Headphones & Amps: Balanced vs Unbalanced

headphone-amp

#21

Thanks for that advice. Having wanted for a while to go balanced with my overears I need to get a balanced Amp. My recent acquisition of the Focal Clears is pushing me further towards balanced, because the Clears came with the balanced cable too. I really want to try my HD650’s via balanced too. I would dearly love to try them on a BH Crack. I guess I will need to let my finances grow a little again before splurging on gear again. I need to enjoy what I have for the moment.


#22

Yeah, I’m using the Jotunheim and driving my HD600s, Eikons or Atticus via XLR, but the input is a Mimby, i.e. SE. With the Jot I drive my THX00 SE, mostly because a long time ago @Torq said it made a small difference with his thx00s, but not huge, so I just decided to not mess with the cable.

I have a zach-built adapter to XLR-to-SE out of his ATMOS cable and I do notice a small difference on the Eikon and Atticus, and there’s more power, but honestly I don’t need it so overall I’m using XLR (balanced) because that’s the cable I have.

On the HD600, I notice the 1/4" to XLR a stitch more, but it could have something to do with the cable too as I have one 1/4" cable and one XLR. The HD600 seems to like the power more so I just leave that XLR.

I’m going to live with this setup the next 6 months and then I’m thinking a RME DAC and some amp I can’t decide on now : ) Great topic!


#23

Ha I just posted about my RME with the Jot! with balanced, really enjoying it!


#24

Impact of Balanced vs. Unbalanced Cables - Round Up Comparison

Paradise By The DAC/Amp Lights: Two Out Of Three Ain’t Bad

Here are my findings from testing three headphones with the standard and balanced outputs of my FiiO Q5 DAC/Amp. This is the only balanced amp I own, so it applies to the Q5’s internal “fully balanced” implementation rather than what you might find in other setups.

Please see background info and the rather involved technical specs of the Q5 here:

http://www.fiio.net/en/products/83

The Q5 is effectively a mobile DAP without local storage, and can accept inputs every which way. It was released in 2018 and has received quite solid reviews. The Q5 can use a variety of interchangeable amp modules, but I have only the standard one (AM3A). This module has both 3.5mm singled ended and 2.5mm balanced outputs.

Music sources:

Nothing fancy here. Ripped Redbook CDs (uncompressed and high quality compressed), downloads from Google Music and Apple Music. I used both Bluetooth (Apple products) and USB (notebooks) DAC inputs for testing.

Headphones and cables tested:

Focal Elex (80 ohms)

  • Factory single ended cable with factory 3.5mm TRS adapter
  • Factory XLR 4 pin cable plus a 2.5mm TRRS balanced adapter ($27 at Amazon)

Sennheiser HD-600 (300 ohms)

  • Factory single-ended cable with 3.5mm TRS adapter
  • Silver plated 2.5mm TRRS balanced cable ($20 at eBay)

MrSpeakers AEON Flow Closed (13 ohms)

  • Factory single ended cable with factory 3.5mm TRS adapter
  • Dyson Audio Mogami OFC 2.5mm TRRS balanced cable ($59 at eBay)

Universal Findings For All Headphones

  • Less background noise on balanced
  • Better channel separation on balanced
  • More precise notes (less blurring) on balanced
  • More power and stronger dynamics on balanced

Headphone-Specific Findings

Focal Elex

  • These are very good headphones but have mid-high brightness or glare on standard cables. It’s intense and can be overwhelming or even painful with bright/poor quality sources. With my balanced setup the notes ‘narrow’ and the glare declines. The general precision increases and moves toward that of the Focal Clear on a singled ended cable. I interpret this as following from less internal noise that requires control by the headphone drivers.
  • Slightly cleaner bass and extension, but it’s not an issue either way.
  • Bottom line: I greatly prefer the balanced output and will likely buy a mobile balanced cable to avoid that huge XLR adapter lump

Sennheiser HD-600

  • I’ve owned this set since 2014, and used them with everything from phone/notebook 3.5mm output on up. They do indeed scale per the signal quality, and can range from mediocre to excellent. However, they previously always had high range noise/hiss and resulted in fatigue within an hour of use.
  • With the balanced Q5 setup, the hiss is GONE, the fatigue is GONE. They were transformed from a shelf-warming reference set to something I actually prefer for many situations. I sat down to test the balanced cable and didn’t stop listening for 3 hours.
  • The HD-600s (for the first time) reveal many 2nd and 3rd order details. With John Williams Spanish Guitar Music (2002), I can hear subtle acoustic air movement that was previously only audible through the Elex. Previously mostly just hiss here.
  • The mid range is fuller on balanced, which exposes either the (1) grain in original recordings, or (2) distortion limits inherent to the drivers. The HD-600s are not as clean or dynamic or deep as the Elex, but that’s sometimes a good thing. Just well-mannered and pleasant all around.
  • I suspect the balanced improvement follows from eliminating electrical noise and cross-talk from the wires in the voice coils – this is more of an issue with 300 ohm drivers (see the CNet link below).
  • Bottom line: No comparison, balanced = a better and completely different experience

MrSpeakers AEON Flow Closed

  • Unfortunately, 1 out of 3 was a horrible transformation. I’m not saying there’s anything wrong with the balanced cable (and actually recommend the seller and physical cable). The problem follows from the inherent characteristics of closed headphones (and perhaps the limited potential of affecting 13 ohm impedance headphones).
  • Closed headphones seal in sound and seal out noise. In the process they create a sensation of compressed air (pop pop in the ears) – it’s both distracting and fatiguing. Closed headphones can sound good for a while, but worse and worse over time as your ears get tired. MrSpeakers did a fantastic job on the tone and controlling the booms and echoes common to closed headphones. The AFC slightly but noticeably compress the dynamics too. Compression is a good thing, as it’s the best experience possible with a small sealed air chamber. My balanced cable throws all of this careful tuning off: the dynamics get out of control and painful, while the narrowing of the notes weakens the low range and makes them too analytical. I want to stop listening within minutes on balanced.
    -Bottom line: Complete waste of money. My balanced cable may be sold and replaced with a shorter single ended cable for mobile use.

That’s all for now. Subject to revision and rethinking.


#25

Nice write-up!

One puzzling thing is that you’re finding that noise/hiss is reduced when using the balanced output. This suggests that the amp module has some design issues with it’s single-ended implementation - or the module is designed for balanced output first and is taking a less than optimum approach to how it provides its single-ended output.

Differential balanced output, which the AMA3 module has, (as opposed to dedicated grounds) involves using twice as many amplifiers per channel vs. single-ended, operating in opposite phase, and by definition this means a doubling of noise (and if any hiss is coming from the amplifier, that’ll double too).

Doesn’t in anyway change what you’re hearing … it’s just a curious design detail with the AMA3, as typically you’ll find single-ended either no noisier than balanced, or quieter if it’s done optimally. Though such noise issues are often well below the threshold of audibility with full-size headphones in either case.


#26

I think this is what distinguishes between objective electrical hiss and human perceptual quirks.

Below the threshold of audibility? I guess my experience forces me to challenge that notion.

  1. There are all sorts of objective electrical noises, ranging from 60Hz AC hum, fluorescent light hum, and static electricity. These are fully measurable and show up in the charts at Audio Science Review, etc.
  2. Humans generate perceptual distortions through internal biology/physiology – as seen with “post-rock concert ears” and individual preferences for warm versus bright profiles.

My guess is that I perceive hiss with audio inputs that cannot be resolved. This is worst with cheaply produced aggressive rock or intentional noise rock, but also occurs with the multiple echoes of live concerts. It occurs within seconds of using Grado headphones. None of this is volume related, and it occurs far, far below the threshold of possible hearing damage. The audio receptors in the brain seemingly get temporarily messed up from neural overload, neural cross-talk, or neurotransmitter depletion.

The Q5 has a generally clean and neutral single-ended output, but maybe I pick up on cross-talk from the shared ground line, the noise floor, etc. With back-to-back comparisons, I do truly hear additional ‘vibrato’ background noise of the Massdrop CTH (but don’t dislike it). Some subtle perceptual difference might explain why I hated the Magni 3 more than any amp ever.
.
.
.
Side evidence on the role of perception and signal chains: The electric guitar ‘tone wood’ debate.

Acoustic instruments use different types of woods, such as spruce, maple, cedar, or mahogany to create a brighter, warmer, or ‘better’ tone. The differences are plainly obvious and beyond dispute. There’s a more ambiguous but persistent debate about wood and construction with electric guitars. They are made with hollow bodies, semi-hollow bodies, and solid bodies. They are made from a dozen different types of wood.

Some people swear there are real differences per each wood while others say it’s nonsense. They say (objectively) that the tone comes from the steel strings and magnetic pickups, so the wood can have little or no impact. But, it is absolutely true that when playing an electric guitar one can feel huge differences between instruments based on their construction. The question is whether this is perceptible to non-players (i.e., anyone else, ever), and whether any such differences can be heard in a performance or recording.

Turning one knob on a pedal, amp, or recording system one millimeter during production can (easily) overwhelm or reinvent nuances introduced elsewhere in the recording chain.

And this is why the subjective vs. objective debates never end…


#27

I’m talking specifically about the noise differences between balanced amplifiers and single-ended outputs taken from them generally being below audible thresholds. It’s not always the case, but it is far more often than not.

Measurements show this quite reliably.

I am definitely NOT suggesting that amplifier hiss, noise or other artifacts are below audible thresholds! All amplifiers make noise, hiss generally, the only questions are at what frequencies and at what level. And how audible that is is then a function of the raw output of the amplifier combined with the relative sensitivity of the headphones (i.e. you’ll get more such hiss with easier to drive units typically).

Those show up in my measurements too. But they have nothing at all to do with balanced headphone outputs vs. single-ended, excepting that they would, by definition, be twice as apparent in a balanced differential amplifier design vs. a single ended version - absent something odd about how that single-ended output was derived (see below).

Though static electricity, by definition, will not affect an audio signal. It has ceased to be static if it does. A static discharge can be audible, but if that’s anything more than random and periodic there are some serious environmental issues at work.

There’s no subjective debate regarding balanced differential amplifiers being noisier than an otherwise identical single-ended topology. When that relationship gets inverted, it is down to how the single-ended output has been derived …

The smart way to do this is to only use a single phase of each channel - but it is sometimes easier and/or less expensive to just put a summer in - which will not only include all the noise from the differential stages but add it’s own - which is probably what FiiO have done here.


I’m not in anyway questioning what you’re hearing - just that it is indicative of a sub-optimal way to put combined balanced/single-ended into a product.


#28

My guess is that they had to hit the $350 retail price point and cut corners. They are also in the business of selling amp upgrade modules.


#29

Definitely a possibility.

Size constraints can be a factor for this sort of thing too, as can feature-priority. If FiiO deemed balanced performance more important, it’s reasonable that they make trade-offs to that effect.


#30

“Oh Yeah” . panic


#31

that’s what she said…


#32

A bold statement that I happen to agree with, especially in regard to the Senn HD600/650 cans.
I’ve caught plenty of flack over the years for stating I hear an objective, discernible difference.
To avoid arguing, I usually just append a post with “IMO” then :zipper_mouth_face: while enjoying the… objectively. discernible. difference! :musical_note:


#33

Great write up, and very detailed. I am with you I like using balanced out of my Dap. I find it a cleaner, better sounding way to listen to my music. Whether or not it’s scientifically provable or not to me it makes a difference. If I went on Reddit I would be scalped for sayin this but hey ho. Horses for courses I guess. Again I really enjoyed reading your write up.


#34

Any human perceptual assessment is ‘subjective,’ but my ears are pretty reliable indicators of sounds they don’t like. The simple measure is listening time before discomfort or wanting to move on to another activity. In this case I went from avoiding the HD-600s outright for the last couple years to 3 hours without wanting to put them down.


#35

Thanks. I tried to make it as repeatable and testable as possible, with a specific list of devices, cables, and sources. There is a ton of snake oil and hogwash in audio, and there are very good reasons to be suspicious. Also note that the greatest improvement came with the cheapest cable ($20) on the cheapest headphones, and the worst outcome was with the most expensive setup. I am not suggesting (and do not believe) that a pricey solid silver miracle cable would make as much of a difference as switching from single to balanced.

As one who’s education is not too far removed from this, I think there’s got to be a way to measure both the electrical/objective and perceptual impacts of a balanced setup in a meaningful fashion.


#36

I agree with you regarding the snake oil. I can only go by what I hear personally. I don’t know why and all the science can sometimes give me a headache but when I use the balanced output on my Dap with my Andromeda’s they sound even better. On some iem’s they don’t sound much better. It could be placebo or it could be the quirkinesses of the Andromeda’s and impedance. But I hear what I hear. Now all I need to do is get a balanced desktop setup for my headphones and I would like to see if there’s any difference besides more power. Although more power is always welcome.:grin:


#37

I’ve done a lot of experimenting with balanced vs not (and everything vs not)… but first in regards to “snake oil”:

This may be controversial, but I truly don’t believe that any major (including boutique) manufacturer (including of cables) wakes up intending to defraud their customers with a bogus product or touts benefits they themselves don’t believe exist. (I think a company that did that couldn’t stay in business – doesn’t mean exaggerations don’t happen though!).

Rather I believe that systems are HIGHLY variable, thus the impact of any given component is somewhat unique to that set-up, in that environment. We don’t have to go much further than any thread where person A has an issue with product X but person B doesn’t: even 2 exact same products don’t perform the same! Now throw in a ton of variables in the total system, all the variables in the environment (like EMI), and biological variables, and boy do you have a lot variables.

For me this means “science” is a guide only - and I say this an engineer who’s worked in two highly regulated fields (due to people dying if we screw up) for over 20 years. A quick example is actually snake oil - for many people it doesn’t have any therapeutic effect, but for some it drives a measurable therapeutic impact (we can argue about why, but an outcome is an outcome).

Further, not a day has gone by in my engineering career where we didn’t have some weird thing happen that nobody could fully explain, but we went with it because it worked. And finally, science is founded on investigating observed phenomena - said differently, you can use science to predict an outcome, but then you experiment and gather empirical data. For engineers, we only care about reliability not the why - for example transistors rely on quantum probability which is reliable but nobody has even the foggiest clue on the why - we simply predict probabilities, yet new components of the standard model are frequently discovered (Higgs boson), and a recent experiment has put the entire model in question … yet we still make transistors and they still power all the stuff we listen to.

Add it all up and what’s left is there is only subjectivity - observed phenomena - because nobody understands the underlying physics well enough beyond predicting a probability.

Anyway … back to balanced, and predictions. In my experience it’s not predictable if balanced will always sound better than not. For example, we can take an Yggy and say that, in general, the balanced outs sound better (based on observed phenomena), but does that mean that balanced will always sound better in any system an Yggy is in? Nope. So we’re left with using the physics as a guide (noise rejection, power, etc) and then just trying stuff to see what pleases us.

Wow, that was a really long-winded to say, “I’m a subjectivist”.


#38

Even in a purely objective world, two instances of any given product will not measure or perform identically. Simple variation in component tolerances ensures this … and while you can attempt to design with this in mind, you cannot avoid it entirely. Short of applying the same kind of binning schemes that CPU manufacturers use.

In any given system you have multiple components that will be at different points on the tolerance curve … two units at the extreme “poor” end of that might result in a very different experience than two on the extreme “good” end.

And even doing incredibly rigorous, repetitive, testing and measurements will show that even in the same system to measurement runs, done back to back, can differ significantly - and often for reasons that are not immediately obvious.

I’ve lost count of the times that I’ve done an outwardly simple measurement run and the results have been totally screwed up - for reasons that are just not apparent.

It can be down to just laying out cables between gear/test tools without paying attention. It’s always funny seeing some things consistently get measured as having AC-related artifacts in one case that do not show up at all in other people’s measurements - something that is highly likely to come down to discipline, attention to detail and consistency in test regimen and practices.

Sticking to the balanced vs. single-ended deltas …

While I tend to use balanced systems (and I mean balanced from end-to-end), there is no escaping the fact that a balanced-differential DAC or amplifier is going to be more expensive than the equivalent single-ended implementation - simply because you require twice the circuitry to pull it off - which drives up your BoM.

The real question then becomes … is the difference bigger/more worthwhile/preferred by a given user over either being able to spend the same amount of money on a single-ended implementation that uses a better fundamental topology and/or better components or going balanced and having cheaper parts/less capable topology.

There’s no single right answer.

Where budget isn’t a concern it’s an easier decision, but when it is it is worth bearing in mind what balanced vs. single-ended really means for DACs, amplifiers and for transducers.


#39

@Torq, @GrussGott, great discussion, I’m inline with both of you in this regard. Everyone has snowflake ears lol, let alone everything else that can factor, I.E. environment, knowledge, capabilities etc


#40

@prfallon69 | @GrussGott | @Torq | @DarthPool

This discussion is veering very very close to my university education and a few sub-specialties of research biology and psychology. For centuries scientists noted individual differences in perception (and honest disagreements about reality). One early observation was that different people marked stars in different locations and navigators first saw lighthouses at different times, etc. This has huge negative consequences for shipping and military combat, so the notion of Signal Detection Theory was born.

People vary with regard to very simple events, to include: (1) seeing/hearing something at all, (2) when it occurred, (3) its magnitude, etc. etc. etc. The tests involve tasks such as: “Did you see the light? Yes/No”

In due course and in conjunction with many research projects, it led to statistical probability models and applications throughout all areas of life. A core finding is that anything is part of a data distribution–but one cannot know the shape of the distribution or where any given event/person falls on the curve without thorough testing. These variable/random distributions apply to physical test equipment (i.e., @Torq above), and people too. People are wildly more fuzzy and inconsistent than electrical equipment.

Psychological scientists analyze the same data in opposite directions: (1) what is universal about all humans, and (2) what is different between individuals. Discussions of objective and subjective differences in audio systems reflects this tension. There are “objective” facts that are “TRUE” in the sense that they can be reliably tested and confirmed to extreme high probability levels (per normal distribution statistical math). There are measurable differences that 95% of humans can detect (p < .05), or 99% can detect (p < .01). These numbers are laughably low for physicists, who look for evidence of 1/10,000 or better. But, humans are messy and complicated animals. One to two percent are generally insane and completely unreliable. You might also have been exposed to these concepts in business training courses such as Lean Six Sigma.

Bringing this back to audio quality and preferences, by the time you get to a playback system there are thousands of physical and biological factors to consider. And these get really messy to test, and there is little financial incentive to test them. However, some people see better than others, some people hear better than others, and some people care more than others. The problem with music research is distinguishing between measurable perception (e.g., signal detection) and qualities (e.g., personal enjoyment and preferences).

The capitalistic world has found ways to quantify preferences and direct resources toward profit. Commercial product development and marketing departments are built entirely on psychological research, and anyone exposed to “mainstream” pop music knows of it too. For example, contrast R.E.M.'s Out Of Time album with Losing My Religion to their earlier work. Several tracks are obvious cross-over efforts meant to receive airplay on different types of radio stations and thereby increase album sales.

And finally, regarding balanced vs. unbalanced systems – the differences might also be tested with probabilistic models that focus on human perceptual limits (i.e., signal detection). In conjunction with preference surveys and measures of what people choose to spend their time doing, hardware manufacturers might improve balanced systems to the point of diminishing returns/human limits. Given past evidence, some people will be able to hear the differences and some people will care. Reliably. So, you can in a fashion “objectively” assess subjectivity.

Note that this formal R&D strategy is VERY different from independent or enthusiast audio developers who use their own ears and training to invent a new playback device. Sometimes it works, sometimes it fails, and sometimes it’s just emotional-commercial cynical marketing (i.e., snake oil).

Scientists take the fun out of everything…sigh…