Headphone and Amp Impedance (2011)
nwavguy.blogspot.com
nwavguy.blogspot.com
I somehow managed to convince myself that a pair of Genelec 8350s would solve all my problems, so I replaced my Presonus Eris E5s with them, and replaced the scarlett 2i4 with a Yellow Tec Puc2 lite, and got the calibration box and mic and the silly volume knob (so I could pour pure PCM at 0dB straight into the monitors DACs).. Of course, audiophilia is a sickness, and it never ends, so I've been looking at a rather pricey stereo-subwoofer setup to go along.. but fortunately, my room can't support that at the moment..
I guess what I hate about gear, is what others love, that there's no one right, no one truth, and that subjective experience has to be taken into account.. I don't want subjective, I want the objective truth, accurate reproduction.. but what is even that..
https://customcans.co.uk/shop/product-category/home/headphon...
Not quite the thousand dollar Arya but I like the dampening/venting work and signal testing he did + nice cables + I switched to a balanced xlr setup with my amp as I have grounding issues in my house.
I personally find the amp the most interesting part over headphones. I really like the tone of tube amps. It's fun testing out the (noticeable) differences. As long as headphones are relatively flat I'm happy. Not sure the planar thing is as dramatic as people say but I only have a modified entry level one.
If I was going this path again I'd probably get a (modified) 990 Beyer again cuz I loved that one and I don't find much value in spending a couple K since it's never just the headphone purchase... I know from experience a new fancier speakers/headphones means you then notice how bad your amp is, or your phono preamp, or turntable, or your DAC, etc and you invest a tone of time/money on upgrading everything.
That's certainly an interesting experience, I've heard that a good pair of speakers costs much more than an equivalent pair of headphones. You can get "audiophile" level Sennheisers/Audio Technicas for a few hundred dollars, but that's close to the starting price for any half-decent pair of speakers, with "good quality" setups probably costing in the thousands to tens of thousands of dollars.
There are a few things you could do to make headphones more speakers-esque - use an external subwoofer or rumbler, add a bit of cross/mix between the channel and perhaps use a dsp plugin for HRTF (Head related transfer functions). Assuming your amp/source is good, it should definitely smoke a fairly more expensive speaker system out of the water.
Well I have some great news for you!
In 2024, you don’t need to spend mega bucks to get a good DAC or amp. You might not need one at all!
One of the better performing and measuring DACs out there right now is the $9 Apple USB-C dongle: https://www.audiosciencereview.com/forum/index.php?threads/r...
I’m not suggesting you need to rush out and buy it, but it’s a good reminder that modern tech is so good that even a $9 Apple part measures so well that nobody could hear the difference between it and a more expensive DAC.
As for amps: The amp story has been exaggerated a lot. A $70-$100 amp could give you more power and better performance than you could ever need (barring exotic headphones that need specialty amps). That’s if you even need one at all.
Don’t drink the kool-aid or take advice from people who learned about Hi-Fi from 1990s or 2000s tech.
> I guess what I hate about gear, is what others love, that there's no one right, no one truth, and that subjective experience has to be taken into account..
Sadly, subjective and placebo are two sides of the same coin in the audiophile world. People will swear up and down that their $1000 amp or DAC or cables are superior to their $100 version, but the claims disintegrate under blind testing conditions.
I built my stack based on the reviews there, and I've been completely satisfied with my Topping D10 (for DAC) and JDS Atom (for an amp), which cost me like $200 total a few years ago. But of course, there's already a newer model of the Atom that seems to blow the old one out of the water: https://www.audiosciencereview.com/forum/index.php?threads/j...
And btw OP, for headphones I'm using DT1990s, which might also be worth a listen if you are looking to upgrade from the 990s :)
The Atom doesn't hold up as well as when it launched in today's market at that price point. That DX3Pro+ is a strongly measuring dac+amp single unit at the price.
As for HD600, they're the time-proven, neutral, uncolored, accurate timbre kings. Great overall balance.
0. https://www.audiosciencereview.com/forum/index.php?threads/t...
And I concur regarding the HD600, it's a great choice for many people. I was trying out both a HD650 and the DT1990 before making my choice, and for me the 'precision' of the 1990s was just too enticing compared to the laid-back sound of the 650s -- but it was a hard choice, because sometimes you don't want precision, you just want comfortable listening, and there the HD600/650 really excel in my opinion.
Shows how bad the entire industry is. Especially on the DAC side. Lots of snake oil that measures worse than very cheap solutions.
In practice, the amplifier matters more, and there the Apple dongle doesn't do so well.
-Which, cynically speaking, is really a requirement. After all, many manufacturers of such gear claim a 'house sound' - but how do you differentiate between sonically transparent gear?
You don't. Hence you need to add some imperfection to the sound in order to claim an (audible) edge over others.
Or, as (signal processing major) me liked to quip to audiophile friends - "Tube sound? Tube sound can be DSP'd!
- 2 of my 3 monitors audio outputs: Good
- Desktop rear audio port: Good
- MacBook Pro audio port: Good
- Third (different model) monitor: Usable
- Desktop front audio port: Usable
- Framework laptop: Usable
- Nintendo Switch: Poor
- Surface Go tablet: Poor
Anything else and as you’ve noticed you’ll always be chasing for something “better”.
When it comes to DACs and amps its just simple: get the one that has the best signal to noise ratio. Usually they should not alter the sound profile at all.
Also most DACs you will find on audiosciencereview for example have such a good signal to noise ratio (>100 or whatever) that is is imperceptible to the human ear anyway. You can go higher, but you won't hear a difference. And some of these cost less then 100$ or even less.
I would go for something that looks ok, is not too large and fulfills your user experience preferences.
In my example I bought separate DAC and AMP to have the best signal to noise ratio, but I have more cables and 2 boxes on the desk and the AMP is huge. I would rather go back for an integrated solution that maybe can also drive my monitors. But here we are...
it's not all about perfect measurements.
then, of course, if you buy into all the snake oil it's a different story ;-)
There is no terminology, that's the problem. What is "airy"? What is "warm"? You're right, subjective. We can only care about what can be measured.
Thank you for helping diagnose one of my obsessions for me.
But you already found the objective truth: people experience it differently.
Maybe it's not the truth that you wanted but rather whatever that you wanted to be the truth
[0]: https://nwavguy.blogspot.com/2011/07/o2-headphone-amp.html [1]: https://nwavguy.blogspot.com/2011/02/headphone-amp-impedance...
https://support.google.com/blogger/thread/226117604/my-blog-...
(also it turns out that these jacks briefly short-circuit as you plug and unplug them)
because 'super simple' excludes power negotiation schemes like usb-c, whatever power i deliver to small devices is also available if someone plugs in a resistor, and earspeakers are pretty similar to resistors at dc, and normally driven at only about a milliwatt
the vast majority of current headphones (and headphone-driving amplifiers) out there today conform to the android specs (though probably approximately 1% of the headphones of interest to nwavguy)
jack: https://source.android.com/docs/core/interaction/accessories...
plug: https://source.android.com/docs/core/interaction/accessories...
relevant appnotes on driving them include david guo's https://www.analog.com/media/en/technical-documentation/appl... and the anonymous https://www.digikey.com/Site/Global/Layouts/DownloadPdf.ashx...
It will cause some noise issues, because a very high impedance front end will pick up stray radio transmissions, but I think it's doable given a small enough bandwidth
noise is pretty irrelevant on the power rail. but if i put a high constant impedance on the power rail—say, a 100Ω resistor in front of 5 volts—none of the devices powered from it can draw much power, 25mA being the limit in that case
(for the data line, sure, resistance is not at all futile, resistance is a perfectly good solution, along with some clamping diodes)
what i have in mind is two things:
1. maybe use a trrs connector and use the second sleeve for the power rail? that way the only thing that ever gets plugged into it by accident is a microphone, which i think is unlikely to burn out at 5 volts (even if normally it's not subjected to more than 2.2 volts)
2. also, limit the current with a linear current source. this design simulates as having about 10Ω impedance over the 0–100mA range and a few kilohms once you exceed the current limit: http://tinyurl.com/23qvuylw
> 'super simple' excludes power negotiation schemes like usb-c
like, i think you're thinking about devices one or two orders of magnitude bigger than what i'm talking about. i want to be able to wire up a 35¢ device like an attiny4 https://www.digikey.com/en/products/detail/microchip-technol... with maybe three 1¢ resistors and two 1¢ diodes for protection http://tinyurl.com/26ngjdzh or maybe just soldered directly to the jack. it has 512 bytes of flash for the code and 32 bytes of ram; that would have to contain both whatever is necessary for the communication protocol and whatever application i want to run on it (in that case maybe it could handle a capacitive touch button or run a couple of 5-volt pwm channels)
i'm not sure the attiny4 is powerful enough. but i'm pretty sure i can make the attiny45, the ch32v003, the pms150c, the attiny202, the stm8s003f3, or the mb95f264hpft work
otoh in usb-c a lot of the power negotiation consists of things putting particular resistors across the line, which you could do even at that small granularity. a bus topology won't let you do that, though; if you had five devices plugged in, all their resistors would be in parallel
(incidentally, despite the name, usb is not a bus; it's a bunch of point-to-point links. that's why it can do that)
That sounds like an all around bad idea.
and, yeah, square. i want to do higher data rates and simpler electronics than an actual audio interface can handle, though
And for people concerned about accidentally shorting a power line when inserting the plug, hypothetically could use a socket with built in switches so power is only sent once the plug is fully inserted into the socket.
it would still probably burn out my earbuds if i put 5 volts at 100 milliamps on the tip connector once the plug was fully inserted
I am hesitant to give advice because I don't know a great way, but a possible method of detecting what type of device (a headphone or USB device) is plugged in could work by doing an impedance measurement right after the device is inserted. USB standard specifies that the signal-to-ground impedance on the data lines is 30 ohms and between data lines is differential impedance of 90 ohms. The right and left channel of headphones meanwhile have an impedance to ground (from 4 ohms to 32 ohms or higher), but aren't set to a differential impedance between the left and right channel. So maybe there is a simple circuit that could measure the differential impedance that is triggered just after a device is inserted, and if it matches 90 ohms, then will treat the device as a USB device, but if a measurement matches a headphone, then treat the device as a headphone. But the specifics on how to do this, I'm not sure.
thank you!
as for usb, i won't be using it
i think maybe i'll just go with trrs to avoid this problem
- cables are available even from hardware stores, and also so are bus adaptors which connect two or three or five cables together, pin for pin; these do exist for 3.5mm jacks but are less common
- spiral 4p4c cables are cheap and easily available
- it doesn't short the power rails together as you're plugging it in
- the cables are all pretty high quality, which is not true at all for the 3.5mm audio cables
they also have some disadvantages:
- the cables are much larger than 3.5mm cables
- the jacks are much larger than 3.5mm jacks
- instead of absentmindedly plugging earphones into my data logger or programmable load and blowing up the earphones, the danger is that i (or a child, say) might absentmindedly plug the data logger into the phone jack on my cable modem, which is something like 50 volts on-hook and 90 vac when it rings
- 3.5mm jacks are an evolution of phone switchboard connectors, so they should be good for thousands of insertions and the plugs for like a million or so, and i'd be surprised if rj-11 or 4p4c connectors made it to hundreds
You could use 4P4C, the handset connector, instead of RJ11/6P4C, the phone line connector. Reasonably common, and no risk of ever plugging in a phone line.
(i assume by 'phone jacks' here you mean rj-11 and 4p4c, but https://en.wikipedia.org/wiki/Phone_jack lists cylindrical 3.5mm and ¼” 'phone connectors' as well, because they originated in telephone switchboards)
you can actually plug a 4p4c plug into an rj-11 jack, or even an rj-45 jack; it enters and holds securely. it may damage the rj-45 contacts, but it will fit. but it's probably true that it's a less likely error for someone to make in practice, and i wonder if maybe the higher on-hook voltages aren't even passed through to the handset
another plus of the modular phone connectors is that they're significantly less work to attach to a cable than a 3.5mm plug, once you have the right crimping tool anyway. (you'll get a shitty connection until you learn to do it right.) otoh, the 3.5mm plug doesn't need a crimping tool
basically in many ways usb c represents the opposite of what i want to achieve
however, it does have some real merits! the other 10% of the market has some really excellent devices which can reliably deliver many watts of power over relatively lightweight, flexible cables; the connectors are quite thin and lightweight, more so even than 3.5mm jacks; most usb-c devices can charge from usb-a chargers and talk to usb-a data ports, where the data protocol is still a fucking nightmare, but at least it's a widely implemented fucking nightmare that delivers multiple megabits; and even if you do have to insert the connector at the right angle, at least now you can do it without looking, unlike usb-a, mini-b, and micro-b
Which part?
>data protocol is a fucking nightmare
How come? I’ve done many projects with type c and never felt that way. For basics (5v/1A) it’s just two resistors. For higher voltages I stick a pd controller that handles it.
one problem with the usb data protocol is that it won't work over a bus; it requires point-to-point links. but, more generally, my complaint about the data protocol is this. i want two devices with no usb hardware, clocked at 8 megahertz ±10%, with a kilobyte or less of code memory, to communicate with each other. there are multiple reasons that is just never going to happen over usb. usb requires timing precision of ±0.5%, so a crystal or at least a ceramic resonator is mandatory (not a reasonable option on an 8-pin micro), and v-usb requires over a kilobyte of code memory to run at all. and even v-usb only provides a slave device! even if all your devices are avrs with crystals, you can't run a wire between two devices running v-usb and have them communicate. and bitbanging a usb master requires, as i understand it, several times more code (i've never done it)
also, v-usb won't let you do anything else while you're communicating; it needs every clock cycle
so, i want something that's about 3% or less of the complexity of the usb protocol, with timing specs that are 40 times more relaxed
i agree that if you only want power and not communication, and your durability requirements are not high, usb-c is a very practical option. and, even with communication, it's an even more practical option for connections at a larger granularity than what i'm talking about, where the devices involved are big enough for one of them to be a usb master. but that's a different order of magnitude: tens of kilobytes of ram on each endpoint, not tens of bytes
This doesn't seem to disagree with what you're saying?
A few years ago, I've built attenuators out of voltage dividers (using around 2-20 ohm resistors). Unfortunately at any given attenuation factor, there's a limit to how high of an impedance you can expose to the audio source, along with how low of an impedance you can expose to the headphones. Additionally, to compensate for stray ground-line resistance causing inverted crosstalk between the stereo channels, I also add a crossfeed resistor or potentiometer of 500 ohms or so, ideally tuned based on the headphone impedance. If you use a potentiometer you can take one side of your headphones/earbuds off, send loud noise to that side alone, and turn the potentiometer knob until no sound is audible in your remaining ear. (To determine the correct fixed resistor value, you can send quiet noise to the remaining ear and adjust its volume until it cancels out crosstalk.) Also since I use TRS cables/sockets, this breaks headphone mics.
I've been meaning to make a PCB version of my attenuator with SMD resistors to reduce space, but don't know where to find a reliable PCB-mount headphone jack (or 3.5mm dongle cable). If anyone has suggestions for a PCB-mount headphone jack (available in the US), let me know!
A. 32-100 ms depending on model.
(This is a joke.)
Edit: Amp types as pointed out by another commenter.
Some headphone amplifiers use cheap integrated circuits that can only drive outputs no lower than around 150 ohms, like the JRC4556 (a device that NavGuy has written quite a bit about, incidentally).
I suspect that's where that idea might come from? If you have a JRC4556 that doesn't go lower than 150 ohms, and the headphones are 32 ohms, you can pad in some series resistance to make up the difference.
The concern with the resonance of the speaker is overblown.
When you boost a frequency range electronically with an equalizer, that is also resonance (electronic rather than mechanical).
Instead of your speaker ringing for a couple of cycles, the circuit is doing it.
The best thing is for the speaker manufacturer to assume zero ohm output impedance from every amplifier, and for them to do their work to get their speaker to have a close to flat response. (Or whatever: some markets prefer a huge bass.) Then it has that designed response in any such amplifier.
Because the effects are negligible and trivial to overcome with even basic attention to design.
The distortion of speakers or headphones is orders of magnitude higher than the distortion from even a basic headphone amp design in the modern era.
My guess is that the 1/8 trick is for the typical characteristics of small-speaker headset. Big speakers may have different capacitance and inductance characteristics.
I don't know though. That's the first time I heard that approximation, and it doesn't seem to align with anything that I learned in electronics.
That 1/8 rate is a property of our hearing, and so will apply to any system that creates sound. But it's a quite hard target to achieve in large bass speakers, so a lot of systems simply don't.
There must be a mathematical description of these different situations but I can't be bothered to work it out. An important part of it is probably that an audio amplifier has feedback, an RF amplifier just emits a wave and the consumer must make itself compatible with it. Lower output impedance = more effective feedback control for an amplifier with feedback.
Power applications (like speakers or the power grid) are generally concerned with maximizing efficiency, which you get by lowering source impedance as much as possible.
For a fixed Zout much higher than zero, if we want to suck the most power out of the amplifier, we should use a matching Zin indeed. E.g. if an amplifier has 16 ohms output impedance, then we will draw the most power out of it with a 16 ohm load.
Loads higher than 16 ohms will result in less current flowing.
Loads lower than 16 ohms will cause power waste in the source impedance. For instance a zero ohm load will mean that all the power is dissipated in the 16 ohm source impedance: the zero ohm load draws current, but no power because it I^2R is zero.
So from there, if we increase the impedance gradually, we obtain more and more power until we get to 16 ohms, and after that less and less power.
In amplifiers we care about efficiency, not with operating at the theoretical point where absolute maximum power is drawn. With a near zero ohm output, we ensure that all the voltage is dropped by the load rather than wasted in the source.
When we have a near zero ohm output impedance, Zin == Zout still applies! E.g. theoretically, the maximum power transfer from a 0.1 ohm amplifier would take place if we connect it to a 0.1 ohm load. And that reflects the trend in low-output-impedance audio amplifiers: the lower the speaker impedance you plug in, the power power you get: 16 ohms, 8 ohms, 4 ohms, ... You just can't go anywhere near 0.1, due to the practical limitations in the amplifier: ability to deliver current without frying itself.
But, so yes, Zin == Zout is relevant, but in the majority of the audio amplifiers built, which have very low output impedance, that theoretical point occurs at low value of Zout/Zin which usually out of reach of the absolute current delivery capability of the amplifier.
Why are you even considering the case of a zero Ohm load with a 16 Ohm source impedance? Modern headphone amps (including really cheap ones) have source impedances in the range of 1 Ohm or less while headphones can range from 30-ish to 300 or more.
If you don’t understand the problem with your post, trying flipping the roles to match reality: Assume a 16 ohm load and a 0 (or realistically, 1 Ohm) source impedance: Now the “no power absorbed” side of this equation is the amplifier. All (or nearly all) power goes to the headphones. That’s exactly what we want.
> Loads higher than 16 ohms will result in less current flowing.
I think this is where you’re confused. Headphone amps are generally voltage limited sources. If you get to the point of current limiting then it’s going to distort intensely and people would turn the volume down because it’s so unpleasant. Any source impedance subtracts from the maximum voltage you can apply across the load, because the source impedance forms a resistive divider. These aren’t high frequency transmission lines where we’re trying to send GHz signals over impedance matched lines.
You don’t maximize power delivered to the headphone by adding an identical source impedance. You actually reduce it massively relative to a modern <1 Ohm source impedance amplifier.
If this still doesn’t make sense, consider why putting a 300 Ohm resistor in your headphone jack before connecting your 300 Ohm headphones isn’t going to improve anything. You’re just burning power in the source-side impedance and fighting against a fixed voltage limit.
> Why are you even considering the case of a zero Ohm load with a 16 Ohm source impedance?
For completeness of the analysis. If we hold Zout constant, and likewise the output voltage of the voltage source, and consider a Zin of zero, in that case, power transfer is mimimal (zero). Likewise power transfer goes to zero for large Zin. In between those is the Zout = Zin point where the maximum transfer is obtained by the load from the source. This is part of explaining of what is Zout = Zin about; what is maximized.
For instance consider a R1 resistor in series with a battery. You may not change the battery or R1. For what value of R2 can you get R2 to dissipate the most power? The solution is R2 = R1.
It is useful to think about what happens if we make R2 zero; why wouldn't we consider it. A certain current will flow through a zero ohm R2, but it will not be dissipating any power. From there as we increase R1, the power dissipation curve rises. At R1 = R2, it turns around, and then converges to zero as R1 grows larger.
Modern tube amps (non hybrid) have significantly higher output impedance because they often have no overall negative feedback and a few special snowflake headphones have impedances as low as 0.1r and 2r.
The considerations are definitively relevant for some cases, as well as in general for analysis of the topic.
That is at odds with what the blog post says.
I'm now confused - for the best sound quality, are we to maximise efficiency of power usage or maximise accuracy of signal reproduction?
I didn't think you can do both.
In the case of a headphone amplifier on AC power, we don't have to think about efficiency, as long as it isn't ridiculous, like drawing 100W.
Class B amplifiers are only efficient when producing no output, or when producing output clipped to the power rails (like a square wave). Due to being almost fully off or fully on.
For a sinusoidal signal, the worst case efficiency occurs at about 40% of the output power delivered to the load, when the transistors will dissipate another 20%.
Source: https://eng.libretexts.org/Bookshelves/Electrical_Engineerin... Section 9.2.4: Class B Power Dissipation
(Yes, I checked a few youtube videos and blogs, but car audiophiles don't really get too scientific and it's hard to tell whether they really know what they are talking about).
As far as matching goes, you want amp power > speaker power handling. If the amp distorts before the speakers do, you're not getting the most out of the speakers, and clipping can be bad for some speakers too, especially tweeters.
Also, the THD figures quoted for amplifiers are usually at moderate output levels; they get worse with higher voltage swing. An amp that is getting close to clipping will generally have worse THD than one which is nowhere near clipping. (As well, THD gets worse with rising frequency.)
https://en.m.wikipedia.org/wiki/Smith_chart
You can put an amp 50ohm output into a 75 ohm cable or connectors but there will be negatives. The further away from output impedance the more negatives. Loss, distortion, and worse roll off, etc. Cables are fun too. for much higher frequencies, the less of the actual copper is used. The signal uses the surface of the wiring. For lower frequencies it uses the entire strand/solid copper wire.
RF mostly uses 50ohm but not always. Cable video (tv) is usually 75ohm.
If you are using a short piece of the cable to convey a voltage signal between two pieces of equipment (like the output of one amplifier into the input of another), it has no meaning. The voltage on the other end rises almost instantly and is reflected back.