About this guide : I will soon be conducting a couple of studies in psychoacoustics that require a listening system that is transparent as possible for the data to be accurate. According to scientific literature, personal experience and external feedback, I am confident that the system I have put together is pretty much the most neutral monitoring system one can have right now (unless you have access to an anechoic chamber and multiple high-end speakers). I've written this guide in order to help people in pursuit of their ultimate monitoring system, and also to get feedback to be sure I'm not missing anything. Note that I'm not saying that this is the "best" monitoring system, because it is based on headphones/IEMs, which produce sound differently from speakers, and some people can't even stand to work on them. "Best", in some way, is subjective.


Also this guide will be quite a long read. Skip to step 2) if you want to skip the science stuff.
​

To many engineers, their perfect monitoring setup is usually a combination of three types of systems :

• A transparent, "main" speaker setup
  
  About this one, we know that there's no such thing as a perfect speaker listening setup, because there's no such thing as a perfect room. If you have a compltely dry, anechoic room, it will sound harsh and unnatural. If you have reflections in your room, it will cover the direct sound in some way, and take away some of the intelligibility of the sound.
  So, even if you manage to perfectly treat the pressure, modal and flutter issues of your room (which is extremely hard and costly), the balance between direct sound, early reflections and late reflections is very personal, and depends on the work you do. For example, those mixing classical music might benefit from a more reflective room, and those mixing a movie might benefit from a less reflective one, because they better represent the acoustics of where the content will be played.

• "Field" listening devices, that can represent where your work will be played
  
  It can be a car, a TV, hifi speakers, a live PA, airpods... This all depends on where do you think your work will be played, and what you have at your disposition.

• Headphones/IEMs
  
  Unlike the other two, there can be such a thing as a perfect pair of headphones or IEMs. It's the closest thing there is to plugging the sound directly to your brain. We're going to focus on how to achieve such a system today (or at least get as close to it as possible). It involves choosing the right pair of headphones/IEMs, driving it properly, and then correcting it.

I know that mixing only on headphones/IEMs is not optimal. They present sound in a very different way compared to speakers (binaural soundstage, no room reverb, lack of tactility). They can sound brutally honest. Maybe too much for some.
However, they do have their advantages. Because of the lack of room sound, they are extremely detailed, revealing the input signal without room reflections. Also, the sound will be consistent regardless of the room you're in. You don't have to plan and buy acoustic treatment, and you can take it anywhere. And remember that a good chunk of people, probably even the majority, listen to music on headphones and wireless buds. You want to know how your work sounds on this type of device.
Also, they are relatively cheaper

Foreword : How can we know what the goal is ?​

Spoiler

You might wonder, how can we know exactly how a monitoring system sounds ? And how do we know that there's an ideal goal to attain ? Well we can know all of that because there are numerous studies about it. It is possible to measure anything about a speaker or a pair of headphones. However it doesn't necessarily translate to our hearing if we don't take into account human audibility. Because our ears and brain can't pick up everything, some measurements are more important than others, and some areas of one graph are more important than the rest. This is a big subject and I won't go into the details here, but for anyone curious, I highly recommend these two videos. The first one is a talk by Floyd Toole, an ex-researcher from Harman that lead studies showing that subjective sound quality (human judgement) is actually very predictable and not very different from objective sound quality (based on measurements) :

The second one is a Genelec 8050B review by Amir Majidimehr, in which he explains how to read and interpret the most common measurements.

So, with that in mind, we can know what the goal is because it is possible to accurately measure an audio system and describe its performance in every area. The goal for headphones/IEMs is to get as close as possible to the target measurements, so that they can best reproduce the input signal. Although there are numerous different types of measurements possible, we want to assure performance in the three different domains that characterize any audio system : Magnitude response, time response, and non-linear distortion.

1) Defining the goal​

Spoiler

a) Time response​

It's simple : We want every frequency to arrive at the same time at our ears. Regarding resonances, most headphones/IEMs are damped well enough. They should be inaudible. You can check spectrograms just in case there's something weird going on, but for most devices, a phase response graph is sufficient. And we want a flat one. Or at least one that is below the human audibility of phase shift. Anything below 1ms of group delay in headphones should be inaudible.

b) Non-linear response​

It's also very simple. We don't want any kind of distortion. We want it to be as low as possible. It is generally accepted that for acoustical transducers played at reference level (around 85dB SPL), a THD below 1% (-40dB below fundamental) is inaudible. The lowest threshold coming from those studies is at 0.7%, so that'll be our threshold too. I know that THD is not a measure of intermodulation distortion, which is the kind of distortion happening when listening to actual music. Unfortunately, it's extremely hard to find multi-tone graphs for headphones/IEMs. Also, devices performing great in THD usually perform great in IMD, as they have the same root cause. So, a simple THD graph is sufficient. And we want it to be below 0.7%.

c) Magnitude response​

That's where things get a little complicated. Although the target magnitude response for speakers is very simple, which is a flat response in an anechoic chamber, it's not the same for headphones/IEMs. This is all because as a sound strikes us, our ears, head, ear canal etc... alter its frequency response, by boosting or attenuating different frequencies. This measured response at the ear drum is called a HRTF. And because the sound produced from headphones/IEMs comes at our ears in a very different way than the sound produced by speakers in a room, or musicians in a concert hall, we need a way for headphones/IEMs to achieve a HRTF that sounds "correct" to our brain. Let's define our target magnitude response curve.

One of the ways to determine such a HRTF curve, is by simply using a head and torso simulator to measure the magnitude response in a reference room. One that has calibrated speakers and good acoustics. This is what Harman did to get the Harman Linear In-room response curve :

​

This is a good representation of the magnitude response of the sound at our eardrums in a studio calibrated to a flat-response. It serves as the baseline for the Harman Over-Ear and In-Ear curves, which are the most popular headphones and IEMs target curves.
However, there is a problem with these curves. This is because when sound is coming from a certain direction, our brain will use the perceived HRTF, as well as the difference in volume and delay between our left and right ear to determine where it is coming from. Then, it will apply inverse filtering to restore the sound's original timbre. Compared to speakers, headphones/IEMs do not feature crossfeed or time differences between the two ears. So their sound doesn't present any kind of localization.
There's a solution to that. This article explains it :

This is called the diffuse-field target curve. It's the HRTF of a sound coming from all directions, or in another way, independant of any direction. It's how our brain localizes the sound coming from headphones/IEMs, so this will be the baseline for our target curve. However, look at how a diffuse-field curve compares to the Harman In-Ear curve from 2019 :

​

You can see the that the diffuse-field target is significantly brighter than the Harman one. This is because it doesn't account for the in-room response of speakers. Because of its directivity and the air absorbtion, a speaker that measures flat in an anechoic chamber will measure with a downward slope in a real room :

​

This is also true for mixing/mastering studios. We need to account for that in our target curve. If not, our headphones/IEMs will sound much brighter compared to the speakers of a studio.

The last step is to adjust the bass profile of our target curve. Because headphones/IEMs lack the tactility that speakers provide, we need to restore some of it by boosting the bass. The amount needed is preferential, but a good starting point is to use the bass boost in the Harman curves, which have been averaged by listeners preferences :

​

Notice how the bass boost is more significant for IEMs, because headphones provide some sort of tactility through the bones around the ears. So, we now know what our target magnitude response curve is : A diffuse-field curve combined with a room response curve and a bass boost to restore the lack of tactility. Here's an example of such a curve without the Harman bass boost (in grey) :

​

If you're paying attention, you should wonder why we use a generic diffuse-field curve, and not a personally customized one. After all, we're all born with different ears and bodies. There are two reasons for that.

• Diffuse-field curves are personal to an individual, but it's the HRTF that varies the least between individuals. There's a good chance that an average DF curve is similar to your own one, and sounds "correct".
• It's very complicated to get your own diffuse-field HRTF measured, without expensive equipment and a dedicated room.

2) Choosing the hardware/software

Spoiler

a) Electronics​

For the digital to analog converter, it really doesn't matter. Most DACs today have ruler flat frequency response, inaudible noise floor and distortion.

For the amplification stage, that's where things get complicated. Headphone amps can have very different output impedances, and because the impedance of a pair of headphones or IEMs can vary across different frequencies, the resulted magnitude response can change from one headphone amp to another. This is called the damping factor. Look at how the magnitude response of the Focal Clear changes between a 0.2Ω amplifier (Red) and a 120Ω one (Green) :

​

Not that this is not the case for all headphones/IEMs. But in general, you should go with amplifiers with a low output impedance (<1Ω). Check the specs of the audio interface/headphone amp you're currently using. If you just need an amplifier, I recommend the Schitt Magni+ (110$) or the Topping L30II (150$). If you want to change your audio interface, there a lot of options. But on low budgets, I recommend any recent model by Motu, like the Motu M2 (200$). They really have best-in-class headphone output performance. You can check this little comparison here :

​

Feel free to suggest any other interfaces, DACs or amps with similar performances

b) Headphones/IEMs​

This is where I'm gonna disappoint a lot of people. Headphones are not optimal for this setup. IEMs are.

Right now, the best measuring headphones according to many is the Dan Clark Expanse. Not only do they cost 4000$, you can also see on those measurements that the timbre of the highs is just bad (on top of poor channel matching).

The problem with headphones is that they are dealing with much more complicated acoustics compared to IEMs. It doesn't matter if it's closed or open-back headphones. Closed ones are dealing with a lot of internal reflections, and open ones are still dealing with some reflections and low acoustic impedance on top of that, making bass response generally thin.
Also, everyone ears are very different, and the way the sound reflects on them differ a lot from one individual to another. So even if the headphones measure perfectly on a measurement rig that mimics an average ear, it might sound different on yours.

For the rest of this guide, we're going to use IEMs. But you maybe have reasons to want headphones instead (comfort, hygiene...), so if you do, I recommend you the Ollo S4X/S5X (419€/489€), mainly because you can request to Ollo the individual measurements of the pair you ordered. I own the S4X, it sounds great out of the box, is amplifier independant and it's very comfortable. You can also buy spare parts directly from their website, so that's always a big plus ! When we'll get to tuning in the next step, you'll have to use a target curve based off the Kemar DF curve, which is the diffuse-field curve obtained from the equipment Ollo use to measure their headphones. The tuning probably won't be as perfect as what we'll do with our IEMs, but it should sound great, better than any other headphones without correction.

​

So, for IEMs, after an extensive research, I determined that the Moondrop Variations (520$) is the closest thing to what we want to achieve.

​

I am absolutely not doing advertisement here. There are A LOT of other fantastic other options out there. Xenns Mangird Top (530$) is probably the second best alternative for this build. I slightly prefer the Variations as they are even closer to the target, so not only it will make the EQing part super easy, it also sounds great without any correction. So if I'm going to a studio where I can't put any DSP correction for my IEMs, I'll still have a great reference. But really, you can actually use any IEM that has low enough distortion and consistent unit variation, and consequently channel matching, so that you can use measurements made by other people, and correct the magnitude response without introducing distortion. Again feel free to recommend other IEMs with similar performances

Here is the magnitude response of the Variations, compared to Moondrop's VDSF target :

​

The VDSF target has been calculated by combining the diffuse-field curve from a B&K 4128 (same rig used to measure the Variations in this graph), and room response curve. Which is exactly what we want. The bass response differ, but we'll get to that in the next step.

Here is the measurement of THD at reference level (in grey). It's under 0.7%, so that's what we want :

​

Here's the phase response graph :

​

We should look at the excess phase response (in white). Here we see a problem. It's flat at -180°. It means that their polarity is inverted. When talking with people online, it seems that it's the case for all models out there, so this is intentional from Moondrop. They're stupid. Thankfuly, it's easily fixable by rotating the 4 pin connector at the jack :

​

It will then get you this phase response, which is perfect :

c) Correction software​

Basically, you can use any minimum phase EQ. I choose to use CraveDSP CraveEQ2 for two reasons :
- It's very light on CPU
- It has a "transparent" mode, making possible to use linear phase filters in the high-end while keeping the latency low. It's useless for the devices we mentioned, but there are some headphones that I corrected that really benefitted from this option, as the Q values could get high, and some dips and peaks weren't minimum phase.

It's also possible to apply correction to the sound of your computer outside of your DAW. On Mac, you can use Rogue Amoeba's SoundSource, and on Windows you can use a combination of VB Audio's Hifi-Cable and Niall Moody's Pedalboard2

3) Tuning​

Spoiler

a) Correcting between 200Hz and 6kHz.​

Below 6kHz, measurements are consistent between measurement rigs and individuals. So, we can assume that the measurements of the Moondrop Variations are very accurate to what we hear. We can confidently correct the differences between the measurement and the VDSF target. We don't want to touch below 200Hz yet, because this is quite subjective (lack of tactility). If you have any other headphones/IEMs, just make sure you're using a target curve obtained from the diffuse-field response of the same measurement rig used for your device's measurements, and that it has been compensated by a room response curve of your choice.

This is what is needed for the Variations :

b) Correcting above 6kHz.​

By inserting an IEM, you're creating a seal between its driver(s) and your eardrum. In such a tiny enclosure, there's gonna be a rise in acoustic impedance in high frequencies. And because we're not all born with the same ears, it will occur at different frequencies :

​

So here, we have to correct for both the differences between the magnitude response of the IEM compared according to the target curve (on the Variations, just a slight 1dB dip at 6kHz (and 14kHz ? The measurement rig used follows a standard that ensures accuracy until 10kHz. Anyway it's just 1dB, and it sounds good for my ears and on my units )), and how it matches to our own ear canal. So the only solution here is to grab a tone generator and sweep a sine wave across the 6-20kHz range, and adjust the EQ so that the response sounds smooth. There should be one or two peaks. Here's the correction for my ears :

c) Adjusting the bass boost and tilt shelf with listening tests​

This is where it gets a bit more subjective. To account for the loss of tactility that speakers provide, we need to compensate the IEMs magnitude response with a bass boost. The most popular one is the Harman one, which is exactly what the Moondrop Variations follows :

​

So play with a bass shelf, and hear what kind of bass profile you like by listening to music you like. The Harman bass boost has been averaged over multiple listeners preferences, so it's a good starting point.

Also, grab a tilt shelf filter and try to see if you prefer different target room curves.

I personally prefer not changing anything and keeping the default tuning, but this is subjective.

4) Speaker emulation (or not)​

Spoiler

In the last few years, we have seen multiple systems that try to emulate the listening experience of speakers on headphones (Waves NX, Slate VSX, Acustica Sienna...). If you've read this post in its entirety, you might guess what I think about these things. I think it's just a gimmick. Because our brain applies inverse filtering to restore the timbre of a localized sound from our own HRTF, those plugins can't accurately recreate the sound of speakers without destroying the timbre. Look at the variations in the HRTFs of different angles, measured across different individuals :

​

The differences are immense. There's no way that the generic, averaged HRTFs used by those plugins can get close to your own ones. The only solution to that is to stand in front of the speakers, capture the IRs of their sound at your own eardrums, and then measure the magnitude response of the pair of headphones at your own eardrums. This is exactly what Sony is doing right now with their customized HRTF program for spatial mixing (capture starts at 7:52):

Anything else is just a gimmick.

​

5) Enjoy !

You now have a perfectly tuned pair of IEMs (or headphones if you went that way) that sounds completely transparent (or at least as close as you can if you don't get your own personal diffuse-field curve). Of course, this is not the same thing as a good set of calibrated speakers. However, it's not inferior, it's just different. I have no problem using this as my main listening system, and using my speakers only to check how hard the bass hits, the stereo soundstage, and the intelligibility in a reverberant room. Things hard to judge without speakers.

Happy producing/mixing/mastering

 

Last edited: Aug 17, 2023

It'd be great to have a transparent monitoring system but totally useless without a "transparent" or tuned room.

 

This is scientific paper stuff

 

but why....if you use IEM the room disappears?

 

Monitoring is more important for quality playback then the room. People say it the other way around nowadays, but it simply is not correct.

The room "only" acts as a multiplicator, the foundation is the monitoring.


For the thread: Everything is correct. The only thing one can argue about is the harman curve. But as far as I know, there isn't a more evidence based curve out there unfortunately.

And the only problem I see with IEM's that is not adequately fixable simply is the bass response. This and that one has to be careful with loudness really are the only reason why one still needs studio monitors.

 

Very well put.

Last year I decided to mix a few projects with a set of cheap bluetooth IEM. The result came out incredibly well.

So when I came back home after my winter trip, I decided to invest in a set of 64 audio u12t IEM (2000$ USD MSRP). Highly praised for it's accuracy and details but my opinion did vary by quite a substantial margin in regards to the balance and the soundstage.

That's when I understood that clearly, IEM where way too inconsistent (because of the different insertion depth every time you put them on ) to rely on them. I obviously sold them but I still have my crappy Bluetooth IEM that I use to reference my bass level a lot.

Later on, I tried a custom fit IEM which helped a lot with the consistency but still felt that everything was thin and busy. Also sold those and I'm back into headphones exploration.

3 or 4 months ago, I bought the VSX bundle which to me is quite difficult to understand. So far, I'm not impressed but I can tell that there's something to get out of it. It's just not your typical precise monitoring system I'm used to in my studio at home. But there's something to get out of it.

So last week I decided to take a plunge and bought a set of Hifiman Ananda Nano. Those are planar headphones that take EQing very well. I'm thinking of sending them to Sonarwork for a custom frequency curve made for SoundID but I'm not sure I'll have enough time to do it before I leave. I'm also worried about the lack of phase correction in Sonarwork as those Headphones are quite a big step up in step response because of the technology used.

I think the IEM idea is an idea someone should venture into with caution. I saw people do great things with them and some others completely rejecting the idea after a few failures. Sadly, I'm of the second batch and I would have seriously loved to be able to get the idea to work properly because of its portability but ....yeah, it just doesn't work for me.

EDIT: You can read my full review of my first High End IEM adventure here:
https://www.head-fi.org/threads/my-first-high-end-iem-64-audio-u12t.967569/

 

Last edited: Aug 17, 2023

Yeah, I know that target frequency response curves are a hot debate. This is a quote from the same Stereophile.com article I linked when I talked about the diffuse-field curve :

There are no study regarding listeners preferences between those tunings, but I still think that because of the inverse filtering our brains do, diffuse-field should be the correct baseline. But honestly, when you apply room response curves, it gets quite similar to the Harman curves :



So when it comes to mundane mixing/mastering and music listening, I think it's fair to say that both tunings should do a good job.

 

I know btw. that you know Sharur. What do you think about his curve?
I never looked deeper into it, is there anything scientific behind or just "good hearing"?

 

It's not easy to look at Moondrop Variations without hearing about Sharur lol. From the videos I watched, I think that he has an insufferable personality, inconsistent hot takes, but his recent videos are quite accurate. I actually learned from him that absolute polarity is indeed audible. So I have some respect for him, I just skip his drama queen attitude of calling everyone else deaf, and calling pretty much every product lo-fi or something I also don't agree with some things he's still saying today, but it's the same for anyone.

There's definitely a scientific consensus that diffuse-field is the most logical baseline, but no one agrees on what is the target that sounds the best. Harman and Sonarworks have both recently published papers about this, and their conclusions were very different. We just know for sure that free-field sounds trash. To be fair, all those studies that aim to determine the best sounding target are using various music recordings, which probably have arbitrary decisions made during mixing. Switching target curves is like changing the EQ preset while mastering So I strongly believe that diffuse-field based targets, like Moondrop's VDSF or Sharur's one, are the most "scientific", as they aren't based on listeners preferences.

 

Hey, maybe put the bulk of that post in a spoiler tag so it's easier to scroll past...

 

Informative post. My take-away...very interesting, but for example, ain't no mastering engineer gonna perform his/her black magic using in-ears. Proper external headphone amp to solve impedance issues, of course - if you aren't familiar with your e.g. 'audio interface's' short-comings. Prep your space, and familiarize your ears with your gear & set-up, you're now ready for anything. Ya gotta just go for it, man.

 

The room doesn't disappear, you just moved to a smaller room that's all.

 

And what happens when the "multiplicator" is greater than the "foundation"?

 

The argument is, that a kitchen radio sounds bad in a good room as well.
Now one can say "Sure, the frequency range is very small". Yes, you can extend the frequency range. It still sounds bad in comparison to a good speaker.

However a good speaker will sound good in any usual reflective listening space.
That is, why it is the foundation.

 

Last edited: Aug 17, 2023

Sure it sounds good for listening no doubt, but I wouldn't mix in an untreated room. Untreated rooms have a volume threshold where the bass or other frequencies will start to build up once you pass that threshold. Hence the need for acoustic panels or treatment etc.

 

I just wanted to point out that the monitoring system is more important then the acoustic environment.

To my experience nowadays a well acoustically treated room is not required at all anymore to get a high quality mix. One just has to understand and apply what's stated in the first post. And all of this does not have to be expensive at all.

 

Last edited: Aug 17, 2023

I think we all agree is a very difficult topic especially in practice so it makes for an interesting debate.

 

There's actually some (rare) examples of top engineers like Glenn Schick who only use headphones to do their black magic. It's still uncommon, but there's no viable reason to believe that you can't do great work on it. I guess that there are practical reasons like the ones @Martel explained, so a lot of engineers have problems with in-ears. It's a shame, because they provide much more consistent frequency responses compared to headphones, which are more widely used in the industry.

That's true, but at least, it's a much simpler room. At those dimensions, the only problems will be pressure issues around 8 to 12kHz. I explained how you can try to fix it on my post.

I think you guys are both right . As long as you have good speakers, reflections above the Schroeder frequency will be ignored by our ears. And below the Schroeder frequency, the room takes over and it will be pressure issues. Those are easily fixable with EQ.
But the liveliness of the room doesn't disappear, so acoustic treatment is still required IF you want your speakers/room to sound tighter, more dry. It can really drag in the low end.

 

While its interesting, a lot of this theory doesn't translate to practice. The measurements are from a perfect environment (i.e the headphone pressed against a piece of metal with a hole in it to emulate the ear).

It doesn't account for air leaks in the system from glasses, hair, the shape of your head and jaw, the pinna, moving your mouth (which changes the shape of you ear canal) not to mention small movements of your head while wearing them. Even the smallest air leak in headphones or an IEM system (as small as .5mm) will result in the creation of an acoustical filter, removing lower frequencies while also creating unwanted frequency humps in the mid and high range.

Indeed any small air leaks will exhibit a significant frequency shift and response in the lower, mid and high frequencies. The air gap created by glasses, for example, can result in an acoustical high pass filter that will misrepresent frequencies below 400Hz to a magnitude of +10dB or above.

In fact even when measuring headphones on a metal plate, any slight movement in position or change in clamping pressure will result in the creation of acoustical filters resulting in loss of low end and significant sharp spikes in the higher frequencies. This is why multiple measurements are typically taken and only the best results are published.

So while the theory and measurements are useful from a scientific perspective, they have very little meaning in a real world application, unless we all have perfectly flat heads, position the headphones completely accurately, don't move our heads whatsoever and don't move our mouth.

 

I had the complete opposite experience until I tried some custom moulded ear tips. The insertion point is a bitch and change the frequency rendering in your ear canal by a huge margin everytime you insert it. Then there was an issue of limitation of the driver size vs rendering and that also didn't help but I think it is going in the right direction. I believe AI and in-head chip implementation could make it all work a lot better especially for any ''in-house frequency curve'' as we all decide about our favourite speakers rendering in a room and not without the room but that's in a few more years. On-ear headphone also provide a tactile experience which I didn't thought was of any importance until I experienced some party friendly IEM at 2K$ a pop. But remembering the experience now makes me think that, as I said earlier, some people will actually love that experience.

In the early 2000's, I met a guy in Holland and he was swearing by his KRK monitoring system. His sub was cranked all the way up and it felt like we were in a club in his studio. This guy also made history in moombahton later on.

So it's really just a preference thing. I like my bass to have less saturation and more definition. That was not the case at all on those 2 High End IEM's.It is also not the case in headphones but the difference in margin with my home studio monitoring system is a lot less.

I often heard people saying how much the bass was lacking on ATC's monitors. I'm one of those that think it is perfectly represented and one of the most transparent one around so when I switch to those type of in-ear solution, I'm not having a good time at all.