A. Are the drivers presented to listeners in random order, in fixed order, or otherwise?
B. So, you are a reformed audiophile. Your mission is now to convince people they can't truly hear any differences ever, or only that it is possible they won't be able to hear a difference under some conditions?
B. So, you are a reformed audiophile. Your mission is now to convince people they can't truly hear any differences ever, or only that it is possible they won't be able to hear a difference under some conditions?
I suspect it is in the low 30's high 20's having been there multiple times. Very quiet room.
Ok, I made an error.
But it's good news. For me. And the test. 😛
Here is some self-explanatory images:
So for some reasons, I took it for granted that my Audio Control Industrial RTA (and my calibrated mic) was able to measure the noise floor. It's good for high SPL (142db I think?) but not for low SPL...
SO, my bedroom reads on the iPhone app dB meter a average of 31db with C-weighting while my RTA reads 54.5db.
Also, as you can see, with A-weighting it's 24db while the RTA reads the same 54.5db.
But let's use the C-weighting since it's more appropriate I believe.
Therefore, my RTA measured an average of 52.5db in the room so the real C-weighting noise floor was 29db.
Post #1 updated.
Thank you Plasmu. 🙂
It's 24db A-weighting or 29db C-weighting.
A) were always presented A first, then B. So you knew how A sounds like and you knew how B sounds like. No trickery of any kind. Then, ''X'' was either A or B, you had to decide.
B) No mission. People are wildly underestimating the psycho-acoustic effects, that's all.
OP I'm all for doing things that are not "normal" my system when finished will have 3 preamps and a xo system that is unconventional... another time about that... but I experiment to see what's different/better in real world listening conditions. It's like you are looking where you don't want to go to find answers about where you do want to go. You must concede a little and admit this trick does not have a practical conclusion. Most people are trying to get more life out of their system not crush it until it sounds the same no matter what driver you use. Challenge yourself to the opposite experiment, in all fairness, to take the same driver and blind test where the threshold is to actually noticing a difference. Don't write everything off. Take a break if you need to.
According to you there isn't any difference in how various drivers sound when you EQ'd the sane FR, that they are all perceived to sound the same, so I'm quite interested in how you're defining a "better" horn within the context of the test results
Edit: and for the record, I mostly agree with you about your conclusions, youre just being incredibly pompous with your replies in this thread and really overpresenting how infallible your methodology is to the point where you even go out of your way to explicitly claim you're quite confident in the accuracy of your psychological diagnosis for every single detractor. It just really drives home the childish vibes.
Last edited:
We tried.
We tried to find the threshold.
What more can we do than having a comparative match between a compression driver and a 8'' woofer, I'm asking you...?
On paper, they're opposites. They don't even live on the same planet.
It has been said before: the differences are in the power response (directivity, beaming), the mechanical limits (bandwith) and the max power output (SPL).
''Practical'' Conclusions are extremely simple:
Since the transducer's mechanical limits is the real bottleneck, and since Frequency Response is king, when designing a loudspeaker we should focus on the 1st and 10th octave, more precisely 25-35hz and 15-20khz. Which are a challenge for many drivers, even the best woofers, even the best tweeters, even with the help of an EQ, sometimes.
Also, I think having headroom is a must. So a sound system that is capable of 105db+ in-room, even if you usually listen at 90db, is better. But a real 105db, from 25-30hz up to 20khz, distortion below audible levels, and flat within 1.5-2db.
Last edited:
Not a big fan of horn per se, but the horn I used was sure not the best to get down to 360hz, I think everyone will agree on that...
So ''better'' in terms of helping go down to 360hz (w/ acoustic impedance)
But anyway, luckily the Radian 950PB is able to sustain torture. I had a pair at home hornless, EQd down to 380hz. Was still able to pull 110db in-room from it, and even more.
Last edited:
Jon,
I want to be sure I understand exactly what you are doing in the miniDSP, since that is key to any analysis or conclusions.
I believe from what has been posted so far that you are making two modifications to the signal.
First, a bandpass filter with 48b/oct slopes at 360Hz and 7200Hz.
Second, an EQ function that smooths out the frequency response within the passband to some +or- db limits. I don't recall those limits being specified, although I may have missed that information.
Can you please confirm if this understanding is correct and provide the EQ limits if so. Thanks.
I dont see any problem with the methodology of the test from op, the problem potentially could be in what room that test was done. ime a untreated room bury resolution.
I'd be curious if you could redo some test in a free reflective space. Nothing too fancy, simply covering every early reflection (ceiling, side panels, back wall and floor).
I could probably come by to participate!
On a slightly different subject. did you ever done blind testing comparing digital eq activated with no eq applied vs no-eq activated? I wonder if you lose resolution doing this test simply caused by the DSP.
I'd be curious if you could redo some test in a free reflective space. Nothing too fancy, simply covering every early reflection (ceiling, side panels, back wall and floor).
I could probably come by to participate!
On a slightly different subject. did you ever done blind testing comparing digital eq activated with no eq applied vs no-eq activated? I wonder if you lose resolution doing this test simply caused by the DSP.
Last edited:
Thanks.
I dunno, to me this entire 'debate' is semantics. Music isn't listened to from 360hz to 7,000hz after being EQ's flat +-0.5-1db.
Just wanna confirm,
As I see it:
2 different transducers each mounted on same style baffle/box each reproducing signal 1 = 2 distinct sounds
1 transducer reproducing signal 1A and 1 transducer for signal 1B, both very similar but very distinct from signal 1, both drivers have same style baffle/box = both transducers produce 1 distinct sound
Correct?
And would you expect the results to hold true for tweeters and woofers? If so, have you done any tests?
It would be interesting to see (rather hear to be exact) if the result would be the same if the testing persons could listen to a speaker that is covering the full audio range. I.e. woofer and tweeter sections that stay the same and only variing midrange.
Regards
Charles
Regards
Charles
A similar test was conducted by the TU-Berlin for tweeters of different types. They came to roughly the same result. Once you EQ out differences and levels it becomes very hard to percieve a difference.
If you read German: https://www.ak.tu-berlin.de/fileadmin/a0135/Magisterarbeiten/Andreas_Rotter_Exp.pdf
If our job was to peer review the methodology then maybe the last 9 pages of posts weren't a total waste.
However, I am afraid that we're missing the fundamental question. If Bocani and the TU Berlin are correct, are we allocating resources correctly when creating our speakers and audio systems? Assuming we don't have measuring equipment and powerful DSP EQs at our disposal, the speaker is undeniably the weakest link and most likely to provide the most immediate changes in sound quality. But, processing power means the game has changed. My guess is that having a powerful DSP in your system gets you much closer to good / accurate sound than very expensive drivers made of exotic high tech materials.
If you read German: https://www.ak.tu-berlin.de/fileadmin/a0135/Magisterarbeiten/Andreas_Rotter_Exp.pdf
If our job was to peer review the methodology then maybe the last 9 pages of posts weren't a total waste.
However, I am afraid that we're missing the fundamental question. If Bocani and the TU Berlin are correct, are we allocating resources correctly when creating our speakers and audio systems? Assuming we don't have measuring equipment and powerful DSP EQs at our disposal, the speaker is undeniably the weakest link and most likely to provide the most immediate changes in sound quality. But, processing power means the game has changed. My guess is that having a powerful DSP in your system gets you much closer to good / accurate sound than very expensive drivers made of exotic high tech materials.
All the people that have passed blind tests listening to dacs/amps/speaker/single-opamps have been completely ignored. I don't know if the number of people who have done that are less than 5% or even less than 1% of the population, but I do know for a fact that such things have happened and continue to happen.
Regarding the specific listening comparison here, its one I have not tried so no opinion on how difficult it might be to pass double blind. In any case there are some things that concern me about way the test was designed and run, the effects of which can only be guessed at now.
Regarding the specific listening comparison here, its one I have not tried so no opinion on how difficult it might be to pass double blind. In any case there are some things that concern me about way the test was designed and run, the effects of which can only be guessed at now.
That's a core point. The test as structured shouldn't be expected to provide definitive data on the limits of human perception. The results are more applicable to common audio reproduction scenarios in the home.
Consider that one lost. Storage and bandwidth costs continue to plummet as capacity skyrockets. Downsampling and lossy compression are quickly becoming irrelevant and unnecessary expenses in the distribution chain. Many artists on Bandcamp appear to not even know the difference and upload hi rez 24 bit FLACs instead of Red Book.
I use a 9 Euro 2" wide band driver on a Selenium 4750-SLF 2" horn. I cross it at 350Hz to a 6.5" in a 70L closed box. It is the best speech intelligibility I ever had. My family speak at least two languages fluently, my son spoke four languages when he was 7 years but that has declined due to circumstances.
Anyway, my efforts to understand speech in movies is essential, having one driver cover almost the whole vocal range is a game changer, despite having experts telling me that it's wrong horn, bad driver, bad x-over point, bad driver size...
Could be that a different horn is better, driver could be better and the list goes on. Fact is, speech is better on these than my Genelec 1030 and music also. Here's an early trial with not so optimal delay with x-over at 350Hz, later iteration have a 35db deep notch at 350Hz when polarity is shifted.
Measurement distance at 60cm
Attachments
Last edited:
Hey Jon, please accept my apology for my previous BS call...
That call was predicated solely on the well known difficulty of getting a CD down to 320 Hz, and a prosound 10" up to 7200Hz.
Your radian CD and kappalite 10", at moderate SPL's sufficient for your tests, meet that difficulty, ime/imo.
I've run bms coax CD's and the new b&c coaxial ring radiator down to 300Hz, at indoor SPL levels. And the kappalite 10" is a driver I'm familar with, as a candidate in synergies I'm building.
I have no experience with any of the other drivers, but they caused me no concern at all.
Because I do agree that you can equalize drivers with sufficient bandwidth and adequate SPL, to most likely sound similar enough to make ABX identification nearly impossible. Especially in a closely held, fairly tight on-axis, listening position.
Further information in this thread makes me tip my hat to the work you've done. 🙂
Like I mentioned before, I've worked a lot at getting different systems / different drivers to sound the same.
I've often found getting dead equal SPLs, to be harder than getting EQs and xovers spot on.
For the EQ / high pass / low pass task, I use FirDesigner and can literally make correction files, and measure near testbook perfect acoustic mag and phase for the exact listening spot in a matter of minutes....for the various drivers under test.
For the SPL task, I run pink as the source and use unweighted metering. If for whatever reason, slow flat Z SPL bounces around any, I use LEQ time averaging.
and fwiw...personally, i've thrown any and all SPL weighting (A, B, C, etc) into the trash...the weightings only exist for regulators to implement/debate ear safety at work and concerts imnsho 😉
It's pretty easy to get drivers to sound the same after that processing and level adjustment, ....at the exact tuning spot.
Although that said, I most often do think i hear slight differences, even at the test spot.. but I don't have a good blind ABX setup, so can't know for sure.
I'm more certain I can hear differences when I move around, or crank volume, or move outdoors.
So I'm very inclined to agree with your assertion in #167,
"Power response, max output potential (SPL) and frequency response (EQ correctable or not). That's about it."
Also agree with flaevor's guess that 'a powerful DSP in your system gets you much closer to good / accurate sound than very expensive drivers made of exotic high tech materials'.
And with Wesayso's comment that FIR filters only do exactly what you ask them to do. I wish more folks would simply drop what 'they know', and play around with dsp, maybe even with FIR a little.
As a related aside, I suspect many of our DIY folks have taken Harman's 'How to Listen' course.
Harman How to Listen
It's good stuff........ for any who haven't bumped into it yet....
That call was predicated solely on the well known difficulty of getting a CD down to 320 Hz, and a prosound 10" up to 7200Hz.
Your radian CD and kappalite 10", at moderate SPL's sufficient for your tests, meet that difficulty, ime/imo.
I've run bms coax CD's and the new b&c coaxial ring radiator down to 300Hz, at indoor SPL levels. And the kappalite 10" is a driver I'm familar with, as a candidate in synergies I'm building.
I have no experience with any of the other drivers, but they caused me no concern at all.
Because I do agree that you can equalize drivers with sufficient bandwidth and adequate SPL, to most likely sound similar enough to make ABX identification nearly impossible. Especially in a closely held, fairly tight on-axis, listening position.
Further information in this thread makes me tip my hat to the work you've done. 🙂
Like I mentioned before, I've worked a lot at getting different systems / different drivers to sound the same.
I've often found getting dead equal SPLs, to be harder than getting EQs and xovers spot on.
For the EQ / high pass / low pass task, I use FirDesigner and can literally make correction files, and measure near testbook perfect acoustic mag and phase for the exact listening spot in a matter of minutes....for the various drivers under test.
For the SPL task, I run pink as the source and use unweighted metering. If for whatever reason, slow flat Z SPL bounces around any, I use LEQ time averaging.
and fwiw...personally, i've thrown any and all SPL weighting (A, B, C, etc) into the trash...the weightings only exist for regulators to implement/debate ear safety at work and concerts imnsho 😉
It's pretty easy to get drivers to sound the same after that processing and level adjustment, ....at the exact tuning spot.
Although that said, I most often do think i hear slight differences, even at the test spot.. but I don't have a good blind ABX setup, so can't know for sure.
I'm more certain I can hear differences when I move around, or crank volume, or move outdoors.
So I'm very inclined to agree with your assertion in #167,
"Power response, max output potential (SPL) and frequency response (EQ correctable or not). That's about it."
Also agree with flaevor's guess that 'a powerful DSP in your system gets you much closer to good / accurate sound than very expensive drivers made of exotic high tech materials'.
And with Wesayso's comment that FIR filters only do exactly what you ask them to do. I wish more folks would simply drop what 'they know', and play around with dsp, maybe even with FIR a little.
As a related aside, I suspect many of our DIY folks have taken Harman's 'How to Listen' course.
Harman How to Listen
It's good stuff........ for any who haven't bumped into it yet....
Mark,
To repeat my response to the assertion in #167, my reply from #186:
Your (JonBocani's) on axis, near field, equalized test effectively nearly eliminated any driver’s major differences to the ears due to :
A) Power response- though power (off-axis) response of each transducer was different, the single listening position relatively near the DUT (driver under test) reduced off-axis room contribution cues. The fixed size of the baffle made LF power response similar for all the DUT.
B) Potential SPL- all DUT were below the SPL where distortion issues would become apparent in the bandwidth chosen.
C) Frequency response-the drivers on-axis response was corrected, and no test of the resulting off-axis response was auditioned.
If the “King” was allowed to move from the location of the “Throne”, the results would have been quite different.
Your (JonBocani's) test protocol eliminated what you regard to be “that's about it” regarding the DUT, with predictable results.
__________________
Comparing drivers equalized for the same response in the near field is similar to comparing the handling and ride of vastly different cars on a very smooth, flat, straight road-we would not be to "feel" the differences. Take those same cars "off axis" by introducing turns or bumps and the difference between them would be apparent.
Art
- Status
- Not open for further replies.