Interesting research about headphone measurements

[mkrzych]

Very interesting article and summary conclusion.

 

Audio Musings by Sean Olive: The Relationship between Perception and Measurement of Headphone Sound Quality

 

One thing also caught my eye: "most preferred headphones in sound quality are least in comfort".

[solidstate]

A good read. I think the headphone industry have made a good progress over the last years, i agree there are some brands that have been marketed extensively but the SQ is poor.

[dalethorn]

Assuming you pick a headphone that's comfortable enough to get you through your listening sessions, and it has a tonal balance that works for the genres you play with it, what matters then is the roughness or unevenness that's caused mainly by undamped resonances, and the destructive effect that unevenness has on soundstage etc. Get a good (and simple) parametric equalizer and fix just those roughness or resonance areas, and you'll start hearing what's possible with reproduced sound - and it can be spectacular.

[mkrzych]

Assuming you pick a headphone that's comfortable enough to get you through your listening sessions, and it has a tonal balance that works for the genres you play with it, what matters then is the roughness or unevenness that's caused mainly by undamped resonances, and the destructive effect that unevenness has on soundstage etc. Get a good (and simple) parametric equalizer and fix just those roughness or resonance areas, and you'll start hearing what's possible with reproduced sound - and it can be spectacular.

 

If you like equalization, me no.

[dalethorn]

If you like equalization, me no.

 

You probably missed my point, not knowing what the goal is I'm soeaking of. I don't refer to ordinary EQ, but to a method of removing the effect of resonances, to restore soundstage and realism. Users are slowly, very slowly catching on, and they are always amazed at the result.

[sandab]

The treble response needs to match the source feeding it, and the less distorted (less bright sounding, but more clear, realistic, and imaged) the signal, the higher the response can be. If the treble is out of phase it won't match up with the mids, won't be part of the timbre, separates, and becomes a voice of its own. Frequently a not very pleasant one. It's like chromatic aberration in a photograph, where the aberration itself creates faux content. Properly lined up in phase, it becomes envelope and isn't perceived as bright, but clear, spacious, imaged, realistic; but there's a limit to every reproduction, too much and it can't keep up and distortion appears (as a separate voice).

 

So yes, when test subjects say they prefer a neutral headphone - of course they do. When attached to a price-no-object source chain carefully selected for its sonic properties. Hook it up to a $100 portable USB-powered Sabre DAC with an integrated op-amp negative feedback output, and they'll grit their teeth and look for ways to mod their headphone with felt, or look for something a little less bright sounding.

[dalethorn]

The treble response needs to match the source feeding it, and the less distorted (less bright sounding, but more clear, realistic, and imaged) the signal, the higher the response can be. If the treble is out of phase it won't match up with the mids, won't be part of the timbre, separates, and becomes a voice of its own. Frequently a not very pleasant one. It's like chromatic aberration in a photograph, where the aberration itself creates faux content. Properly lined up in phase, it becomes envelope and isn't perceived as bright, but clear, spacious, imaged, realistic; but there's a limit to every reproduction, too much and it can't keep up and distortion appears (as a separate voice). So yes, when test subjects say they prefer a neutral headphone - of course they do. When attached to a price-no-object source chain carefully selected for its sonic properties. Hook it up to a $100 portable USB-powered Sabre DAC with an integrated op-amp negative feedback output, and they'll grit their teeth and look for ways to mod their headphone with felt, or look for something a little less bright sounding.

 

This is a great observation, and relevant photographic analogy. The only argument I would stipulate is, take what you have - or have settled for in a particular listening circumstance, and apply just the minimal EQ to correct an obvious resonance (sudden peak or recess), then toggle the EQ on/off to hear the difference. In most cases with most headphones, that EQ makes a very positive difference. But apply EQ broadly to 'balance' the bass-mids-treble on the average sub-$500 headphone, and your observed behavior could very well outweigh any advantages that EQ can offer.

[fiaoyujia]

good,Users are slowly, very slowly catching on, and they are always amazed at the result.thanks

[firedog]

You have DSP correction starting to come on board for headphones:

 

Every headphone actually distorts the sound it's own drivers make - it's just a function of the sound bouncing off the structure of the headphone itself, and is unavoidable, whatever the headphone. DSP enables the headphone maker to measure the distortions caused by the headphone itself, and correct for them.

 

What you then get is "perfect" reproduction from the drivers reaching your ears. Vastly improves the sound of the headphone for those that use it. The catch, of course, is that it has to be individually measured for each model of headphone and either added to the headphone itself (expensive ) or added to playback software.

[dalethorn]

You have DSP correction starting to come on board for headphones: Every headphone actually distorts the sound it's own drivers make - it's just a function of the sound bouncing off the structure of the headphone itself, and is unavoidable, whatever the headphone. DSP enables the headphone maker to measure the distortions caused by the headphone itself, and correct for them. What you then get is "perfect" reproduction from the drivers reaching your ears. Vastly improves the sound of the headphone for those that use it. The catch, of course, is that it has to be individually measured for each model of headphone and either added to the headphone itself (expensive ) or added to playback software.

 

Yes - DSP's like Dirac that were created for earpods and some IEMs were very helpful. One caveat, if you can't modify the DSP yourself, you may have to do it yourself anyway with some headphones, due to sample variances. Most of the top brands probably won't vary much during a particular production run, but with time a given model may see its signature change without creating a clear alternative item number. The Sennheiser HD-650 is one example of that. I paid $320 for the new Sennheiser HD-26, and I hear some huge left-right swings in volume with mono test tones in the treble range - which I don't hear at all with 25 other top-quality headphones. But most equalizers don't have separate left-right channel settings, so that problem might not be fixable on most systems.

[sandyk]

If you like equalization, me no.

 

I'm with Krzysztof here.

You can often make very worthwhile improvements simply by varying the series output resistance value of the amplifier, provided that the amplifier has sufficiently high supply rail voltages. I don't subscribe to the analogue EQ using additional opamps, or further digital processing schools of thought. Even the venerable AKG 701 can be made to sound very good indeed when matched with a good SS amplifier having say + and -20V supply rails, and a typical 120 ohms output impedance, without the need to fiddle around with additional EQ.

[Skeptic]

While it's good to see that people seem to agree that a smoother frequency response is better, a flat headphone response does not sound the same as a calibrated (flat) speaker response due to interaction with the ear.

 

 

While a flat response scores well, a smooth "U-shaped" response is preferred, and is a closer match to the speakers.

 

 

 

I'm with Krzysztof here.

You can often make very worthwhile improvements simply by varying the series output resistance value of the amplifier, provided that the amplifier has sufficiently high supply rail voltages. I don't subscribe to the analogue EQ using additional opamps, or further digital processing schools of thought. Even the venerable AKG 701 can be made to sound very good indeed when matched with a good SS amplifier having say + and -20V supply rails, and a typical 120 ohms output impedance, without the need to fiddle around with additional EQ.

I completely disagree. The impedance of the AKG K701 is 62 ohms.

That means the output impedance of your amplifier should be no more than 8 ohms, ideally as close to 0 as possible.

 

I don't know of any dynamic headphone with a high enough impedance that you could use a 120 ohm headphone output and get good results.

You would be much better off with even a cheap headphone amplifier like a Schiit Magni/Vali than that. ($99/$119)

With a 120 ohm output, I have to imagine that you're simply connecting your headphones up to an AVR or Integrated Amplifier.

 

And avoiding DSP is simply hurting the potential of your system, no matter what it is:

 

 

 

Frankly, I am hesitant to upgrade my current headphones - which I am very happy with - to something higher-end like the LCD-3 because we are only now seeing the return of true flagship headphones in the last few years.

Unfortunately that means we also have average headphones being sold as flagships (Beyer Tesla, AKG K812, Sony MDR-Z7 etc.) but the true flagships like the HD800 and LCD3 are excellent products.

 

However these headphones are not based on the latest research into headphone design (target response curve) nor do they take advantage of what modern DSP can do.

There are a number of Dirac-optimized headphones on the market now, and that DSP makes a huge difference to what would otherwise be mediocre or even bad headphones. I'd love to see that technology applied to a flagship.

 

The best demo for it would be if you have an iOS device and a pair of the original Apple Earbuds. (not the EarPods)

The Dirac App makes a significant improvement to how those sound, the main problem is that you can only get that quality via their app on one of those devices.

 

However without the equipment to measure and calibrate these headphones yourself, it's probably never going to happen for any existing flagship.

It's more likely that we will see a flagship with a built-in DAC/DSP (similar to the non-flagship Sony MDR-1A-DAC) rather than releasing an impulse file or convolver config file for your specific headphones. (something that companies like Audeze should do)

[firedog]

This is what promises to be a quality headphone (not flagship) with DSP. Hopefully it will be very good, as I funded the kickstarter and should have one in about a month.

 

While it's good to see that people seem to agree that a smoother frequency response is better, a flat headphone response does not sound the same as a calibrated (flat) speaker response due to interaction with the ear.

 

 

While a flat response scores well, a smooth "U-shaped" response is preferred, and is a closer match to the speakers.

 

 

 

 

 

I completely disagree. The impedance of the AKG K701 is 62 ohms.

That means the output impedance of your amplifier should be no more than 8 ohms, ideally as close to 0 as possible.

 

I don't know of any dynamic headphone with a high enough impedance that you could use a 120 ohm headphone output and get good results.

You would be much better off with even a cheap headphone amplifier like a Schiit Magni/Vali than that. ($99/$119)

With a 120 ohm output, I have to imagine that you're simply connecting your headphones up to an AVR or Integrated Amplifier.

 

And avoiding DSP is simply hurting the potential of your system, no matter what it is:

 

 

 

 

Frankly, I am hesitant to upgrade my current headphones - which I am very happy with - to something higher-end like the LCD-3 because we are only now seeing the return of true flagship headphones in the last few years.

Unfortunately that means we also have average headphones being sold as flagships (Beyer Tesla, AKG K812, Sony MDR-Z7 etc.) but the true flagships like the HD800 and LCD3 are excellent products.

 

However these headphones are not based on the latest research into headphone design (target response curve) nor do they take advantage of what modern DSP can do.

There are a number of Dirac-optimized headphones on the market now, and that DSP makes a huge difference to what would otherwise be mediocre or even bad headphones. I'd love to see that technology applied to a flagship.

 

The best demo for it would be if you have an iOS device and a pair of the original Apple Earbuds. (not the EarPods)

The Dirac App makes a significant improvement to how those sound, the main problem is that you can only get that quality via their app on one of those devices.

 

However without the equipment to measure and calibrate these headphones yourself, it's probably never going to happen for any existing flagship.

It's more likely that we will see a flagship with a built-in DAC/DSP (similar to the non-flagship Sony MDR-1A-DAC) rather than releasing an impulse file or convolver config file for your specific headphones. (something that companies like Audeze should do)

[Skeptic]

This is what promises to be a quality headphone (not flagship) with DSP. Hopefully it will be very good, as I funded the kickstarter and should have one in about a month.

As with all the others I've seen using Dirac, it's running the DSP in a mobile app rather than being built into the headphones, which severely limits its usefulness in my opinion.

 

Building it into the headphones would allow it to work with any source, but that means you're stuck with the DAC/Amp built into them and I've never heard a headphone which does that well. Not even the high-end Sennheiser wireless headphones that use lossless transmission instead of Bluetooth.

 

I'd rather have a flagship headphone and have an individual convolution file that can be used with any compatible software, allowing me to stick with my DAC/Amp of choice.

[sandyk]

I completely disagree. The impedance of the AKG K701 is 62 ohms.

 

Many earlier headphones, including the AKG701 and my own Audio Technica W1000, complied with IEC61938 which specified that all headphones, irrespective of impedance, should be driven from a 120 ohms source impedance.

These days with the proliferation of portable devices having low supply voltage rails, this earlier specification has fallen into disuse.

 

With a 120 ohm output, I have to imagine that you're simply connecting your headphones up to an AVR or Integrated Amplifier.

 

My own Class A HA/preamp with a bias of 100mA and + and -20V supply rails is more than capable of accomplishing this. There are quite a few other commercial amplifies available with similar capabilities.(Panda etc.)

[dalethorn]

There are a number of Dirac-optimized headphones on the market now, and that DSP makes a huge difference to what would otherwise be mediocre or even bad headphones. I'd love to see that technology applied to a flagship. The best demo for it would be if you have an iOS device and a pair of the original Apple Earbuds. (not the EarPods) The Dirac App makes a significant improvement to how those sound, the main problem is that you can only get that quality via their app on one of those devices. However without the equipment to measure and calibrate these headphones yourself, it's probably never going to happen for any existing flagship. It's more likely that we will see a flagship with a built-in DAC/DSP (similar to the non-flagship Sony MDR-1A-DAC) rather than releasing an impulse file or convolver config file for your specific headphones. (something that companies like Audeze should do)

 

One problem with Dirac in headphones is that its stated design is very specific in correcting the effect of resonances and so on, moreso than creating a new signature that Dirac thinks has the right bass-treble-mids balance for most users - and the user doesn't to my knowledge have the ability to tweak the DSP. But the biggest problem with a canned (no pun intended) DSP is it assumes that all of the target model sound the same, with no variances in where the response deviates from "flat" or target, or how much. Sample-to-sample variations happen, and sometimes they're very significant.

 

Foobar2000 is available for free - a decent music player with plugins for nearly everything, and the optional 30-band equalizer is not only good for low distortion, but with some experimenting and coaching most users could probably get a better correction than with a canned DSP. In this case I'm not suggesting EQ correction for personal taste or hearing anomalies, merely to correct the more obvious bumps in the response - and do so accurately based on that sample's actual response.

[Skeptic]

Many earlier headphones, including the AKG701 and my own Audio Technica W1000, complied with IEC61938 which specified that all headphones, irrespective of impedance, should be driven from a 120 ohms source impedance.

These days with the proliferation of portable devices having low supply voltage rails, this earlier specification has fallen into disuse.

Driving headphones with an impedance of 62 ohms from a 120 ohm output gives you a damping factor of 0.52

 

You are seriously compromising your headphone performance if you use that output.

Damping factor should be at least 8 for dynamic headphones.

 

The Sonic Advantages of Low-Impedance Headphone Amplifiers

Headphone & Amp Impedance

 

One problem with Dirac in headphones is that its stated design is very specific in correcting the effect of resonances and so on, moreso than creating a new signature that Dirac thinks has the right bass-treble-mids balance for most users - and the user doesn't to my knowledge have the ability to tweak the DSP. But the biggest problem with a canned (no pun intended) DSP is it assumes that all of the target model sound the same, with no variances in where the response deviates from "flat" or target, or how much. Sample-to-sample variations happen, and sometimes they're very significant.

Yes, that's why I would much rather see flagship headphones come with factory measurements from the headphones so that you can create your own "room correction" filter for them, than in-built DSP or a preset config in a mobile app.

 

Sennheiser measures every HD800 that leaves the factory and provides the buyer with their measured frequency response. It would not be difficult for them to take an impulse measurement so that things could be further improved via DSP.

I would hope that most true flagship headphones, such as the Audeze LCD3, also go through a similar testing process.

 

Foobar2000 is available for free - a decent music player with plugins for nearly everything, and the optional 30-band equalizer is not only good for low distortion, but with some experimenting and coaching most users could probably get a better correction than with a canned DSP. In this case I'm not suggesting EQ correction for personal taste or hearing anomalies, merely to correct the more obvious bumps in the response - and do so accurately based on that sample's actual response.

This is the "old fashioned" way of doing DSP and unless you are a trained listener with years of experience, I don't trust anyone's ears to do a good job of optimizing the sound with a multi-band EQ.

 

Measure the headphones on a dummy head to smooth out the response, and then allow the user to apply a target curve in software.

This could also be done if the DSP was inside the headphone, but I've explained above why I prefer to keep the DAC & Amp outside of the headphones.

[sandyk]

Driving headphones with an impedance of 62 ohms from a 120 ohm output gives you a damping factor of 0.52

 

You are seriously compromising your headphone performance if you use that output.

Damping factor should be at least 8 for dynamic headphones.

 

 

The fact remains that many earlier headphones complied with IEC61938, and they sound more tonally balanced when driven as required by the specification.

There were quite a few HAs constructed by members of another forum to my Class A design and a previous modified version of a Silicon Chip magazine design, (>300 units for the latter). Many members owned AKG K701 headphones and found that they did indeed sound far better driven from a 120 ohms source impedance. AKG K701 are notorious for their harsh and fatiguing treble when not correctly driven.

 

At a 0.01-Ohm input impedance, both sets of headphones received a clean electrical signal (as show by the blue traces). Nevertheless, in our listening tests, the HD-650 headphones always seemed to be cleaner than the MDR-V6 headphones. This implies that only some of the transducer distortion can be removed with a low-impedance headphone amplifier. Again, this is an area of ongoing study.

[Skeptic]

Many members owned AKG K701 headphones and found that they did indeed sound far better driven from a 120 ohms source impedance. AKG K701 are notorious for their harsh and fatiguing treble when not correctly driven.

It is primarily low frequencies that are affected by the output impedance. An area where the K701 is often said to be lacking.

[donunus]

Its not always the low frequencies that get affected by the output impedance. It depends on where the big impedance peaks are on a particular headphone.

[sandyk]

There is an old rule of thumb that says if the HF is a little too dominant, try increasing the amplifier's output impedance.

Conversely, if the LF is too dominant, try decreasing the output impedance of the amplifier.

Even a 20% variation is often all that is needed for a worthwhile subjective improvement in tonal balance.

A good starting place for many of the earlier headphones was an output impedance of around 68 ohms.

Until recently, there were many reports that damping factor with headphones was of far less importance than with speakers due to the diaphragm size and characteristics.

[donunus]

Ill give an example here... a grado headphone when driven from a high output impedance receiver will sound very similar balancewise compared to being driven by an ipad. A Sennheiser HD555 however will sound bloated out of that same receiver vs an ipad just because the headphones impedance goes up to about 300 ohms at 100hz while is only about 50 ohms at other frequencies in the spectrum.

[dalethorn]

I've had some speaker cables that had a lot of capacitance (or so some reviews said), and it had a major negative effect on highs. Perhaps headphone cables could also play a part?

[sandyk]

I've had some speaker cables that had a lot of capacitance (or so some reviews said), and it had a major negative effect on highs. Perhaps headphone cables could also play a part?

 

According to a certain EE from The Netherlands it doesn't matter, but some people love their low capacitance headphone cables.