Be it resolved that there is no such thing as achieving sonic perfection - technically at least based on objective ideals, but "perfect" can also be a subjective state of mind.
In domestic homes, we're simply not going to be able to achieve near-perfect rooms. Likewise, no sound system will be absolutely transparent although the digital electronics side will get much closer to "perfect" than analogue playback and loudspeakers. Furthermore, beyond what we can say about our domestic playback systems, despite all the care and detail-orientedness put into a recording studio, they're not identical even if they followed published standards (like ITU-R BS.1116, or EBU Tech 3276), and besides, is there ever such a thing as a perfectly engineered recording? Or a perfect performance from our favorite artists for that matter?
While perfection might be out of reach, let's not turn away from the pursuit of what is reasonably accurate which is something we can all do to some extent whether it's choosing better-engineered gear, working on small room acoustics, or using DSP to improve frequency response and/or time-domain performance at our listening position. Regarding the specifics of DSP room correction, let me refer you to previous articles covering various methods here, here, here, here, and here.
The intent of this post is to focus on the topic of Room Target Curves which we define and apply when doing EQ and DSP correction. Let's explore a little history, what they are, what they look like and considerations for what to try at home.
[I know there are people who want to instill fear in audiophiles against the use of DSP but know not what they speak of nor provide evidence - like this fellow, who prefers to have his customers waste time and money on Synergistic Research stuff instead of doing something useful. He sells the stuff of course so is financially incentivized.
Modern reputable DSPs including EQ are done at very high quality/precision. Assuming you already have high quality gear and a good room, when set properly within reason, they can improve clarity and restore neutrality, enhancing subjective enjoyment without distortion.
I suspect most audiophiles can benefit from good DSP correction. You just needs to know what you're doing; apparently not that guy! š¤£]
1. What Is a Target Curve?
A target curve (or "house curve") is a specified frequency response that defines what the combined output of a loudspeaker system should look/sound like when measured at the listening position in a room. It describes the end result of the entire chain: the electronics, the loudspeaker's inherent response, its directivity, the room's acoustic properties, and any equalization applied.
The key insight is simple but profound: an "ideal" loudspeaker that measures perfectly flat in an anechoic chamber will not sound or measure flat in a real room. As discussed previously, in a typical small room, boundaries reinforce bass, reflecting surfaces add midrange energy, air absorption and furnishings attenuate high frequencies. The net result is a downward slope from low to high frequencies. Research consistently shows most listeners prefer this natural slope over truly flat in-room response which would sound overly "thin" or harsh.
.png) |
| Fig. 1: Simulated anechoically flat speaker down to 40Hz (blue) produces a tilted in-room response (red) due to room gain, along with modal behavior, SBIR, and HF absorption. The exact amount of room effect of course will vary depending on the space, speaker positioning, furniture, seating position, etc. |
Therefore, from the perspective of human experience inside rooms enjoying music, whether it's a small room at home, a cinema, or the local concert hall, what sounds "natural" in our listening rooms is not a perfectly flat frequency response.
So, while we generally do want our DACs and amplifiers to be capable of flat frequency response and not impart their own "color" to the sound, and want loudspeaker anechoic response to be reasonably flat, we generally do not want our room frequency response at the listening position to be ruler-flat.
2. Why Do We Use Target Curves?
Well, it allows us to control what kind of tonal quality we want our sound systems to have in the room. The shape of the target curve acts as room correction guidance for DSP systems such as Dirac Live, Audyssey, Audiolense, Acourate, etc. Since a flat response would be unnatural as discussed above, we therefore would want the shape of the target curve to correlate with a euphonic, "natural" experience.
[Even without using these DSP systems listed above, you can measure your existing frequency response in REW and compare what you're getting with the research over the years in psychoacoustics we'll be talking about.
REW is very capable. With the measured frequency response, you can still define the "house curve" you want to aim for and have the program calculate the appropriate correction EQ parameters to apply. These can then be programmed into hardware options from manufacturers like miniDSP, Behringer (discussed awhile back, also here) or plug into playback software like foobar, Roon, and AudirvÄna.]
To summarize, here are 3 main reasons to use a target curve:
• Standardization: Ensures recordings translate predictably (not necessarily perfectly) between calibrated rooms. This is important especially in studio control rooms to ensure that they're at least listening to the music "accurately". Sonarworks SoundID is a good example of a calibration and virtual monitoring system targeted to sound creators for speaker and headphone monitoring.
• Preference optimization: Just because we have typical target curves that most people seem to like doesn't mean you have to follow them! Target curves allow us to have some power over the tonality we subjectively prefer. As discussed previously, you can be your own "tonmeister".
• Compensating for deficiencies: Beyond the overall shape of the frequency response, applying DSP (multi-thousand-tap FIR filters much more precise than a handful of parametric EQs) provides us with the opportunity to smooth out the frequency response, taming bass room modes, while adjusting speaker tonality. In a typical small room that's not extensively treated, 15-20 dB range between modal peaks-to-nulls is to be expected, and usually we can tame that down to ~10dB total range, with ±3dB through most of the frequency range using DSP/EQ. This is audibly significant.
Implied in the points above, there is no single universally correct target curve. What is "best" could depend on room properties, listening distance, and to a very large extent, personal preferences. The various target curves we'll be talking about below provide for us scientifically, typically empirically, informed starting points.
3. A History of (Potential) Target Curves
3.1 Bruel & Kjaer Measurements (1974)
Among the earliest systematic measurements, B&K documented what they called the "optimum curve for HiFi equipment measured in the actual listening room" using their pink noise QR 2011 test record and 1/3-octave analysis. The curve ends up being flat from 20 Hz to approximately 160 Hz, then begins a gradual downward slope reaching roughly -6 dB by 20 kHz. This gentle, continuous roll-off starting in the upper bass/lower midrange region rather than a sharp knee at any single frequency would later be echoed in the shape of the Toole/Harman research findings and Dolby Atmos Music curve.
 |
| Fig. 2: The B&K (1974) "optimum curve for HiFi equipment" traced from the original paper. |
3.2 The SMPTE X-Curve for Cinema (1975)
The first standardized "electroacoustic response guideline" target is SMPTE ST202 (see review article - note some earlier work described from back in 1948) also referred to as the "Academy Curve", first proposed in 1975. This curve was derived also from analysis of pink noise averaged over 1/3-octave. Instead of a hi-fi room, this curve was for the cinema venue with typical 1970's era projection and sound systems; typically 35mm reels with a mono soundtrack along with all kinds of compromises and difficulties in measuring the frequency response as documented in that review.
It served as the EQ standard for cinema "B-chain" playback including the amps, speakers, and room effects, measured at 2/3 of the distance from screen to back wall, at the lateral midpoint of the room. A microphone array (at least 4) would be used, clustered around that target position.
The X-Curve specified flat response to 2 kHz, then -3 dB/octave roll-off above, compensating for low and mid-frequency accumulated energy in large cinema spaces. This -3dB/octave roll-off would be too aggressive for small rooms where it would sound too "dull".
A compensation for small rooms (<150m³ or 5300ft³) that I've seen, which can be applied in home theaters, is to reduce the roll-off to -1.5dB/octave as plotted below:
 |
| Fig. 3: The SMPTE X-Curve: flat to 2 kHz, -3 dB/octave for large cinemas, and -1.5dB/octave for small rooms. |
3.3 Floyd Toole and the NRC/Harman Research (1986–2015)
Through decades of research, Toole established that listeners in typical domestic listening rooms preferred speakers with flat, smooth anechoic response and directivity. A key contribution was demonstrating how room reflectivity determines the steady-state in-room response when the loudspeaker itself has flat direct-sound output.
The figure below, traced from Toole's paper, shows estimated steady-state curves for home and professional-monitor loudspeakers in three acoustic environments. The highly reflective small room curve shows a steep downward slope of roughly 8 dB from bass to treble, driven by a bass-heavy reverberant sound field that's stronger than the direct sound. Bass frequencies are radiated omnidirectionally where it strikes walls and reflects as opposed to the progressively more directional high frequencies as drivers start to beam.
A typically reflective small room curve (like an untreated living room with furniture and carpet) with a bit more absorption shows a more moderate slope of approximately 5-6 dB; and an acoustically dead space shows essentially flat response since without reflections, only the flat direct sound is measured.
Real rooms and real speakers will fall within the shaded region between these extremes in this graph:
 |
| Fig. 4: Estimated steady-state sound in small rooms, traced from Toole's published figure. All loudspeakers have flat direct-sound response. The slope of the steady-state curve is determined primarily by room reflectivity and the speaker's directivity characteristics. |
3.4 Olive, Welti & McMullin: The "Harman In-Room Loudspeaker Curve" (2013)
In this study from 2013 (AES Convention Paper 8994) which I've seen referred to a bit, the authors ran an experiment with 11 listeners (8 trained with Harman How To Listen software achieving level 8+, 3 untrained - obviously small sample size) allowing them to adjust the bass and treble dial of a calibrated system to their subjective preference. This test was using Revel Performa F208 speakers, 24' x 21' x 9' treated reference listening room, RT60 at 0.4s constant above 125Hz, with selections from Jennifer Warnes' "Bird On A Wire", Steely Dan's "Cousin Dupree", and Estelle & Ye's "American Boy". Already, I think we can imagine the limits of this experiment and how it might or might not reflect the context of our own sound rooms, equipment, and choice of music.
On average, these 11 listeners liked a room curve with +6.6 dB bass (below 105 Hz), and -2.4 dB treble (above 2.5 kHz) - a spread of 9dB. As you might imagine, there was a wide range found between individual preferences: 17dB range for bass and 11dB for treble!
What's most interesting I think is that the 8 trained listeners preferred a more gradual slope whereas the untrained (only 3 listeners!) preferred a steeper bass-heavy shelf.
Naturally, dear audiophiles, the question then becomes, what kind of listener are you? š¤
%20preferred%20in-room%20loudspeaker%20response.png) |
| Fig. 5: The Olive/Harman (2013) preferred in-room loudspeaker response. The bass shelf transitions smoothly from the low-frequency plateau through the 80–250 Hz region into the midrange as defined by their experimental procedure where they used a 2nd order filter at 105Hz. |
3.5 The Dolby Atmos Music Target Curve (2020s)
A recent addition to the target curve landscape is the Dolby Atmos Music Target Curve, documented in Dolby's Best Practices for Dolby Atmos Music Studios. This curve evolved from the Cinema X-Curve heritage but was specifically refined for music production in smaller, controlled studio environments. (Dolby still recommends using the X-Curve for large rooms >125m³ - Section 2.7.2.)
The curve's key specifications are:
• Bass: +1 dB shelf below approximately 160 Hz
• Midrange: Flat reference region from 160 Hz to 1.6 kHz
• Upper treble roll-off: -1.5 dB/octave from 1.6 kHz to 10 kHz
• High treble roll-off: -3 dB/octave above 10 kHz (steepening)
• Tolerance: Speakers should extend 40 Hz to 18 kHz within ±3 dB
The Dolby specification does not explicitly define the response below 40 Hz, but as I previously discussed, a smooth sub-bass extension is desirable for systems with capable subwoofers. The graph below uses the sub-bass extension, which smoothly rolls the +1 dB bass shelf, reaching approximately -3 dB at 20 Hz. I believe this provides a realistic and pleasant target for audiophile systems with full-range bass capability that rhymes with the X-Curve sub-bass roll-off.
 |
| Fig. 6: The Dolby Atmos Music Target Curve with sub-bass extension to 20 Hz. SMPTE X-Curve is overlaid for historical comparison. |
The Dolby Atmos Music Curve I think matters for audiophile listeners for several reasons these days as we're seeing more Atmos 'spatial' streaming content from sites like Apple Music, Tidal, and Amazon Music:
1. Standardization benefits translation: if Atmos-certified studios mix to this target, home systems calibrated similarly should reproduce the intended tonal balance. There would still be variation of course since there is ±3 dB tolerance, but at least higher end studios and home systems can seek to tighten the system's accuracy.
If you believe there is such a thing as sound quality "as the artist intended", then this is probably as close as you can get.
2. The curve addresses intra-system consistency: in a multichannel Atmos layout, all speakers should match the same timbral target so that audio objects moving through space do not change tonality as they pan between speakers around and also vertically.
Some studio professionals have expressed a preference for more bass emphasis (around +3 dB instead of +1 dB below 160 Hz). This echoes Harman's finding that many listeners prefer more bass. For home listeners, I think the Dolby Atmos Music Curve can serve as a very reasonable starting point that may benefit from a modest additional bass shelf of +2–4 dB seasoned to taste. As I noted above, there's nothing locking us hobbyists into a specific curve although I think it would be important for professional studios to calibrate to the Dolby standards tightly.
Software tools like Sonarworks SoundID Reference already include the Dolby Atmos Music Target as a built-in calibration option.
[As far as I can tell, point 2 above about intra-speaker consistency and using the same target across all speakers, has been a recommendation across the documents I have seen for doing DSP/EQ. This is why I don't quite understand all this stuff from Magic Beans Audio and their claims about a unique "True Target" - see videos such as this - with target for each speaker in a system based on measurements (also see the per-channel filters video here).
Hmmm... Doesn't doing that and not using a specified target, same for each speaker, potentially result in some wild tonal variability between the speakers? For example if we have small surround channels vs. large front mains, or if we mix and match speakers of different brands in a surround system?!
This discussion video with Erin remaining almost silent through it all seems awkward and I find even more confusing. š¤ For example, when it comes to nomenclature, my understanding is that the "correction curve" (which can vary by magnitude depending on deviance from the target), is not the "target curve". The correction curve is referring to the inverse adjustments to be applied to the channel in the room in order to achieve the target response (in practice, represented by "8. Room Macro 3: Inversion" step in this post).
From my perspective, the "Target Curve" is something we set, there is an act of will here that determines the shape of this, not something we just measure to find; thus there is no such thing as a "true", single, target.
To be clear, I'm not saying that we don't respect the natural frequency response of the speaker. We should try not to design target curves that push the hardware in ways that it cannot handle. For example, here's what a target curve design in Dirac Live looks like for me overlaid on the natural measured frequency response of my speakers:
I'm also not saying that in a dynamic system playing music that the sound itself wouldn't be different than the steady-state sweep.However, I don't get the sense that what's being done by this Magic Beans system clearly makes the correction "better" sounding, and probably would lead to more tonal irregularities.
Let me know if you've tried Magic Beans and found it worthwhile or have a better explanation for why there's such a thing as a "true" target curve, or even how it's supposed to work!
Permit me to walk back my opinion of Magic Beans now that I understand how it works! As you can see in the Comments, there was a long discussion with Joe N Tell / Joe Mariano. He has been graceful in offering an opportunity for me to give the app a go. Well folks, the Magic Beans True Target article is up. It's actually pretty cool. ☺️]
4. Comparing Target Curves
From the curves above, published across five decades, they all share the same fundamental principle: the preferred in-room response slopes downward from bass to treble. They differ in steepness, breakpoint frequencies, and intended application.
The Toole "typically reflective room" curve provides a physical reference for the slope that arises naturally from speaker-room interactions in a room that's not specially treated. Another point of interest is the fact that the black Olive/Harman 2013 curve based on the average preference of all 11 subjects tested has quite a notable bass bump which I think is telling of what "average" listeners in-the-wild prefer - an accentuated bass (like Beats headphones back in the day š«£)! This could be quite different from the preferences of "trained listeners" and I suspect experienced audiophiles who like a more neutral sound.
The Dolby Atmos Music curve as a recent evolution inherits a bit of the X-Curve's cinematic lineage while incorporating small-room characteristics for music production plus a small bass bump which is typically empirically preferred by listeners.
5. Applying Target Curves in Home Hi-Fi; some practical ideas...
Assuming you have the pre-requisite system to apply room correction (eg. computer playback with convolution engine like AudirvÄna or JRiver; plug-in like Hang Loose Convolver; server-side DSP like Roon; built-in Dirac capabilities in your receiver, etc.), here are some general points to keep in mind when designing and applying your target curve.
5.1 Start with good speakers and hopefully a good room
5.2 Correct more aggressively in the Modal and Transition Zones
5.3 Be more gentle above the Transition Zone
6. Summary
From B&K's empirical results in the 1970s through Dolby's specification, the science of loudspeaker target curves has converged on a consistent finding: listeners prefer an in-room response that slopes gently downward from bass to treble. The cinema X-Curve formalized this for large venues; Toole and Olive quantified the preferred slope in domestic settings; and the Dolby Atmos Music curve provides a modern, standardized reference for immersive music production that bridges the gap between studio and home.
• Start -1dB @ 20Hz then up to +2dB @ 35Hz
• Flat accentuated bass: 35Hz to 80Hz
• Gentle downward slope from 80Hz: ~2 dB decline from 80Hz to 250Hz. (Reduce energy by the 200-500Hz midrange "Mud Zone".)
• Flat midrange region: ~250 Hz to 1.5 kHz
• Gentle HF roll-off: ~5 dB from 1.5 kHz to 20 kHz
• Total bass-to-treble difference: ~7 dB
 |
| General Audiophile Target Curve: I've been using this one in place of the Harman Trained Listener curve in my system for the last few months. Quite enjoyable for a wide range of music. š Of course if you don't have much sub-bass, obviously don't force it, just let it roll-off below 60Hz. |
[Implementation Note: I'm not a "bass-head". While I can certainly enjoy the deep vibrations especially in movies, I find that a lot of modern music production accentuates the sub-bass too much which is why I like to attenuate the frequency response down by 20Hz, otherwise it's just unpleasantly "rumbly" for my taste.
As for infrasound <20Hz, I usually will roll off steeply down to at least -12dB @ 15Hz even if subs capable of more. As noted in comments, I find visceral/vestibular infrasonic pressure in the room uncomfortable so personally do not feel a need to amplify that stuff even if present in the soundtracks or music.
A target curve like the average preference curve from Olive, 2013 with +6.6dB at 20Hz (0dB at 1kHz) would be unpleasantly bass-heavy for my taste! Yuck.]
Another target curve you might want to try reflects my attitude that these curves are also determined by subjective preference and there's no "true" target - just a combination of knowing what is "natural" sounding and what one prefers.
I will often use this one for playback of acoustic albums and vocal tracks. This is what I'll call the Vocal Bloom Target Curve:
 |
| Again, we don't overload the 200-500Hz zone so as to reduce "mud". And roll-off above 3kHz (especially 3-6kHz) to reduce vocal harshness. |
Specific data points for implementation:
# Frequency Level Description
1 20 Hz 0.0 dB Sub-bass roll-off
2 35 Hz +2.0 dB Sub-bass rise
3 60 Hz +3.0 dB Bass shelf
4 100 Hz +2.5 dB Mid-bass warmth
5 200 Hz +0.8 dB Upper bass
6 300 Hz +0.1 dB Low midrange transition
7 500 Hz +0.5 dB Lower midrange body
8 700 Hz +1.5 dB Male vocal fundamental
9 1000 Hz +2.0 dB Vocal bloom peak
10 2000 Hz +1.25dB Vocal presence
11 3000 Hz 0.0 dB Upper presence / articulation
12 4000 Hz -1.0 dB Presence taper
13 12000 Hz -3.5 dB Upper treble roll-off
14 20000 Hz -5.0 dB HF limit
In my Integra receiver these days, I have three Dirac Live Bass Control target settings to toggle between:
I can switch between these easily depending on the music and mood. And yes, switching between them results in clearly noticeable changes in tonality as you can imagine if we overlay the three curves:
Addendum:
Sean Olive's post: "The Subjective and Objective Evaluation of Room Correction Products" from 2009 showing preference for room correction being applied compared to no EQ for a majority of the products 3/5 (one of them rated about the same as no-EQ so 4/5 at least better or equivalent to no correction), double blind, 8 trained listeners.
 |
| 10 octaves between 20Hz to 20kHz; so a simple 0dB @ 20Hz to -10dB @ 20kHz "curve" will approximate this. |
--------------------
Some parting music: let's start with retro badass Joan Jett & The Blackhearts - "I Love Rock 'N Roll" (1981):
Recently the Joan Jett Essentials (2026, DR12 multichannel) compilation has been released in surround/spatial audio on Apple Music (I think the Atmos mix currently is an exclusive to Apple, not on Tidal or others). Pretty good remixes - turn it up!
As for new music, Ludwig Gƶransson's soundtrack for The Odyssey (2026, DR6 stereo mix, DR12 multichannel/Atmos) is pretty interesting. Here's "Zeus's Law":
Very strong deep bass if you're using sub(s). I heard that Christopher Nolan didn't want a symphonic suite like they did for the movie Troy (2004, DR15) because there were no orchestras back in those day. But did they have such strong dynamic range compression (for stereo) back in ~1200BC during the Trojan War? š¤
[That difference between the DR6 2-channel stereo and DR12 multichannel/Atmos mix of The Odyssey soundtrack as you might imagine is profound and I think an interesting compare/contrast opportunity! The loud, compressed version sound harsh and menacing compared to the multichannel version which maintains nuance, coming across to me as more emotionally subtle.]
Track 2, "Ithaca" is an interesting mix of menacing sweetness:
I hope you're having a great season, dear audiophiles! And having fun with your music and audio systems.
Hey Arch, nice writeup as usual! Unless I missed it in your article, don't forget about Sean Olive's work on https://seanolive.blogspot.com/2009/11/subjective-and-objective-evaluation-of.html and the slides: https://drive.google.com/file/d/0B97zTRsdcJTfY2U4ODhiZmUtNDEyNC00ZDcyLWEzZTAtMGJiODQ1ZTUxMGQ4/view?hl=en&resourcekey=0-uSgBBin0zcCtmpgvEsdskQ Especially slide 24 with that flat, but tilted response - a personal favorite of mine. Keep up the good writings!
ReplyDeleteHey there Mitch!
DeleteNice hearing from you. And thanks for the contributions :-). I'll add an addendum to make sure others have a look at that Olive blog post from 2009 showing the subjective preference for room correction (mostly) and also that slide with the simple tilted response!
Have a great summer... Hope we'll get a chance to meet up this year. Cheers!
Arch, looking forward to meeting up! We are due!
DeleteAnother aspect about room listening curves is listening sound pressure level (SPL). Bob Katz has a good description of critical listening levels, while targeted for mastering, it applies to mixing and critical listening: https://www.digido.com/portfolio-item/level-practices-part-2/
I calibrate my system between 77 to 83 dB SPL as per Katz using the preferred DRC straight line tilt as per Olive's findings. Sounds great at reference level, but for background listening or lower levels, I add a loudness control to compensate for the fact that our ears frequency response changes with volume. JRiver has a good handle on it: https://wiki.jriver.com/index.php/Volume#Loudness plus you can get loudness control plugins.
My point is, the room target curve is really only valid at reference level. As the SPL decrease our ears sensitivity to low frequencies drops quickly - ergo equal loudness contours.
While there is considerable discussion on room target curves, very little about loudness controls, especially when not listening at reference level.
Makes we wonder about some of the room targets with bass boost. Are folks listening to that at reference level or is that loudness compensation for a lower SPL?
Have an awesome summer, Arch!
Indeed Mitch would be great to touch base again soon!
DeleteThanks for bringing up the listening level and Bob Katz's article as well. It's amazing to think that that article came out ~2000 as we entered the era of SACD and DVD-A; how great it would have been if music producers seriously implemented the K-System instead of our ongoing situation of overly "hot", distorted recordings - including apparently the way Christopher Nolan likes it. š
At least it's good to see that Dolby standardized on -18dB LKFS with maximum -1dBTP for Atmos mixes. Hopefully they remain steadfast on this loudness target and maintain higher level of dynamics! Standard CDs and 2-channel material including dynamically compressed "hi-res" (in name only) seems to be a lost cause in the 2020's. š„²
Yeah, I typically aim for ~80dB SPL for calibrations these days (good to be aware of that 83dB SPL magic number). And usually listen at average 75-80dB SPL. As my main playback system, it would be awesome if Roon featured an ISO 226:2003 equal-loudness contour compensation volume control. Sadly, no such luck currently. The RME ADI-2 products are cool and have this. On the AVR side, Audyssey MultEQ Dynamic EQ is able to do this as well.
Part 1:
ReplyDeleteFirst of all, thank you for mentioning my app, True Target by Magic Beans Audio. We just refer to it as MB for short. I think this discussion deserves more than going off vibes from Erin being quiet and making it "awkward" according to you. His demeanor does not prove/disprove anything. You could've reached out to me to try the app yourself. But you didn't, so here we are.
I prefer the term house curve to target curve because it more accurately describes the fact that it's the room itself affecting the sound of the speaker. The issue is that people started using the house curve as something prescriptive, a "target", to fix the sound of their speakers based on how ideal speakers measure in an ideal room, rather than descriptive; to show how the room has affected the sound of their speakers. Simply copying that response won't automatically make bad speakers good. Blindly using any "target curve" response will not lead to the results most people think because a single mic doesn't properly represent how we hear with two ears and a brain as Dr. Floyd Toole has stated numerous times.
In this article you've shown numerous target curves that vary greatly and yet you still felt the need to hedge that the final target response might just be subjective. In car audio, there are even more target curves people use and yet most professional calibrators still feel the need to tune by ear. Why?
Because no single in-room, listening position-based response curve is correct for every situation.
Hi Joe,
DeleteNice meeting you virtually! It's OK, we can talk here openly. This blog is about open discussion and exploring things relevant to the hobby, so it's good to chat about this.
I tend to use "target curve" and "house curve" interchangeably as I think that's the convention in most discussions I've seen on forums.
When I think of the term "house curve", I usually will connect that with the idea of a "house sound", so for example, Dirac offers a target curve download for the NAD "house sound" which had a pretty strong low-bass "thump" down at 30Hz and a typical treble roll-off.
My point with these target curves is actually this this is not "blind". Because we already kind of know what "natural" tends to be like in most small rooms as per the research since the '70s. As we can already see in the paper by Olive (2013), each of the 11 listeners tested identified different "curves" they preferred for the music they're listening to. Is there any evidence there that there is only 1 "true" target?
So when you comment that:
"In this article you've shown numerous target curves that vary greatly and yet you still felt the need to hedge that the final target response might just be subjective. In car audio, there are even more target curves people use and yet most professional calibrators still feel the need to tune by ear. Why?
Because no single in-room, listening position-based response curve is correct for every situation."
I don't see why this is a "hedge". It's just a reflection of my belief that:
1. The tonality of sound that we each prefer is a subjective one and might even change with mood, music we're listening to, etc.
2. There is no standard when it comes to how music is produced (as discussed with Mitch above), so it's good to have options to control for times when some music can use a bit of "warmth" (like the sterile '80s pop), others a bit of bass accentuation, etc.
Different target curves can provide for these customizations we can each enjoy and in my case, I flip between as I see fit.
I can certainly agree with the comment: "Because no single in-room, listening position-based response curve is correct for every situation.
But not because I think it's mainly a room effect. I think there are greater issues and subjective preferences at play.
Part 2:
ReplyDeleteI see things very differently from you when you say that using the same target curve will result in a more standardized listening experience. I know it's not just you, many people say that. Dolby makes the same claim. But I don't think they're correct either and I've told them. Talk to some mixing studio calibrators and see how many of them tune to the Dolby music curve to get the room certified, only to revert it to their own tune later. Here's why it doesn't work that way.
A mic is usually placed at the listening position. It captures the steady state response (combination of direct and reflected sound). What the mic captures is different from what our binaural hearing can distinguish: the difference between direct and late reflected sound. Our brain already adapts to the various acoustics of the room and can extract the direct sound. So does that mean we should do nothing?
No. We should! What MB True Target does is speaker response correction, not room correction. Only treatment can truly fix the room. If you really want the sound to be more coherent from speaker to speaker, then one place to start is to make the direct sound more linear. This is the "true target." We already know from Harman research that a neutral on-axis sound is preferable. Unfortunately, not all speakers are neutral, but DSP can help with that. Many studio monitors do exactly this, so why not with passive speakers?
But what about directivity you say? MB takes directivity into account to not boost in areas where there are directivity mismatches. No DSP can fix directivity errors caused by physical issues such as diffraction, but at least we should not make things worse by boosting in those areas.
Also, MB determines the transition region in the room and corrects for the nearfield response above it, and based on listening position (steady state response) below it. How can MB do this?
Yeah, fair enough, and I guess my tendency is towards what Dolby is claiming when it comes to the use of the target curve to standardize the listening experience.
DeleteAgain, I think there are more powerful factors at play here in terms of subjective preferences and that's I suspect the reason why calibrators might tune to the Dolby Atmos Music Target and then revert to what they're used to or prefer. Again, I'm not sure this proves that it has to do with there being a specific target that will universally result in approval.
For example, assuming the end-user did everything correctly, have you not had any situations where someone used MB to calibrate their room and then decided that they were going to tweak the results or go back to their previous target?
Indeed, it's more complex than steady-state frequency response analysis at any one spatial point. I suspect most of us use multi-point room measurements already. The application of FDW and ability for excess-phase correction are ways to overcome risks of overcompensation (ie. the ability to isolate direct sound particularly in the higher frequencies) as well as correct for multi-way driver temporal offsets. I would argue that these advanced capabilities are more useful than defining the target curve itself.
I think it's cool that MB determines the transition region and builds the target curve based on that awareness.
Part 3:
ReplyDeleteMB takes more measurements. We actually take nearfield measurements to get the direct sound. We take farfield measurements to see the effect of the room. Most other correction software only takes farfield measurements. (I'm using the terms NF and FF loosely here.)
You mentioned using different speakers. We do have an option to use a unified target curve so the bass target is the same for all the speakers, but the target response above the transition may vary based on distance and dispersion characteristics of each speaker. Despite the name, we cannot do magic with DSP. Speakers with different dispersion characteristics and low frequency extension will always inherently have differences in sound, but to make them have a similar direct sound is a step in the right direction.
So back to target curves. Using the same target curve will not result in the same perceived listening experience in every situation. If you take an ideal speaker (flat on-axis, smooth directivity) and listen to it in an ideal room from a typical listening distance, then maybe you will see a resulting response curve that looks similar to the ones Harman found in their tests. But take that same speaker and listen to it in a car and that same target curve with the 8dB bass rise and falling treble response will sound like it doesn't have enough bass and the treble will sound too rolled off for a nearfield listening experience.
The "true target" for that ideal speaker in a car would be how that speaker measures in that car since the speaker itself is already ideal. It's the environment that isn't. In the ideal room we applied no EQ to achieve the resulting target response. If you also apply no EQ to that same speaker placed in a car, you will experience it as tonally similar even though the likely in-room response has a higher bass rise that starts higher in frequency due to the small cabin (related to RT) and a high frequency response that isn't downward sloping due to being physically closer to the speaker and having more nearby reflective surfaces.
Two different response curves, but similar perceived tonal response. If you forced the car to have the same steady state response (target curve), it would actually sound less tonally similar because you are sacrificing the "ideal" direct response to make it match a response that is for a different room.
Hopefully, that helps clarify a bit.
- Joe
Thanks for the ongoing discussion Joe.
DeleteYes, I'm aware that you're doing averaged near and far moving-mic measurements and I see that there is the option of a unified target curve applied to all speakers (I see that in the Techno Dad video but he used per-channel target).
So, despite the imperfections of measurement microphones correlating to human hearing, are you saying that using per-channel targets from your software with very disparate speakers (like those smaller surround and tiny height channels in Techno Dad's video) would result in better tonal coherence in an Atmos system when sonic objects pan around the room? Do you have any evidence that this improvement in "direct sound" is actually what's happening? Curious what kind of metrics did you use in the R&D phase of the software development to show that benefit?
Of course I do not expect "magic"; I'm just wondering what evidence there is that it works.
Hmmm, I guess the car example is interesting but extreme because there's simply no way to take "the same (ideal) speaker" and put it inside. In reality, we're always talking about very different speakers (small, mounted against places like doors and dashboards), orientations (R-L asymmetry at seating positions), and acoustics (something like 3m³ cabin, highly reflective windshield, reverb time, etc.) causing all kinds of alignment and phase issues compared to domestic listening spaces that need to be addressed!
Having said that, let's talk about this one...
"The "true target" for that ideal speaker in a car would be how that speaker measures in that car since the speaker itself is already ideal. It's the environment that isn't. In the ideal room we applied no EQ to achieve the resulting target response. If you also apply no EQ to that same speaker placed in a car, you will experience it as tonally similar even though the likely in-room response has a higher bass rise that starts higher in frequency due to the small cabin (related to RT) and a high frequency response that isn't downward sloping due to being physically closer to the speaker and having more nearby reflective surfaces."
We cannot magically wave away physics in our thought experiment because we are talking about room/car cab acoustics. Why do you believe the speaker would be "tonally similar"? Even if we could take a hi-fi speaker meant to be used in listening rooms into the small space, more than likely, these would not be "ideal" any more because the small space will not allow for proper driver integration. At best we might have some small low-frequency restricted coaxial drivers sitting on our dashboard nearfield which could sound OK but the SBIR likely would be awful with smeared midrange including some deep nulls because of that placement inches away from all these surfaces!
Applying DSP/EQ measured at the head position of the driver's seat might improve things - at least you'd get better L-R balance for imaging and maybe some relatively improved frequency response depending on how the drivers are placed and pointed. It would worsen the sound for others sitting in that car!
Regardless, I think the listener should still have something to say about the frequency response they want to hear while driving down the freeway. Those bass-heavy guys who like to announce their presence should be allowed to play with their EQ "target", right?
I'm still not sure why the concept of a specific "true" target is a meaningful thing.
FYI, I have placed a concentric speaker in a car as well as a bookshelf speaker with a known anechoic response, so what I say is not hypothetical. I also do car audio calibration and in my second car audio SQ competition, I use the MB derived target curve and did not tune by ear. I won first place. I've done this for others. Yes, anecdotal evidence, but this is a blog response, not a white paper.
DeleteIt sounds like you subscribe to the research from Harman. To me, the entire thing can be summarized by saying that in blind listening tests, people prefer a speaker with low frequency extension, a flat listening window response, and smooth directivity. We cannot control directivity but a speaker with good directivity but a poor LW response is typically highly EQ-able. So the "True Target" that fixed the poor listening window response issue by making it flat. I don't know why that's not easier for most people to comprehend. If you want to question how well the nearfield measurement we take corresponds to anechoic measurements, you will see it corresponds closely with the LW response.
Making the speaker more neutral with respect to its directivity is the optimal EQ for the higher frequencies. We smooth the natural bass response. I'm simplifying since my responses are already long.
I think what you think is listener preference is related to the circle of confusion. But we can close that loop when we play known source content such as pink noise, and we use a calibrated mic. We know what we expect the response to be if we're talking about HiFi reproduction. But even with a reference system, the mix might have too little bass and too much treble. A target curve isn't the right tool to tweak the per-track tonality. Toole recommends a simple bass/treble tone control for this.
Going back to the tonal response of different speakers, I've already addressed that in my previous response. Directivity and bass extension differences can't be fixed with EQ. But yes, making the direct response more similar, despite the listening position response possibly being less similar, will make them sound tonally more similar because the direct sound is what we are able to separate out with our binaural hearing. If you make the MLP response identical for two different speakers, you might make the direct sound more disimilar and you will hear that discrepancy. Human hearing is amazing. That's why I find psychoacoustics personally more fascinating than just acoustics. It's about how humans perceive sound.
Thanks again Joe,
DeleteI appreciate your detailed responses. While I do not do car audio, I have had my share of tight spaces and experiencing tonal anomalies in such places.
I don't see what's wrong with subscribing to Harman's work since as you know they do have a very significant presence in the car audio market as well! Having said this, I don't necessarily agree to everything they might say (like some disagreement with Toole and treating lateral reflection points).
Congrats for the kudos in the car competitions...
Let's focus on that last paragraph a bit, specifically this:
"... But yes, making the direct response more similar, despite the listening position response possibly being less similar, will make them sound tonally more similar because the direct sound is what we are able to separate out with our binaural hearing. If you make the MLP response identical for two different speakers, you might make the direct sound more disimilar and you will hear that discrepancy."
I do not disagree with this concept that it's the direct sound (first wavefront) we want to correct. That is why in the article I indicated the importance of being gentle with the frequencies above the transition point. And that is why I brought up the use of FDW as a way to isolate the direct sound for analysis in those mid-to-high frequencies.
Using FDW would more accurately allow the DSP system like Dirac Live, Audiolense, and Acourate to then correct each speaker to the same target which I agree with Toole is basically going to be equivalent to a subjective tone control, at least consistent with each speaker so as not to consciously mess up tonal balance! (Obviously the same Target Curve typically means different Correction Curves being applied to each speaker unless the speakers are absolutely matching and room absolutely symmetrical.)
The more we talk, the more curious this seems to be!
1. Beyond anecdotes and testimonies, by what metric are you using in software development to show that your system is indeed capable of more accurately reproducing the direct sound? If the answer to this is the nearfield response, then how do you know the person measuring it has accurately captured proper multi-driver integration when they did that averaging? For example, I saw in the recent TAS review of your product that Steven Stone needed to measure one of his speakers from "over a meter away". Hard to call something a "true" target if it can potentially be fraught with uncertainties?
2. We know different DSP systems correct differently. Since your software is only able to create a target curve based on nearfield and MLP measurements, how do you know that the target you created is applicable for basic parametric EQ vs. something more sophisticated like Dirac Live? Is there a "best" room-correction system you would recommend this be matched with?
3. What is the definition of the "true" target from a tonal perspective? Is it basically following whatever the speaker's natural tonality is? So if a person were to listen to MB-corrected LS3/5 speakers, how would it sound compared to say MB-corrected KEF Blade speakers in the same position? Would they sound tonally very different? So if we take that a step further and say that guy wanted to use the KEF Blade as fronts, and LS3/5 as rear in a quad position, would MB-correction help maintain surround sound tonal coherence? If not, with per-channel target applied (like what Techno Dad did in his video with those small surrounds and tiny height speakers), doesn't that mess up the multichannel sound?
1. We recommend a set distance for most speakers, but some taller speakers, we recommend a starting distance for a NF measurement that is 1X the distance between the bottom of the lowest driver and the top of the highest driver. That is usually enough to get integration of all the drivers. Also, our MMM is different in that we move in a straight line back from that starting distance about 1.5ft. This is the instruction video embedded into the app: https://youtu.be/uPjQRlXpyqs You can also see how the measurement compares to one taken on a Klippel NFS. Yes, there will always be user error, or just edge cases. That doesn't take away from the validity of the method we use.
Delete2. Some DSP has higher resolution, some lower. Our app works with many systems and we can only work within the constraints of the hardware. We have 10-band GEQ and convolution filter exports. The intended result is the same, but the quality and resolution of the corrections will vary. We like Dirac, but even then, we have a 2-pass method designed to undo some of the corrections Dirac does. We would prefer something like the MiniDSP, where it doesn't take any measurements and only takes correction filters. Really, car audio DSP is way more advanced. Helix for example.
3. No, the true target isn't the speaker's natural tonality because some speakers have terrible responses. I would simply say it's as flat as possible with respect to that speaker's directivity. I have a version of an LS3/5a and although the directivity is decent for such an old design, there's definitely a directivity mismatch at the crossover point. That was the entire purpose of the BBC dip to minimize that mismatch. MB is smart enough not to boost that area because it recognizes the DI mismatch. It will never sound exactly like a Blade 2 Meta, because it's DI is not the same. But, the direct sound, therefore the perceived sound will be more similar with MB than without.
Okay. Lots to talk about and debate which might or might not be fruitful. For example, there are some clear limits to the FIR filters you can create based on the moving-mic technique you're using.
DeleteFor those curious about your software and want to mull over what the meaning of the "true" target is, do you have some technical discussions on this topic?
Yes, there are limits because this is not meant to be a replacement for something like Dirac. To add to your statement, our MMM doesn't measure time/phase information, so it operates solely on the magnitude response. MB is supplemental and works with other software solutions. The primary purpose is to determine the proper target curve to use for each speaker. Remember that target curves is what most software expect to see. I would much rather import correction curves like we do on a MiniDSP because it's easier to understand that we are doing speaker response correction.
DeleteThe hardest thing is to make complicated things simple. That was one of our goals with the software. We are more than willing to discuss the technical merits if someone asks. Most of the time, these discussions happen within our Discord group for owners or people interested in purchasing the software. https://discord.com/channels/1158188313182752851/1158196561617105046
We also have an official AVS forum post for people with questions. Also on ASR and in the comment section of my videos and FB groups. Basically, anywhere people are asking questions about MB. Even here on your blog.
Yeah, I've seen the exchanges at AVS Forum thread and the ASR thread.
DeleteCurious, did you ever get to set aside time to chat with Floyd Toole since late 2023 about this (as discussed here)? If so, what did he say after all was discussed?
To be honest, while the YouTube videos are great for instruction, draw viewers, and to help those using the product already, they don't really explain the basis and argue for the benefit(s). Again, it would be great to actually have something like even a "white paper" showing results of different speakers, various rooms, what metrics were used to show benefit. Maybe how this is beneficial in multichannel theater layouts compared to existing solutions. Would also love to see whether different DSP systems (Dirac Live, Audyssey MultEQ, FIR convolution, etc...) perform differently when importing the target curves the app calculates.
I know that could be a lot to ask but to be honest, your videos and comments invite these questions given the complexity of the topic, the tantalizing dive into the technical discussions of improved direct sound, detecting speaker directivity issues, comparisons with Klippel data, piggybacking on existing technologies (therefore suggesting your results better than straight MultEQ, Dirac, etc.).
BTW, I see that the price for MB is $357 for the non-pro version (plus need to buy the UMIK mic, test tone BluRay probably, tri/monopod).
That's not unsubstantial cost and it sure would be nice to get a better grip on the value of this technique beyond the strong potential of spending money and reporting post-purchase rationalizations that it sounds "better" because of the cost and time one has invested into it.
All along, I personally believe the shape of target curves are just as much a subjective choice as I expressed above. We treat our rooms, find good speakers (and ancillary hardware), use good DSP software for analysis of the sound that can improve frequency and time-domain issues whether it's toward euphonic pleasure or maybe a technical target (like a recording studio aiming to be compliant).
Ultimately, I think having fun and tweaking the target curves (even just as a type of tone control) to our heart's desire is the prerogative of the hobbyist audiophile if they care about this and have the capability unless there is good evidence of benefit and the existence even of any single "true" target - which I have not seen thus far.
In that ASR post, the OP misrepresented what MB does and Floyd responded to that misrepresentation. I have spoken with Floyd a couple times over the phone since that post. The discussion centered more about measurement precision and use-case and ended with him telling me to keep rockin'.
DeleteYes, a white paper would be good, but don't overestimate the general public's
interest in the technical details. How many people don't even understand Floyd's research despite him writing a book and the research being over 30 years old? The technical videos I've done typically don't perform extremely well for my channel. I still do them though
https://youtu.be/8bRGKazxgXY Why generic target curves don't work (1 of 2)
https://youtu.be/-1L8ipgj8Ng Why generic target curves don't work (2 of 2)
I don't think the target curve is as subjective as you may think That would go against the Harman research regarding listener preference towards well performing speakers.
I tune car audio SQ systems where they are judged by people I don't know. I trust in the target responses from MB. My clients usually rank top 3 with many 1st place wins. It could be luck. It could be good application of psychoacoustic research.
Just wanted to add the $357 I quoted above was CAD, USD$299 I believe for the non-Pro app.
DeleteGreat. Congrats on the car audio competitions. Nice.
I think from the discussions in the post, it's pretty clear that Harman doesn't have just one "true" target which is my main contention because surely you can't insist that subjectivity doesn't matter or that 100% of your users believe whatever MB calculated was the absolute best for them!
I think having a white paper would really help solidify your claims into one document; for now some of what was said in the various videos are inaccurate or lacking in clarity over time. I don't think I care too much about what the general public thinks. Audiophilia and even the target audience who would bother to buy MB I suspect was never the general public.
Nothing wrong with "keep rockin'", especially in the summer... Cheers.
Standard license is $249 USD.
DeleteI've never claimed 100%. Even Harman's research only shows 85%-90% of listeners preferred the neutral on-axis, smooth off-axis. That means some people liked other speakers more.
But, I see that as a far cry from being "mostly subjective." There's enough statistical significance to show we have an idea what is more or less correct. We all agree that the sound from an amplifier should be linear. Why should the on-axis/listening window/direct sound from the transducer be any different? From there maybe we can debate about whether narrow or wide dispersion is preferred, but even that is likely room dependent.
I will concede that as Floyd says, that too much bass is the forgivable sin in audio. š¤£
Oops thanks for the price correction, Joe.
DeleteI obviously don't mean "mostly subjective" as anything one's heart desires although it's a free world and anyone can listen to whatever curve they prefer! š
I imagine all music lovers and audiophiles know that extreme loss of bass or early high-frequency cut-offs would be "bad", likewise most audiophiles know not to go extreme with the "smiley face EQ"; so I suspect reasonable audiophile would choose normal curve designs. Just as with any normal distribution, +/-1 standard distribution would already capture ~70%. Within a reasonable balance of the full frequency response, exactly which of the "normal" looking curves I discussed above in the article is something I think we choose for ourselves with the freedom to tweak.
Like I said earlier, given the wide potential preferences - recall the 17dB bass range and 11dB treble range being chosen by the Olive (2013) study as their preferred tonality among just 11 listeners, I fail to see the need to get too carried away with any one target for any one listening situation.
I prefer understanding of self, our rooms, the hardware we own, and the DSP technique we have than the idea that any one app is going to somehow spit out a "true" target for the system. It's OK that you think your software can do that; I'm just not so sure as I look more into these videos and discussions.
Perhaps you haven't seen how many people in the forums use a flat curve or advocate for the x-curve on AVS or FB. I have talked with someone who worked at Audyssey who I've had long discussions with because he believed flat at MLP was the most correct.
DeleteAll I'm saying is the fluctuations in bass targets might be due more to the size of the room and what the natural rise is in that room. The treble roll-off target variation is more likely due to distance from the speaker, speaker directivity in the higher frequencies, and whether the listening position is off-axis to the speaker. These are all objective and measurable aspects that I think should be taken into account. There might still be some aspect of subjective preference beyond that, but I think this gets you much closer than choosing from the variety of curves you mentioned as a starting point.
Fair enough Joe.
DeleteI can understand the argument about distance/directivity/off-axis for the treble response. But also, I can see how myself and others with sensitive treble hearing might just find a lot of modern studio productions we like too harsh (especially the very compressed pop!). Therefore, to choose to tone that down for pleasure is just a practical form of "trusting your ears" beyond whatever the room or speakers might want to dictate - can't be a bad thing even for myself as an objective-leaning guy, right?!
Likewise, you said above that from Floyd's perspective, too much bass is a forgivable sin, absolutely, regardless of how we each prefer sound, nothing here is irredeemable and that's as much a subjective choice as the angle/amount of treble tilt IMO!
Like with most things, I think there are "many ways to skin a cat" as they say (although that idiom seems a bit grotesque š³) so long as we enjoy our music and systems.
Cheers...
I've enjoyed the discussion.
DeleteI think what you're describing is the circle of confusion Floyd mentions in his book and the reason he advocates for wideband bass/treble tone control to account for differences in a mix. I don't think the target curve is the right place to change things on a per-track basis, just because it's too slow to change. Even if you're sensitive to treble, what do you do for a mix that is mixed by someone with HF sensitivity as well, so they made the treble low already? Then what do you do when the mix is bright? To me, the best solution is to have a neutral response based on known signals such as pink noise or sweeps. Them you can use tone control to change on a per-track basis.
I am just curious which parts of the method or philosophy behind MB you object to. It may help me better understand what I need to explain better. At the end of the day, I think the results speak for themselves. If you want to reach out to me and try it for yourself, then I think you can better judge the results rather than guessing based on Erin's body language. š I think MB deserves a fair shake and that wasn't it.
Hey Joe, likewise this has been a stimulating discussion! This discussion is I think a good start to a "fair shake" at least to try to understand MB, so thanks for taking the time.
DeleteYeah, I think tone control can certainly be useful to address Toole's "circle of confusion". In terms of adjustments, I'm OK with using the 3 Dirac presets as discussed depending on the situation (General Audiophile for stronger bass, Vocal Bloom for midrange, Dolby Atmos Music Target for a more even studio target sound) - not the continuous option like playing with GEQ settings, but I trust nobody would have settings for every song. š
As you know, these days a lot of audiophiles don't have tone controls so the room-correction DSP with target curve is a way to get it done with the ability for the user to tweak the specifics of the curves and have the software perform more advanced bass correction (like DLBC, or ART as the case may be), and excess phase correction for each of the settings. For those who want even more options, systems like Hang Loose Convolver and Roon's MUSE allow even more slots to be loaded at the click of a mouse button or on an app.
When it comes to the Magic Beans procedure and philosophy, I guess there are a couple of things at the heart of my concern/uncertainty:
1. I don't believe there is any single "true" target curve. This seems antithetical to your claim that generic targets "don't work" - I think they do. To me there exists simply a "family" of curves that listeners could experience as natural, within that Toolean "circle of confusion". "Natural enough" playback is all we can reasonably achieve.
Since you claim that there is a "true" curve which optimizes direct sound, addresses directivity issues, and different measured targets can even be applied per-speaker to multichannel systems, I'd love to see the white paper or technical discussion on the definition of such a curve and what you accomplish. I of course don't expect you to reveal all the proprietary algorithms.
2. I am concerned that your methodology with the moving-mic measurement technique might result in a less ideal filter than modern advanced DSP system. As you already noted, MMM is a way to spatially average frequency response magnitude but not for time-domain accuracy. Also, there's an element of error depending on where and how it's done - already the videos have shown various ways from circular movements to moving backwards, then there's distance. In contrast, advanced DSPs such as Dirac Live and Audiolense, already are spatially averaging using multipoint measurements, maintain temporal accuracy, and utilize FDW to model the direct sound.
Up above, you said that: "We like Dirac, but even then, we have a 2-pass method designed to undo some of the corrections Dirac does."
This is an example of where a good technical discussion would come in useful. What does that mean? Knowing that the MMM system itself adds a fair amount of "circle of confusion", lacks temporal resolution and since Dirac corrects for temporal characteristics and already has a mechanism to window for the direct sound, what are you "undoing"? In what way is Dirac doing something suboptimal and your technique better?
Again, I'm glad that the results have been good for you and your customers. I can certainly reach out. Realize of course that if I give MB a try and review, I'll be honest, and already there are those concerns above which might or might not be resolved or resolvable when it comes to how they could impact my opinion.
I hope it's obvious that I believe in the method and the results. I feel like I've addressed most of those things you brought up numerous times in various ways, so I'm not sure how else to explain it. I will happily explain it over the phone if you want. We seem to be going around in circles.
Delete1) The true target is any correction that makes the nearfield direct response more neutral. Pick any of the other curves you've recommended and the corrections required achieve those targets are likely to make a neutral speaker sound and measure worse. If you took a great measuring speaker in a good room, and ran MB, the correction it would recommend is probably no correction. So, if you think that another target response for the direct sound is more desirable than flat/neutral/accurate (wrt directivity) then we just simply disagree.
2) We purposely take a spatial average with more measurement samples than most software with the intention of not overcorrecting. In 10 seconds we get approximately 60 measurements per speaker for the nearfield and farfield. With Dirac, you get maybe 12 total only from the MLP? Time domain only requires 1 measurement position and Dirac does that well. We don't need to do that with MB. But at the very end, it asks for a target curve. Which do you you use?
Regarding the need for 2-pass. Theoretically, FDW should isolate the direct response but have you verified whether it does that well, or are you just taking their word for it? Do they consider directivity or determine the transition region? I've had their former CEO and senior engineer at my house. I was a beta tester for ART before it was released for Storm. I don't know the answers to those questions, but I have seen that their corrections sometimes overcorrect in the higher frequencies. I think that's why a lot of people prefer to apply a 500Hz curtain.
Hey Joe,
DeleteClearly lots of discussions already and probably at the limit of what can be spoken of at this point. Clearly you believe in the technique and great to hear that others have found benefit.
Alas, words can only go so far when it comes to evidence of benefit. I have found benefit on my end with Dirac Live (as shown here and elsewhere); likewise Audiolense and Acourate. These are results with FDW and time-domain correction. As suggested earlier, I would love to see the results you get using MB, not as a video with brief cuts, but as an actual written analysis - I trust this should not be too difficult, right?
Since I cannot find any write-up on the benefits of a separate target curve app like MB, I do not know whether what you speak of is truly beneficial and in what context.
For example: "The true target is any correction that makes the nearfield direct response more neutral." to me sounds incomplete because the best thing to do then is just to take the nearfield measurement (hopefully accurately capturing the direct sound) and apply a subtle correction EQ that makes it flatter but not too extreme? Is that all it is? If not, would love to hear what else and why you measure from the MLP - even better if you can demonstrate with some graphs of course!
"Theoretically, FDW should isolate the direct response but have you verified whether it does that well, or are you just taking their word for it?" š¤ Okay then Joe, let me reach out and if you're good I'll give your software a fair shake... Deal?
Yes, reach out. If you want to review it, that would be great. And I expect nothing less than a critical review. I would love that!
DeleteRegarding simply taking a nearfield measurement and making that flat, I think that's a great start if you have a speaker with great directivity. Then when you measure at MLP, that's just resulting target curve. Now you can see why every target curve might be different despite the nearfield being the same. For bass, they should be the same hence why we have a unified EQ. The issue with only measuring NF is you're flattening without any regard to directivity. If you boost where there's a mismatch, that could sound bad.
Sounds good Joe,
DeleteWill reach out...
Hey, Arch,
ReplyDeleteThe most telling of the of the images supplied is #4. A room curve is unnecessary in a property treated room (a room with no reflections).
Which brings me back to the LEDE (live end/dead end) room acoustics design. By eliminating all of the reflections in the hemisphere in front of the listener, you not only eliminate the need for a “room curve”, but also eliminate room modes in the low bass.
All of the “fog” caused by reflections surrounding your speakers that diminish their spatial resolution are mitigated. You hear your speakers and not your room. That, if possible, should be your goal.
The room curve is a cheap fix for rooms that reflect lots of off-axis treble energy. Yes, attenuating the treble will give you a perceptually flat response, but this doesn’t solve the problem with your room’s acoustics. It’s a band-aid fix that only addresses the total treble energy in the room.
Hey, if you want the speakers in your living room to sound better, do a room curve. Just accept the fact that if you need a “room curve”, it’s not a listening room.
Hey Jeffrey,
DeleteI certainly don't have any qualms with that position in that if we apply room treatments to the best of our ability, then just let the speakers sound the way they sound. Better speakers with flatter FR, smoother directivity will sound even better!
Having said that, for my main sound room which is a multichannel system with speakers that have different tonality, some EQ/DSP is beneficial for surround coherence. Then there's the element of subjective "tone control" preference (eg. I want to be able to control bass and treble roll-off). And finally there are the benefits of modern DSP systems for time-domain optimization for the listening position (as discussed here).
One extra bit for clarification: a room with no reflections is not necessary to have a great sounding space. If you strategically and effectively treat 66-75% of the hemisphere in front of you, you’ll have a true listening room.
ReplyDeleteHeavy carpeting takes care of some of this. Strong absorption at all frequencies on that first reflection zone of the ceiling is paramount and obviates the need for total coverage.
Because our ears and brain are designed to predominantly discern sound directivity on a horizontal plane, lateral reflections have to be your primary focus when designing a listening room. The entire side walls from the front corner to 90° from front center has to be treated on both sides.
The front wall needs treatment from each corner to a foot or two past the speakers towards the center.
On the walls, this is most effectively and economically achieved by using a combination of sound absorbing blankets and heavy curtains. They act as limp mass absorbers that suck up all the sound energy coming towards them, and bouncing off of the walls behind them. This type of acoustical treatment is much cheaper than foam and much more effective at all frequencies. I’ve found this out through experience, not by reading research papers
Also, it can be done for less than $1k. No tweak to your audio system will make this much of a positive effect on your listening experience.
Once again, this might not be a set up you want in your living room. This is what you need to have a listening room. If you do that, you don’t need a room curve. This is why I make the distinction between the two,
A great automotive analogy for it might be this: you can stick a 1500 hp engine into a ‘69 Chevelle, but that doesn’t mean it’s going to take corners any better.
Yup Jeff,
DeleteAgree that a lot of this won't need to cost an arm and a leg! Cheers.
"On the walls, this is most effectively and economically achieved by using a combination of sound absorbing blankets and heavy curtains. They act as limp mass absorbers that suck up all the sound energy coming towards them, and bouncing off of the walls behind them. This type of acoustical treatment is much cheaper than foam and much more effective at all frequencies."
DeleteThis statement is factually incorrect - it just contradicts the physics. Your typical "audio" blanket https://www.vocalboothtogo.eu/p/acoustic-blanket-vb71g-black-black-200-x-200cm-acoustic-panel-for-audio-film-production
will do much more harm than good with its non-linear absorption. I would use max for very large spaces in order to control overall reverb.
I would also question LEDE [in the sense you describe it] suitability for residential listening - as it requires large space to work properly [and a lot of precisely designed acoustic treatments] and if you have that space available you have lot of better options. If you want to stay in pro world - you will get much better results with RFZ [Reflection Free Zone] concept.
For those interested in theory and bit of technical details https://downloads.bbc.co.uk/rd/pubs/reports/1995-04.pdf gives
Thanks for another excellent post. I use Focus Fidelity to create Convolution Filters for my system and have found that the Harman Target with 3dB less boost in the bass is just right for me.
ReplyDeleteThanks for the note Dave,
DeleteHad a look at Focus Fidelity and it looks like a well-put-together program with flexible target curve design, mixed-phase correction among other features.
Yup, I concur with the -3dB bass boost on that "average" Harman curve!
Hi Arch,
ReplyDeleteThank you for your great summary on the Room Target Curve aspect, and also to Mitch for bringing two other important articles into this equation.
One point that I personally also take into account is the RT60 curve of the room.
When I’m at customers’ places optimizing their loudspeaker/room interaction with DSP, I start with my target curve—similar to the actual Dirac target curves—and “multiply” it with the room’s RT60 curve. The goal is to make the speaker’s sound in that room have some similarities to the natural sound of the room itself.
Of course, it would be better to optimize the room acoustics (and as you know, I have three different AES-recommended listening rooms), but many customers are a bit shy about doing this, as they want to preserve the look of their homes. Just a small addition from my side.
But as always, thank you for your great article.
All the best,
Juergen
A friend of mine and former engineer at Harman, had a discussion about RT60 and room responses and there seems to be a direct correlation between bass rise in certain rooms and the RT60. It kind of makes logical sense that a frequency that reverberates for longer will have more energy in the room over a given time period.
DeleteYes, taking the RT60 curve into acount, does really make the sound of the speaker more "natural" into the different rooms.
DeleteAnd what I also take into acount is the invers characteristic of the IMD (intermodulation distortion) of the specific loudspeaker. Because I noticed, the the same frequency response of an specific loudspeaker with the same radiation behavior (Spinorama Measurements) does sound different, when at some frequency areas the IMDs are higher or lower than the frequency bands close to it. So when the IMDs are higher, I reduce the level in that band a bit, and when they are lower, I increase the level in that band a bit.
Interesting addition to the discussions Juergen!
DeleteCool, I'll have to have a look at these fine-tuning points with tweaking curves covarying with RT60 and IMD. Neat ideas.
BTW, I saw the recent articles with you leaving MBL - "seismic shift" indeed. Hope this also means more time in your studio and sound rooms creating content. š
Have a great summer and congrats on the retirement!
Juergen,
Deletegreat points - re RT60 per frequency bands and its impact on perceived energy. I try to get my room sounding "linear" in terms of absorption. Calculators like this one: https://www.troldtekt.com/web-tools-downloads/acoustics-calculator/ and data from companies like GIK helped me a lot in terms of acoustic design. I prefer more lush sounding room [Musikverein type of acoustics] so I use lot of diffusion and range limited absorbers.
New interesting area are new DRC algos, that aim for active decay control in bass area - Waveforming or Dirac ART - I can see people having senselessly hot bass curves, in order to compensate for significantly shorter decay below 100Hz. Especially ART can sound unnatural and detached.
This is my place, all panels from GIK [and I am sucker for electrostats ;-) ]
https://www.audiosciencereview.com/forum/index.php?threads/show-us-your-listening-room.65837/post-2447631
Arch,
just an idea for next post, while you are at it. I think lot of people meanwhile are educated re Schoreder etc. What is not so common knowledge is room behavior below first modal frequency, where there are no standing waves and no room modes. Room starts to behave like pressure vessel, interesting stuff.
Neat looking space Fidji! Impressive work to have all that in the room and achieve the tight reverb time.
DeleteCurious with your 2-seater loveseat, where is the "sweet spot" for you when listening critically? Do you sit between the 2 seats to be directly centered?
Interesting idea about the frequencies below 1st modal frequency which for my room would be below 30Hz and indeed will be experienced as "pressured", pulsations as we get even lower, rather than auditory. I'm pretty sensitive to some of the infrasonic visceral/vestibular effects and find it quite uncomfortable especially when there's prolonged high infrasound; this is why I almost always make sure to roll off from at most 0dB @ 20Hz to something like -12dB @ 15Hz.
Would love to hear your thoughts on this and anything in particular you're thinking of at those low frequencies below the 1st mode?!
Hey,
DeleteRe: Room - my room is trapezoidal and asymmetrical, with front wall 3 ft longer, so MLP is actually the right seat. I have also couple of presets optimized for seat-to-seat consistency, but not using them too often.
Challenging acoustics in some areas [it has some great advantages, e.g. high ceiling, insulated floor [neighbours] etc] made proper treatments necessary. Currently I am with what I call v4.0 - gradually finding balance between absorption and diffusion, and getting better results by increasing decay times. v2.0 was really too dead with 0.19s, now at 0.32s and +-10% deviation.
Most of acoustically treated rooms, that I had opportunity to listen to were in reality over-damped, especially in 300-1kHz region and of course in treble. Normally furnished rooms and speakers with good directivity do not usually need to fix mid and treble [except ceiling/floor for point sources].
Best advice I was given - control the bass, diffuse everything else, finish the job with some absorption here and there.
Re: Bass and Infra-Bass
DeleteBoth of this is true ;-) - I am NOT a Bass head and I LOVE Infra-Bass. My general attitude towards bass systems is "Only Overkill will do" - just to clarify.
In my experience lowest frequencies convey 2 information [in music] - venue size and ambiance and sub-fundamental transients. If both done correctly they do this big, realistic sound full of energy, percussions, bass guitar having live snap with volume.
Problem is, that it is rarely done without compromises.
Hardware -> too little air displacement, too little and too few drivers with underpowered amps, and stretched with DSP to get to some fancy "10hz"
spec.
Headroom -> adding house curve with like 5-6dB hotter bass, cutting another 6-9dB with filters, bass managing speakers adds another couple of dB of lost headroom.
DRC - most of the algos work introduce phase shift in lowest frequencies, while using min. phase IIR filter. Also they try to iron out FR to comply to TC, ignoring everything else. And I will be honest - I think DLBC is not very good in what it is trying to achieve, especially in time domain.
And last but not least - room - especially in US - drywall resonances and more importantly power leakage, that does not allow for proper room gain.
All this, combined results more often, than not in highly distorted and even more critical - you get lot of timing issues - Group Delay, Peak Energy etc. Especially GD is critical for high fidelity. You can see it in your spectrograph, that Peak Energy is bit all over the place [as very often with DLBC and MSO] and this sounds like what you describe, especially as you have also some destructive cancellations between 40-60Hz, that make lowest bass feel detached, as second harmonics of whatever below 30Hz is suppressed and those sounds have quite often 2nd harmonics even more pronounced than fundamentals.
What could you look at - subwoofer placement - experiment with one sub placed nearfield, behind your head. Gives different sensation.
Look at optimizing DRC filters in order to improve GD - I would experiment with significantly flatter TC in bass region, so you get as little as possible additive filters. Then afterward you add gain on subs to get desired bass volume. Use All-Pass-Filters to get Phase adjusted etc.
Thank you Arch for all these interesting topics and thoughts.
ReplyDeleteAnalysis of the post make me wondering if the concept of high fidelity still make actually sense.
Contrary to specific correction of either loudspeaker flaws or room behaviour or both combined, the use of a generic curve/preset, no matter how good or pleasing to the ear it could be, means that we leave the pursuit of true to recorded sound for sort of a tone control on steroids.
It is the negation of the entire music production chain which aims at producing a sound satisfying the artists themselves and/or typical for a given style of music, for better or for worse.
Then why all that jazz about fighting every sort of distorsion from our electronic stuff? (loudspeakers are another story)
Why not letting some euphonic distorsion from our amplifiers?
Hey there Fmplayer's pensƩes & thoughts - cool name,
DeleteIndeed that is a good question. For me, I believe always the intent of "high fidelity" is just to make sure the hardware is capable of transparency (as argued long ago when folks were talking about nonsense like "measuring emotional connection to music").
So long as the hardware is good enough in staying out of the way of the sonic reproduction within the limits of what and how we're listening (including our own hearing acuity), then to me that's all that matters. DACs do not 99.9% of the time need THD+N values better than -110dB regardless of how cool that might be to have "the quietest DAC in the world" at -126dB or whatever impressive number. Amps likewise probably would be just fine with say a dynamic range of 80dB with low distortion (<0.5% peak THD maybe?) for the vast majority of home listeners depending on the noise-floor in their sound room. (And everyone will need to figure out for themselves how much power they need from that amp for their speakers, space, listening distance...)
This is why I've always said of myself as a "mostly objective" hardware audiophile. If I'm evaluating/reviewing gear, then the central question is one of measurements to determine the limits of the device's fidelity. But when it comes to listening to music, hey, so long as the hardware gets me all the way I need to experience what's on the recording accurately in that space and with those speakers, that's all I can ask!
Ergo:
1. Room acoustics, noise floor, placement, listening position, and loudspeakers are more than likely the limiting factors when it comes to achieving highest playback fidelity for most of us who already have decent quality electronics.
2. We need to be mindful, insightful, and honest about our hearing abilities especially as we get older. Let's not assume of ourselves to be "golden ears" who can hear down to the capacitance of speaker cables. š¤£
3. There's nothing wrong with having a preference for coloration. Tube DACs and amps with extra harmonic distortion. Devices with non-flat frequency response. A little bit of background hiss might all add a little something for some folks (like euphonic nostalgia). While technically it's not "hi-fi", it's what I've called "euphonophilia" over the years - an absolutely legitimate type of the audiophile hobby for those who prefer that path rather than examine objective test results. Realize though that this path usually does not lead to cumulative knowledge, might result in increased Industry influence, and preference-wise can end up highly idiosyncratic.
The "bloom" of the vocals from a nice tube amp, or the high crosstalk, high noise floor of a vinyl recording can sound great or evoke feelings, but that's not technical "high fidelity" and for those analogue lovers, please don't pretend that it is!
4. When it comes to love of music, appreciation of artists, tapping toes, and emotional well being, we're all subjectivists. I'm totally comfortable with that. š
There's no such thing as an in room "Target curve" https://www.audiosciencereview.com/forum/index.php?threads/what-is-your-favorite-house-curve.2382/page-13#post-1470714
ReplyDeleteThe only "target curve" is the anechoic on/off axis at 1-2m. In room, below 500hz +/-, "target" peaks judiciously. The End.
Hi Arch: This is great stuff. I broke down and bought a pair of ASCI A6b's to upgrade from the LS 50 Metas which are now my surrounds, and I upgraded by Dirac from the Beta Version of DLBC and ART to the full version, and now I'm looking for target curves for my upgraded system with its Dual SB 2000's and Kef HTC 3001 SE center. I guess my first question is whether given the system is 5.1, does it make sense to use the ATMOS target curve for it, or just follow something like a regular Harman with maybe your vivid as a change of pace?
ReplyDeleteAlso there is the issue of MB. I have a 5.1 system with two subs located in different areas of the room. As I understand MB, it requires the user to take a MM measurement at the listening position and a near field at the location of each channel. Perhaps you see the problem already. My LFE channel does not have one unique location but rather two. I certain many set ups do the same thing and have multiple mono subs. In fact such set ups are widely advocated as a way of evening out bass response, but they don't have a uniques nearfield location to measure as a result. So how would that work with MBs? Perhaps Joe Mariano is still with us and would like to weigh in on this issue?
Obviously, it's an important one. The entire raison d'etre of programs like Dirac and especially ART is to provide a way for the room to evenly and consistently play frequencies below the Schroeder frequency where these subs account for a large portion of what is loaded into the room. It would be nice to know how MB could work in a multi-mono subs use case.
Thanks Arch.
MB can only apply a corrections to the single LFE channel. So, whether you have 1 sub or 20 subs, MB comes up for the response for the summed response. Things like Dirac DLBC, Dirac ART or MSO can do fancy things to come up with that summed response, but the target curve MB comes up with is only for that single LFE channel.
DeleteTo answer your other question, MB does not ask for nearfield measurements for the sub(s). We only take the MLP measurement of the summed response. The target response for the LFE channel is based on the average room response from all the ear-level speakers. If you think about the purpose of subs, it's to augment the natural response of the other speakers, so the target curve for the LFE should be whatever the full extended response of full-range speakers would be if they could play down to 10Hz (for example).
Hope that helps.