[Currawong]

Quote:

it's a deliberately stupid example

Yes, it is. 
 

 

[Willakan]

In some ways, smearing cake on the PCB would be a better tweak if it actually worked. Balanced amps tend to need rather more parts, so you're paying a premium for...what exactly?
 
Anyways, here's an engineer from Benchmark on the subject, who knows vastly more about the matter than me. Anything I added is marked in bold:
 
Hello!
My name is Elias Gwinn, I'm an engineer at Benchmark Media Systems.
We've been getting a lot of questions lately about balanced headphones. We are interested in the debate, but I can't say we agree with any technical explanations about the benefits of the set up.
So far, there are 4 major points mentioned so far (that I have heard, at least):

1. Unshared common conductor reduces crosstalk

2. Two amps (per channel) increases slew-rate

3. Two amps provide better damping

4. Balanced cabling provides better common-mode rejection

If I may, I'd like to add my thoughts on these points:

1. Most headphones (at least those of decent quality) do not share a common conductor through the length of the cable (as opposed to what was said in 6 Moons). Most headphones have a separate wire from each negative terminal that remain isolated through the length of the cable. In other words, most headphone cables are effectively balanced inherently. If they were sharing a common through the length of the cable, the impedance of the cable may cause some of the signal to show up on opposing channels. However, they are not connected until the plug, and therefore have a minimal impedance to ground.

2. Any headphone amp that is struggling with slew-rate is a poorly designed headphone amplifier. The HPA2 headphone amplifier on the DAC1 has a bandwidth of 55 kHz, and it doesn't even approach any slew-rate limitations even at those high frequencies. Indeed, there is no advantage to an amplifier providing more slew rate over what it needs for any signal it could possibly have to deal with. 

3. Two amps provide WORSE damping. This is why power amplifiers run better in normal mode vs. bridged mode. A balanced (dual-active) headphone amplifier is exactly analogous to a bridged amplifier driving one speaker. The only advantage is increased power, but it comes at an expense of increased distortion, decreased damping, and altered frequency response. This is common knowledge for bridged amplifiers. Output impedance is doubled: see point 5.

4. Headphones don't need any help with common-mode rejection because they inherently will not respond to common mode signal. If, for example, you apply a signal to both terminals of a speaker, it will not move at all. A speaker only responds to differential voltages.

5. There is another cost incurred by dual-active headphone amps that is not addressed. Headphone amps should have as low of a source impedance as possible. If you are using two amps to drive a channel, you are doubling the source impedance. This will cause the headphones to suffer in frequency response, distortion, and ringing.

Please continue the great discussions. It is important to resolve these debates so that product manufactures can respond to provide the best audio solutions possible.

Thanks!
Elias Gwinn
 
Invariably, the benefits of balanced drive are defined in terms of "rich soundstage" and similarly meaningless terminology.

 

[Currawong]

If you're going to quote that, you should be fair and quote AMB's reply from his forum, since it was discussion of the Beta 22 that prompted Elias to write those comments. 
 
 
 
Quote: http://www.amb.org/forum/benchmark-engineer-on-balanced-v-unbalanced-headphone-amps-t326.html

While Benchmark's statement that having two amplifiers drive the headphones (per channel, in balanced mode) doubles its effective output impedance is true, for any AMB headphone amp, the intrinsic output impedance is so low, that the output impedance difference between balanced and unbalanced is effectively nil. A miniscule fraction of one ohm, when doubled, is still a miniscule fraction of one ohm. When driving a headphone load of tens of ohms to hundreds of ohms in impedance, the damping factor is huge either way.

Mr. Gwinn's remark about common-mode rejection is also misleading. When he said that feeding an identical signal to both sides of the transducer would result in no sound, he is is precisely describing the benefit of common mode rejection! This is why XLR balanced cabling is most commonly used in pro settings (concert/stage, recording studio, radio/broadcast, sound reinforcement, etc., where noise immunity in long cable runs is of paramount importance). Noise interference will affect both phases of the balanced wiring, thus the difference between them is zero (i.e., canceled).

That said, I do agree that CMR is not an important factor in headphone (or speaker) wiring, because the output impedance of the amp is so low that noise interference through the wiring is not going to be a problem. If the source is balanced, it makes sense to preserve the balanced mode of operation through the entire chain (i.e., with balanced amplification and balanced headphones too). Benchmark's own DAC1s have balanced XLR line outputs for driving balanced gear. Those outputs are not good for driving headphones directly (they don't have the cojones to drive low-Z headphones satisfactorily, and their high-ish output impedance is a manifestation of that), but they could be used to drive a balanced amp very well indeed.

The statement about distortion and noise being additive in a balanced setup is also misleading. As "seen" by the headphone load, if the hot and cold side amplifiers are identical, then the distortion products cancel each other.

With all due respect to Benchmark, I own a DAC1 and think it's a very good product, but I believe Mr. Gwinn's remark is dumbed-down to appease an audience at head-fi who are not amp designers or engineers, because there is quite a bit more to the case for balanced amplification than what he mentioned. Perhaps he is justifying Benchmark's lack of a balanced headphone output.

I had posted in the past on other forums about how supply rail currents in a balanced class A amp (as well as the 3-channel active ground configuration) cancel out, resulting in a net zero current flow. This, along with the fact that the headphone load is not referenced to ground, makes the signal as seen by the load to be totally immune to distortion products caused by ground pollution (which occurs in conventional 2-channel passive-ground amps). Such ground pollution occurs when the headphone transducer's return current flows through the ground wiring and ESR of the power supply bulk caps, inducing voltage wiggles in the ground. Due to the fact that real world loads are reactive, the return currents are phase-shifted relative to the amp's output voltage swing. So the "ground voltage wiggle" is also phase-shifted to the original signal, and when you sum the two, you get distortion.

As for bandwidth and slew rate, have a look at the specifications of any AMB amplifier, and you'll see how Benchmark's 55KHz compare. 

We could generalize as Mr. Gwinn did, and it would not prove one thing or another. However, one look at the actual measured results of the group build balanced β22 (For pics, visit the β22 website gallery section, and click on "group-build"), will show you that it gives up nothing in performance compared to any amp, balanced or otherwise:

RMAA test: balanced β22 with 330 ohm load
RMAA test: balanced β22 with 33 ohm load
RMAA test: balanced β22 with 8 ohm load

Note that the actual performance of the amp is better than these figures and graphs because the results approach the limits of the measurement sound card (E-MU 0404 USB). For example, the frequency response rolloff of the amp is actually flat from 0Hz to 2.5MHz -3dB, much of the distortion products are of the sound card itself, and the rising stereo crosstalk graph at high frequencies is also an artifact of the input attenuator pots on the E-MU 0404 USB.

 
 
I'd suggest you talk to engineers who design and build amps for a living and discuss topologies with them, as well as building some DIY amps yourself, rather than trying to railroad your beliefs on other people through this web site.

 

[RexAeterna]

i like only using balanced headphones personally. professional headphones as well has always been balanced with 4-core wiring since the 70's. it was mostly low consumer headphones and consumer gear that was not balanced. i think also if your gonna dish out certain amount of money for a headphone it should already be balanced with 4-core wiring. i find other benefits of 4-core/balanced headphones as well as options to drive them off balanced sources,preamps and even drive them off of any power amp you want. there is tons of benefits beside what has already been said. with headphones though i would like to see 4-pin xlr ends more often cause i find dual 3-pin xlr just odd for headphones cause of the extra unnecessary weight. akg had the right idea with the k1000 and hi-fi man headphones as well with the 4-pin xlr.

 

[Willakan]

Quote:

If you're going to quote that, you should be fair and quote AMB's reply from his forum, since it was discussion of the Beta 22 that prompted Elias to write those comments. 
 
I'd suggest you talk to engineers who design and build amps for a living and discuss topologies with them, as well as building some DIY amps yourself, rather than trying to railroad your beliefs on other people through this web site.

I didn't quote AMB because he was going on about his favourite non-existent spectre of "ground contamination." I also felt he was being disingenuous in several places. Lest you accuse me of railroading (I'm deeply sorry for not having time to quote counterarguments and annotate them should I want to say something.)
 
While Benchmark's statement that having two amplifiers drive the headphones (per channel, in balanced mode) doubles its effective output impedance is true, for any AMB headphone amp, the intrinsic output impedance is so low, that the output impedance difference between balanced and unbalanced is effectively nil. A miniscule fraction of one ohm, when doubled, is still a miniscule fraction of one ohm. When driving a headphone load of tens of ohms to hundreds of ohms in impedance, the damping factor is huge either way. You can argue, quite justifiably, that the reduced damping factor is insignificant, but it still makes it worse.

Mr. Gwinn's remark about common-mode rejection is also misleading. When he said that feeding an identical signal to both sides of the transducer would result in no sound, he is is precisely describing the benefit of common mode rejection! This is why XLR balanced cabling is most commonly used in pro settings (concert/stage, recording studio, radio/broadcast, sound reinforcement, etc., where noise immunity in long cable runs is of paramount importance). Noise interference will affect both phases of the balanced wiring, thus the difference between them is zero (i.e., canceled).

That said, I do agree that CMR is not an important factor in headphone (or speaker) wiring, because the output impedance of the amp is so low that noise interference through the wiring is not going to be a problem. If the source is balanced, it makes sense to preserve the balanced mode of operation through the entire chain (i.e., with balanced amplification and balanced headphones too). Benchmark's own DAC1s have balanced XLR line outputs for driving balanced gear. Those outputs are not good for driving headphones directly (they don't have the cojones to drive low-Z headphones satisfactorily, and their high-ish output impedance is a manifestation of that), but they could be used to drive a balanced amp very well indeed.

The statement about distortion and noise being additive in a balanced setup is also misleading. As "seen" by the headphone load, if the hot and cold side amplifiers are identical, then the distortion products cancel each other. And if they produce distortion products which are not completely identical...

With all due respect to Benchmark, I own a DAC1 and think it's a very good product, but I believe Mr. Gwinn's remark is dumbed-down to appease an audience at head-fi who are not amp designers or engineers, because there is quite a bit more to the case for balanced amplification than what he mentioned. Perhaps he is justifying Benchmark's lack of a balanced headphone output. It would be lovely if you could actually lay out this case...

I had posted in the past on other forums about how supply rail currents in a balanced class A amp (as well as the 3-channel active ground configuration) cancel out, resulting in a net zero current flow. This, along with the fact that the headphone load is not referenced to ground, makes the signal as seen by the load to be totally immune to distortion products caused by ground pollution (which occurs in conventional 2-channel passive-ground amps). Such ground pollution occurs when the headphone transducer's return current flows through the ground wiring and ESR of the power supply bulk caps, inducing voltage wiggles in the ground. Due to the fact that real world loads are reactive, the return currents are phase-shifted relative to the amp's output voltage swing. So the "ground voltage wiggle" is also phase-shifted to the original signal, and when you sum the two, you get distortion. As others have said on DIYAudio and other places, ground contamination is not a problem when you deal with grounding properly. It is AMB's favourite nonexistent issue to pull out of the woodwork to justify three-channel topologies and can easily be expanded to gibber about balanced amplifiers as well.

As for bandwidth and slew rate, have a look at the specifications of any AMB amplifier, and you'll see how Benchmark's 55KHz compare.

 Pointless excess slew rate is rarely a good thing. As I said and the engineer from Benchmark, the required slew rate for a given signal is mathematically defined. You don't get clearer treble or something from excess slew rate.

We could generalize as Mr. Gwinn did, and it would not prove one thing or another. However, one look at the actual measured results of the group build balanced β22 (For pics, visit the β22 website gallery section, and click on "group-build"), will show you that it gives up nothing in performance compared to any amp, balanced or otherwise:

RMAA test: balanced β22 with 330 ohm load
RMAA test: balanced β22 with 33 ohm load
RMAA test: balanced β22 with 8 ohm load

Note that the actual performance of the amp is better than these figures and graphs because the results approach the limits of the measurement sound card (E-MU 0404 USB). For example, the frequency response rolloff of the amp is actually flat from 0Hz to 2.5MHz -3dB, much of the distortion products are of the sound card itself, and the rising stereo crosstalk graph at high frequencies is also an artifact of the input attenuator pots on the E-MU 0404 USB. Largely irrelevant. "I can't sustain an argument so here's some highly limited measurements of my amp with some cheap equipment. Aren't they pretty?"
 
I'll add some more annotations later/correct existing if neccessary.
 

 

[Currawong]

 
Quote:

Note that the actual performance of the amp is better than these figures and graphs because the results approach the limits of the measurement sound card (E-MU 0404 USB). For example, the frequency response rolloff of the amp is actually flat from 0Hz to 2.5MHz -3dB, much of the distortion products are of the sound card itself, and the rising stereo crosstalk graph at high frequencies is also an artifact of the input attenuator pots on the E-MU 0404 USB. Largely irrelevant. "I can't sustain an argument so here's some highly limited measurements of my amp with some cheap equipment. Aren't they pretty?"

As usual, you try and back up your opinions with a personal attack. I'm leaving this quoted as an example of what science is not.
 

 

[xnor]

Willikan is right though. We've already seen that something is very wrong with his mini3 measurements as confirmed by independent measurements showing far worse performance.

 

[Willakan]

 
Quote:

As usual, you try and back up your opinions with a personal attack. I'm leaving this quoted as an example of what science is not.

It's not a personal attack. He gave those measurements as part of his argument. They are not applicable to his argument, yet comprise a large proportion of his post and are referred to as a real world example. They are far too limited both in scope and accuracy to even come close to serving that role, and hence their use there is actually damaging to what little exists of his point. I didn't put it particularly politely, but still...doesn't constitute a personal attack. I'm afraid I can't think of a stupid example to illustrate my point, but rest assured I will add one it if one pops into my head.
 
It's an assessment of the facts as I see it, with an opinion that I would say is relatively reasonable derived from those facts (the measurements are both inadequate and irrelevant, therefore their inclusion demonstrates that the creator of the post is likely having difficulty arguing his position (in my opinion, of course).)
 
Do you have anything to say that has some bearing on the matter at hand as opposed to merely being condescending?
 

 

[Currawong]

Quote:

Willikan is right though. We've already seen that something is very wrong with his mini3 measurements as confirmed by independent measurements showing far worse performance.

The Mini 3 isn't a balanced amp and the measurements haven't, to my knowledge, been independently verified.
 


Quote:

 
Quote:

As usual, you try and back up your opinions with a personal attack. I'm leaving this quoted as an example of what science is not.

It's not a personal attack. He gave those measurements as part of his argument. They are not applicable to his argument, yet comprise a large proportion of his post and are referred to as a real world example. They are far too limited both in scope and accuracy to even come close to serving that role, and hence their use there is actually damaging to what little exists of his point. I didn't put it particularly politely, but still...doesn't constitute a personal attack. I'm afraid I can't think of a stupid example to illustrate my point, but rest assured I will add one it if one pops into my head.
 
It's an assessment of the facts as I see it, with an opinion that I would say is relatively reasonable derived from those facts (the measurements are both inadequate and irrelevant, therefore their inclusion demonstrates that the creator of the post is likely having difficulty arguing his position (in my opinion, of course).)
 
Do you have anything to say that has some bearing on the matter at hand as opposed to merely being condescending?
 

If you think that attacking someone suggesting they are saying "I can't sustain an argument" is "an assessment of facts" and is appropriate manner in which to have reasonable discussions anywhere, not just here, then there is nothing more to say.