[KMann]

Headphones have degrees of openess and closedness

Is there a way to tell this from measurements? or measure the degree of openness?

They also may have different dynamics and handle bass differently

If an electrostatic headphone, a planar and a dynamic headphone are all eqed to have the same bass will they all sound equally 'dynamic' is there a way to measure this?

An oversampling DAC should be able to give you transparent sound within the audible range.

Are there any measurements that help one judge how transparent a DAC or a headphone is?

I am not disagreeing with anything you said, the science to judge these objectively may already exist, it is not something that is researched enough and published. The alternate possibility is it is all audiophile myths, they should all sound the same.

 

[bigshot]

Is there a way to tell this from measurements? or measure the degree of openness?

Not unless you’re willing to measure from a standard dummy head. It can vary from person to person. How you mic it might make a difference too. In any case, the measurements are only going to be comparable if they were measured using the same rig and techniques. Two different people’s measurements are going to be different.

If an electrostatic headphone, a planar and a dynamic headphone are all eqed to have the same bass will they all sound equally 'dynamic' is there a way to measure this?

It’s difficult to do blind testing with headphones because the test subject can feel how they fit and identify them from that. The fit and construction will make audible differences. You could get the response close enough to make yourself satisfied for the purposes of listening to music though.

Are there any measurements that help one judge how transparent a DAC or a headphone is?.

If all the aspects of sound reproduction (ie response, distortion, noise, timing, etc.) fall below the established thresholds of perception, it’s transparent. Most DACs and amps should be transparent. If they aren’t, something is wrong.

No headphones are transparent. The same make and model can vary from one copy to another, and the same set can sound different to two different people. Transducers are the wild card in any system. You have to EQ them all one at a time if you want them to adhere to a target curve, and even then, there can be audible differences. But if they can reproduce a full range of frequencies clean and loud, you can get close enough to your target response for the purposes of listening to music in the home.

 

[KinGensai]

@KMann
In regards to that Estat vs PM vs DD question, it's a combination of impulse response, THD, and HD profile is it not?

The LCD-2 has a distinct feel compared to something like a DT 1990 pro because DDs have exponentially increasing THD as they approach infrasonic frequencies with an even order harmonic profile, whereas planars and electrostats tend to have remarkably low and clean THD, with a few spikes here and there on specific models. There's also the cleaner decay in the bass that comes with a shorter impulse response due to the lower amount of mass required by the PM and Estat operating principles.

 

[VNandor]

On the contrary, that is exactly my point, but most of the analysis and measurements are done to highlight the frequency response. In theory yes you can go back and forth an infinite number of times, but for measurements, a lot of information is lost as FFTs are done with fixed window sizes and applying averaging and choice of the windowing functions and duration of the window affects how the frequency response looks and the phase response becomes useless at this point as it is time averaged and there is no way to just go back from there.

I think I see where you are coming from. This can be often the case for audio measurements but this isn't how a digital filter's frequency and phase response is being evaluated. There's very little room for ambiguity when measuring a digital filter. All that has to be done is to pass a single sample to the input, then analyzing the series of samples at the output. Only a single pass is needed. If the filter doesn't respond to the impulse the same way every single time or it reacts differently to a complex signal than to a single tone, then there's something wrong with the filter (or the filter is specifically made to be used as a creative effect). This isn't something that can happen by accident, in fact, it's much harder to make a filter that's time-variant and nonlinear than a filter that isn't. The window size and windowing function has an influence on what's being seen of the response so they do have to be chosen well though.

In practice, the spectrum of white noise is still going to be different compared to an impulse ... I don't understand why you would suggest that these two would be the same either in theory, or in practice.

 

[KMann]

@KMann
In regards to that Estat vs PM vs DD question, it's a combination of impulse response, THD, and HD profile is it not?

The LCD-2 has a distinct feel compared to something like a DT 1990 pro because DDs have exponentially increasing THD as they approach infrasonic frequencies with an even order harmonic profile, whereas planars and electrostats tend to have remarkably low and clean THD, with a few spikes here and there on specific models. There's also the cleaner decay in the bass that comes with a shorter impulse response due to the lower amount of mass required by the PM and Estat operating principles.

Yes impulse response holds the key to some of the questions, which is why I have been talking about the time domain response. I have looked at impulse responses from different headphones and overlaid them too, once you EQ the headphones and look at the impulse response you know they are not the same but at the same time they are very hard to interpret and determine what to expect. Change in position of the headphone even if EQd to have the same response will look different, the EQ filter used to alter the response will it self have an effect too.

It is also possible to alter the impulse response to correct for some of the time domain errors and alter the sound of the headphone not just tonally. Of course the ideal impulse response for a true transparent sound is the Dirac delta

 

[KMann]

I think I see where you are coming from. This can be often the case for audio measurements but this isn't how a digital filter's frequency and phase response is being evaluated. There's very little room for ambiguity when measuring a digital filter. All that has to be done is to pass a single sample to the input, then analyzing the series of samples at the output. Only a single pass is needed. If the filter doesn't respond to the impulse the same way every single time or it reacts differently to a complex signal than to a single tone, then there's something wrong with the filter (or the filter is specifically made to be used as a creative effect). This isn't something that can happen by accident, in fact, it's much harder to make a filter that's time-variant and nonlinear than a filter that isn't. The window size and windowing function has an influence on what's being seen of the response so they do have to be chosen well though.

In practice, the spectrum of white noise is still going to be different compared to an impulse ... I don't understand why you would suggest that these two would be the same either in theory, or in practice.

I was commenting on only using frequency response to form an opinion. Yes the frequency domain response includes the phase response but a phase response is rarely shown during measurements and it is even harder to interpret or use to form any kind of onion. I view measuring the digital filter as the ideal/easy case as like you said, they are very easy to measure and they are predictable (they better be).

I was providing white noise and impulse as simplified example to make a point that two signals can have the same energy in different frequencies and yet sound different.

 

[bigshot]

Frequency response is clearly the most likely aspect to be a significant problem with headphones, even more so for speakers.

In regards the comment about Dirac delta being transparent (as opposed to others), what threshold of audibility would you be using to determine whether a measurement is an audible difference or not? Or are you pointing to listening tests? Are we talking about audibility with some sort of test signal or are we talking about in music? Because the transients in music are an order of magnitude broader than what is talked about in a technical sense. No one is going to hear square waves in their Mendelssohn.

 

[KinGensai]

Frequency response is clearly the most likely aspect to be a significant problem with headphones, even more so for speakers.

In regards the comment about Dirac delta being transparent (as opposed to others), what threshold of audibility would you be using to determine whether a measurement is an audible difference or not? Or are you pointing to listening tests? Are we talking about audibility with some sort of test signal or are we talking about in music? Because the transients in music are an order of magnitude broader than what is talked about in a technical sense. No one is going to hear square waves in their Mendelssohn.

Impulse response cleanliness matters the more complex the waveform gets and the larger the dynamic range of the music. Single driver IEMs have more and more trouble rendering music at the proper levels as the number of tracks/stems in the mix goes up because that single diaphragm has to reach specific amplitudes at each sample, so approx 23μm intervals. The longer and sloppier the IR is, the higher the likelihood the driver misses that target amplitude in each sample, effectively reducing the dynamic range the IEM/HP can render.

Companies are taking two different strategies to address this: improve the technology in that driver to improve it's response further and further (sennheiser with their IE series, beyerdynamic with their Xelento, Dita with the Perpetua, etc), or use multiple drivers in a multi-way crossover network to split the frequency range into bands to hit the encoded amplitudes with less error (my V14, subtonic storm).

DDs have messier IR that can extend close to 3ms while BAs have a cleaner IR that effectively stop at 1.7ms. They have their pros and cons, but IMO controlling IR and THD with multi-driver setups has the best results.

 

[KMann]

In regards the comment about Dirac delta being transparent (as opposed to others),

Transparent means you cannot see it is there and in the case of audio, you should not hear it and all you hear is the source (i.e., the sound that was recorded)

If a system's response is a Dirac delta function it means it simply reproduces the input at the output, that is both frequency and phase/the time domain response. So it vanishes and becomes truly transparent. For the purpose of sound reproduction, unless the intent is to color the sound, trying to get closer to a Dirac delta is a good objective, though in practice it is close to impossible.

Current technology is not at a point where one is listening to a system and the sound feels like it is originating out of thin air. You don't need to think in terms of threshold of hearing, it is very easy to tell whether the sound is produced by a transducer or if it is a live source. Yes, you can reach a point where it is hard to distinguish between two playback chains, but it does not mean they are both transparent, it just means they are very similar.

 

[bigshot]

Impulse response cleanliness matters the more complex the waveform gets and the larger the dynamic range of the music.

How do you determine transparency? Through measurements, or through listening tests? What is the audible threshold?

KMann, I'm not talking about directionality or realism of sound. I'm talking about how similar two sounds sound. Do the two sound samples sound different or the same? Since it is an issue with headphones, I don't know how you would reduce the variables to know if what you're hearing is caused by timing error, response variation, or the physical acoustic properties of the headphone enclosures. My question is how is it measured, and how would one isolate it to be able to say for sure that one thing is audibly transparent and another thing isn't?

Until you can test for transparency, I don't know how you can say that something isn't transparent. There's no such thing as a sound that is "more transparent". I suppose you could say that in theory something is more likely to be transparent, but that doesn't mean the thing you are comparing it to isn't transparent too.

I'm trying to lay down a logical path for understanding this, and I may not be articulating it clearly. Maybe the answer is no transducer is transparent, and we can't determine what aspect is making two transducers sound different because the aspects are all mixed together with trade offs and compromises. That's been my operating principle with transducers. I don't use the term transparent with them. I just listen to it and see if it works for my purposes. I know that isn't very scientific, but I don't know how I could determine things as definitively as I can with electronics.

 

[KinGensai]

How do you determine transparency? Through measurements, or through listening tests? What is the audible threshold?

It's just what I noticed as the most obvious change going from single driver IEMs to a quad driver IEM (FH5s), then to my 14 driver IEM. I would provide my own measurements if I had a measurement rig, but unlike most of the changes claimed to be due to cables or whatever, this change is the most obvious and logically sound improvement you can get because transducer accuracy is the biggest bottleneck in the chain.

I don't know if you have tried listening to the range of IEMs available in the market, but IMO this is the most self-evident improvement you can get. It's not obvious at all if you listen to a jazz trio on a single driver IEM, but turn on some modern big band arrangements or speedcore and it's much more obvious how compressed the range becomes compared to even midrange multi-driver IEMs.

 

[bigshot]

I don't know anything about IEMs, but I know with speakers there are all kinds of theories about one design that is better than another, but it always seems that there are trade offs. It might do one thing better, but a different design will do another thing better. For instance, I have KEM speakers with nested drivers. The tweeter, mid and bass are all stacked inside of each other. They say that this makes phase error better, and maybe that is the case, but when I compare them to traditional box speakers, towers or horn loaded, there are so many variables, I can't say that the radial design is better or not. I can only say it's theoretically better, which in audiophile circles is a slippery slope.

So I just listen to them and judge the speaker as a whole and I don't try to optimize any one aspect.

 

[KMann]

Until you can test for transparency, I don't know how you can say that something isn't transparent. There's no such thing as a sound that is "more transparent". I suppose you could say that in theory something is more likely to be transparent, but that doesn't mean the thing you are comparing it to isn't transparent too.

That just brings it back a full circle and my original concern that there is not enough focus on time domain response and more work needs to be done in measurements and understanding. We know what true transparency in the theoretical sense is but a true test for the degree of transparency does not exist.

In lab conditions, it is relatively easy to control the number of variables for comparison purposes. For example, in the case of planar drivers, it is easy to measure in an open IEC baffle, keep the magnetic circuit the same, keep the trace pattern on the film the same, and vary only the mass of the film. Due to the change in the inertia of the film, the impulse response will be different and it is easy to keep the assembly (earcups, ear pads, headband etc. the same) and just switch the transducer and they will sound different because as lower mass can accelerate faster, it can move keeping in pace with the rapid change in music.

So everything else being the same in the above simplified example, I will call the more nimble transducer as being the more transparent one as it will be closer to a an ideal impulse.

 

[bigshot]

So everything else being the same in the above simplified example, I will call the more nimble transducer as being the more transparent one as it will be closer to a an ideal impulse.

Would a more nimble transducer be better for response and distortion too? There would also be practical issues related to reliability and durability to consider.

With transducers, I find that all of the aspects affect each other and there are balances and tradeoffs to any particular design. (That may be true of the practical application of physics in general.)

Since the threshold of transparency is a fixed line where one side is audible and the other isn't, you can't have something be "more transparent". Once it's transparent, it's fully transparent.

 

[KMann]

Would a more nimble transducer be better for response and distortion too? There would also be practical issues related to reliability and durability to consider.

No it is just one parameters, which is why I said everything else being equal. You can have a nimble transducer, but if you have a poorly designed magnetic circuit and introduce local modes, then distortion is assured. If you have two of the same design with the same distortion issues, the nimble one will be more transparent in spite of the flaws.

Since the threshold of transparency is a fixed line where one side is audible and the other isn't, you can't have something be "more transparent". Once it's transparent, it's fully transparent.

Why is it so?