  How did you do that one.?
 
All well and good on your comment except that is a 300 hz square wave, not 30 hz.  Sawtooth was 100hz. 
 
He could do the test again to see if the low end stays elevated on balanced.  But it is just what I see doing speaker measurements in homes.  A fridge being on or off will do just about that.

 

[Head Injury]

  Hey love that dancing spectrum.  How did you do that one.?
 
All well and good on your comment except that is a 300 hz square wave, not 30 hz.  Sawtooth was 100hz. 
 
He could do the test again to see if the low end stays elevated on balanced.  But it is just what I see doing speaker measurements in homes.  A fridge being on or off will do just about that.

Made a GIF in Gimp. I copied the screenshots of the spectrums into layers in Gimp to make cropping them easier, so I just saved the result as a GIF too.
 
OP measured both a 30 Hz square wave (which shows the larger dip in the middle on balanced) and a 300 Hz square wave. I was discussing the 30 Hz one because the 300 Hz ones look practically the same.