Hi, I went through all this with very serious blind testing, with my DSP-equalized system (true dipole in the bass, 15" PA coax driver) where I can introduce arbitrary additional phase shifts further to the minimum phase bandpass target equalization (50Hz...18kHz, sloping down 3..4dB overall, so much for the "flat" part again not being acoustically preferable in a specific setup). There are definitely differences when I phase-EQ a 4th(6th)-order rolloff to minphase as a BR(with HP) would do, compared to a mixed phase rolloff of resulting second order nature (not to be confused with a second order amplitude rolloff which sure sounds different per se).
Yet best balanced in timbre and perceived "compactness" of bass signals (drums, bassguitar, etc), as I found, is a fully linphase phase "rolloff", eg, none) together with a 4th-order amplitude rolloff at least at 50Hz or any lower/flatter the better. The changes are subtle but not completely irrelevant to me, and indeed differences from other reasons, may still be more significant (when I phase-linearize my small 50Hz BR monitors with their 6th-order rolloff they still don't get the bass right from inherent design problems, it remains just less bad, compared to my main system).
A problem with the full linphase highpass target is the leading "pre-echo" slope in the (symmetrical) step response -- see pic -- which sometimes makes itself audible with very strong, very low frecuency signals (in electronic music). Therfore I could use a phase-EQ that gradually reverts back to minphase on the way down from f3 to DC. Additionally I could try to generally roll back the phase around f3 more than linphase to compensate a bit for any typical minphase low-cut in the recording process... I fully agree a level of rapidly diminishing returns is reached here (except for the fact that changing absolute polarity then even becomes more noticable than before -- related in the same way to the result as phase shifts do : related to the half-wave rectifying fashion the hair cell's neurons seem to fire).
Other aspects of a speaker make more to good or bad than that.
- Klaus
Attachments
Last edited:
KSTR
I am not clear on what you are saying, but it does seem that you tests contradict the posted paper. Admittedly they are not looking at exactly the same thing.
You said that your tests were blind but yet it sounds like you did them yourself. How is this possible? Since you admit that the effects are subtle, only a fairly rigid double blind test could be significant.
I am not clear on what you are saying, but it does seem that you tests contradict the posted paper. Admittedly they are not looking at exactly the same thing.
You said that your tests were blind but yet it sounds like you did them yourself. How is this possible? Since you admit that the effects are subtle, only a fairly rigid double blind test could be significant.
Earl,
I fully agree that such a rigid test is hardly affordable by a single person when the full academic set of rules are to be applied, but you might agree that one can come quite close to a point of personal evidence without extremely rigourous testing, at least when the person is somewhat skilled in the art and has some experience to manage such things, which I do humbly claim for myself. With any digital audio workstation piece of software or the foobar software player with ABX-plugin, you usually can arrange for blind gapless A/B-switching between versions of manipulations, especially convolvers are very useful for their universality. One must take care to control secondary clues, like time offsets between versions and artifacts of switching (best used are fast crossfades, also I sweep levels a bit to find out if and where the point comes when the sligtly louder version sounds better regardless of what it is, and so on...
As a programmer, I can also mix up (with book-keeping) the convolution kernels to arrange for randomized and placebo tests which are required to control learning effects from the switching act, in an effort to find out my personal levels of Just Noticable Differences, and to find out if some manipulation does or does not lead to a long term increase of listening satisfaction. My setup is also mobile (Laptop) so I can easlily test at other places and found some findings hold there, some don't.
This all is a set and setting sort of thing, no hard lines, and obviously I cannot fully exclude that some unrecognized side effect does always bias me to the manipulation that I think should sound better (but with the polarity issue, this is impossible anyway), still I think it can be seen that I'm trying hard and serious to find out what is going on, not being a greenhorn instead I'm doing this for years. For the way I listen to music of most any genre (closed eyes always, and mostly meditative style, even with heavy rock music, always one complete album in one go, no zapping, "letting myself be taken away by the music" and the auditory scene which transports it), the JND-levels are really astonishingly low to my own surprise while mostly not really significant in overall listening experience (no preference). Some others sound consistently more convincing when I apply rigourous listening satisfaction standards. Best polarity and low group delay in bass / low mids are among them, for me, allpass phase responses at higher frequencies much less so.
I would welcome rigorous research on that topic, of course, and I am open to any discussion (perhaps not in this thread, side-tracking it too much).
Last edited:
Blind self-testing can have merit. I know that when the A/B involves a switch, you can distract yourself and hammer at the switch until you're no longer sure which setting it is toggled to.
Psychologically if you are trying to convince yourself of something you are probably just as concerned about disproving it, which makes you try to listen for true indications.
Psychologically if you are trying to convince yourself of something you are probably just as concerned about disproving it, which makes you try to listen for true indications.
Hi KSTR
Given that your conclusions contradict many other studies and that you claim that the effects are indeed small, this would require the most highly controlled testing before you could conclude that your results were significant. My point is that your test was not stringent enough given the particulars of this situation to be conclusive. Yes, single blind testsing does have merits, but it also has a low level of meaningful resolution.
Given that your conclusions contradict many other studies and that you claim that the effects are indeed small, this would require the most highly controlled testing before you could conclude that your results were significant. My point is that your test was not stringent enough given the particulars of this situation to be conclusive. Yes, single blind testsing does have merits, but it also has a low level of meaningful resolution.
Hi Earl,
Again I certainly have to agree.
Some comments, though :
Professional double blind audio testing of things like JND-levels of group-delay phenomen is IMHO a very difficult task in providing full (maximum, that is) support for many a listener's own preferred and "best-case" listening environment : at home, in their daily use (where generally the best JND-resolution levels are achieved and many will agree). Preferably people who have PC-based DSP-driven optimised setups with arbitrary variable transfer functions (quite a few around, in this forum), which could be hiddenly handled online to account for the double-blind test protocol. Only prerequisite would be an agreement that knowledgable participants do not reverse-engineer or manipulate in any other way the listening test (ok, that is a tricky part). More details might be needed to think through but to me it pretty obvious that a home-based professional double-blind long-term test is readily doable (also because we -- in a german forum -- already conducted a first "semi-pro" group-delay blind test via intenet-distributed disguised test tracks, and using impulse kernels instead is an obvious procedural improvement), IMHO the results might be surprising compared to the classic "guinea-pig sitting in a lab"-situation, as I found it documented in some papers. We have entered the third millenium and scientific test methods and practices should continue to evolve accordingly. Anybody need a topic for their master thesis?
- Klaus
Again I certainly have to agree.
Some comments, though :
Professional double blind audio testing of things like JND-levels of group-delay phenomen is IMHO a very difficult task in providing full (maximum, that is) support for many a listener's own preferred and "best-case" listening environment : at home, in their daily use (where generally the best JND-resolution levels are achieved and many will agree). Preferably people who have PC-based DSP-driven optimised setups with arbitrary variable transfer functions (quite a few around, in this forum), which could be hiddenly handled online to account for the double-blind test protocol. Only prerequisite would be an agreement that knowledgable participants do not reverse-engineer or manipulate in any other way the listening test (ok, that is a tricky part). More details might be needed to think through but to me it pretty obvious that a home-based professional double-blind long-term test is readily doable (also because we -- in a german forum -- already conducted a first "semi-pro" group-delay blind test via intenet-distributed disguised test tracks, and using impulse kernels instead is an obvious procedural improvement), IMHO the results might be surprising compared to the classic "guinea-pig sitting in a lab"-situation, as I found it documented in some papers. We have entered the third millenium and scientific test methods and practices should continue to evolve accordingly. Anybody need a topic for their master thesis?
- Klaus
Last edited:
HI Klaus
I have thought about doing internet studies many times, but I always come up against the fact that there is no way to control the playback and so this would always be a confounding variable. You could try and standardize on, lets say, a certain model of insert earphone, but could you trust that this was always done?
I have not found a solution to this problem.
I have thought about doing internet studies many times, but I always come up against the fact that there is no way to control the playback and so this would always be a confounding variable. You could try and standardize on, lets say, a certain model of insert earphone, but could you trust that this was always done?
I have not found a solution to this problem.
I have found a location where helper woofers 'help'. They are placed part way between myself and the mains on the perimeter of the listening triangle closer to the speakers. I believe that perhaps at least half of the sub 300Hz anomalies are floor and ceiling related.
Subjective thoughts are that this is better. EQ or no EQ, it simply sounds better. No amount of EQ could seem to fix the mains only result to sound as good.
Subjective thoughts are that this is better. EQ or no EQ, it simply sounds better. No amount of EQ could seem to fix the mains only result to sound as good.
Attachments
What is the crossover frequency of your woofer? Does the biggest cancellation frequency half wavelength distance (5.6 feet for 100HZ for example) correlate with the distance between the main speaker, the listening position, and/or any particular room boundary, especially a corner? It makes sense to me that the lower midrange and bass cancellations could be largely filled in by "flanker" speakers, which because of their different distances to these things, would have different cancellation frequencies, and therefore when added at the listening location, would largely fill in each others cancellations. It might be worth trying adding a 3rd woofer in a corner or against a wall, some distance up the wall, to accommodate the 3rd axis in the room. Wouldn't it be nice to get back "acoustic fidelity" in the lower mid and bass at the listening position, and perhaps everywhere in the room, so when we turn up the bass on our preamp (which of coarse has tone controls), it would only sound warmer and better... No real boom problem...
These are crossed at 300Hz. The subs are off so ignore below 100Hz. The issues here are the dip at 140 and 250. The 250 dip is composed of the 1st floor and 2nd ceiling cancellation. The helper woofers happen to be a half of a wavelength closer to me than the mains at 250Hz.
As you are saying, it does seem to return fidelity. There is no noticeable boom...although you can see on the plot there is a peak at 200Hz and I can hear that and I'm thinking of notching it.
As you are saying, it does seem to return fidelity. There is no noticeable boom...although you can see on the plot there is a peak at 200Hz and I can hear that and I'm thinking of notching it.
I think what we perceive as boominess is really the uneveness of the lower mid and bass region. Any high points can and should be EQ'd down, but if you try to bring up a cancellation, you end up causing peaks in other locations of the room, so that is not recommended. I think flankers of some sort may be the only way to even out those particular boom causing cancellations.
You know Allen, my response used to look similar. But with more placement trial and error, you should be able to do much better.
from:
To:
Is possible through placement. I ended up turning the bass knob up a bit since that graph. It was down almost all the way.
Dan
Since we were discussing dipoles and I was voiceing a concern about extra full range energy that would bounce off the front wall, while others were saying "so what", I thought I'd do a reasonable simulation of it. Dipoles 5 ft in front of a wall were suggested so that would be a reflection delay of about11 ms. Rear output is out of phase (which sounded like it lowered all the resonance pitches when I did that in the simulation).
The only other variable would be the attenuation of the back bounce which is a combination of extra distance, actual radiation angle and wall reflectivity. I think the 5dB reduced reflection is realistic for a dipole in front of a non-absorbtive wall.
Cool Edit Pro
David S.
This is a rather old post, but I remembered it while reading Toole's book just now. He claims delayed sounds from the same direction as the primary sound have a significantly higher threshold of audibility than reflections from other directions. There is no mention of the gross coloration audible with your sound samples. Interesting; maybe dipoles should not be disregarded too soon.
Toole - Sound Reproduction (p. 82) said:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.