The 10 stages you’ll go through when learning to read headphone measurements

hi guys this is andrew with headphones.com many of you guys might be familiar with the reviews that we produce at headphones.com and of course we conduct and produce headphone measurements to go along with them now if you're totally unfamiliar headphone

measurements specifically frequency response are extremely useful for getting a sense of how a headphone is going to sound before you buy it but at the same time headphone measurements are an extremely complicated and deep

subject one that's impossible to fully explain within the context of an evaluation like the kinds that i do like some of the videos that you might have seen i try and do my best but it's very difficult more importantly it's also

impossible to know how familiar the audience you guys are i don't just like putting the measurements out there with no context so in this video i'm going to do the self-serving task of expressing what i found to be a common trajectory

for people when it comes to learning about this stuff and learning how to interpret headphone measurements now that said don't take the order of this too seriously people come at this information in all kinds of different

ways but with that out of the way let's get right into the 10 stages of interpreting headphone measurements alright so stage 1 deviations from flat this stage is the most basic

interpretation of a headphones frequency response it's probably the first thing you come across when looking at audio measurements of any kind or researching headphones online now this category is also where the majority of people fall

there's a prevailing wisdom that it's good if headphones have an even tonal balance without any particular region being elevated over the rest and it would be represented as a flat line that would be seen as neutral or good or

generally desirable now if you're at this stage you likely know what elevated bass looks like or recessed or boosted trouble and so forth and you're able to get a general sense of what the broad regions of headphones frequency response

affect when it comes to you know what you hear so you're able to kind of correlate what you're seeing there on the graph generally speaking with with what you hear another way to think about this would be like the general tilt or sound signature of a headphone but if

you go a little bit deeper into this topic you'll learn that deviations from flat line on a graph when looking at headphone measurements are really just deviations from a particular target that the review site in question has deemed

neutral and this brings me to stage two at stage two you have a grasp of raw versus compensated measurements and what that means once you've learned that deviations from flat are actually compensated measurements and the result you end up with is highly determined by

the target being used as the compensation you're bound to be more interested in raw measurements or at least i was so that you can better understand and predict how something is going to sound you know irrespective of

the target that's being used at this point you've likely asked yourself well what other targets are there out there and this leads you down the path towards you know the differences between diffuse

field free fields harman and i think more realistically how other targets differ from you know the famous harmon target and then maybe you're asking yourself well why is

harman commonly used as the standard and that leads to stage three hating the harmon target and i know a lot of people who are there and probably some of you guys watching are there right now it's at this point where you will at least be somewhat familiar with the harman target

maybe you've watched some of my videos on the topic but people at this stage typically don't fully grasp the extent of the research partially because of the generally cursed piecemeal publication model that requires reading multiple

articles behind the audio engineering society paywall which is super annoying and so this is perfectly understandable without doing that there's really no way to get a complete picture i think maybe more commonly there are situations where people realize the headphones they love

deviate from the harmon target in a number of places and this will lead them to conclude that they just don't like the harmon target and they walk away thinking it's just a tuning for the unwashed masses one that's been overly influenced by those who don't know any

better or they don't know anything about good sound quality another reason you might feel this way about the harman target is if you've ever used an eq profile like the many that are available um you know for a headphone that you own that's intended to achieve the target

and what you find is that you don't really like the way that it sounds and then you think i guess i just don't like harman but this is also an extremely limited and even somewhat incorrect view of the research and i'll talk more on

this later but this leads me into the next stage stage 4 discovering other metrics beyond frequency response after rejecting harmon it's pretty common for people to delve into the deeper or more complicated measurements you might be

saying to yourself time-based views like square wave and impulse response should tell me if a headphone is fast enough or if it's resonating badly why didn't the harman research include these clearly there's more to headphone measurements than just frequency response right or

maybe another example how can a headphone with high harmonic distortion be as detailed as one with almost no distortion that's clearly another thing we need to measure and indicate to be able to decide which headphones sound

best and further the cumulative spectral decay or waterfall plots what better way to look at how accurately a headphone will render the decay than to literally plot the change in level against time and after learning about these different

metrics it might be difficult to believe that the harman target or anything that's based on you know just frequency response could matter that much given how many additional metrics exist when it comes to headphone measurements

or just audio measurements in general and this leads to stage five which is realizing that those other headphone measurements and metrics don't really matter all that much this is where you have to do a little bit more rigorous analysis and it helps if you've actually

had the opportunity to listen to the headphones that you're testing and measuring um and when you do that you'll realize that there's something not quite right about those other metrics and what your assumptions might be or the conclusions that you may have drawn from

those other metrics that you're getting for example the hifiman says vara has much higher measured harmonic distortion than the odyssey lcd2 so why isn't it less detailed if you actually get to listen to those headphones you'll know

what i mean and there are even way more extreme examples than that you might also keep seeing people say that you know headphones are minimum phase and that means that csd and impulse response don't matter so what's that all about

this is one of the things that i actually initially got wrong as well and it's taken me quite a bit of time to kind of come around on that but if this is you you're starting to run into cases where those other metrics that seemed like they'd explain what was missing

from frequency response aren't lining up with what you hear or the technical explanations that you're seeing and maybe you also had an opportunity to see those plots where an eq that changes the frequency response of a headphone removes the csd decay

with it maybe you found that old headphone post that showed a minimum phase eq with the same response as the headphone produces the same square wave and you might know that a paper studying the audibility of headphone distortion

found pretty negligible differences as long as it wasn't awful and there's actually a video on this channel that also talks a little bit more about that i'll leave that link in the description as well so you look a little bit deeper

into the other metrics and you realize that unless things are very wrong they really aren't the thing to explain what you're hearing and unfortunately they don't explain what it looks like they're explaining but now you're back to where you started and things make a lot less

sense than before and now we get into stage six loving the harman target and that's probably because you'll have begun to peer around the audio engineering society paywall to dive deeper into the harm and research

to understand it more thoroughly to get a more complete picture of it it's at this point that you may think that while it's a generally decent starting point you still dislike the idea of frequency response targets that are based on listener preference but the more you

read the more you learn how the headphone target was developed and that while preference is a contributing factor to the end result it's also anchored to what's been generally agreed upon as good sound in speakers you start

to learn what the actual adjustments or preference elements were in the research and why the base shelf is where it is for example and this is also where you realize just how thorough and comprehensive the research is as you see

how the various different pieces all fit together you see how each study is corroborated or validated by a different study that tests additional things and provides a solid and interconnected foundation for what constitutes good

sound in headphones you also learn that there's so much more to the harman research than just the targets we commonly use to evaluate headphones like for example the paper on segmentation and this is when you really start to appreciate the harman research and you

realize the reasons why many reviewers like myself use it and find it extremely valuable on to the next stage stage seven understanding the differences among measurement rigs at this point you will likely have come across

measurements from a wide variety of measurement rigs you'll probably already know the limitations of measurement equipment like the mini dsp ears and the you know various different knockoff

grass systems that are out there in the wild but you'll also start to become acquainted with the differences that exist among the various 7-eleven standard compliant measurement rigs and their various features for example this

is where you will have learned about the 8 to 10k concha dip on the kb5000 anthropometric pina used in the 43 ags and 45 cas and how that's different from the features found in the bnk 4128 and

the hms systems more substantially you'll likely have also encountered measurements done on the more advanced b k 5128 like the ones done by jude over head fi and a number of other places i think most importantly that headphone

measurements done on these are not comparable to the ones that are done on the more widely used 711 couplers another important thing to note with this is that targets devised on the older 711 couplers like the 2013 or 2018

harman target are not compatible with this new system and shouldn't be used as a reference point on it so harmon is really only applicable to the older 7-eleven couplers at the moment and they haven't said that they're going to make new research public on the new one and that

brings us to stage eight hating the harman target again many who have gotten to this stage have a love-hate relationship with a harmon target we love it because of how useful and thorough the research is and particularly for me as a headphone

reviewer because it makes my job a lot easier but we also hate it for several different reasons the most significant of these is that the target is heavily smooth while the headphones frequency response typically isn't and generally shouldn't be because we want as much

information in a measurement as possible this means that there are bound to be meaningful deviations in a headphone's treble response above 5k that the target can't possibly be used to evaluate against now there is a sense in which a

zoomed out analysis provides a general understanding of a good or bad result but for those of us who want a more fine-grained analysis the one-half or one-third octave smoothing of the harmon target simply isn't good enough and

we're left to sort of interpret that as best we can and if that doesn't make sense let me show you what i mean by this so what i'm showing you guys here for the red line is the frequency response of the sennheiser hd800s which is a very well known headphone the

dotted line here is a version of the harman target now the red line the frequency response of the headphones that are being measured here is a very fine grained result by contrast the dotted line here is a much more coarse grained or smooth target

so let me show you what it looks like when i take the headphones frequency response and do an apples-to-apples comparison and apply the same smoothing as the target suddenly all those features that you saw earlier in the treble are gone

let me show you once again this is the fine grain result you can see there's various different peaks and dips going on in the treble and when that suits the same degree suddenly it looks a lot closer to what the target looks like now

let's apply that same logic to the target on its own because the target is highly smoothed here we have no idea what those various different features in the treble should be for any individual now what's worse is that we also know

that most people are not evaluating it this way either and it gets increasingly frustrating to see certain regions of a headphones frequency response being so heavily scrutinized when the target doesn't tell us exactly what it should

be anyways it's just not at all obvious from looking at common headphone measurements that there are limitations to how certain regions should be evaluated in addition to that the segmentation paper i mentioned earlier

from olive welty and konsarpur back in 2019 it also shows one of the problems with overly focusing on the harman target we commonly use in practice so within the harman research there's a reasonable clustering of preferences

around that target some people like more bass some people like less some people like more trouble some people like less and so you know there is actually a cluster analysis that's done when you go through and read the rest of it but we

never use the rest of that so a headphone with a frequency response that deviates quite strongly from harman probably won't sound very good but even with the amount of work that has been done it can't be as simple as matching

closer will sound better to me and then you see people developing eq profiles to match the target with the expectation and indeed the kind of fervor like you know this is how it should sound that it will sound better to them all i can really say there is that confirmation

bias shows up in all kinds of places but this gets us to stage nine probably the most advanced stage and that's the importance of the head related transfer function this is a very deep and complicated subject but in short the

head related transfer function or hrtf can be thought of as the way your head and ears impact incoming sound the way your ear changes the sound based on where it's coming from which is also part of how we place sounds in space

before you knew that you know diffuse field and free field were headphone target response curves at this stage though if you know about hrtf you know that they are also head related transfer functions and that they're individual to

every head and every ear this on its own should be a massive revelation for anyone trying to evaluate headphones by looking at graphs with any of these standard targets being used or any of them being shown as the the reference

point so maybe you've gone on to read a paper from a particular danish university about the human hrtf variation and there will be links to that in the description as well if you're curious but this is also where a number of additional questions arise you

start to think that maybe it's possible to have a higher resolution headphone target response you know maybe by applying the harman in room target adjustment to a high resolution diffuse field had related transfer function or even something more complicated of

course now you're upset that more graphing places don't let you upload your own target but in any case you've likely realized that your own hrtf differs to some degree from that of the measurement fixtures and you're trying to figure out how you can extrapolate

those eardrum responses to your own eardrum right because you're ideally trying to figure out what the optimal response would be for you you might even be looking into buying a set of in-ear microphones or getting someone to make

them for you so you can measure speakers and headphones on your own head from your own actual ear response really at this point if you aren't eq'ing your headphones religiously it's almost inevitable that you start probably to a

custom target that you've come up with and the reason you might want to do this is that you know you ultimately want to get the best possible sound quality for you and not just rely on you know eq profiles or something matching harmon or

not because that's not necessarily going to be the best for you and this leads us to stage 10 being the worst person to talk to at parties i think the point here is that if you're ever considering or have considered asking someone else

to let you measure headphones on their head with inner microphones make sure they're at least somewhat interested in this topic but there is also something else to consider and it's that this doesn't end as much as there are a lot of things that have been figured out

there's a lot more still to figure out there's a lot of stuff where we just can't predict what that's going to sound like for individual people i've often talked about how we're only really interpreting frequency response in one

particular way right now that being of course you know generally total balance and while that is super important i really think that there's so much more contained within frequency response that we have yet to uncover and figure out

what's going on there so that we can predict how something is going to sound because without that you simply don't know how something's going to sound just by looking at a graph unfortunately you still have to actually listen to the headphones to get a sense of that but

anyways that does it for this video thanks for taking the time to watch it and i will see you guys in the next one bye for now