Trying To Understand Neutral / Reference

[DesEsseintes]

I am trying to better understand the inter-relationships of frequency response, loudness, perception. My objective I suppose is to be able to identify and understand a reference or neutral tuning (particularly with headphones / earphones). I'm not necessarily sure if "reference" and "neutral" are truly interchangeable, or if some distinction that I am not aware of exists. I apologise in advance for some errors in syntax, or confusion of some terms, corrections are most welcome.

I somewhat entered into a state of confusion yesterday reading about equal loudness curves. But I think they are a good place to start. So, an equal loudness curve, as I understand it, should help to explain the difference between measured loudness and perceived loudness. So a microphone will measure one thing, and identify some relative difference, but our perception would differ more or less according to these curves (depending on the state of the listeners own hearing). It is probably true to say we all have a unique (to some degree) equal loudness curve.

Then we get on to the subject of frequency response. The frequency response, these graphs that are quite common online (particularly for headphones/earphones), are a measurement of loudness across a range of frequencies for a given sound reproduction device, tested under a certain condition. The raw curve, essentially represent what the microphone "hears", a compensated curve will simply describe how to response deviates from some target.

Once we get onto the subject of targets, it seems to me that there are some models which are based on trying to reproduce the effect of a "flat" speaker in some room condition, and some other models, which start from this point and add some preference adjustments (maybe yet others should belong to another category).

My understanding of a flat speaker is where I start to get a little hazy. Should a flat speaker be measured flat on frequency response by a microphone or sound flat for example on a sine sweep by a listener. (i think it is the latter). So, for example, I have some cheap monitors on my desk (Tannoy 402). When I perform a sine sweep, it sounds quite flat (at least in the mids/treble, the bass is somewhat uneven).

Then I think, well, this sounds flat to me, but to someone else with different hearing, maybe not. I guess it should be more reproducible to measure by microphone "flat".

I may of course be missing some obvious point here, such as microphone calibration. Are microphones compensated for this equal loudness curve? Or maybe it is my understanding of how the sound changes in a room.

I think on the subject of headphones / earphones, we have some further issues with unique anatomy, and let's say targets which have additional layers of complexity in how they were constructed. Nonetheless, a reminder of my objective, to better understand what a reference or neutral tuning should sound like and how I could identify such tuning.

An example, when i perform a sine sweep using ER2SE, it sounds mostly flat, some bass rolloff, but mostly the response is smooth and of equal perceived loudness. This, to me seems the most likely candidate for this reference sound. The frequency response of this earphone bears some inverse relationship to the equal loudness curve, as well as some built-in considerations for earphone resonance etc.

My HD600, which some might consider "relatively" flat has comparably more variation in a sine sweep, particularly past 5k. So for instance, if I wanted to EQ these headphones to a more reference tuning, I would surmise that I should attempt to correct these variations, and get some result which is closer (in perception) to the ER2SE, though not in frequency response, because the conditions of measurement would be significantly dissimilar.

I apologise for the stream of consciousness on this one. Please point out if you can, the obvious points which I have missed or failed to understand.

My thanks.

 

[NTK]

Welcome to ASR!

Equal loudness is built into our auditory system. Everything we hear is filtered through it. If we apply EQ that mimics equal loudness to the reproduction, the "effects" will be applied twice.

Let's say we are in a room with a musical instrument playing, the instrument radiates its sound as pressure waves. Our hearing system turns the pressure waves that reach our eardrums into "sounds", and somewhere in this process our hearing system applies the loudness curve (relating the acoustic pressure levels to loudness).

The role of a high fidelity audio reproduction system is to recreate the pressure waves that reach our ear drums as close as possible to the original ones. That's why the reproduction system should be neutral (flat frequency response is one of the requirements).

When we apply EQ to the reproduction to give us an overall non-flat frequency response, the pressure waves will be quite different. Our hearing system may detect that there are deviations from what we have expected, and we may find that the reproduction is no longer a natural reproduction of the original.

There is a separate but related topic of loudness compensation when we listen to music reproductions at different (usually lower) loudness levels from the original. But that would be a different discussion.

 

[ADU]

My understanding of a flat speaker is where I start to get a little hazy. Should a flat speaker be measured flat on frequency response by a microphone or sound flat for example on a sine sweep by a listener. (i think it is the latter). So, for example, I have some cheap monitors on my desk (Tannoy 402). When I perform a sine sweep, it sounds quite flat (at least in the mids/treble, the bass is somewhat uneven).

Howdy DesEsseintes,

Your questions above cover a fairly broad gamut of topics, which are to some extent all inter-related. I think NTK (as usual) has done an admirable job though of focusing in on one or two of the key points.

To give a more direct answer to the above question, a loudspeaker with a "flat", or "neutral", or "accurate", or "reference" response would, imho, be one which actually measures the same amplitude in dB SPL across most of the frequency range, at some given volume. This applies only to its direct/on-axis response though, which is only one component of a speaker's sound in a room.

Until recently, there were only a couple ways to accurately measure a speaker's direct response. And the primary method was to use an echo-free room, where most or all of the reflections from the room's walls, floor, and ceiling have been removed or deadened through the use of damping materials. This is why a speaker's direct or on-axis response is sometimes also (incorrectly) referred to as its "anechoic response".

There are measurement systems, such as the Klippel system used by ASR, which have been developed more recently though which can also calculate these kinds of anechoic measurements in a regular semi-reflective room. This is accomplished by measuring the speaker's response at a number of different points in the room. And then constructing a mathematical model of the speaker's response through time, and using that to separate its direct and reflected sounds in the room. (That is probably a gross oversimplification of how the Klippel works, but it hopefully gives you at least a vague idea of the process.)

This is unfortunately not something you can do very easily yourself with a simple omni-directional mic... unless you happen to have a degree in physics and/or acoustic engineering, and alot of patience. And maybe also some skills in the area of robotics. What you can measure though with your mic fairly easily is the speaker's in-room response, which contains a mixture of its direct, reflected and diffuse sound in the room.

Since it's rather late here, I'll have to stop there for now. And try to pick this up again tomorrow.

 

[DesEsseintes]

Thank you both (ADU and NTK), starting to convert from nebulous understanding to a more condensed one. Though, your comments I will need time to digest fully. Currently going through Floyd Toole's Sound Reproduction, which is a surprisingly good read. I think I did confuse matters for myself, but no doubt I will again soon, so I shall revert.

 

[ADU]

I imagine you've probably already seen this by now. But this video by Dr. Toole gives a pretty good synopsis of some of the more important aspects of a loudspeaker's frequency response. And also offers some good ideas on the best ways of interpreting some of those different characteristics to assess the quality, accuracy and neutrality of a speaker's frequency response.

 

[ADU]

The first caveat that I'll add to some of the info already posted above is that some reviewers and experts will use a speaker's listening window as the reference for a flat response, rather than its direct/on-axis response... They are similar things though.

As Dr. Toole explains in the above video, the listening window is essentially the speaker's direct response, if it was computed as an average over several different listening or seating positions. The direct response (or listening window) is always the first sound to reach your ears though, when listening to speakers in a room.

For an audio system to be accurate, and faithful to a source, the output of every component in the system needs to be proportional to its input. (I hope that's correct anyway, though this concept is still a bit confusing to me as well.) And this needs to be true across all of the frequencies in the audible spectrum that are contained within the program material.

If the system does not have a flat or linear measured response, as NTK mentioned, then the content will be timbrally or tonally distorted or compromised. This applies not only to the static components in the audio system (such as the DACs, receivers, amps, cables, and so forth). But also to some of its kinetic components as well... Meaning the speakers, in this particular case. Or more specifically, the direct response of a speaker's transducers, which can be thought of as basically just extending the linear response of your other audio gear into the mechanical realm.

This even applies, to some extent, to a speaker's sound in a room. And to a pair of headphones. Though it's a bit more involved in how we get there in these cases, especially with the latter. For a speaker to be perceived as "neutral", "accurate" or "reference" though, the first requirement is for it to essentially reproduce or carry-over the same flat amplitude across most of the audible frequency spectrum as your other audio gear, via its direct response.

 

[kongwee]

Go to listen audiophile show, you listen dozens of different setup. You will build up your own preference. How speakers behave in better treated room and non treated. Feel the SPL from different speaker size and driver size. Listen through with your favourite play list. You will know it is more than flat response and low noise. Not surprise your playlist will sound good on 300B. Then you will confuse yourself about measurement good amps. It is whether a system to entertain you or you entertain the system. Everyone has the unique approach what they like to hear.

 

[DesEsseintes]

I imagine you've probably already seen this by now. But this video by Dr. Toole gives a pretty good synopsis of some of the more important aspects of a loudspeaker's frequency response. And also offers some good ideas on the best ways of interpreting some of those different characteristics to assess the quality, accuracy and neutrality of a speaker's frequency response.

Yes, nice video. Thank you. Hadn't seen it already. Long, so watched it in 2 sittings. Quite a bit of crossover with the book. I think both the book, and this type of lecture help in at least setting me up to better understand the terms, and some overall context.

I think if I were to better define my objective here, I would state it thusly: How do I, as an average consumer relate my perceptions (again I will focus on headphones, as that is my preferred listening method) to some baseline? How can I tell if my headphones are good? Sure, I can buy a bunch of headphones and make some preference determination, those impression will be largely gathered in a non-repeatable, non-scientific approach. Perhaps with lots of experience I can more readily make these determinations. But again these are always relative impressions. Such and such headphone compares to some other headphone in some way. We get into circle of confusion territory (I think), as if I am testing with recorded music, I cannot know how it should sound. A frequency response curve could be a useful tool if I can relate my perceptions to the various frequencies. I can have expectations of performance, and make purchasing decisions based on this information. This also relies on a assumption that the measured frequency response is reliable. They most certainly are not always reliable due to variance in consistency of manufacture (and measurement). This variance itself will vary with manufacturer, and some will be more consistent than others.

So for me, what can give a reference, a constant or anchor in all of this? I think a sine sweep or pure tone generator can. Pink noise could also. Although we have to control for volume, which is not trivial.

It seems to generally be the case, for me, right now that a headphone or earphone which, when tested with a sine sweep produces and output which I perceive as "flat", (ie equal amplitude at all frequencies) sounds good. Whether this is optimum, or even comparable to a "flat" speaker in some condition, I cannot answer with much confidence right now.

/streamofconsciousness

 

[Mart68]

This may not be of any use to you but my method is to use a 'best of' album that has songs spanning decades (Bowie 'Changes' is a good one).

If the differences in production across the years with different studios, equipment, techniques, engineers are blatantly obvious you're probably close to 'neutral' with the replay. If it all just 'sounds like music' then you're probably not.

 

[ADU]

Yes, nice video. Thank you. Hadn't seen it already. Long, so watched it in 2 sittings. Quite a bit of crossover with the book. I think both the book, and this type of lecture help in at least setting me up to better understand the terms, and some overall context.

I think if I were to better define my objective here, I would state it thusly: How do I, as an average consumer relate my perceptions (again I will focus on headphones, as that is my preferred listening method) to some baseline? How can I tell if my headphones are good? Sure, I can buy a bunch of headphones and make some preference determination, those impression will be largely gathered in a non-repeatable, non-scientific approach. Perhaps with lots of experience I can more readily make these determinations. But again these are always relative impressions. Such and such headphone compares to some other headphone in some way. We get into circle of confusion territory (I think), as if I am testing with recorded music, I cannot know how it should sound. A frequency response curve could be a useful tool if I can relate my perceptions to the various frequencies. I can have expectations of performance, and make purchasing decisions based on this information. This also relies on a assumption that the measured frequency response is reliable. They most certainly are not always reliable due to variance in consistency of manufacture (and measurement). This variance itself will vary with manufacturer, and some will be more consistent than others.

So for me, what can give a reference, a constant or anchor in all of this? I think a sine sweep or pure tone generator can. Pink noise could also. Although we have to control for volume, which is not trivial.

It seems to generally be the case, for me, right now that a headphone or earphone which, when tested with a sine sweep produces and output which I perceive as "flat", (ie equal amplitude at all frequencies) sounds good. Whether this is optimum, or even comparable to a "flat" speaker in some condition, I cannot answer with much confidence right now.

/streamofconsciousness

Thanks for this update, DesEsseintes. I was gettin to some of the above (rather slowly I'm afraid). Because imo it's easier to understand what a neutral response is in a headphone after you have a better grasp of what neutral is in a pair of speakers. Since you've watched the Toole vid and have already read some of his book on this though, I'll try to skip a little ahead on some of this.

Since headphones are measured from inside the ear (unlike loudspeakers), the baseline for a neutral response on a pair of headphones should, imo, be the same as the measured in-ear response of a pair of neutral loudspeakers. This is difficult to achieve in practice because the in-ear response of a loudspeaker is complex, and the result of a number of overlapping effects in the time domain. And basically contains a combination of all the different frequency response components discussed above in the Toole video (ie, the direct sound, early reflections, and diffuse sound power), as well as the resonant effects of the ear, head and body on those different components.

The closest that we can probably get to approximating something like this on a standard pair of headphones, with no DSP for the time domain effects, is to simply measure the steady-state response of a pair of neutral loudspeakers in an average room at the eardrum reference point. Which is the same place that a headphone is typically measured. Most headphone review sites do not currently do this though. Nor do the manufacturers of speakers. So obtaining this kind of in-ear data on loudspeakers can be somewhat difficult.

There are some other methods of estimating or approximating this type of response at the eardrum though. And one method that seems to work fairly well is to combine the speaker's diffuse sound power with the in-ear measurement system's response to a spectrally flat diffuse sound field.

Some in-ear measurement systems are better for doing this, and also for taking actual in-ear measurements of loudspeakers than others though. Because accurately approximating the resonant effects of the head, body, and ears on a diffuse sound field, or on the steady-state sound of loudspeakers in a room requires some kind of a mannikin. And the in-ear measurement rigs that many headphone review sites (including ASR) currently use do not generally include that.

 

[ADU]

One way of assessing the accuracy or neutrality of a headphone's sound by ear would be to compare its tonal quality/response directly to a pair of well set up neutral loudspeakers with the same recordings, using some type of A/B switching. If you were comparing the sound quality and responses of several different headphones, this might help you to narrow down which set of headphones does the best job of approximating the speaker's sound, for example.

And if you wanted to achieve a better match between the headphones and speakers, then I suppose you could also listen to a variety of tones on the two systems at different frequencies, and try to EQ the volume of each frequency on the headphone to more precisely match the same tone on the speaker. This would probably take a fair amount of work though, and be more difficult than simply comparing the in-ear measurements/frequency response plots of the two systems, and making the adjustments that way.

If you have the time, desire and equipment to try something like this though, it might be an interesting experiment.

 

[ADU]

I haven't read Dr. Toole's book on Sound Reproduction btw. Though I'd very much like to at some point, because it's considered by many to be one of the definitive works on the subject of frequency response in loudspeakers and rooms.

One area that the above video could perhaps have done a little better job explaining though is the subject of a speaker's dispersion or directivity. Because this is apparently just about as important (if not more so), as the neutrality or linearity of a speaker's direct response.

If Dr. Toole's assumptions are correct, then a neutral loudspeaker really only needs two things in terms of its frequency response to be considered neutral. And those are a flat direct response across most of the audible spectrum, and a uniformly increasing (or decreasing, depending on the way you look at graphs of this function) dispersion/directivity. Because the latter of these characteristics is, to a large degree, what seems to determine the tonal balance of the speaker's off-axis response in a room.

If you have a speaker with both of these qualities, namely a flat direct/on-axis response, and reasonably smooth/uniform/linear dispersion/directivity and off-axis response, then it should be capable of delivering what I'd describe as a reference response.

 

[DesEsseintes]

One way of assessing the accuracy or neutrality of a headphone's sound by ear would be to compare its tonal quality/response directly to a pair of well set up neutral loudspeakers with the same recordings, using some type of A/B switching. If you were comparing the sound quality and responses of several different headphones, this might help you to narrow down which set of headphones does the best job of approximating the speaker's sound, for example.

And if you wanted to achieve a better match between the headphones and speakers, then I suppose you could also listen to a variety of tones on the two systems at different frequencies, and try to EQ the volume of each frequency on the headphone to more precisely match the same tone on the speaker. This would probably take a fair amount of work though, and be more difficult than simply comparing the in-ear measurements/frequency response plots of the two systems, and making the adjustments that way.

If you have the time, desire and equipment to try something like this though, it might be an interesting experiment.

Yes, this could work. I could even maybe figure out how to listen to pink noise effectively for such a test.

 

[ADU]

I'm not quite sure how to answer your question about using equal/perceived loudness with a sine sweep to accurately tune or assess the neutrality of a pair of headphones or speakers, without some kind of a neutral reference for comparison.

What I mainly use equal loudness for is to adjust the relative levels of the bass, treble and mids on my headphones to have a more neutral or natural-sounding response for listening at different volumes. Because I think you need a little more bass and maybe treble (or less mids) for a recording to sound approximately the same at lower volume as you do at a higher or reference volume.

I don't know about the new ones, but alot of older receivers used to include a loudness control, which would boost the bass and treble a bit versus the midrange for lower volume listening. And according to the Wikipedia article on this subject, this was based on the concept of equal loudness (aka as the "Fletcher-Munson effect"). The $100 Bellari HA543 headphone amp that I use also has a feature like this, called "Enhance". But I prefer to keep this control turned off, even though I generally listen at lower volumes, because the headphones I'm using at the moment (a pair of 250-ohm Beyerdynamic DT-770's) already have a fairly U-shaped sound and a bit too much emphasis in the treble to begin with.

I tend to like the sound of speakers that are a little more recessed in the mids for listening at lower than reference volumes for the same reason. But also because alot of the content that I listen to is probably mastered on inferior gear, which is somewhat compromised in its off-axis response in the upper mids, at the cross-over of the speaker's tweeter and midrange driver. Where there is often an increase in some speaker's directivity (or a decrease in its dispersion), and a corresponding dip in its sound power/off-axis response.