Jump to content


Photo

Time to kill the myth that "flat" bass is "correct" bass.


  • Please log in to reply
64 replies to this topic

#21 SME

SME

    Super Bass Overlord

  • Members
  • PipPipPipPipPipPipPipPip
  • 983 posts

Posted 05 March 2017 - 02:42 AM

It's been a while since I updated, but I've been experimenting with variations of bass boost with music and have studied the situation with movie soundtracks a bit more closely.

 

I'll discuss movies first because there's not much to talk about here.  The short answer is that cinema calibration standards are fundamentally broken, so you can't calibrate for a reference response.  The standards aren't just broken for home playback.  They are broken in cinemas and dub stages too.  Recent studies have proven that the subjective responses of various cinemas and dub stages are all over the map.  As a consequence, theatrical mixes are all over the map themselves.  There's no fix for this other than to re-EQ for each feature, and because we don't have data on the dub stages used for each production, the re-EQ can only be optimized by ear.  In principle, home mixes have the potential to be better, but I'm sure a lot have been ruined  by using the same broken calibration method, albeit without the X curve treble roll-off.  Either way, they are broken.  As of 2014, the industry is aware of the problems with cinema calibration standards and are working, albeit slowly, to some kind of resolution.  Until then, the only option is re-EQ.  I've been doing this myself a lot lately.

 

With that said, let's talk about music again.  My earlier point about "flat in-room bass" being a myth is absolutely relevant with music where the reference is an anechoic flat speaker that extends down to at least 40 Hz or so.  Below there, subwoofers are typically in use, and a common practice is to level match the subs with the mains at the crossover point.  This reference has essentially been proven by Harman's research.  However, it is an incomplete reference.  It is a much better reference than flat in-room response, but room effects still alter subjective response in the bass region, at frequencies up to a few hundred Hz or so.

 

The best we can do when calibrating a system for music is to anticipate the consequences of the room on an anechoic flat speaker response in a typical mastering studio.  In the studios, the existence of a bass lift of 3-5 dB is practically universal.  It arises because the woofers are close enough to the floor to make the direct sound and floor reflection indistinguishable.  Even if the room was otherwise acoustically dead, that reflection would still boost the bass.  The transition frequency depends on the particular speaker design and on listening distance.  An engineer working on a particular system will tend to cut frequencies below that transition frequency, wherever it is.  Therefore, in order to hear the music the way the artist intended, we need a matching bass lift in our own response.

 

If we run high quality passive speakers (i.e., anechoic flat) without EQ, we will get the bass lift naturally, but it may not occur at the same frequency that it did in the mastering studio for whatever content we are listening to.  So as a consequence, some recordings will sound very good (similar transition frequency and gain), and others will seem to have flaw such as muddiness (if our transition frequency or gain are higher than in the studio) or thinness or harshness (if our transition frequency or gain are lower than in the studio).

 

For several weeks now, I have experimented with using an adjustable bass lift of 3-5 dB with an adjustable transition frequency to allow me to optimize playback for a wider variety of content.  My experience is that adjusting the transition frequency for different music allows me to hear the best sound with a variety of material.  Whereas in the recent past, I tended to use a -1 to -3 dB high shelf centered around 2 kHz to tame thinness and harshness with a lot of music, I am finding that upward adjustment of the bass boost transition frequency works better most of the time.  (I still use the 2 kHz shelf with movies from time to time.)

 

To re-iterate, the typical transition frequency ranges from 100-500 Hz with gain ranging from 3-5 dB.  Typically higher gains are appropriate where the turnover frequency is higher.



#22 SME

SME

    Super Bass Overlord

  • Members
  • PipPipPipPipPipPipPipPip
  • 983 posts

Posted 05 March 2017 - 02:50 AM

Funny, I was going to say something very similar.

How I choose to setup my system to my preferences is the ONLY thing that is relevant. I prefer a lot of bass and mid bass. But, not so much to where is sounds bloated or muddy. However, I think a flat system sounds awful. I find a middle ground between the two. I suspect I would walk out of your room due to your preferences. Does that make yours wrong? Certainly not. It's YOUR preference; I just happen to prefer something different. Isn't that what this hobby is all about? Sharing our own preferences without someone coming in and shitting on them?

 

Exactly.  If your sound is bloated or muddy, then there's probably too much bass or too much low frequency energy somewhere. Otherwise, more bass is usually a good thing.  Strangely enough, almost all listeners would agree with this point.  Furthermore, the thresholds for muddiness and bloatedness tend to be quite consistent between listeners.  So the fact that flat sounds bad to you should tell you that flat is not correct.  For music, this is absolutely true for in-room response as well as outdoor response.  A boost of 3-5 dB is likely best for outdoor response.  The boost for indoor response may be higher because your in-room response may include a lot of reflected energy that is not necessarily heard as being part of the speaker innate voice.



#23 SME

SME

    Super Bass Overlord

  • Members
  • PipPipPipPipPipPipPipPip
  • 983 posts

Posted 05 March 2017 - 08:29 AM

Skipping the other silliness for the moment, In the rooms you have calibrated, has every room of differing size, construction, and acoustics subjectively sounded to have the same spectral balance just by adjusting for a flat magnitude response as measured by a sine-sweep?  How about measured by RTA?  If the answer is yes they sound the same beyond loudspeaker/subwoofer deficiencies, then there's not much to discuss.  I don't know anyone with experience calibrating more than a dozen rooms who would answer yes to that question, and that gives us something to discuss.

 

There's an even bigger issue lurking in this discussion and in many others about frequency response flatness.  A sine sweep measurement yields an impulse response, which can be converted using the Discrete Fourier Transform to a frequency response consisting of magnitude and phase vs. frequency.  That data contains an enormous amount of information, yet it looks like a total mess and bears little direct relevance to what the system actually sounds like.

 

So the usual approach taken is to apply some kind of smoothing.  When people say they have a flat in-room frequency response, they almost always refer to the frequency response data after some kind of smoothing has been applied.  However, there is more than one way to smooth the data, even after selecting a bandwidth (e.g. 1/3rd octave) for the smoothing filter.  Smoothing is effectively the application of a low pass filter to frequency data, and obviously there are countless possibilities for such filters.  I'm not aware of any standard that specifies which kind of smoothing filter should be used for audio data.  Believe me, I tried to track down such a standard.  I can't even find documentation detailing the filters used in programs like REW beyond some ambiguous description like "moving average filter".

 

The only real standard for smoothing data is the cinema X curve, which relies on measurement with a 1/3rd octave RTA using full-band pink noise.  This data is binned rather than being smoothed, and the averaging is totally different from the averaging that's applied to smooth frequency data.  In any case, it is a broken standard that produces widely variable subjective results on different systems and in different rooms.  It is believed, with good reason, that the variance is due to the fact that RTA measurements fail to distinguish between the sound from the speaker versus the sound contributed by the room.  Almost all frequency response smoothing methods suffer the same deficiency.



#24 Bossobass Dave

Bossobass Dave

    Bass Overlord

  • Members
  • PipPipPipPipPipPipPip
  • 914 posts

Posted 05 March 2017 - 02:47 PM

Skipping the other silliness for the moment, In the rooms you have calibrated, has every room of differing size, construction, and acoustics subjectively sounded to have the same spectral balance just by adjusting for a flat magnitude response as measured by a sine-sweep?  How about measured by RTA?  If the answer is yes they sound the same beyond loudspeaker/subwoofer deficiencies, then there's not much to discuss.  I don't know anyone with experience calibrating more than a dozen rooms who would answer yes to that question, and that gives us something to discuss.

 

Of course not. Since the bandwidth includes quite a bit of infrasonic content, and boundary rigidity and transmission losses are not trivial, the reaction of the environment  can completely change the presentation.

 

I think the point of the OP is calibration method and the results therefrom.

 

Back to Olive quote, with a different emphasis this time:

 

 

Accuracy is Not Applicable to Most Recordings Made Today
Most recordings made today are not intended to sound like the live performance. Anyone who heard Taylor Swift's live performance with Stevie Nicks at the 2010 Grammy Awards understands why. (Note: you can relive the magical moment on Youtube. Warning: this may be offensive for the musically-inclined). About 90% of commercial recordings are studio creations consisting of a series of overdubs, processed with auto-tuning, equalization, dynamic compression, and reverb sampled from an alien nation. For these recordings, there is no equivalent live performance to which the recording/reproduction can be compared for accuracy. The only reference is what the artist heard over the loudspeakers in the recording control room. If the important performance aspects of the playback system through which the art (the music and recording) was created can be reproduced in the home, then the consumer will hear an accurate reproduction of the music, as the artist intended. It is possible to achieve this if we adopt a science in the service of art philosophy towards audio recording and reproduction.

 

The experiments I've looked into, since both of my sons are musicians who both record and mix their own audio creations and who both mix in the nearfield with a subwoofer, is along the lines of what Olive says in the quoted article.

 

If my sons mix a snippet of sound, music or otherwise, and accompany that recording with a FR at their ears and recording level, it is notable that when you match that FR and level in the playback environment, the experience is far more accurate than the randomness of what you find in modern recordings.

 

That would be the reference from which the "silliness" can feel free to roam, whether it be the spawn of any of the thousands of listening experiments done in the 20th century or the car sub enthusiasts-turned-home theater buffs.



#25 Bossobass Dave

Bossobass Dave

    Bass Overlord

  • Members
  • PipPipPipPipPipPipPip
  • 914 posts

Posted 05 March 2017 - 03:09 PM

Exactly.  If your sound is bloated or muddy, then there's probably too much bass or too much low frequency energy somewhere. Otherwise, more bass is usually a good thing.  Strangely enough, almost all listeners would agree with this point.  Furthermore, the thresholds for muddiness and bloatedness tend to be quite consistent between listeners.  So the fact that flat sounds bad to you should tell you that flat is not correct.  For music, this is absolutely true for in-room response as well as outdoor response.  A boost of 3-5 dB is likely best for outdoor response.  The boost for indoor response may be higher because your in-room response may include a lot of reflected energy that is not necessarily heard as being part of the speaker innate voice.

Exactly the problem, IMO. One man's "bloated" is another man's "thin". They are both meaningless terms to the reader.

 

Mark mentioned decay earlier. As a matter of physics, generally speaking, the lower the frequency, the longer the decay, the more weight in the presentation. If a recording is brick wall filtered at 30 Hz, no amount of SW trim boost will add the weight that will be perceived when the same content is unfiltered.

 

Subjective silliness aside (I'll use that word as no one has built a cross on which to hang Seaton for using it) "I like the filtered version better" "I like the unfiltered version better", Tastes Great", "Less Filling"... irrelevant.

 

Mark then brings the differences in rooms into the mix. I've studied the FRM measurements of 15 forum members who can be relied upon to produce fairly accurate data. I've compared those to the known ground plane measurements of their subwoofers by Josh and Ilkka. I've found the room gain curve, which is derived from those differences and then averaged, to be very similar. FRM (frequency response magnitude) is predictable and has little to do with room size. I've argued this before in what the kids today call "contention", but the results are what they are, silliness notwithstanding.

 

Rooms differ in how the boundaries are constructed and with what materials they are decorated. This has been known since the beginning of audio hardware and is discussed in detail in several books in my library, circa 1945-1960.

 

I think you are correct (as Olive and others have suggested) that calibration begins with standardization on the production side. Absent that, FRM and level via the placement/phase method can be a practical one and an effective one.



#26 maxmercy

maxmercy

    Super Bass Overlord

  • Moderators
  • 1,839 posts
  • LocationNW FL

Posted 05 March 2017 - 03:44 PM

Bosso lives!

 

JSS


  • MemX likes this

#27 SME

SME

    Super Bass Overlord

  • Members
  • PipPipPipPipPipPipPipPip
  • 983 posts

Posted 05 March 2017 - 10:56 PM

Exactly the problem, IMO. One man's "bloated" is another man's "thin". They are both meaningless terms to the reader.

 

I totally disagree.  As long as we agree on the meaning of the terms, we will mostly agree on what's "thin" and what "bloated".  This is born out in studies by Harman and is also validated by the enormous success of lossy digital audio compression algorithms.

 

By employing a detailed, experimentally-verified model of human auditory perception, these algorithms are able to discard 80% or more of the information contained in a digital audio stream with almost only very minor consequences on perceived sound for a wide variety of listeners.  Where some audible artifacts are acceptable as in cellular voice service, substantially more information can be discarded.

 

In an information sense, the bandwidth of the human ear/brain system is very narrow.  One of the challenges faced by a mixer is to ensure that every element or part of the mix is balanced so that the listener actually hears it all.  But of course, this only works if the tonal balance of the production and playbacks systems are consistent.  Where there is inconsistency, relevant auditory details may be buried or lost completely.

 

The problem is that we haven't fully solved the problem of calibrating for consistency between different systems.  Only two firm standards exist for doing so.  One is the X curve standard for cinemas in which EQ is used to adjust the level of full bandwidth pink noise filtered in each 1/3rd octave RTA bin to fit the X curve target.  This standard has been proven to be seriously flawed.  The other is to listen to anechoic flat speakers (except for a UHF roll-off) at least 5 feet from nearby walls in a small to medium size room.  This isn't really a documented standard, but it is completely consistent with how most music is mixed and mastered.  And unlike X curve, Harman's research indicates that this approach yields a preferred sound for both trained and untrained listeners.  The approach doesn't fully solve the translation/consistency problem, but it works way better than X curve does.

 

That's it.  Many people will claim that a "flat frequency response" is best, but this is only close to true if the flat response was of a speaker measured in anechoic conditions.  Otherwise, it's most likely false.  And I can practically guarantee that for music that the smoothed, in-room bass response should measure at least 3-5 dB hotter than the treble because of the floor reflection.  In most situations, the difference will be even greater because of the power response difference and its impact on smoothed frequency response independent of first arrival sound.  Harman's data suggests a 10 dB gap is not unusual.  I believe the gap for my system as calibrated and with the usual boost applied is about 7-8 dB.



#28 Bossobass Dave

Bossobass Dave

    Bass Overlord

  • Members
  • PipPipPipPipPipPipPip
  • 914 posts

Posted 06 March 2017 - 02:13 PM

I totally disagree.  As long as we agree on the meaning of the terms, we will mostly agree on what's "thin" and what "bloated".  This is born out in studies by Harman

 

 

 

I believe the gap for my system as calibrated and with the usual boost applied is about 7-8 dB.

 

Define the terms.

 

Cite the studies.

 

How do you calibrate your subwoofers to the satellites?



#29 Bossobass Dave

Bossobass Dave

    Bass Overlord

  • Members
  • PipPipPipPipPipPipPip
  • 914 posts

Posted 06 March 2017 - 02:14 PM

Bosso lives!

 

JSS

Hey, DOC!



#30 Bossobass Dave

Bossobass Dave

    Bass Overlord

  • Members
  • PipPipPipPipPipPipPip
  • 914 posts

Posted 06 March 2017 - 02:37 PM

My opinion is that every recording should come with a FR taken at the mix desk with mix level noted.

 

Rooms are the largest variable and should be treated accordingly.

 

Ham fisted use of parametric EQ has been the bane of audio playback.

 

Studying the distortion preferences of the general public has nothing to do with proper calibration.

 

Relying on Harmon is akin to asking Pfizer if drugs are good for your health. I read a Harmon study years back and saw this FRM graph. And when I normalized the graph to what we're used to studying... the house curve-looking graph is actually a flat response from a ported sub tuned to around 18 Hz with a typical 80 Hz LR4 LPF.

 

art-895641.jpg

 

The problem is that the sub used is covering 2 octaves of the 5-1/2 octave RB+LFE+10dB summed channel. The results are useless in my experience.

 

Most people on this forum are enthusiastic about dBSPL.

 

If you have a jazz quartet playing live in your room, you don't typically tell the upright bass player to move his instrument to a different position in the room, dampen the D string and pluck his strings a LOT harder.



#31 lukeamdman

lukeamdman

    Bass Overlord

  • Members
  • PipPipPipPipPipPipPip
  • 833 posts
  • LocationBurnsville, MN

Posted 06 March 2017 - 04:16 PM

Preference-led manipulation thereafter is generally unlistenably distorted, although enjoyed, celebrated and defended regularly in these forums.

 

 

That would be the reference from which the "silliness" can feel free to roam, whether it be the spawn of any of the thousands of listening experiments done in the 20th century or the car sub enthusiasts-turned-home theater buffs.

 

 

Most people on this forum are enthusiastic about dBSPL.

 

If you have a jazz quartet playing live in your room, you don't typically tell the upright bass player to move his instrument to a different position in the room, dampen the D string and pluck his strings a LOT harder.

 

You frequently mention the members of this forum in conjunction with "SPL wars" and similar comments to the ones I've highlighted above, but outside of you attending a GTG of some of the members here, where 99% of the time the bass is cranked because people are drinking and just having fun, how many of our theaters have you auditioned at their "normal" playback levels (typical movie watching or 2ch listening levels)?  

 

Yes some of us occasionally have have fun with SPL, but at least in my HT, that probably only constitutes about 1% of the actual playtime.  The other 99% is spent watching movies with my wife and kids with the main volume around -15 without touching the sub trim.  Just last week we watched a movie at our normal listening levels and it wasn't enough to wake up our 9 month old from his nap.  

 

At the many GTGs I've hosted, and probably 50+ individual demos I've given to members of this board and AVS, they always start at normal listening levels with 2ch music (Livingston Taylor, Jessica Pigeon, Nils Lofgren, etc.).   For those not interested in "bass head" stuff, sometimes the subs are never even turned on and the mains are run "full range".  Typically by the end of a big GTG though most people are wanting to hear some crazy bass so I oblige, which is no different than what you did for me when you hosted a last minute GTG for me.  

 

I've also asked all those folks when listening to 2ch music or movie clips at -10db without the sub trim boosted if it seemed too heavy in the bass or out of balance and I have yet to hear a single person say the bass was too loud/heavy or out of balance compared to the mains.  



#32 Ricci

Ricci

    Super Bass Overlord

  • Moderators
  • 1,088 posts
  • LocationLouisville, KY

Posted 06 March 2017 - 05:06 PM

My opinion is that every recording should come with a FR taken at the mix desk with mix level noted.

 

 

 

That would be interesting but of course it will never happen. The playback level often varies quite a bit during mixing or depending on the project to gauge how it sounds at high middle and low volume playback. Often there will be a best or "big" room / speaker setup, a smaller room / smaller speaker but still quality setup and usually a consumer grade "typical" speaker with virtually no bass and limited HF playback similar to what you'd see with a radio, tv speaker or Bluetooth dock. Clearly we're more concerned with the main system playback during mixing but there's a lot of variation there. Not to mention differences in room decay rates, tactile response and dispersion characteristics.

 

Are there any recordings that give this type of information? Perhaps some of the audiophile recordings? I like the idea of it even if it would be simplified quite a bit. I've had this urge for a long time to record a raw stereo acoustic drumset track with very high quality condenser mics and somehow calibrate the playback level of the track to reproduce it with full dynamic range. I'd probably have to include a pink noise track or something.



#33 Bossobass Dave

Bossobass Dave

    Bass Overlord

  • Members
  • PipPipPipPipPipPipPip
  • 914 posts

Posted 06 March 2017 - 06:45 PM

That would be interesting but of course it will never happen. The playback level often varies quite a bit during mixing or depending on the project to gauge how it sounds at high middle and low volume playback. Often there will be a best or "big" room / speaker setup, a smaller room / smaller speaker but still quality setup and usually a consumer grade "typical" speaker with virtually no bass and limited HF playback similar to what you'd see with a radio, tv speaker or Bluetooth dock. Clearly we're more concerned with the main system playback during mixing but there's a lot of variation there. Not to mention differences in room decay rates, tactile response and dispersion characteristics.

 

Are there any recordings that give this type of information? Perhaps some of the audiophile recordings? I like the idea of it even if it would be simplified quite a bit. I've had this urge for a long time to record a raw stereo acoustic drumset track with very high quality condenser mics and somehow calibrate the playback level of the track to reproduce it with full dynamic range. I'd probably have to include a pink noise track or something.

 

No, I'm not aware of a single production disc that includes the mix desk FR, but some do include the mix playback level, which goes to the influence of equal loudness curves.

 

I'm aware that some mixers try their product through different systems (most don't do that as a mix tool), but they rarely adjust the final mix for those results, in my recording studio experiences. And, the mastering stage is where the final mix is actually done, where none I've ever known left their seat during the mastering process.

 

The next time you're involved in a studio, run the FR measurement yourself and note the playback level for your own reference. Like I said many years back, when the actual bass player is sitting in your sweet spot listening to a recording he played the bass on (something I did years back) and turns to you and says "THAT'S what I'm talkin' about!", how the hell do you argue with that? I could have told him that I prefer the bass bumped +10dB and I certainly have the right to do that in private listening experiences but that wouldn't have any relevance to how he told me it should sound when played back. That experience and listening to recordings on which I was the bass player opened my eyes. When you match the FR used to mix and the playback level, the result is very accurate, other parts of the discussion notwithstanding.



#34 SME

SME

    Super Bass Overlord

  • Members
  • PipPipPipPipPipPipPipPip
  • 983 posts

Posted 06 March 2017 - 09:24 PM

Define the terms.

 

Cite the studies.

 

How do you calibrate your subwoofers to the satellites?

 

It's in the article I linked to, but maybe you disregarded it because it is associated with Harman, even though it's written by Sean Olive who you have twice quoted here.  But just to summarize:

 

    full vs. thin describes the relative level of low frequencies (bass and low mids) with "full" describing excess and "thin" describing deficiency.

    bright vs. dull describe the relative level of treble with "bright" describing excess and "dull" describing deficiency.

 

I will note that these attributes most accurately apply to particular notes or particular sound effects.  In a presentation, it's entirely possible to have thinness with some notes and fullness with others.

 

As for my sub calibration, the answer requires a bit of detailed discussion.

 

Basically, I calibrate so that the impulse response for each input channel (except LFE) measured with a -20 dBFS sine sweep and analyzed with a 1/3rd octave frequency dependent window (FDW) in REW is 80 dB SPL.  Acttualy, I bend this rule very slightly at the top end, the bottom end, and at the sub/mains XO.  At the top and bottom end, the effective LPF and HPF introduce phase shift that delays the first arrival enough that it slips past the FDW.  A similar issues happens around the sub/mains XO.  For these, I consider the response with up to 1/6th octave FDW, but note that if I use 1/6th octave FDW or higher universally I end up with subjectively imbalanced mid-range due to a very early reflection from the ceiling that my brain apparently ignores when assessing the tonal balance of the source.

 

After the calibration, I apply additional adjustments that are needed for typical content.  Pretty much everything I play gets a filter on ultra high frequencies that is consistent with the effect of in-air absorption and primarily diminishes the response at the very very top.  Almost all professional monitors have such an adjustment built-in, particularly those meant for near-field use.  The second filter is the bass boost I describe here: 3-5 dB somewhere between 100 and 500 Hz or so, again depending on content.  These two filters are sufficient to optimize the presentation of most music I own.  For movies, I also have an adjustable X-curve like roll-off of up to -3 dB/octave starting at 2 kHz.  I also occasionally use a high shelf at 2 kHz of -1 to -3 dB, which is similar to the "high trim" switch available on many monitors and is helpful at with times with both music and movies.  In the case of movies, the need for this upper mid shelf may arise due to some upper mid range energy accumulating and inflating the RTA measurements in the particular dub stage used.  With music, it may arise due to the use of the "high trim" switch during the mix or it may arise due intentional EQ of the monitoring system by the mix engineer.  Apparently many professionals believe that the response above 1 kHz should diminish by -1 dB/octave regardless of room effects and whatnot.  Needless to say, some stuff is just too thin and bright without the upper-mid shelf.  Lastly, there are occasions in which I disable the bass boost, most likely because the mixer used an in-room measurement for calibration here as well instead of relying on the informal reference of "anechoic flat speaker placed in-room at least 5 feet form adjacent walls".

 

Because cinema mixes are done in dub stages that are calibrated using in-room measurements, the bass boost may be inappropriate here as well, but it really depends.  I believe most of movie soundtrack production is done using anechoic flat monitors that may not be configured using any additional EQ.  Only the dub stage is actually calibrated to the X curve, and given growing awareness of the flaws of the X curve calibration, re-recording mixers are probably a lot less likely to make EQ adjustments once the mix gets to the dub stage.  Home mixes are also likely done in an environment without X curve calibration and may sound very good on a system that is optimized for 2 channel music playback if done correctly.  Hence, the bass boost is still often appropriate.  Since the floor bounce arrives relatively late and is somewhat diminished under near-field monitoring conditions, I tend to set the boost at around +3 or +4 dB at 100 Hz for most movies.  With most recent releases, this approach works very well, but there are still exceptions.

 

The end result of this calibration process is left and right mains that measure about 85 dBC using the -20 dBFS pink noise that's band-limited to 500-2000 Hz.  So these days, my master volume "0" is essentially compatible with the cinema reference standard.  Based on my own "first arrival" hypothesis for EQ and SPL calibration, I should be playing at MV "-4" to achieve equivalent loudness to a cinema.  And in fact, this is the level that I prefer on average after correcting for tonal balance issues with most recent film soundtracks.  Older tracks are more likely to come in at around "-6", presumably because of the early confusion about the SPL calibration standard concerning the proper meters to use and the fact that Dolby's pink noise test signal was actually -18 dBFS instead of -20 dBFS.

 

The pink noise test signal for my sub (-20 dBFS and band-limited to 40-80 Hz, IIRC) yields 85 dBC only if I have about +4.5 dB of bass boost active over my baseline calibration.  My result is probably not typical and arises because of my use of near-field MBMs.  The directivity of the MBMs in the 50-100 Hz range is effectively a lot higher than for my mains in the 500-2000 Hz region.  Below 50 Hz, directivity is fairly high too because the room response is dominated by modal behavior where global de-reverberation via EQ becomes possible.  So ironically, "boosting" by 3-5 dB gives me a sub response that's balanced and correct according to cinema standards but oftentimes sounds too hot.  I can't emphasize enough the fact that cinema calibration standards are completely broken.

 

 

My opinion is that every recording should come with a FR taken at the mix desk with mix level noted.

 

This information is mostly useless unless the full impulse response measurement data is included.  I've said many times and will say again that 1/6th octave smoothed frequency response magnitude is most useless for characterizing how a system sounds.  If the full impulse response is provided, then there is an opportunity to run the kind of analysis that's actually needed.

 

As far as what that analysis should be, I'm still working on that problem and will be probably for a long while.  The 1/3rd octave FDW worked extremely well for me in my room, but I know it is will need refinement to deal with crossover delays and what not at the very least.  It's still an open question in my mind just how much first arrival dominates perception vs. contributions from early reflections.  With good speakers like I have, flat first arrival at the MLP tends to provide very smooth response of the reflected sound as well, so I can't rule out that the reflected sound isn't at least partly important.  And this will be important for achieving accurate calibration of systems whose speakers have less-than ideal off-axis response.

 

There's also some question as to how perception works in the bass region where room effects, particularly floor bounce in quality mastering studios but also other boundaries in typical home environments where speaker placements are more subject to compromise.  I've seen Harman suggest that below 300 Hz or so, perceived tonal balance is dominated by power response rather than the anechoic response of the speaker as is the case above 300 Hz.  My guess is that this is not correct, but rather appears to be true because of characteristic of the particular speaker and room used for testing.  Most speakers lose their baffle step directivity at roughly that frequency, and the wavelengths there become long enough that some early reflections (particularly the floor) become indistinguishable.  So under typical conditions, 300 Hz may be the approximate frequency where subjective tonal balance perception is no longer dominated by anechoic response, but that may be due to the particulars of speaker and room rather than any perceptual tendency.  What happens if one is using a floor-to-ceiling I.B. array speaker in which all room reflections arrive late enough to be completely distinguishable from the direct sound?  My guess is that "anechoic response" would still dominate then, even though such a system would need a 3-5 dB bass boost to sound good with most content mastered on typical floor-stander or near-field monitor speakers.



#35 Bossobass Dave

Bossobass Dave

    Bass Overlord

  • Members
  • PipPipPipPipPipPipPip
  • 914 posts

Posted 07 March 2017 - 01:10 AM

Your 3-5dB boosted sub sounds hot because it is.

 

FR is never useless.



#36 SME

SME

    Super Bass Overlord

  • Members
  • PipPipPipPipPipPipPipPip
  • 983 posts

Posted 07 March 2017 - 10:26 AM

Your 3-5dB boosted sub sounds hot because it is.

 

FR is never useless.

 

No, you misread.  The 3-5 dB boost sounds balanced with most music and many movies too.  Only with a minority of content does that boost sound excessive.

 

To repeat something I said in my previous post:  I have to have 4.5 dB boost from my baseline calibration for my subs to level match the mains with the cinema pink noise test.  So by cinema standards, I'm not running my subs hot at all!

 

FR is not completely useless, but the frequency response magnitude plots people post online mostly are useless when they are based on in-room measurements.  No one hears frequency response.  If music was nothing but unchanging sine waves, then listeners would hear the unsmoothed frequency response.  But real life content is always changing.  Even long continuous tones in music tend to flutter a bit in pitch or be subject to echo or reverb.  That's a good thing too.  Most people's unsmoothed frequency responses look like garbage.

 

To understand what happens to content other than sine-waves, you have to perform a mathematical convolution of the content with the impulse response, which can be obtained from the unsmoothed frequency and phase data.  Smoothing *does not* do this.  Smoothing does improve the appearance of the data, but most methods effectively degrades or destroys the information that's needed to do the convolution analysis described above.  Therefore, all smoothing really does is improves the visual appearance of the system's response to sine waves.  That's not very interesting.

 

Also, human hearing does not operate with regard to frequency alone.  Crucially, hearing has a time aspect to it.  So does the response of a speaker in a room.  Using the original unsmoothed frequency and phase response data, a time-frequency transform can be performed to reveal the response of the speaker and room in terms of time and frequency together.  This is entirely analogous to using Spec Lab to visualize auditory content in movies.  Without the time aspect of the in-room response, it's hard to gain insight into how your speaker and room are actually affecting what you hear.  Frequency response smoothing almost always degrades or destroys this information.

 

So to give a short answer: "when you you say frequency response, I do not think it means what you think it means."  I wish it were simpler, that we could all just calibrate our 1/6th octave smoothed in-room responses to flat, but that just isn't how things work.  On the positive side, if things did work that way, then we'd probably find listening to be very difficult and frustrating in different acoustic environments or even different places in the same room because the "frequency response" changes so dramatically.  But what happens instead is that things sound much the same in different environments, despite the dramatic changes in frequency response that result from the room.  I'm not saying that acoustics don't matter, but they don't matter as much as one might assume from looking at a smoothed frequency response measurement.  This is a crucial point to grasp.



#37 lukeamdman

lukeamdman

    Bass Overlord

  • Members
  • PipPipPipPipPipPipPip
  • 833 posts
  • LocationBurnsville, MN

Posted 07 March 2017 - 02:19 PM

No, you misread.  The 3-5 dB boost sounds balanced with most music and many movies too.  Only with a minority of content does that boost sound excessive.

 

To repeat something I said in my previous post:  I have to have 4.5 dB boost from my baseline calibration for my subs to level match the mains with the cinema pink noise test.  So by cinema standards, I'm not running my subs hot at all!

 

FR is not completely useless, but the frequency response magnitude plots people post online mostly are useless when they are based on in-room measurements.  No one hears frequency response.  If music was nothing but unchanging sine waves, then listeners would hear the unsmoothed frequency response.  But real life content is always changing.  Even long continuous tones in music tend to flutter a bit in pitch or be subject to echo or reverb.  That's a good thing too.  Most people's unsmoothed frequency responses look like garbage.

 

To understand what happens to content other than sine-waves, you have to perform a mathematical convolution of the content with the impulse response, which can be obtained from the unsmoothed frequency and phase data.  Smoothing *does not* do this.  Smoothing does improve the appearance of the data, but most methods effectively degrades or destroys the information that's needed to do the convolution analysis described above.  Therefore, all smoothing really does is improves the visual appearance of the system's response to sine waves.  That's not very interesting.

 

Also, human hearing does not operate with regard to frequency alone.  Crucially, hearing has a time aspect to it.  So does the response of a speaker in a room.  Using the original unsmoothed frequency and phase response data, a time-frequency transform can be performed to reveal the response of the speaker and room in terms of time and frequency together.  This is entirely analogous to using Spec Lab to visualize auditory content in movies.  Without the time aspect of the in-room response, it's hard to gain insight into how your speaker and room are actually affecting what you hear.  Frequency response smoothing almost always degrades or destroys this information.

 

So to give a short answer: "when you you say frequency response, I do not think it means what you think it means."  I wish it were simpler, that we could all just calibrate our 1/6th octave smoothed in-room responses to flat, but that just isn't how things work.  On the positive side, if things did work that way, then we'd probably find listening to be very difficult and frustrating in different acoustic environments or even different places in the same room because the "frequency response" changes so dramatically.  But what happens instead is that things sound much the same in different environments, despite the dramatic changes in frequency response that result from the room.  I'm not saying that acoustics don't matter, but they don't matter as much as one might assume from looking at a smoothed frequency response measurement.  This is a crucial point to grasp.

 

A lot of this reminds of me of when tuxedocivic turned my world upside down a few years ago with the proper method of calibrating HF (1khz+):

 

http://data-bass.ipb...age-7#entry9037

 

The concept of how the first arrival dominates how we perceive sound is fascinating, and the proof was in the pudding.  Calibrating at the seats may have sounded semi-decent at that specific seat alone, but calibrating close mic with a gated response sounded worlds better in that primary seat and every other seat!  The new FR didn't visually look nearly as flat at the seats, but mics can't ignore the reflections that impact FR differently than the way our ears perceive them.  

 

Something else I've been confused about lately though is whether mic calibration files are valid for "close mic" HF measurements.  If you look at the calibration data and the distance they are taken from, that distance definitely impacts the amount of HF roll-off because of how the air effects the upper end response.  Anyway, that's off topic and a discussion for another thread.  

 

Quick question though. When you say you calibrate to -20 dBFS which ends up being about 85dBC, is that with an input headroom of 105db for the L/C/R and 115db for the sub?  


  • deepthoughts likes this

#38 SME

SME

    Super Bass Overlord

  • Members
  • PipPipPipPipPipPipPipPip
  • 983 posts

Posted 07 March 2017 - 10:35 PM

A lot of this reminds of me of when tuxedocivic turned my world upside down a few years ago with the proper method of calibrating HF (1khz+):

 

http://data-bass.ipb...age-7#entry9037

 

The concept of how the first arrival dominates how we perceive sound is fascinating, and the proof was in the pudding.  Calibrating at the seats may have sounded semi-decent at that specific seat alone, but calibrating close mic with a gated response sounded worlds better in that primary seat and every other seat!  The new FR didn't visually look nearly as flat at the seats, but mics can't ignore the reflections that impact FR differently than the way our ears perceive them.

 

Technically speaking the mic captures almost all the information that the ear has to work with.  The information needed to isolate the speaker response from the room contribution is contained within a measurement taken at the listening position.  It just needs to be extracted with the right analysis.  Smoothing the frequency magnitude response as is typically done in measurement plots is the wrong kind of analysis.  Measuring up close is another way to get that information, but it only works as long as there are no near-field effects, which mostly limits its usefulness to high frequencies.

 

So in principle, you should be able to calibrate using measurement data at the seats.  The hard part is figuring out which analysis is most consistent with the processing that the ear and brain are doing.

 

Something else I've been confused about lately though is whether mic calibration files are valid for "close mic" HF measurements.  If you look at the calibration data and the distance they are taken from, that distance definitely impacts the amount of HF roll-off because of how the air effects the upper end response.  Anyway, that's off topic and a discussion for another thread. 

 

The simple answer is:  Yes, the mic calibration files are valid for both close and distant mic measurements.  If the calibration was done correctly, then it should be valid at all distances.  You *want* your measurements to reflect the effects of distance on the UHF response.  The roll-off of UHF is a real physical thing.

 

The more complicated answer is that almost all measurement mics are also quite directional in the UHF.  You should have calibration files for multiple mic angles, and you should use the calibration file that is consistent with the direction the mic is pointed, relative to the speaker.  When measuring up close, almost all the energy arriving at the mic is directly from the speaker, so the calibration will be completely accurate.  But when measuring at a distance, some room energy will arrive from other angles, and depending on the mic angle and calibration file used this reflected energy may be under or over emphasized in the measurement.

 

Quick question though. When you say you calibrate to -20 dBFS which ends up being about 85dBC, is that with an input headroom of 105db for the L/C/R and 115db for the sub?  

 

I don't know what you mean by input headroom.

 

I would also add that 105 dB and 115 dB are actually kind of arbitrary figures for max output from L/C/R and sub with reference level content.  Don't forget that the pink noise calibration signal is continuous and substantially energizes the room.  So it's not really accurate to say that cinema standards allow for 105 dB peaks on each channel or 115 dB peaks for LFE (not sub).  On a realistic cinema system in which a fair amount of reverb is present, digital full scale peaks may not reach those SPLs at all because the peak level measurements won't include the build-up of room energy.  Hypothetically speaking, if you created 0 dBFS RMS pink noise sample that's band-limited to 500-2000 Hz, then that pink noise would measure at 105 dB when played on each channel of a cinema system, but as you can imagine, program material looks nothing like that kind of signal.

 

Another thing is that cinema SPL calibration uses band-limited signals and is largely ignorant of what happens outside the pass band.  It is the X curve standard that prescribes that the system response be EQed to the X curve target using band-limited pink noise.  If we ignore the X curve, it's entirely possible for -20 dBFS pink noise to measure quite a bit higher or lower if it's limited to bands other than the 500-2000 Hz for the calibration.



#39 Bossobass Dave

Bossobass Dave

    Bass Overlord

  • Members
  • PipPipPipPipPipPipPip
  • 914 posts

Posted 07 March 2017 - 10:37 PM

No, you misread.  The 3-5 dB boost sounds balanced with most music and many movies too.  Only with a minority of content does that boost sound excessive.

 

Not sure how anyone would know if content sounds bloated or whatever adjective means the sub is hot? I think that's an impossibility.

 

Also, human hearing does not operate with regard to frequency alone.  Crucially, hearing has a time aspect to it.  So does the response of a speaker in a room.  Using the original unsmoothed frequency and phase response data, a time-frequency transform can be performed to reveal the response of the speaker and room in terms of time and frequency together.  This is entirely analogous to using Spec Lab to visualize auditory content in movies.  Without the time aspect of the in-room response, it's hard to gain insight into how your speaker and room are actually affecting what you hear.  Frequency response smoothing almost always degrades or destroys this information.

 

Do you have any examples without words?



#40 SME

SME

    Super Bass Overlord

  • Members
  • PipPipPipPipPipPipPipPip
  • 983 posts

Posted 08 March 2017 - 12:15 AM

 

No, you misread.  The 3-5 dB boost sounds balanced with most music and many movies too.  Only with a minority of content does that boost sound excessive.

 

Not sure how anyone would know if content sounds bloated or whatever adjective means the sub is hot? I think that's an impossibility.

 

Also, human hearing does not operate with regard to frequency alone.  Crucially, hearing has a time aspect to it.  So does the response of a speaker in a room.  Using the original unsmoothed frequency and phase response data, a time-frequency transform can be performed to reveal the response of the speaker and room in terms of time and frequency together.  This is entirely analogous to using Spec Lab to visualize auditory content in movies.  Without the time aspect of the in-room response, it's hard to gain insight into how your speaker and room are actually affecting what you hear.  Frequency response smoothing almost always degrades or destroys this information.

 

Do you have any examples without words?

 

 

How would one know if the content sounds bloated?  By using one's ears.  I realize that evaluation using one's ears is kind of heretical on a forum like this.  I can't blame anyone for being skeptical of claims that listening can be used to evaluate sound quality in any objective sense, given that so much of the audio industry is completely fraudulent and successful in so far as placebo effect influences subjective judgments.  However, until we have a complete experimentally validated model of human hearing, our ears remain the best judge of what things sound like.

 

The trouble with the whole audiophile industry is people's obsessive focus on the subtle.  "Did those new speaker cables make a difference?"  "Yes, of course.  It was subtle, but it definitely made a difference."  Uh huh.  I'm far more interested in aspects of audio that are *not subtle*.  If dialog sounds too muddy, that is not a subtle problem, especially it prevents you from understanding what is being said.  If the individual notes in a bass-line can't be discerned because of too much deep bass from the subwoofer, then that is *not subtle*.  I can demo this on my system for any interested listener who can come here to Denver.  I can shelve the bass from 50 Hz and below by +2 dB, and switch the filter in with only a fraction of a second of gap while a song is playing.  A bass line that was once snappy and a kick drum that once punched the chest turn into mush and rumble.  That's *not subtle*.  I'm convinced that everyone who heard the difference would agree that the latter sound was muddy and inferior.

 

You want some pictures?  Check out the first two frequency response plots I posted under the "Speaker / Room Calibration" section in the first post of my system/room thread.  One is processed with 1/3rd octave smoothing and the other with 1/3rd octave FDW.  The latter is a much more accurate approximation of the first arrival sound coming from my speakers and is also much closer to what I think my speakers sound like.  I don't have any good time-frequency plots to show, for a variety of reasons, chief among them is the difficulty of visualizing everything that's relevant to hearing in a single plot.  I believe this to be a problem with Spec Lab spectrograms as well.  The choice of window size always involves a compromise between time and frequency resolution.  To really understand what's going on, you need to view the data with a variety of window sizes.






0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users