I agree. Duke (AudioKinesis) did this on one of his better systems, basically incorporating a built-in flanking sub. Great idea, in my opinion.
I thought it was all down to #2. I'm not so sure about the "dipole action". It is only on the downslope (running down the 24dB per Octave falloff) where the front and back contributions would be similar in strength and out of phase. You are getting differences well above that range.
David S.
Sorry about the delay, but I was flying back from China and that takes awhile all things considered.
No, I have not used ports on my speakers in a long time. I did once but ended up delateing them as they served no purpose.
I don't follow your last sentence, but maybe thats because your assumption was incorrect and then maybe this sentence is incorrect and not clear.
Basically on the "room gain" argument, Marcus was completely correct. Defining room gain to include all modes is not how it was initially defined. As with so many terms in Audiophileland, definitions seem to change to suite the writer.
And the bandpass subs are all monopoles, so they can pressurize a room. There are three of them and there is no "room gain" - there is nothing to pressurize.
I'm surprised that you would think that I missed the lack of a dipoles ability to pressurize a room since I wrote about this back in the early 80's, nearly 30 years ago. Thats precisely why we used closed boxes in all our car audio systems. The closed box could excite the "room gain" (it acually exists in cars), but a ported box could not.
Last edited:
I could be wrong, but I think he was referencing your use of Bandpass subs. (..and I'm not sure that bandpass designs would apply at really low freq.s in this context.)
BTW, oddly enough the first time I read use of the phrase "room gain" was from Martin Colloms (decidedly an audiophile personality) - and I think he made a point of distinguishing room gain from modal gain.
Ah yes. The BP woofer only has output from the port so RP is possible.
I will say one thing, I have tested dipole, monopole, and cardioid woofers in my room. All were designed to have exactly the same free field response. The dipole woofer rolled off below my room's fundamental much like the free field response. The monopole and cardioid clearly showed room pressurization effects with boosted output below the woofer cut off and room fundamental. Not than any of this really matters. I spend more time listening to music than worring about floor bounce or wall bounce. Just move thge speakers around until it sounds good at the listening position. I'm not looking for wide seating.
gornir's post here is good illustration of the frequency response trade offs for on and off axis for 2-ways, and why a blind adherence to flat on axis is often a bad idea:
http://www.diyaudio.com/forums/mult...dspeaker-measurement-study-4.html#post2561697
Dave
http://www.diyaudio.com/forums/mult...dspeaker-measurement-study-4.html#post2561697
Dave
DANGER DANGER DANGER
To NOT use the term "room gain" or persons from car audio (all of whom, I believe own tiny well-sealed cars) will pop up in this thread and CUT TO PIECES anyone who denies that "room gain" isn't the most important factor in bass response in the history of audio reproduction.
DAMHIK.
Joking aside, I don't think anyone can object to me saying that room gain in almost any reasonably sized, reasonably designed/constructed music room, is an insignificant factor at all but the very lowest frequencies, at which it is a very minor factor.... unless you play music inside a giant gas pressure vessel.
Even if you could assess the influence of room gain in your room at one time with the doors and windows in one state of closedness, you couldn't deal with the effects any too well either theoretically or practically for other conditions.
To NOT use the term "room gain" or persons from car audio (all of whom, I believe own tiny well-sealed cars) will pop up in this thread and CUT TO PIECES anyone who denies that "room gain" isn't the most important factor in bass response in the history of audio reproduction.
DAMHIK.
Joking aside, I don't think anyone can object to me saying that room gain in almost any reasonably sized, reasonably designed/constructed music room, is an insignificant factor at all but the very lowest frequencies, at which it is a very minor factor.... unless you play music inside a giant gas pressure vessel.
Even if you could assess the influence of room gain in your room at one time with the doors and windows in one state of closedness, you couldn't deal with the effects any too well either theoretically or practically for other conditions.
Last edited:
Hmm,
So if the conclusions of that thread are to be believed we have yet another cookie cutter 2 way design with a dome tweeter that sounds wrong when equalized flat on axis.
No big surprise to those following this thread ? Surely it suggests the power response is the main culprit, and most likely the sudden increase in power response above the crossover point ?
What's particularly telling in this case is that the design is actively crossed over with a linear phase digital crossover, where the author is using 60dB / octave or more slopes. (96dB/oct is also suggested as an option, and the unit itself can support slopes up to a ridiculous 300dB/octave)
To my mind that's the single biggest flaw of the design (albeit one that can be easily corrected in this case at the press of a button by changing the slopes of the digital filter) and I shudder to think how discontinuous the power response will be when crossed over at 60+dB/octave.
In skimming through the thread it doesn't appear the author has tried listening tests with the same drivers crossed over at more conventional 12, 18, and 24dB/octave slopes.
I might be in the minority but I honestly believe that with rare unusual driver specific exceptions, any speaker that needs a crossover with a slope of greater than 24dB/oct is indicative of drivers which either have insufficient overlap or serious out of band problems. (obnoxious resonances etc)
With a simple dome tweeter and small midbass driver like that anything steeper than 24db/octave is just "guilding the lilly", and only serves to worsen the discontinuity in the power response that is somewhat smoothed by the gradual transition of a lower order conventional filter.
Just because modern digital filters have the capability to produce up to 300dB/octave slopes at the press of a button doesn't mean it's a good idea to do so 🙄
I think all that thread serves to illustrate is that the discontinuous power response between a dome tweeter and a cone driver is a significant problem and making the crossover unusually steep serves to accentuate the problem.
I've mentioned it before in this thread but I don't slope or shelve down my ribbon tweeters - I aim for as flat as possible measured on axis response between the ribbon and midrange driver and the mid/treble sounds perfectly well balanced to me both at near ~2 metre listening where I'm largely in the direct field region as well as at 3+ metres where I'm largely in the diffuse field of the room.
Then again the short wave-guide in the ribbon controls the directivity compared to a flush dome tweeter - horizontal directivity (-6dB) measures almost exactly a constant 90 degrees from 2Khz to 10Khz aside from a slight hump increasing to about 110 degrees over 5-6Khz, and a reduction from 90 to about 60 degrees from 10 - 20Khz. Vertical dispersion narrows considerably in the vertical plane in the top octave as well.
I believe this controlled directivity especially at the bottom end of the treble range serves to avoid power response bloom above the crossover frequency as well as avoid excessive overall levels of treble in the reverberant field, whilst still keeping a dead flat on axis response.
Why more designs don't make use of controlled directivity tweeters is beyond me - it really does seem the correct solution for a large part of the whole "flat on axis is not correct" conundrum, and as a nice bonus it drastically reduces tweeter cabinet edge diffraction problems (to the point where it's a non-issue) and improves mid-tweeter time alignment as well.
Last edited:
Here's a paper on the perception of bass decay. Maybe, just maybe Dr. Geddes is not crazy. http://www.genelec.com/documents/publications/aes116th_2.pdf
Interesting,
Dan
Interesting,
Dan
Yes, quite interesting.
So to summarise - for bass below 100Hz only flatness of amplitude response matters - even if significant ringing at room modes remains in the time domain after amplitude response correction it is imperceptible. (And by extension I assume this means it's ok to apply minimum phase PEQ correction below 100Hz even if the room response below 100Hz is not minimum phase, since perceptually only the amplitude response matters and residual ringing is not noticable)
Above 100Hz the time domain response (decay of resonances) suddenly becomes very important, at least as important as the amplitude response. For bass from 100-200Hz it's not enough to simply equalize the amplitude response flat, ringing in the time domain has to be dealt with as well.
Assuming this study to be accurate I can think of two interesting questions posed by these results, one speaker related, one room related.
1) Sealed boxes are often claimed to provide superior "quality" bass due to better transient response at the cut-off frequency compared to bass reflex systems which are sometimes criticised due to "ringing" near the box tuned frequency which is "inherent" in the design.
I'll leave aside the fact that a bass reflex design is a 2 dimensional matrix of possible alignments depending on the choice of box volume and tuned frequency, and that many of the possible alignments (especially over damped early roll-off ones) can have very good transient response, and therefore all bass reflex alignments shouldn't be tarred with the same brush. I will accept however that on paper a 2nd order system does still have a better transient response.
The results of this study seem to suggest that provided both the sealed and bass reflex designs have cut-off frequencies that are well below 100Hz, (always the case, excluding very small bookshelf speakers) it doesn't actually matter whether the bass reflex design rings slightly longer in the time domain than the sealed box as it will not be perceptible - all that matters is that the amplitude response is flat without peaking at resonance.
Yes, they studied room modes not speaker resonances, but they did so by adding electronically synthesised resonances, (after electronically cancelling the real room modes) so whether you synthesise a resonance electronically or add one mechanically in the speaker shouldn't matter, a resonance is a resonance, with a centre frequency and a Q.
Does this cast some doubt on the theoretical importance of faster resonance decay in sealed boxes vs bass reflex, or will those who are convinced sealed is inherently better in this way and that it matters, continue to believe so ?
I've always favoured bass reflex designs for efficiency / extension / headroom / distortion reasons, but I do tend to go for early gradual roll-off alignments that sacrifice some F3 on an efficient driver for a more gradual roll-off and a smaller cabinet, rather than going for a maximally flat alignment with an abrupt cut-off.
2) This study suggests that time domain decay of resonances above 100Hz (so basically 100-200Hz if we are considering bass) is very important and perceptible - it's not just enough to flatten the frequency response if decay signatures of resonances are still present in the time domain.
I wonder how this fits in with spatial diversity for solving the amplitude response issues in the 100-200Hz range. (Flanking subs etc) Will improving the amplitude response between 100-200Hz by providing additional spatially displaced helper woofers (rather than EQ) improve, worsen, or make no difference to the time decay signature of room resonances ?
Also, a lot of the response anomalies (perhaps the dominant ones, in some rooms) in the 100-200Hz range are non-resonant in nature - for example a boundary cancellation from the wall behind the speaker whilst causing peaks/dips in the response due to constructive/destructive interference, is not in itself a resonance, and won't have a continuous decay signature in the time domain. (Only repeated reflections back and forth between opposing surfaces will form a resonance, not a single bounce from a boundary...)
Would using EQ to try to correct boundary cancellation response anomalies make the amplitude response better at the expense of introducing resonances that show up in the time domain, making the time domain decay worse ?
For example if you had a peak in response at 160Hz that is caused by a boundary reflection behind the speaker, would using a PEQ with a negative gain to pull it down actually introduce a resonant decay in the time domain that wasn't there to begin with ? (Of course you then have the question of whether the improved amplitude response perceptually outweighs the worsened time domain decay - in other words flattening the amplitude response may still sound better, but not as good as improving both amplitude and time domains at once)
Would using a displaced helper woofer to mitigate the same boundary interference improve the amplitude response without worsening the time domain decay ? If so, this seems to be a good endorsement for flattening the frequency response between 100-200Hz through spatial diversity rather than applying EQ, as much as possible.
So to summarise - for bass below 100Hz only flatness of amplitude response matters - even if significant ringing at room modes remains in the time domain after amplitude response correction it is imperceptible. (And by extension I assume this means it's ok to apply minimum phase PEQ correction below 100Hz even if the room response below 100Hz is not minimum phase, since perceptually only the amplitude response matters and residual ringing is not noticable)
Above 100Hz the time domain response (decay of resonances) suddenly becomes very important, at least as important as the amplitude response. For bass from 100-200Hz it's not enough to simply equalize the amplitude response flat, ringing in the time domain has to be dealt with as well.
Assuming this study to be accurate I can think of two interesting questions posed by these results, one speaker related, one room related.
1) Sealed boxes are often claimed to provide superior "quality" bass due to better transient response at the cut-off frequency compared to bass reflex systems which are sometimes criticised due to "ringing" near the box tuned frequency which is "inherent" in the design.
I'll leave aside the fact that a bass reflex design is a 2 dimensional matrix of possible alignments depending on the choice of box volume and tuned frequency, and that many of the possible alignments (especially over damped early roll-off ones) can have very good transient response, and therefore all bass reflex alignments shouldn't be tarred with the same brush. I will accept however that on paper a 2nd order system does still have a better transient response.
The results of this study seem to suggest that provided both the sealed and bass reflex designs have cut-off frequencies that are well below 100Hz, (always the case, excluding very small bookshelf speakers) it doesn't actually matter whether the bass reflex design rings slightly longer in the time domain than the sealed box as it will not be perceptible - all that matters is that the amplitude response is flat without peaking at resonance.
Yes, they studied room modes not speaker resonances, but they did so by adding electronically synthesised resonances, (after electronically cancelling the real room modes) so whether you synthesise a resonance electronically or add one mechanically in the speaker shouldn't matter, a resonance is a resonance, with a centre frequency and a Q.
Does this cast some doubt on the theoretical importance of faster resonance decay in sealed boxes vs bass reflex, or will those who are convinced sealed is inherently better in this way and that it matters, continue to believe so ?
I've always favoured bass reflex designs for efficiency / extension / headroom / distortion reasons, but I do tend to go for early gradual roll-off alignments that sacrifice some F3 on an efficient driver for a more gradual roll-off and a smaller cabinet, rather than going for a maximally flat alignment with an abrupt cut-off.
2) This study suggests that time domain decay of resonances above 100Hz (so basically 100-200Hz if we are considering bass) is very important and perceptible - it's not just enough to flatten the frequency response if decay signatures of resonances are still present in the time domain.
I wonder how this fits in with spatial diversity for solving the amplitude response issues in the 100-200Hz range. (Flanking subs etc) Will improving the amplitude response between 100-200Hz by providing additional spatially displaced helper woofers (rather than EQ) improve, worsen, or make no difference to the time decay signature of room resonances ?
Also, a lot of the response anomalies (perhaps the dominant ones, in some rooms) in the 100-200Hz range are non-resonant in nature - for example a boundary cancellation from the wall behind the speaker whilst causing peaks/dips in the response due to constructive/destructive interference, is not in itself a resonance, and won't have a continuous decay signature in the time domain. (Only repeated reflections back and forth between opposing surfaces will form a resonance, not a single bounce from a boundary...)
Would using EQ to try to correct boundary cancellation response anomalies make the amplitude response better at the expense of introducing resonances that show up in the time domain, making the time domain decay worse ?
For example if you had a peak in response at 160Hz that is caused by a boundary reflection behind the speaker, would using a PEQ with a negative gain to pull it down actually introduce a resonant decay in the time domain that wasn't there to begin with ? (Of course you then have the question of whether the improved amplitude response perceptually outweighs the worsened time domain decay - in other words flattening the amplitude response may still sound better, but not as good as improving both amplitude and time domains at once)
Would using a displaced helper woofer to mitigate the same boundary interference improve the amplitude response without worsening the time domain decay ? If so, this seems to be a good endorsement for flattening the frequency response between 100-200Hz through spatial diversity rather than applying EQ, as much as possible.
Last edited:
Thats not what the paper says at all:
I tried to copy the conclusions to post here, but the doc is locked. Basically it says that below 100 Hz temporal resolution is zero and that above 100 Hz it is degraded from the midband down to 100 Hz. There is no "sudden" change and the region just above 100 Hz is not "very important", just "more" important than below 100 Hz, which is zero. Temperal resolution does not reach its most significant until above 500 Hz, peaking at about 2 kHz and then falling after that. This study completely supports that hypothesis.
Well, "suddenly" might have been a poor choice of words on my part, but I really think you're splitting hairs based on your interpretation of the word "suddenly". It's pretty unambiguous from the summary of the study that a loss of "temporal resolution" as you call it is pretty rapid below 100Hz, while much more gradual from midrange down to 100Hz. To me "suddenly" implies a rapid change in a trend, like a knee on a curve.
Since modal problems in rooms are only important below about 200Hz and we are talking specifically about bass, it doesn't really matter what our "temporal resolution" is at high frequencies. I'm sure it is greater at 2Khz, but it's not relevant to the discussion about bass.
From the summary:
"The experiment shows clearly that the human auditory perception of temporal details in the given conditions degrades rapidly below 100Hz. Such tendency may be slightly noticeable already from mid frequencies to 100Hz. It can be concluded that down to about 100Hz the temporal properties of modal decay are somewhat critical." (my emphasis)
[....]
"Below 100Hz, particularly at or below 50Hz even very long delays, up to 2 seconds and above may not be noticeable as far as the magnitude response is equalized well enough."
Where exactly does this contradict what I said ? In what way does it influence the two questions I posed ?
One thing it would do is correct the early response at the listening position at the expense of the power response. My subjective opinion is that trying to EQ a large boundary peak (floor or ceiling) makes things worse.
I am contemplating this myself at the moment as I push a mic around the room and watching the comb filtering change frequencies with distance.
Surely we can look at a helper woofer like an array. We can aim for a fortunate combination response at the listening position, and any incidental comb filtering/lobing created within the room will likely be dispersed, such that the power response would be relatively flat, when looked at over the span of the room, and spatially averaged.
Suddenly to me means discontinuos. I imagine that the resolution starts to degrade somewhere below 300-500 Hz, does so at a fairly slow rate and then continuosly increases this rate such that at 100 Hz it has become quite significant - the slope always being continuous and monotonic, but increasing steadily as the frequency drops. There is nothing in our hearing system that would cause an abrupt change with frequency in performance characteristics like this, particularly at LFs. But "sudenly" seems to mean something different to you. As long as the reader understands that it is not "abrupt", then I'm fine with it.
I found terms like "somewhat critical" in the paper to be ambiguous.
Last edited:
Poor choice of words I agree. You got me 😀 I didn't mean to imply discontinuous, just a rapid change in the slope. Another example would be if you plotted distortion vs excursion of a driver - up until xmax there will be a fairly linear increase in distortion with excursion and then "suddenly" the slope of the line will shoot up. It's not instantaneous (what is, in an analogue world) but the change in slope is very rapid departure from the prior trend.
Regardless of how good the hearing mechanism is, at lower frequencies you're going to be limited by the Time-Frequency uncertainty principle, "temporal resolution" inherently gets worse at lower frequencies. I wonder if this is sufficient to explain the falloff as you go down from midrange down to 100Hz, or whether our hearing system is still significantly worse than the theoretically obtainable best case.
Are you sure about that ? Depending on how tightly you define abruptly a fairly abrupt loss of temporal resolution between 100Hz and 50Hz is exactly what the report is suggesting.
I wonder if it's something quite simple though. Two things occur to me:
1) I notice the test was done at fairly moderate SPL's, where the bass notes would be relatively far down the fletcher munsen curves. This means that as the resonance decayed at a certain rate based on its Q the apparent decay would be much more rapid than it actually is due to the ear's rapid loss of bass sensitivity at lower SPL's - the decaying note would more quickly drop below the threshold of hearing at such low frequencies.
They actually bring up exactly this point in the article but perhaps don't give as much weight to it as they should have. Without reading the whole article again I don't remember them saying that they tested at more than the one playback level. Sensitivity to slow decay of low frequencies may be significantly greater at high SPL's.
2) Most typical bass notes have strong 2nd and 3rd harmonic overtones and it's these overtones that provide nearly all the "definition" and attack of such a bass note, so you could say most of the "temporal resolution" of the perception of such bass notes comes from their harmonics, and harmonic structure, not the fundamental.
Thus it's not surprising that a slow decay of the fundamental, accompanied by normal fast decay of the harmonics (assuming the harmonics didn't also fall on room modes) still sounds ok. The "fast" part of the bass note was always the harmonics in the first place.
I don't think they did any tests with a pure sine-wave as the bass instrument (as used in some electronic music) - that may have given quite different results, although generally a sine-wave bass instrument (below 80Hz or so) always sounds "slow" regardless of an absence of room modes - completely in agreement with their findings of a rapid loss of temporal resolution below ~100Hz.
Last edited:
What are the implications as far as the audibility of group delay in a reflex box? If I am understanding this correctly it seems to imply it would be a non issue with boxes tuned below 50 Hz as many are.
Rob🙂
Rob🙂
As I said, I found some of the wording ambiguos and there isn't really a graph of the effect with frequency, just some observations, and sketchy conclusions, so its hard to tell what the results actually show. Don't get me wrong, its a good piece of work, but I don't think it quantifies the effects to the point of settling our discussion.
Lower Mid smoothness
At midrange frequencies in a typical listening room, there may be many different reflection paths filling in each others cancellation frequencies, since each path mixed with the direct sound path (and all other reflection paths) will have cancellations at different frequencies. When you get down below 200 HZ, cancellations happen when a delayed path is several feet or more. Since there may be fewer substantially effective reflection paths in this frequency range, the cancellations may not get filled in as much by other reflections. I would think they would, but apparently they don't always somehow. It sounds like below 200 HZ, it may be best to have flankers that are several feet away from the main woofers and lower mid drivers, relative to room boundaries and listening position. I doubt if any temporal distortions below 200HZ will sound as bad as uneven amplitude in the lower mid/bass region. Even if it adds some coloration (another general word), it may be as much an enhancement as not. Maybe I should move my crossover frequency from 100 up to 200, and have four woofer boxes instead of two?
It's also interesting to me that our sensitivity to temporal distortions tapers off above 4kHZ. When I added a rear firing tweeter to my system, the cymbals sound way better, more 3-D and more present. Apparently each section (perhaps quarter) of the frequency range has a different set of dominantly effective psycho-acoustic properties and room acoustic mechanisms, which should be handled in their own way.
I think that the implications are that in the modal region of a small room the steady state is all that matters - just as I have been saying. Group delay or any of time domain concepts such as "reflections" etc, simply have no meaning or at least are not relavent. Closed boxes may sound better in a small room, but its not because of the group delay. Some say that dipoles have better bass because they excite fewer modes, but that is patently false, so thats not the answer, if indeed the premise is correct.
The only thing that I do know is that multiple subs, no matter what type, have lower variances both in frequency and spatially and this measures and sounds better, so this lends a lot of credence to that concept.
- Status
- Not open for further replies.