Audibility of subwoofer distortion

[SME]

This thread was started in response to a series of posts made here.

 

If a bass player is sitting in your sweet spot listening to a recording of a tune in which he wrote and played the bass line and says, "Now THAT'S what playback of my bass should sound like", you gonna tell him he's full of shit and drone on about THD and equal loudness studies done in the 50s? Probably, and good luck with that.

 

Hell no!  I would thank her for the complement and extoll the benefits of listening on a flat, low distortion system and having good room acoustics.  Either the recording was mixed and mastered on a flat, low distortion system in a room with good acoustics by competent engineers who produced the track with careful attention to the musician's preference, or the artist just got lucky and ended up with a recording that sounds good anyway.  Now, if I am the mix or mastering engineer and the artist is telling me she likes the sound with "100% THD" added, then I'll added it to her track.  I won't add 100% THD to my signal chain!

 

Here's a link to an interesting exercise that focused on HD by Axiom done a decade ago. Not saying I agree with the study methodology or results, just saying it's interesting and worth a read:

 

http://www.axiomaudio.com/distortion

 

I have read the article at least a few times and even went through the trouble to thoroughly debunked it in a post here.  To summarize, they tested the audibility of pure tones against a broadband background.  This is totally different than testing the audibility of distortion arising from the nonlinear response of a system to a broadband source.  This design error leads to results that are actually contrary to their stated conclusions.  There's other nasty sleight of hand here too.  They are very inconsistent with their frequency axis scale.  It's neither linear or log.  They actually don't present any data points between 1 and 5 kHz (over two octaves!) where the ear is actually the most sensitive.  Was this omitted by accident?

 

Your example of 5dB room gain, etc., is purely hypothetical. I deal in reality. When the <20 Hz part of the effect hits, I immediately know it for 2 reasons. One, I'm acclimated to the experience. It's a recent phenomenon that a human can routinely experience playback of recorded source <20 Hz. Since there is no frame of reference, it's a learned experience. Two, it isn't because my infrasonic hearing is abnormally acute. The reaction of my environment is the cue, not human hearing of steady state sine tones.

 

So what.  I'm talking not talking about the audibility of the fundamental tones against the distortion.  As you have said, the environment gives you cues that help you to better perceive the lower fundamentals that might otherwise be masked.  I'm talking about the audibility of the distortion against the signal that induced it.

 

I chose 20 Hz in my example because that's an area where I'd expect distortion to be especially audible because (1) room gain is not as high there as it is down lower; and (2) the equal loudness contours are very steep there.  The equal loudness contours (ELCs) are steeper (dB for dB) in the 20-80 Hz than typical room gain curves implying that the same percent distortion is actually more audible at lower frequencies.  Evidence suggests that once you get to 10-15 Hz or so, the ELCs start to flatten out again.  Even though ELCs are based on sine tone tests, similar trends exist with real world material.   The study of these trends has led to important technologies like lossy audio data compression algorithms that can effectively throw out 75% or more of the information content while retaining enough information to sound surprisingly close to the original.  What's lacking is detailed study of masking in the lowest frequencies, so the ELCs are probably the best thing we have to go by.  Not coincidentally, the information content down there is much lower than in the rest of the spectrum, so such studies wouldn't have much to contribute to lossy digital audio data compression technology anyway.

 

Your proof in the form of spectrographs is not black and white at all.  It is very hard to see where distortion is occuring the way your data is presented.  If you subtract the digital data from the miced data, we can at least see the difference more easily, even though we still can't assess nonlinear distortion separately from room effects.  To do this right is a bit ugly.  We want to consider how the room amplifies or attenuates the distortion relative to the signal because this affects the distortion at the listening position.  However, we wish to exclude the effects of the room on the signal itself because the room effects are mostly linear and are independent of the distortion being produced by the system.  In both of your pictures, the response over 60 Hz deviates quite a bit from the digital version.  How do I, you, or anyone else know which differences have to do with (linear) room effects and which have to do with nonlinear distortion in the system?  A better approach that might work (I haven't tried it) is to use a sine sweep measurement at a much lower level to determine the room response under the assumption of negligible distortion.  Then, deconvolute the miced response with the room response before subtracting the digital data to determine the distortion.  Finally, convolve the distortion with the room response again to consider the effect the room has on the distortion.

 

FWIW, if we apply the ELCs theory to the 10 Hz fundamental and the distortion products at 40 Hz and 60 Hz separately, we find that each harmonic likely sounds as loud or louder than the 10 Hz fundamental.  In other words, the psycho-acoustic distortion is over 100%.  To reach that result, I extrapolated the 60 phon curve to 120 dB @ 10 Hz, which accounts for some of the flattening that's observed to occur in the low teens.  The 40 Hz component is at -22 dB (8%) which is 98 dB SPL.  Note that 98 dB SPL falls well above the 60 phon curve, meaning that that component can be expected to actually sound louder than the fundamental.  The 60 Hz component is at -30 dB (3%) or 90 dB SPL, and again we see that 90 dB SPL falls well above the 60 phon curve.  No, that distortion won't necessarily mask the 10 Hz fundamental, but I would be surprised if that distortion wasn't audible.  I would expect the perceived timber of the sound to change noticeably when run at a somewhat lower playback level.  Yet another problem is if a distortion component occurs where the room has a strong peak, then the distortion will be strongly amplified even if that peak is EQed out. 

 

I'm not saying that this distortion will always be objectionable to the listener, but of course many people out there routinely clip and distort the signals in their subwoofers or playback chain without objecting to it or even realizing that there's anything wrong.  Often this is solely due to ignorance, and just about anyone who hears the difference and appreciates accurate sound reproduction will make a reasonable effort to minimize such distortion.  I've been there and done that, having learned of the output limitations of my AVR sub-outs the hard way, and from what I'm seeing, others on this forum including yourself have been through that experience as well.  Why argue that distortion is irrelevant when it absolutely is relevant?  Why not concede that distortion is relevant, even below 20 Hz?  If distortion under 20 Hz didn't matter, then why don't we just sacrifice linearity completely and build woofers that are largely indestructible.  Forget CEA.  Maximum output rules!  In reality, the dose makes the poison.  So what dose of distortion is acceptable at a given frequency?

[Shredhead]

 

Hell no!  I would thank her for the complement and extoll the benefits of listening on a flat, low distortion system and having good room acoustics.

Bass player... her? Nope.

[Bossobass Dave]

This thread was started in response to a series of posts made here.

 

 

Hell no!  I would thank her for the complement and extoll the benefits of listening on a flat, low distortion system and having good room acoustics.  Either the recording was mixed and mastered on a flat, low distortion system in a room with good acoustics by competent engineers who produced the track with careful attention to the musician's preference, or the artist just got lucky and ended up with a recording that sounds good anyway.  Now, if I am the mix or mastering engineer and the artist is telling me she likes the sound with "100% THD" added, then I'll added it to her track.  I won't add 100% THD to my signal chain!

 

 

 

I'm saying you wouldn't know either way. How the hell do you know what it supposed to sound like? You don't. Arrogance rules with golden ears.

 

THD is irrelevant.

 

Low E string of a 4 string bass guitar:

 

 

That's before processing, mixing and mastering, which includes an infinite universe of distortion possibilities.

 

Low frequency effect:

 

 

Whoa, do you see that 50% THD of 11.3 Hz!!?? Man, that's gonna sound like shit. Oh, wait... never mind.

[Bossobass Dave]

Here's a 20 Hz ported sub vs Raptor. Same room, same placement, same mic placement, same source.

 

Please scrutinize them for harmonic distortion of any kind at any frequency while you ignore the 900 pound gorilla riding the elephant in the room and make the case for the audibility of harmonics.

 

[maxmercy]

I'll bite.  The SpecLab graphs are probably not the best way to show harmonic or IM distortion.  But make no mistake, it is audible, or at least I found it annoying.  I would never have highpassed my old horn subs if that distortion was inaudible.  I have heard several people's "HT's" with undersized sealed subs that were distorting away while the owners thought it was a terrific experience.  My wife would even notice.

 

Unfortunately, I have no graphs or pics to post, but ELC (assuming no aural masking) predicts that THD from ULF signals should be definitively audible.  Assuming 2nd and 3rd harmonics will be the highest power ones, signals between 10-20Hz would be the hardest to reproduce cleanly given ELC curves.  Once under 10Hz, the 3rd harmonics begin to reach ranges which require more dB for the same perceptive loudness.  Under 5Hz, low-order THD is nearly inaudible, as the threshold of detection rises quickly.

 

Room gain really helps with that 10-20Hz region, decreasing distortion.

 

JSS

[Bossobass Dave]

I make no claims regarding subwoofers that aren't up to the task. Pushing hardware well beyond it's capabilities is a fool's errand and deserves no explanation. That includes those who believe their ported sub has "useful output to well below tune".

 

AVTalk member Slartibartfast put this subject into proper perspective back when. Upon request, he measured FR and THD indoors at his seat and outdoors @ 2M GP for direct comparison. We have the GP FR, GP THD result, in-room FR and in-room THD result:

 

 

THD is harmonic distortion AS A %. Notice how room gain completely eliminates THD below 30 Hz, but also notice the huge spikes where there is a dip in response because the fundamental is lowered by the dip but the harmonics are not and may even be boosted by a peak.

 

So, all those sports fans who are mesmerized by the THD graphs <30 Hz are completely chasing a ghost while ignoring the much more audible harmonics >30 Hz caused by the frequency response distortion at their seats.

[3ll3d00d]

Are those measurements from a single dd18? Do you have the frequency response for each output level?

There is no doubt that any distortion is amplified by the room though so the audibility of that distortion will end up being room specific. Put another way, the worse the native response of your room, the more important it is not to overload your sub(s) (or correct the response with multiple subs). My room shows this v well as it looks v ugly due to positioning constraints (will post a graph later). The room shown above has quite small variation (+8dB peak) in comparison to mine so the corresponding THD measurements are similarly muted.

I don't see how this equates to sub 30Hz distortion being unimportant though (if that is what you mean).

[MKtheater]

Distortion is very important except what was said is that the distortion from outside will not matter in room because in room the sub will be using so much leads power to hit the same SPL that the distortion will be low. I believe this started when people were saying the 24 inch driver was not as good as the LMS because it had twice the distortion but people don't realize it was 6 dB higher as well. You play them in room and the distortion is so low on both that the difference becomes the higher playback ability of the 24.

[3ll3d00d]

Here's an example from my room, dual UXL-18 in 190L powered by a speakerpower SP1-6000. This was an initial sweep so sub not fully setup but enough to illustrate I think.

 

No EQ response, 1st axial mode is a ~22dB peak

 

 

My estimate of room gain shows v little gain until ~20Hz.

 

 

95dB sweep with some EQ to take out that peak

 

 

105dB sweep with some EQ to take out that peak

 

 

the 40Hz peak amplifies what 2nd/3rd/4th HD there is as expected & the shape tracks that of the 40Hz peak. If you look at the level of 2HD below that peak so at ~12-13Hz then you get an idea that this isn't the sub doing that much but it is working harder. The distortion rises here by about 7.5dB more than the fundamental, similarly the 2HD peak at 20Hz rises by 7.5dB more than the fundamental. This suggests the sub itself is going from 1% to 2% distortion but the room takes that and turns it into a rise from about 6% to ~15%.

 

Obviously the stats will look worse as the output level goes up so say the next sweep produced 6% from the sub then the room would turn that into 30% 2HD around 20Hz. 30% sounds bad, 6% not so much.

 

As to audibility of this in my room, it's impossible for me to be sure (as I don't have a way to reduce that distortion at present) so I can only compare my current sub against my old sub. My current sub is pretty flat down to ~115dB at ~8Hz in a sweep whereas my old one could do the same to ~19Hz so I've added about an extra octave. The new one doesn't sound much different for normal content (e.g. music) but has a noticeable heft/weight with certain content, whether that heft is augmented by the room amplified HD is impossible to say but it doesn't sound "unclean" to my ears (though honestlyI don't have enough experience of other such systems to really be able to judge that).

 

I probably should repeat these now that I've finalised this setup to see what it looks now at this point.

[SME]

I'm saying you wouldn't know either way. How the hell do you know what it supposed to sound like? You don't. Arrogance rules with golden ears.

 

THD is irrelevant.

 

Low E string of a 4 string bass guitar:

 

 

That's before processing, mixing and mastering, which includes an infinite universe of distortion possibilities.

 

Low frequency effect:

 

 

Whoa, do you see that 50% THD of 11.3 Hz!!?? Man, that's gonna sound like shit. Oh, wait... never mind.

I don't know what it's supposed to sound like, but the more accurately (defined in a psycho-acoustic sense) my system reproduces the sound or music the more likely what I'm hearing is "what it's supposed to sound like".  Thus far, the closer I've gotten things to "how it's supposed to sound", the more detail, clarity, and immersion I experience.  The tricky part is figuring out what is psycho-acoustically relevant.

 

You seem to have missed the fact that I make a major distinction between THD (or really, just HD because the orders of harmonics are a big deal) measured in playback equipment using sine sweeps and HD that is present in the content that gets played through the equipment.  The two are completely different.  The goal in producing music or sound effects is to make something that sounds cool.  How much distortion is too much is purely an aesthetic concern, and in fact, I routinely enjoy content with sounds that have been distorted quite extremely.  The goal in the playback system is to accurately reproduce the content, and accuracy demands consideration of (among other things) non-linear sources of distortion and the psycho-acoustics involved.  Non-linearity causes HD in pure tones and sets of harmonically related tones and also causes intermodular distortion (IMD) when multiple non-harmonically related tones are present.  The IMD is more objectionable, but both forms of distortion potentially rob the presentation of its accuracy.

 

Here's a 20 Hz ported sub vs Raptor. Same room, same placement, same mic placement, same source.

 

Please scrutinize them for harmonic distortion of any kind at any frequency while you ignore the 900 pound gorilla riding the elephant in the room and make the case for the audibility of harmonics.

 

I'd rather not, actually.  As I already said, the spectrogram is a poor tool for visualizing non-linear distortion unless you're just playing pure tones.  In my previous post, I suggested an analysis method that could present distortion in a much more effective way.  I may actually give this a shot when I get my subs upgraded as a method for finding their in-room limits and optimizing accordingly.  I'll post some pretty pictures here if/when I get around to it.

 

Oh, and the whole sealed vs. 20 Hz port sub comparison is kinda silly being that this conversation started with THD versus frequency comparisons of two sealed systems.

[SME]

Unfortunately, I have no graphs or pics to post, but ELC (assuming no aural masking) predicts that THD from ULF signals should be definitively audible.  Assuming 2nd and 3rd harmonics will be the highest power ones, signals between 10-20Hz would be the hardest to reproduce cleanly given ELC curves.  Once under 10Hz, the 3rd harmonics begin to reach ranges which require more dB for the same perceptive loudness.  Under 5Hz, low-order THD is nearly inaudible, as the threshold of detection rises quickly.

 

Room gain really helps with that 10-20Hz region, decreasing distortion.

This is along the lines of what I'm thinking.  That is, the 10-20 Hz range is often the most important due to a confluence of factors:

1. Hearing sensitivity drops off faster than room gain makes up for, according to ELCs.
2. Any signal shaping in use may demand more out of the sub here than higher up.
3. Native subwoofer distortion frequently increases here also.
4. A room response dip is often present in this range, according to here.

And yes, without room gain, the situation would be far worse.  First of all, the fundamental would be a lot quieter relative to the overtones, and second it would take more aggressive boosting to achieve a flat response.  It's interesting the room gain seems to enter the picture twice.

[Bossobass Dave]

Are those measurements from a single dd18? Do you have the frequency response for each output level?

 

There is no doubt that any distortion is amplified by the room though so the audibility of that distortion will end up being room specific. Put another way, the worse the native response of your room, the more important it is not to overload your sub(s) (or correct the response with multiple subs). My room shows this v well as it looks v ugly due to positioning constraints (will post a graph later). The room shown above has quite small variation (+8dB peak) in comparison to mine so the corresponding THD measurements are similarly muted.

 

I don't see how this equates to sub 30Hz distortion being unimportant though (if that is what you mean).

 

Those measurements are from a single DD-18. although it doesn't matter, does it? People push their hardware beyond it's "clean" GP limits regardless of the number of subs they own, making that question irrelevant. Slarti ran the same levels inside as outside. If you can't see how that simple experiment makes THD <30 Hz measured at GP outdoors unimportant then I really have nothing more to say on the subject. Maybe you can show me how THD measured outdoors at GP is important, if that's what you mean?

 

 

As to audibility of this in my room, it's impossible for me to be sure (as I don't have a way to reduce that distortion at present) so I can only compare my current sub against my old sub. My current sub is pretty flat down to ~115dB at ~8Hz in a sweep whereas my old one could do the same to ~19Hz so I've added about an extra octave. The new one doesn't sound much different for normal content (e.g. music) but has a noticeable heft/weight with certain content, whether that heft is augmented by the room amplified HD is impossible to say but it doesn't sound "unclean" to my ears (though honestlyI don't have enough experience of other such systems to really be able to judge that).

 

I probably should repeat these now that I've finalised this setup to see what it looks now at this point.

 

If it's impossible to be sure of the audibility of harmonic distortions in your room, that's a pretty darned good indication that harmonic distortions in your room are unimportant.

 

Look at the speclab cap of the low E string of the bass guitar. If the musician were in your room (or in anyone's room) playing that low E string, the room would affect the sound the same as a recording of that low E string being played back, with one difference... you can change the position of the point source of the sub that's playing the recording as well as the frequency response, but the live bass player is stuck wherever he is with whatever FR is hitting your ears.

 

Since the instrument already contains over 100% "Harmonic Distortion" to at least 6HD, how will the room affect the sound? As a bass player who's played in every conceivable venue, including gandma's living room, I can tell you the problem is the room, not the HD from your dual UXL-18 + SP-6K system. Harmonic distortion (of the in-room amounts we're discussing from subwoofers such as your own) is irrelevant.

 

Proof to the contrary is welcome.

 

I don't know what it's supposed to sound like, but the more accurately (defined in a psycho-acoustic sense) my system reproduces the sound or music the more likely what I'm hearing is "what it's supposed to sound like".  Thus far, the closer I've gotten things to "how it's supposed to sound", the more detail, clarity, and immersion I experience.  The tricky part is figuring out what is psycho-acoustically relevant.

 

You seem to have missed the fact that I make a major distinction between THD (or really, just HD because the orders of harmonics are a big deal) measured in playback equipment using sine sweeps and HD that is present in the content that gets played through the equipment.  The two are completely different.  The goal in producing music or sound effects is to make something that sounds cool.  How much distortion is too much is purely an aesthetic concern, and in fact, I routinely enjoy content with sounds that have been distorted quite extremely.  The goal in the playback system is to accurately reproduce the content, and accuracy demands consideration of (among other things) non-linear sources of distortion and the psycho-acoustics involved.  Non-linearity causes HD in pure tones and sets of harmonically related tones and also causes intermodular distortion (IMD) when multiple non-harmonically related tones are present.  The IMD is more objectionable, but both forms of distortion potentially rob the presentation of its accuracy.

 

I'd rather not, actually.  As I already said, the spectrogram is a poor tool for visualizing non-linear distortion unless you're just playing pure tones.  In my previous post, I suggested an analysis method that could present distortion in a much more effective way.  I may actually give this a shot when I get my subs upgraded as a method for finding their in-room limits and optimizing accordingly.  I'll post some pretty pictures here if/when I get around to it.

 

Oh, and the whole sealed vs. 20 Hz port sub comparison is kinda silly being that this conversation started with THD versus frequency comparisons of two sealed systems.

 

Accuracy and harmonic distortion are mutually exclusive. Save for sitting in your HT space playing pure sinusoidal tones, you have posted no evidence to the contrary. And, again, the harmonics relative to any given instrument are what make a piano playing the low E string easily distinguishable from the upright bass playing the same note. I would say that when that distinction becomes too muddied or lost, then distortion is too high.

 

Oh, and the 20 Hz ported sub comparison is "silly", but accuracy of playback is the crux of the debate? Silly me.

 

Paul Klipsch said (paraphrased): "If it measures bad and sounds good, you're measuring the wrong thing. If it measures good and sounds bad, you're measuring the wrong thing."

 

If accuracy in playback were so paramount, then house curve, ported subs, passive radiator subs, horn subs, bandpass subs, running subs hot, 6,000W amplifiers tripping shut down protection circuitry, power line sag and things of this sort that represent grotesquely distorted playback would be top priority with HD being way, way, way down the list.

 

Like James said above, saying that the HS 24 is not as good as the LMS-U because it showed higher HD in a corn field is not just silly, it's a heaping, steaming pile of bovine excrement. Proof to the contrary is welcome.

[3ll3d00d]

Those measurements are from a single DD-18. although it doesn't matter, does it? People push their hardware beyond it's "clean" GP limits regardless of the number of subs they own, making that question irrelevant. Slarti ran the same levels inside as outside. If you can't see how that simple experiment makes THD <30 Hz measured at GP outdoors unimportant then I really have nothing more to say on the subject. Maybe you can show me how THD measured outdoors at GP is important, if that's what you mean?

No that's not why I asked, I was just curious to what extent the sub was compressing as the level went up. It's a shame, for the historical record as it were, that those avtalk tests are lost to the web.

 

I wouldn't say outdoor THD data is completely useless though, it at least lets you compare driver performance in a somewhat apples to apples fashion. I don't think it has much to say about in room performance though.

[Bossobass Dave]

No that's not why I asked, I was just curious to what extent the sub was compressing as the level went up. It's a shame, for the historical record as it were, that those avtalk tests are lost to the web.

 

I wouldn't say outdoor THD data is completely useless though, it at least lets you compare driver performance in a somewhat apples to apples fashion. I don't think it has much to say about in room performance though.

 

Yes, all things must pass, but it's really a sad note for me. I really enjoyed the differences in CJ's/Slarti's posts and the extra-curricular data he provided was golden. You'd think someone would have offered to host the data.

 

I just think in-room data is far more relevant. He did the in-room vs GP HD comparisons and group delay as well. That stuff was a real eye-opener for me. Apparently, not so much for the rest of the HT enthusiast crowd.

 

The outdoor GP data is a comparative tool. If THD happens to be your personal penchant, comparing driver A to driver B is helpful, but, as I've mentioned many times before, different driver parameters vs same box volume for both drivers and the monstrous amp on the monstrous mains in the skillful hands of Josh are serious considerations when making those comparisons.

 

For example, people negatively compared the HS-24 to the LMS-U, GP, citing much higher measured THD. One might reason that if one LMS-U has lower THD then a pair of them would have even lower THD. But, the HS-24 utterly dominated the pair in-room with same room, source, signal chain and amplification. This indicates to me (actually, it's pretty solid proof, IMO) that a flat Qtc is universally preferred to a lower-than-critically-damped Qtc and THD is irrelevant.