Headphone and Amp Impedance (2011) | Hacker News Reader

Headphone and Amp Impedance (2011)

nwavguy.blogspot.com

117 pointspmoriarty57 comments

Open articleView on HN

This guy also built and open-sourced a universally admired headphone amplifier and DAC.

His O2 amp: http://nwavguy.blogspot.bg/2011/07/o2-headphone-amp.html

His O2 DAC: http://nwavguy.blogspot.bg/2012/04/odac-released.html

JDS Labs and others sell them commercially for those who don't feel like building them by hand.

There is an interesting article about him, his amp and the resulting problems due to his disappearance:

http://spectrum.ieee.org/geek-life/profiles/nwavguy-the-audi...

He/She has also inspired some other manufacturers to rethink their products. IIRC Schiit was formed when some engineers thought they could improve the O2/ODAC design on their own[0]. I own an original Schiit Stack (Magnus & Modi) and would definitely recommend them.

[0]: https://news.ycombinator.com/item?id=8834176

Not exactly, you may want to actually read the source you linked :)

The company existed before O2 and M/M were independent designs.

Thanks. I had read that someplace else and went digging for the link and simply assumed that was it.

I wonder what nwavguy does for a living these days. He disappeared from the Internet in that persona shortly after the release of his designs and dropping some serious allegations at some manufacturers for a transient overvoltage issue. He's a solid engineer and I hope he's just quietly improving the state of audio somewhere.

On a tangentially related topic:

When I used to listen to music on my previous phone (a Nexus 4) I would have to constantly adjust the phone's volume: all the way to maximum when using an aux cord in my car, and way back down when using headphones. Recently I upgraded to a Nexus 5X, and noticed a surprising new feature. Somehow, the phone detects which device it's plugged into, and remembers a separate, previously-set volume level for each device.

Does anyone know how this works? It's really convenient, which makes it surprising that I can't find it documented or even mentioned anywhere online. I'm assuming the audio hardware is somehow sensing the impedance of whatever it's connected to, but is that really a reliable indicator of headphones vs. line-level inputs?

Line-level inputs will be in excess of 1 KOhm (10KOhm if properly designed) Headphones will be up to 600 Ohm (I guess some boutique brands could have examples higher but I digress, the grand majority is between 16 and 250 Ohm).

In my opinion as an EE there would be some reasonably reliable ways to detect either a line or headphone connection. Headphone impedances are typically between 16 and 600 Ohms, and line impedance is typically in the 10's of kOhm range. That's one way to differentiate. There has to be an output signal (presumably audio) first, however. Another way is to apply a very small DC current (that would certainly not damage headphones, so I would guess in the micro-ampere range) and check if there is continuity or measure the DC resistance of the load outright. For line inputs, as they're usually AC-coupled, DC resistance would be infinite, and for (DC coupled) headphones it is very low. It's an interesting subject and I wouldn't be surprised if there are some patents involved.

Detecting AC vs. DC coupling would be obvious to a practicing EE, if presented with the problem, yes? I'm just a software guy with an electronics hobby and I think I could have figured that one out.

That said, DC vs. AC coupling detection could probably be granted a patent if nobody had "disclosed" the idea prior to filing of such a patent, based on my experience with software patents.

Sorry to vent, but patents on this sort of thing drive me crazy. I don't have any gripe with you or your post.

I did a quick search on google patents: https://www.google.com/patents/US9084035 - it uses a combination of AC and DC methods.

> For line inputs, as they're usually AC-coupled

Headphone amplifiers often have DC-coupled 10k potentiometer from input to ground with washer feeding a DC- or AC-coupled gain stage. So much for AC/DC detection.

I think impedance measurement is used in practice. I can hear a quiet beep when I connect headphones to one of my laptops, for example.

A uA DC current can leak through a shitty AC-coupling capacitor, actually measuring the load is probably best.

Most devices just distinguish between TRS (Tip-Ring-Sleeve, or regular stereo mini jack) and TRRS (with an extra ring) connectors, with a separate volume level for each.

At least some MBP's and iPhones do this too (e.g. mine). In my opinion (though less reliable than the sibling's because I'm an ME not a EE) impedance detection is probably the easiest way to do this.

This guy is fantastic. Having spent a little time around professional audio, it doesn't surprise me that he did amazing stuff for 18 months (anonymously) then vanished from the web.

This piece on his site has lots of interesting insight into the audio industry. It also relates to objective, critical thinking in general.

http://nwavguy.blogspot.com/2011/05/subjective-vs-objective-...

He had an interesting mind. Absolutely too objective, in my (subjective) opinion. It was almost as if his entire purpose was to remove the validity of all opinion from the discussion completely, and frankly, it was sad and inhuman—despite his amazing technical contributions.

I think his viewpoint is spot-on. The goal of most audio systems should be to accurately reproduce the original waveform. If you want to add "warmth" or "color" or whatever to the sound, it's better to quantify what that means, then purposefully distort the audio in a dedicated device (or in software).

Think about what life would be like if displays were built and reviewed like audio tech. Imagine companies like NuForce and Schiit sold screens that didn't accurately reproduce colors. Owners of these displays would praise them for their "warmth" or talk about how much green colors "pop" or whatever. It would be absurd.

We can actually measure contrast ratio, color accuracy, viewing angles, etc. That's how we evaluate displays. People who prefer a different gamma or white point change it in software. They don't buy defective monitors. But switch the topic to audio and people forget all that. It's bizarre.

I greatly respect NwAvGuy and agree with almost everything he says, but I think in practical terms that's a slightly over-simplified perspective. Look how inaccurate even these headphones are [1], arguably the best or one of the best ever made.

Getting the inaccuracies to be the right inaccuracies is quite subtle and very important in my opinion. See for example this review of his [2] and in particular the paragraph near the end:

>AN EQ EXPERIMENT: I tried using highly configurable parametric EQ to tweak my Denon D2000s to be similarly forgiving and I failed. When I filter out equivalent amounts of sibilance and harshness with the Denons, more of the music is gone and the result sounds veiled and artificial. The HD 650, to my ears, somehow works more like polarized sunglasses. They largely remove what I don’t want while revealing more of what’s left. It’s not just a simple frequency response difference. My objective side wants to pin down what’s going on but I don’t have a full explanation besides the HD 650’s lack of distortion.

[1] http://www.innerfidelity.com/images/StaxSR009.pdf [2] http://nwavguy.blogspot.co.uk/2011/06/sennheiser-hd-650.html

I was mostly talking about DACs and amplifiers, not headphones. Headphones have to deal with more trade-offs: comfort, size, isolation, and the fact that it's really hard to get them to accurately reproduce sound. Still, so what if current headphones suck at reproducing sounds? That doesn't mean we should ignore quantitative measurements. If the "right" inaccuracies are so subtle-yet-crucial, it's important to quantify them! Then we can evaluate headphones based on how accurately they create those inaccuracies.

And I think you're missing the point of the EQ experiment. NwAvGuy wanted to see if there was more to quantifying headphone quality than a frequency response curve. He was right. That just means that objective evaluations of headphone quality will need to include more measurements. That doesn't mean we should disregard measurements and go with subjective impressions. Otherwise, something as mundane as a warm drink can skew the reviewer's opinion.[1]

1. http://news.bbc.co.uk/2/hi/health/7688315.stm

With speakers and headphones, low distortion probably is more important than a flat response. Bad response can be EQed away, but distortion can't.

I think it's because with displays, you can put two of them side by side and say, "look, the shadow detail is gone in this part" but with audio you have to rely on your memory.

I think you just nailed why judging audio quality is so difficult. You can compare video frame by frame, or look at a still image for as long as you need to, but audio is always disappearing into the past as you listen to it, and the closest you'll get to freezing a moment in time is an annoying buzzing sound.

Or, you pop one in your left ear, and the other in the right ? Never tried, but I suppose this could work ?

Then again our brains play soooo much with sound-proccessing ...

The waveform you are trying to reproduce, though, was mixed and mastered by a guy on studio monitors or similar, and in such a way that it would sound appropriate on the systems his audience is going to use.. or some version thereof. He's not pushing things to the limit of what his studio setup can produce, because that's going to sound like crap anywhere else.

What you listen on to master isn't necessarily what you'd listen on for fun.

If you were taking raw recordings of something with mics with really flat response curves, and wanted to play it back as accurately as possible, this would hold - but that's not the case. It's all about what sounds good to you.

Again, consider displays: Similar to music, movies are mastered on high-end gear. Staff use properly-calibrated displays. They often view scenes in dark rooms. Similar to music, movies are mastered so that they won't look terrible on the audience's displays. Some dark scenes are made a little more contrast-y. Some colors are tweaked. Maybe even the white balance is altered (though this is often more stylistic than corrective).

Now if you're buying a display, are you going to ignore measurements and just go with whatever looks good to you? Hopefully not. Even if you prefer bluer tones or higher contrast, you'll want to know numbers for white point, color accuracy, and contrast ratio. You won't just eyeball it in a store. You definitely wouldn't trust subjective reviews where someone says, "This display seems more contrast-y and blue." It shouldn't be any different with audio gear.

A counter argument I'd give is has anyone ever described a TV as being too harsh, as in they can't watch it for an hour without experiencing "fatigue" that causes them to need to stop. Of course with a TV this sounds ridiculous we just want the most life like picture possible so these measurements are a great way to accomplish that.

With headphones though while something may sound more accurate it might not be enjoyable to listen to. It can be too harsh to our ears and cause listening fatigue. While the Sennheiser HD800 are a great pair of headphones I find them very difficult to listen to on a neutral amp for long periods of time. The only way I could really enjoy them was to get a warmer tube amp. In the end I replaced them with the Audeze LCD-3 which I would argue are less accurate headphones but are far more enjoyable.

Heh... Folks use software like f.lux to purposefully distort their displays in order to reduce fatigue.

The ability to adjust the output to one's liking should not be confused with an inability to faithfully reproduce input.

That's a pretty good comparison. In both cases you are adding warmth to a source one via software and the other via an amp.

I posted this 5 days ago. Read it and tell me that an objective-to-a-fault voice isn't cause for celebration :)

https://forum.audiogon.com/discussions/anyone-own-an-audiodh...

Should have known better than to expect HN to have any respect for humanity, art, or subjective experience.

Bigger point, the human ear is one of the most varied and wildly inconsistent of our senses, while at the same time being highly accurate and able to discern tiny details fairly well. It tricks us, and as such it's difficult to stick to science alone to describe what we "hear" because "hearing" is a whole lot more complex than simply "accurately reproducing." And that should be celebrated.

In other words, I feel sorry for all you heartless computer minds, even as a computer scientist myself; you have no humanity in your unabashed technocracy, and you are the worse for it.

To an outsider like me it looks to me like all the basic things about audio have been solved and competition is only really about things you hardly can perceive or total hocus-pocus. Still some people seem to still earn quite some money, so it's no wonder that there's a toxic environment of people trying to protect their income source when it's not really backed up by offering a product that's really better than something much cheaper.

Circuit theory becomes much more elegant if you start to use complex numbers (conceptually, the Laplace transform [1]). But for an introductory lesson, this page seems nice, though perhaps a bit long.

[1] https://en.wikibooks.org/wiki/Circuit_Theory/Laplace_Transfo...

(If you grasp this reference, then you understand, I'd say, about 25% of what electrical engineering is about)

But only if you use linear components (meaning L,C,R). Introduce just one diode and the system of diff equations needs to be solved numerically. In which case you have switch to simulations like spice anyways. Therefore there is no reason to even bother with Laplace transforms nowadays. For hobbyists (and the vast majority of professionals) I would consider it completely useless knowledge.

> Introduce just one diode and the system of diff equations needs to be solved numerically

Not really. Maybe that's why nobody invented anything with a diode in the signal path before computers were commonplace right? Like radio and guitar distortion pedals

> Therefore there is no reason to even bother with Laplace transforms nowadays

Sure, let's forget the math and make engineers totally dependent on numerical simulation without knowing the basics, I'm sure that'll work just fine

> For hobbyists (and the vast majority of professionals) I would consider it completely useless knowledge.

LOL

Sure, in the same way that sorting algorithms don't are useless for computer science students, right? Just call .sort() and you're done!

Also don't forget Digital Signal Processing, that also relies on that "useless knowledge"

> Not really. Maybe that's why nobody invented anything with a diode in the signal path before computers were commonplace right? Like radio and guitar distortion pedals

No, those problems were solved via ultra simplified models. E.g. The current through a diode is described as I = I_0 \Theta(U - 0.7 V). And the rest was done via try and error.

> Sure, let's forget the math and make engineers totally dependent on numerical simulation without knowing the basics, I'm sure that'll work just fine

How about, lets learn how to solve every differential equation you will ever come across in electronics using e.g. Runge-Kutta.

> Just call .sort() and you're done!

If you want to get shit done you do exactly that. This is a great problem of many fresh CS graduates. Instead of doing the actual work they would rather first implement their own sort function. In 99.9% of the cases it is a complete waste of time.

> Also don't forget Digital Signal Processing, that also relies on that "useless knowledge"

In Digital Signal Processing you mainly need to Fourier transform and know the properties of the sin(x)/x function, Nyquist, Shannon etc... Learning how to solve a very special case of systems of differential equations using the Laplace transform won't help you there at all.

>Instead of doing the actual work they would rather first implement their own sort function.

Strawman much. On the otherhand I regularly come across people who call library functions without understanding their runtime consequences, well those are the good ones, the bad ones 'get shit done" by calling sort just to find the maximum. Yeah they do get exactly that done, that is, shit.

I could have done with reading this article a week earlier.

I just built a converter to allow earphones to be used in aircraft audio panel outputs which are typically 150-350 Ohms. Another bizarre thing I learnt is that 1/2" and 1/4" stereo jacks are wired opposite ways.

How did you learn that 1/2" or 1/4" head phones would be wired "opposite ways"? I would doubt that this is the case.

Not sure if we are talking about the same thing, but on a flight with a Boing 777, I found the throw-away headphones that they hand out upon boarding with the very strange connector (dual mono jacks) to be so bad as to be nearly useless.

I tried to cut the cable, remove the isolation and connect my Apple earbuds, which worked somehow but not very well because the (very thin) wires were individually isolated. Did you come up with a solution?

There are adapters you can buy for like 3$.

Nice, good and cheap amp ($19) mentioned at the bottom of the article, in case you need immediate boost over stock audio from your phone - http://nwavguy.blogspot.bg/2011/03/fiio-e5-headphone-amp.htm...

Or when you are annoyed by a shitty soundcard in your business notebook, get a Behringer external DAC/soundcard for $29 ( http://nwavguy.blogspot.de/2011/02/behringer-uca202-review.h... )

I have a UCA202 (and a UFO202 - AFAIK it is identical but with a phono stage on the line ins) to use with my Surface (the build-in card is very noisy) and it is a noticeable improvement over the built-in card. Definitely recommended if you need a USB DAC without spending a fortune.

I use an Acer C720 Chromebook running Linux to drive my virtual piano software. I had no complaints with it's on-board audio performance, but after reading multiple exhortations on the internet about how much better external soundcards are, I bought a $5 USB soundcard widely available from multiple vendors on ebay. OK, I can just barely hear an improvement over the on-board. It's better in the sense of "significant" (ie definitely there), but it just isn't a material difference to me.

If I'm that deaf to the differences between these two sources, I'm curious if people think the UCA202 will change my life?

Really depends on the quality, noise floor, and hiss of your on-board sound-card. I've had motherboard where I tried the on-board card, didn't take me long to put in my old PCI card again.

Also depends on how "hard" your audio equipment to feed. You don't go hanging big speakers directly to your PC kind-of-thing.

The noise floor is the main issue with most cheap sound hardware. It's often loud, and not spectrally flat. You'll be able to hear more of your music at lower volumes if you get a high quality DAC with a low noise floor.

The other big issue is distortion/nonlinearity. This is much less of an issue than the noise floor, but can still be noticeable on cheap DACs.

It's likely that you can do a lot better than a $5 DAC on both of these factors.

I currently use a Schitt (http://schiit.com/) Modi and Magni stack with my Beyerdynamic DT880 600 ohm headphones. When I tell you that music is a completely different experience at home with these cans, believe it. I'm a DJ and honestly prefer staying in and listening to mixes than going to my mate's parties to hear what they're spinning because they absolutely obliterate any live system (and I've been to some of the best venues in North America.) I know it's a poor comparison of course, but I'm never going back to monitor speakers for personal use again. If you're looking for a decent AMP/DAC combo, I highly recommend Schiit.

NwAvGuy has some not-very-nice things to say about Schiit. Basically, that they don't know how to design audio gear. He discovered that one of their amps could damage your headphones. For this warning, he was banned from the Head-Fi forums: http://nwavguy.blogspot.com/2011/07/banned-at-head-fi.html

Whoa, never spotted this. Thanks for the relay

Not only banned him, but tried to deny his findings at first.

NwAvGuy has his side of the story, Schiit has theirs.

http://www.head-fi.org/t/701900/schiit-happened-the-story-of...

Personally, I'm extremely satisfied with the variety of Schiit gear I own, from my work Modi/Vali setup to my home Ragnarok/Yggdrasil

I have some issues with Schiit's accounting.

First, that much DC is simply unacceptable, even when removing power from an amp. Amps don't just lose power because they're deliberately turned off. Sometimes people trip over a cord. Sometimes, a circuit breaker trips. Sometimes, the power goes out. Sometimes, batteries die. None of this circumstances should endanger downstream equipment.

Second, the fix is extremely simple. As Schiit admitted, it took less than a day to develop and ship boards without this issue. There is only one reason why a company would sell a device with this problem: incompetence. If they'd cop to screwing up, I'd respect them. Instead, they dissemble and equivocate.

Third, so what if their own experiment didn't manage to destroy a pair of expensive headphones? "I can't reproduce this failure." doesn't mean, "This failure can never happen." The headphones only survived because they were designed with extremely robust safety factors. Even then, it's quite likely that Schiit's amp damaged the headphones. 1-2V of DC is way out of spec and has –in some cases– damaged and destroyed equipment. There is video of Schiit's amp distorting headphone drivers. That's crazy. No equipment should be doing that.

Fourth, "other audiophile gear also does it" is not a valid excuse. As NwAvGuy has shown, Sturgeon's law[1] applies to this stuff. Just because there's lots of terrible gear out there, that doesn't mean it's OK to produce more terrible gear.

After reading Schiit's accounting, I'm certain they lack the engineering knowledge to build safe, effective audio equipment. I think their intentions are good, but there's simply no excusing such a glaring oversight. In my book, Schiit is an anti-brand.

1. https://en.wikipedia.org/wiki/Sturgeon%27s_law

Exactly. The horse parent is beating has been dead for years now.

It's been fixed and they've sold thousands of devices since then.

On the one hand, any loudspeaker amplifier with feedback is going to have very low output impedance (minimum acceptable damping factor of 20 with an 8 ohm load or .4 ohms max). So for at home, AC powered use just hook your headphones to the speaker outputs.

On the other, things do get interesting with low power / battery powered amplifiers, for example where they use a low power OP-amp for headphone drivers. For this case, the article is very relevant.

Heh, or get headphones with nearly flat impedance, for example Sony MDR-v600: http://www.cliftonlaboratories.com/headphone_impedance.htm

Good grief, headphones do not need electrical damping. This is relevant to large speakers, and only at around their resonant frequency (as mounted in the cabinet). If you're getting some funny resonance in your headphones, electrical damping isn't going to do anything; it's a mechanical issue.

one further point to keep in mind is the rule for the inductive reactance of a device (either an amplifiers output or a speakers input) is stated for 1000Hz which is an old rule for the mid range of audio for the clear hearing of intelligible speech, and of course the most efficient transfer of energy is when the output of an amp matches the input of the device it is driving.

To understand a given amp and speaker you also need to know the inductive reactance for the full range of frequencies you want to use.. so amp and speaker maybe matched at the rated impedance is at 1000hz but way off at 15khz

Reply on news.ycombinator.com