nwavguy vs NuForce

[Tyll Hertsens]

It seems like there's more to this whole output impedance thing than what's been stated or that it's not as black and white as it's been made out to be here. If an output impedance higher than two is universally thought of as being bad then why would a company publish a higher than optimal spec especially when it looks as if the majority of others companies don't publish it at all?

This issue is WAY worse than you'd think. Even the fucking IEC is screwed up on it.

Back in the day, before headphone amps existed, most headphone jacks sucked. On cosumer gear, headphone outs would be driven eithe by a shitty op-amp (whoch would have a fairly highish output impedance), a shitty 1 or 2 Watt clock radio power amp chip (which would have a lower output impedance, but would be noisy), or (on recievers usually) a resistor devider off the power amp out to pad down the signal voltage from the power amps (which would up the output impedance). In the first and last case (most common) output impedances would be fairly high.

Headphone designers knew that they didn't have any choise to design headphones assuming the amp would have a low output impedance. That kinda sucks because it means they can't rely on the amps feeding the cans to provide much damping, so they would design cans with the need to damp through acoustic means (ports and fancy materials covering the drivers to control air). It also made less difference back then because most cans were high impedance(300-600 ohm).

Then the Walkman came along and slowly but surely impedances went down and efficiencies went up in order to deliver cans that would play on the lower output voltage portable players. Problem is that that made interaction problems worse. Anyway, the IEC (the group writing these specs are all engineers from the various big headphone makers) decided that they needed some sort of level playing field somewhere, and settled rather arbitrarily on 120 ohms as the roughly average headphone jack output Z as the "standard" at which to measure cans. Actooly, in the spec it says that headphone should be measured using the manufacturers specified amp output impedance, but good luck finding that number.

From this Stereophile article:

Although it doesn't always appear in the specifications of headphone amps, output resistance has a nontrivial effect on frequency response, and hence on tonal balance. Many headphone amps have output resistances in the range of 20–50 ohms, the principal justification being that this helps even out the substantial differences in sensitivity between different headphones. Output from low-impedance/high-sensitivity models is attenuated more than that of high-impedance/low-sensitivity types, making it less likely that users will inadvertently expose themselves to potentially damaging sound-pressure levels. Insertion of a resistor in series with the amplifier output also helps the designer ensure short-circuit protection and unconditional feedback stability. The IEC 61938 standard goes even further in recommending an output resistance of 120 ohms, noting that "For most types of headphones, the source impedance has very little effect on the performance."

Whoever wrote that must live in a fantasy world. Many of the headphones currently available in the hi-fi market have a medium nominal impedance of 30–60 ohms and typically display a quite wide variation of impedance with frequency. Fig.7 shows overlaid graphs of modulus of impedance vs frequency for the four aforementioned headphones, and fig.8 a simulation of the effect of using them with a source impedance of 120 ohms. The model with the largest impedance variation, the AKG K530 (orange trace), would suffer a total response error of almost 5dB.

Bottom line: the IEC spec is a clusterfuck. In it they describe four different, acceptable ways to measure headphone frequency response (two objective and two subjective), and clearly state that the measurements will not yield similar results. And that there is no clearly understood reason why the results differ.

So, the IEC settled on 120 ohms because they didn't have the balls to try to influence CE gear makers to build better headphone amps into their gear (I wouldn't be surprised to hear that they tried, but the really big boys (Sony, et al) told the headphone geeks to STFU and go play with their cans). The IEC 120 ohm spec is a totla bullshit compromise from headphone engineers stuck in a corner.

When I visited Sennheiser about 6 years ago, Axel Grell (their chief designer) said he would design high end cans differently if he could know for sure the amp driving them would have <2 ohm output Z, but he couldn't rely on that.

Then, Meyer comes out with a 120 ohm switch on the front claiming it's because that's the spec the designers have in mind, and the HF rumor mill goes into full swing, and a missconception is born.

Basically, lower output impedance is always better because it causes less interaction between the cans and amp due to better damping factors. And headphones for home use oughta have higher Z 'cuz you've got potentially higher voltages and that makes for better dampin as well.

Problem is there's always some douchebag wanting to plug in their 800s to an iPod and whining that they don't get loud enough (if the impedance on the cans are too high) or sound as good (if the headphone iedance is too low and interacts with the amp too much) so they really are stuck between a rock and a hard place. As a result, our cans are designed with some serious compromises as manufacturers keep feeling the presure to do all the damping acoustically.

Fucking iPods ... Ya gotta love/hate them.

[NwAvGuy]

As Tyll said, output impedance is a complex topic. He knows more than I do about what headphone designers aim for. I've talked to several vendors at trade shows and, when I could find someone who even knew what I was talking about, they usually told me they design for "zero ohms" with their consumer headphones. I think this is because as balanced armature and multi-driver designs have become more popular, it's forced a lot of sources to lower their output impedance to deliver good sound with those headphones (i.e. partly the "iPod effect" Tyll mentioned).

Even considering just conventional dynamic headphones (which usually have a relatively constant impedance with frequency) the damping changes significantly with different output impedances. As Tyll implied, the headphone designer has to balance the electrical vs mechanical damping. And it's a delicate balance if you want tight, clean, bass with deep LF extension. If you screw up that balance with the wrong output impedance, the quality of the bass gets worse. If it ends up over-damped the bass rolls off too soon and sounds weak. If it ends up under-damped you get an upper bass peak and lousy transient response making the headphones sound boomy with "flabby" bass. (EDIT: fixed "under" and "over" being swapped)

I'm a fan of "zero ohm" output impedances for the reasons I've already mentioned. And I'm also a fan of more headphone designers aiming for that target. The more we get of both, the more the world will be a happier place.

You can have your cake and eat it to. First, some of the better amps/dac/etc give you a choice of internal gain settings. This lets you better match the amp to high or low impedance cans. And a few companies let you select the output impedance as well. I'm surprised more haven't done so. That would let people experiment with what works best with their cans and their subjective tastes.

You can find lots more in my article on this very subject:

Headphone & Amp Impedance

Edited March 18, 2011 by NwAvGuy

[elnero]

Thanks for the replies, I think I'm starting to get a handle on things. Who says you can't teach an old dog new tricks.

[CarlSeibert]

The clipping occurs in the DAC SoC, but at the analog output section of it, so not in the digital domain.

The output gain of the ES9023 chip used in the uDAC can be decided by the implementer by changing a feed-back resistor.

What appears to have happened in the case of the uDAC is that the gain have been set so high that it makes the DAC output section clip at input levels above -1dBFS.

Makes sense. That's basically what I thought I understood.

So, the effect in question would occur when music contains a great number of peaks in the last dB, which would generally be if the music was significantly compressed. Or it's so hot that the peaks in the last dB are mixed with digitally clipped ones that go to the zero line and get whacked. Either way, that recording would sound pretty awful even if it wasn't causing the DAC's output to clip. On the one hand, one could say that this flaw kills music that is already mortally wounded anyway and is thus more or less academic. (I did say that, actually.) Or, on the other hand, for a device that will generally be used to power high efficiency headphones and IEMs, it strikes me that this maybe wasn't a very smart trade-off. And so the designer should be called out for some answers.

[NwAvGuy]

on the other hand, for a device that will generally be used to power high efficiency headphones and IEMs, it strikes me that this maybe wasn't a very smart trade-off. And so the designer should be called out for some answers.

Bingo. The trade-off buys a whopping 1 dB of dynamic range out of about 90 dB which is really questionable at best--especially when the NuForce is already plenty quiet for most. A few on HF accused NuForce of "taking the loudness wars to their hardware".

I do agree the recordings that most expose the problem are more likely to already be in trouble to begin with. I found that out the hard way in my first listening test.

[Currawong]

The admins at HF have become increasingly hostile over my saying bad things about one of their sponsors.

This is a lie. You were told, specifically, multiple times, that the issue was with you coming to Head-Fi and promoting your blog. I'm all for revealing bullshit with genuine science, but that means, if you're going to get any respect, you need to stop bullshitting as well.

[NwAvGuy]

This is a lie. You were told, specifically, multiple times, that the issue was with you coming to Head-Fi and promoting your blog. I'm all for revealing bullshit with genuine science, but that means, if you're going to get any respect, you need to stop bullshitting as well.

It's obviously a matter of opinion, but I think if I shared what has been said to me on HF, what I've been accused of, what's been done with my posts, what's been done with many other members posts who dared try and stick up for me, many would consider it "hostile". So I don't think it's a "lie" or "bullshitting" to use that word.

Yes, the official reason given is the "blog promoting". But that policy is very inconsistently applied at Head-Fi and there's a lot of discretion going on. Hundreds (thousands?) of other HF users do the same thing. But they haven't come out against a HF sponsor as hard as I have. Hell, how many threads are on THIS forum about someone taking on a HF sponsor? This thread was started before I even joined here. It's apparently not an everyday occurrence. So is it such a stretch to think the same controversy might have also somehow influenced Jude's actions against me?

Jude is certainly entitled to run his site how he chooses. But there are plenty of examples where he exercises considerable discretion in doing so. I'm one of those examples. I was "promoting my blog" in exactly the same ways with very similar HF threads about Cowon, Behringer, Sansa, and other topics but none of them are HF sponsors. And there wasn't a peep from any HF admins or moderators about any of those which came first.

And Jude has flat lied in what he's posted to my blog. I have the screenshots to prove it. So if you want to call someone a liar, you might start with him.

And the thing none of the HF admins, Jude included, have answered is what's wrong with directing people to my blog? Especially when I do it as "new window" link so anyone following the link isn't even leaving HF? To me it's no different than linking to a Wikipedia page as a reference.

And, of my 145 posts on HF, only a few dozen have any links in them. So how does that qualify as my "primary purpose for being there" as Jude puts it when the other hundred plus posts in 40+ different threads don't even reference my blog?

I'm not selling anything. I have no advertisers. Hits to my blog don't count for shit and I'm not pushing some weird religion or porn. So, really, why all the paranoia over linking to my blog when hundreds of other Head-Fiers are allowed to do the same thing? Please explain that to me?

[luvdunhill]

I'm not pushing some weird religion or porn.

It made me a bit moist, I must admit.

[jvlgato]

Yeah, this is all porn to me.

[Tyll Hertsens]

i know I'm getting fucking hot.

[Currawong]

because Currawong is an asshole, would be my guess.

Yes, I am, because I have seen all the PM convo and I'm in a bad mood. He was told, by jude, he was welcome to post everything on his blog on Head-Fi if he wanted. But signing up to just promote his site wasn't welcome.

*Cough*

[Tyll Hertsens]

Yes, I am, because I have seen all the PM convo and I'm in a bad mood. He was told, by jude, he was welcome to post everything on his blog on Head-Fi if he wanted. But signing up to just promote his site wasn't welcome.

*Cough*

I read it as him promoting his point of view, and using his site to back up his mouth. Head-fiers need a fucking education. We all do. What the fuck is so wrong with him linking his site!?

What's Jude's fucking agenda? To build walls around the fucking hobby? Or would it be better to use teh intrawebble to open lines of communication!

Are you a mod or a lacky to agendas you don't have the balls to see? There's good people there to try to bring balance, but Jude's hurting more than he's helping in this case, IMHO.

Please Jude, please, lighten the fuck up, man. Head-fi will be better off collaberating with informed input from experts. We need people like NwAvGuy in this hobby to have strong voices. PLEASE LET HEAD-FIERS REAP THE BENEFITS OF THE LARGER COMMUNITY OF AUDIO ENTHUSIASTS! Head-fi will be far more valuable if it learns how to play well with others.

Look how nice it is to link back and forth between Guttenbergs postings. Or are those linkings advantagious because Cnet is a traffic monster and nets bigger numbers for Head-Fi? Its not all about the numbers, man, it's also about hobbyists learning. Having folks temporarily click over to another site doesn't mean they are abandoning Head-Fi.

Fuck me, I'm sick of this shit.

[NwAvGuy]

Thanks Tyll and Rex. I'm sick of this shit too. I expected to be attacked by NuForce, but I didn't expect Head-Fi to act like some communist dictatorship and delete multiple posts of THEIR OTHER MEMBERS that stick up for me. So the whole Jude/Currawong bullshit argument of "it's only about him promoting his blog" falls apart right there. I have no idea what Jude's real reasons are, but it's gone well beyond just censoring links to my blog.

[CarlSeibert]

for a device that will generally be used to power high efficiency headphones and IEMs, it strikes me that this maybe wasn't a very smart trade-off. And so the designer should be called out for some answers.

..... in a moment of lucidity I recalled that when I looked at review threads prior to buying the uDAC2, there were complaints of the uDAC, and the uDAC2 for that matter, not having enough gain to drive this that or the other probably inappropriate headphone.

I just did a search on Head-Fi and now too-much-gain posts outnumber too-little-gain posts about ten to one.

There you have it. Damned if you do. damned if you don't.

[NwAvGuy]

I just did a search on Head-Fi and now too-much-gain posts outnumber too-little-gain posts about ten to one.

There you have it. Damned if you do. damned if you don't.

It's hard for me to imagine the uDAC-2 having too little gain unless you want an obscene amount of extra gain to crank it way up for improperly ripped tracks. So I'll have to side with the current majority. I have no idea about the original uDAC.

Even the line outputs of the uDAC-2 clip at about 2 O'clock on its volume control and the headphone outputs clip at about 1 O'clock with 2.2 volts of output into 150 ohms--about 32 mW (or 50 mW into 16 ohms). Few will get it past 12 O'Clock with normal tracks and that leaves 10+ dB of surplus gain beyond that point available to compensate for quiet tracks or whatever.

But, the damned-if-you-do/damned-if-you-don't argument is still a reason to either have an internal gain option like many other DACs/amps do, or something I haven't seen, offer the same DAC in two versions--low gain and high gain.

[elnero]

So another question, how does an amplifier manufacturer go about lower the headphone output impedance?

[NwAvGuy]

So another question, how does an amplifier manufacturer go about lower the headphone output impedance?

That depends. Some raise the output impedance intentionally simply by putting a resistor in series with the headphones. Some use op amps that can't provide enough current to properly drive a 16 ohm load directly, so they're forced to add series resistance to keep the op amp "happy". Some IC's/designs don't have any short circuit protection and might self destruct without some series resistance. Still others might become unstable with certain headphones without series resistance.

So the cheap and easy way out of all of these problems is just toss in some series resistance and be done with it. Unfortunately that has plainly audible consequences.

There are plenty of IC op amps that are already short circuit protected, and will happily drive 16 ohm headphones directly. And there are plenty more discrete output circuits (using output transistors) that can do the same. So it's really not that difficult to have a low output impedance.

As I pointed out, even the little $20 FiiO E5 headphone amp has a near zero output impedance. If FiiO can do it at that price in a tiny battery powered device, it's hard for anyone else to justify it's difficult or expensive. They either have a higher output impedance on purpose, they don't care, or they're being sloppy/cheap/etc.

And to geek out a bit more, some put the series resistor inside the amplifier's feedback loop. This has the effect of making the resistor sort of magically disappear and you get a much lower output impedance while still having some protection. But it also creates a whole different set of problems to overcome having to do with large amounts of non-linear feedback, phase shift, etc. when driving real world headphones.

[digger945]

But it also creates a whole different set of problems to overcome having to do with large amounts of non-linear feedback, phase shift, etc. when driving real world headphones.

I followed you up to this point. I'm not seeing how a resistor, either inside or outside of a global feedback loop, would alter the linearity or phase.

[Tyll Hertsens]

I think its because the reactive nature of the cans becomes part of the feedback network.

[NwAvGuy]

I followed you up to this point. I'm not seeing how a resistor, either inside or outside of a global feedback loop, would alter the linearity or phase.

There are two things going on here:

I'm not sure how familiar you are with feedback loops, but they take a portion of the amplifer output and feed it back to a phase inverted (i.e. negative) input of the amplifier. This is typically done to reduce the distortion and effective output impedance of the amplifier. If the phase of this signal shifts significantly it starts to look more like positive feedback to the amplifier. And positive feedback turns amplifiers into either oscillators or they slam into a supply rail and stay there--not good. So an amplifier using NFB has a certain "phase margin" which helps assure it will remain stable under normal use.

If we toss a series resistor into the amplifier output, and it's inside the global feedback loop, any drop across that resistor is now effectively seen as a source of "error" from the amplifier's perspective and the feedback loop will attempt to correct the error. That's why the output impedance is still near zero ohms despite the series resistor. Driving a resistive test load, the amplifier will likely behave great as long as you don't try to ask for too much current from it.

But if you connect a real pair of headphones--especially balanced armature or multi-driver types--things change. Now the load is non-linear and has a significant reactive component. The load impedance changes with frequency and the current is out of phase with respect to the voltage. That phase shifted current creates the voltage drop across the series output resistor in the feedback loop while the feedback error signal is based on that drop. So now you have a significant portion of the feedback signal phase shifted from the amplifier's output signal and it also varies with frequency. This is less than ideal.

With say a 6 ohm resistor and a 16 ohm load, nearly a third of the amp's output signal is dropped across the series resistor. So that's a significant amount of the "error signal" in the feedback loop. While it's possible to make an amp work with such an output resistor in the feedback loop, it's not, IMHO, the best of designs. You would likely have to make some significant compromises to assure it would be stable with any possible worst case headphone. NFB partly gets its bad reputation for what happens when it's on the edge of creating instability. So even if the amp doesn't oscillate or fail in any obvious way, it may well still sound bad playing real music with certain headphones as the relatively large phase shifted error signal may cause some unpredictable results.

Edited March 24, 2011 by NwAvGuy

[dsavitsk]

As grawk said, you can't have something higher than 0dBFS in the digital signal and if that portion is perfectly reproduced you'll get that splat.

What I was talking about, I suspect is a function of the DAC chip's reproduction capabilities and the following analog circuit adding distortion because it is being pushed into its worst-case scenario.

It's easy for a DAC to reproduce 0 dBFS without breaking a sweat let alone clipping.

The Wolfson WM8741 DAC chips have a voltage output of 2Vrms with a full scale digital signal. 2Vrms is about 5.7Vp-p. The DAC chips can only handle a PS of up to 5.5V, and are really recommended to run at 5V (actually, they can take 7V, but Wolfson will tell you never to actually run them over 5.5V). Anyhow, suffice it to say that the chips clip. They have an option to cut the signal by 2dB which results in 4.5Vp-p (~1.59Vrms) which they do just fine with.

But the problem with any output impedance over 2 ohms is you get an unpredictable sound that varies widely with different headphones. What benefit is that to the Average Joe reading reviews online? Yeah, if you can even find a retail brick-and-mortar audio store with all the gear you want available for demo, and they let you try enough different combinations, you might find some combo that's "synergistic", but how realistic is that?

To Elnero, please see above. I'd say anything under about 8 ohms is (marginally) acceptable with conventional dynamic headphones. Under 2 ohms is best for balanced armature, multiple driver, or other more esoteric cans. There are more examples on my blog such as this one which shows the Behringer with it's original 50 ohms and after being modified at around 2 ohms (note the 14 dB of total deviation for the original):

Different headphones perform differently with different amplifiers. Full stop. The notion that one perfect amp can perfectly drive everything is simply nonsense. For instance, with a traditional voltage amplifier, I think Grados do best with 10-15 ohms of Zout, balanced or SE. With an amp with near 0 Zout, they sound terrible. And this is a large part of the reason for the "Grados work well with tubes" myth in that tube amps tend to have a non-0 Zout. However, I also think they really shine with a Gm amp with a Zout in the hundreds, if not thousands, of ohms. Those same amps may not do well on other brands/designs. So, finding just the right synergistic combo would be a lot easier if we started matching gear in reasonable ways, rather than treating every headphone as a black box to be driven the same as every other one. Nobody expects a 45 amp to drive their B&W 801's, and nobody expects their Krell monstrosity to drive Loethers, so why do we expect this in the headphone world?

Anyhow, this is a great thread so far

[The Monkey]

Anyhow, this is a great thread so far

Agreed. This is an issue that I've never understood, and though I barely understand it now, at least I know more than a few days ago.

[luvdunhill]

nobody expects their Krell monstrosity to drive Loethers

[digger945]

(ouch)

You could always get another amp.

[NwAvGuy]

Different headphones perform differently with different amplifiers. Full stop. The notion that one perfect amp can perfectly drive everything is simply nonsense. For instance, with a traditional voltage amplifier, I think Grados do best with 10-15 ohms of Zout, balanced or SE. With an amp with near 0 Zout, they sound terrible. And this is a large part of the reason for the "Grados work well with tubes" myth in that tube amps tend to have a non-0 Zout. However, I also think they really shine with a Gm amp with a Zout in the hundreds, if not thousands, of ohms. Those same amps may not do well on other brands/designs. So, finding just the right synergistic combo would be a lot easier if we started matching gear in reasonable ways, rather than treating every headphone as a black box to be driven the same as every other one.

I can't argue with the above, beyond the subjective aspects of things like "grados do best with 10-15 ohms", etc. One's man pleasure is another man's poison in that regard. But I can see how there are some general preferences the majority might agree on.

I've said all along, the impedance thing largely comes down to what the headphones are designed for (assuming you agree with how the designer wanted them to sound). The sad thing is, unless it's zero ohms, hardly anyone will end up using those headphones with the designer's ideal impedance target.

It really is possible for all practical purposes to design a "one-size-drives-all" headphone amp. And it doesn't even have to be expensive. Its output won't change in audible ways with any headphone you want to throw on it. That doesn't mean all those headphones will be a great match, but the amplifier part is easy. So, from my perspective, it's frustrating this is way more complicated than it needs to be.

If someone back in 1970 at the IEC said "thou shall make headphone amps with an output impedance less than 1 ohm" we would have nearly all headphones today designed for such a source. And this whole "synergy" thing would be way more simple. From Grado to Sennheiser they would all work as designed on any source that met the standard which would be nearly all of them. Wouldn't that be huge improvement from the compromised mess we have today?

As Tyll has said, if the designers could count on high electrical damping from a low impedance source they could optimize their designs around that. But sadly they can't. Few designers have any freakin' idea what their cans will get plugged into. It might have a zero ohm output, 10 ohms, 250 ohms, or whatever. So it's all a big compromise. And when you're trying to design high-end ultimate gear compromises suck.

I suppose some headphone enthusiasts may enjoy the endless variations they get due to impedance mismatches when they swap gear around. But surely there has to be a better way to fine tune the sound than random impedance mismatching?

Speakers have been designed since the 70's to work with a zero ohm source. There's absolutely no reason the same thing can't be done with headphones. It would make for much more consistent results. Personally, I hope we continue to head in that direction.