JonBocani, you win. 🙂
I know that I also would have failed the test, as I said. I still wonder if the result is because of the threshold of the condition or threshold of the sameness. Maybe combination of both. I performed a very similar test here, and I barely passed the ABX test, and that is why I'm challenging your result. I probably could not match the speakers as precise as you did.
How did you determine the target EQ curve? Something like Sonarworks? I demoed it but I didn't it like it. (I don't remember the reason why...)
I know that I also would have failed the test, as I said. I still wonder if the result is because of the threshold of the condition or threshold of the sameness. Maybe combination of both. I performed a very similar test here, and I barely passed the ABX test, and that is why I'm challenging your result. I probably could not match the speakers as precise as you did.
How did you determine the target EQ curve? Something like Sonarworks? I demoed it but I didn't it like it. (I don't remember the reason why...)
I'm using 2 differents mics: a CM-10 for SPL (good for up to 136db) and for FR I am using a Earthworks M50 (I have a matching pair) which is the best combo of mics I have worked with so far.
I have found that correction files for mics are unreliable, they are probably lacking a solid reference to start with... Especially above 15khz and below 35hz, but probably in the mid frequencies too. More importantly, microphones are transducers too.. They are limited in SPL and they sure are not perfectly flat themselves. According to few reviews and comments, the M50 is the most reliable tool, even if not ''perfect''.
1.65m distance on-axis was pretty easy to correct. If I recall correctly I started with sweep tones and finished with a pink noise. Probably ended with 1.5-2.0db differentials on some EQ bands, on some of the drivers tested, but overall flat. SPL-matching, however, was precise: 0.1db is the max resolution of my RTA and I usually don't exceed 0.2db, but then again it's by using a pink noise so if a driver has a peak of, say, 0.8db at 1khz, we might end with a 1.0db differential @ 1khz from that driver and the average of other drivers... So, yeah, not perfect EQ as I mentioned before, but precise enough, it seems, to have made them all impossible to identify.
The M50 I am using:
M50 — Earthworks Audio
Entirely satisfied. Works great and even without any correction file, I feel it reflects ''what we hear''.
Got also a printed FR graph for each of my M50's and you can download the ECF if needed.
M50 — Earthworks Audio
Entirely satisfied. Works great and even without any correction file, I feel it reflects ''what we hear''.
Got also a printed FR graph for each of my M50's and you can download the ECF if needed.
Highly saturated sounds are usually the ''best'' for these tests, that's why I use as much as possible music from artists such as Pan sonic, COH and Alva Noto. They like to play with sounds at the extremes, and it shows things that we sometimes cannot notice with ''normal'' music. Very high frequencies, very low frequencies, saturated mids, de-phasing, etc...
We were able to spot the 128 or 160kbps MP3 threshold in the digital formats blindtest by using such music...
But we couldnt find that kind of threshold this time with the driver's blindtest.
About harmonic distortion: I believe, now, that our hearing capacities are limited enough to let HD get in the way much more than we think.
However, we might spot HD more easily in the highest frequencies and/or in the lowest frequencies.
Never tested HD per se, so I have no idea how the human hearing would react about HD in a blind test...
Also I'd like to point out an important technical aspect of the test:
With a noise floor of only 29db C-weighted, with an on-axis distance of 1,65m, sailing between 86db and 95db (peaks) ALL the drivers were obviously well below they thermal limits, including the weakest of the group, the FR10.
With the highpass being 360hz @ 48db/octave, the electrical energy consumption could not exceed 2 or 3 watts (at worst) while playing normal music (not pure tones or pink noise) even with peaks up to 95db at the listening position.
So what's left is the mechanical limits of said drivers. I would assume the Radian 950PB-beryllium was the one that was testing those limits the most. Especially since the horn was not made to go that low, so I had to EQ quite a lot (FYI: instead of boosting below 500hz I negative-highshelf it).
But, anyways, we couldnt spot the Radian.
AND if that would have been the case, rest assure that I would have try another crossover point: 400hz or higher.
With a noise floor of only 29db C-weighted, with an on-axis distance of 1,65m, sailing between 86db and 95db (peaks) ALL the drivers were obviously well below they thermal limits, including the weakest of the group, the FR10.
With the highpass being 360hz @ 48db/octave, the electrical energy consumption could not exceed 2 or 3 watts (at worst) while playing normal music (not pure tones or pink noise) even with peaks up to 95db at the listening position.
So what's left is the mechanical limits of said drivers. I would assume the Radian 950PB-beryllium was the one that was testing those limits the most. Especially since the horn was not made to go that low, so I had to EQ quite a lot (FYI: instead of boosting below 500hz I negative-highshelf it).
But, anyways, we couldnt spot the Radian.
AND if that would have been the case, rest assure that I would have try another crossover point: 400hz or higher.
As I said previously, Identification blind test must be done before an appreciation blind test.
You MUST prove that you are able to differentiate A from B before pretending you better like A than B.
That's pure logic.
The error that we are making (and I include myself) is we ASSUME that we can differentiate even the slightest differences... while we actually.. CAN'T.
That would lead us back to psycho-acoustics. Plain and simple. Lots and lots of studies about it.
As underlined numerous times, the drivers must be used within their thermal and mechanical limits. I cannot stress that point too much as it's extremely important.
I am NOT pretending that a FR10 will sound the same as a bigger PA driver if they are both compared at 105db... For starters, the FR10 probably won't be able to achieve 105db, even with 10% THD...
If I recall correctly, the upper 95db limit was a limit for a good reason. We couldnt go much higher. Not to mention that people would have not enjoy a non-stop 95db level. Most music excerpts and participants preferences were more in the 88-90db.
Sonce, yes it all makes sense, but I think you are a bit optimistic about the ''all manufacturers''... I can assure you that some manufacturers are NOT measuring their drivers after 2 hours at nominal power. They should, but they don't.
The nominal power ratings, for starters, are often optimistic themselves and manufacturers are sure not all following standards such as AES2-1984 (revision 2003) like Faital Pro (by exemple) is doing...
http://diy-audio.narod.ru/litr/AES2-1984-r2003.pdf
Or they pretend to do but no verifications were made, so you end up with drivers that are not delivering what it's promised.
Or the prototypes works as per AES testing but the production's drivers are not the same.
We're talking audio industry, not medical industry. It's not as verified and as trustworthy.
Also have to take into account if the tests were done in free air or in box. Better to plan conservatively if possible.
But those test must be done unfiltered with a wide variety of music to have any meaning. What you have done is to impose a narrow set of severe restrictions on the test. Restrictions that are not normally encountered by the average listener.
You have not proven that if those restrictions are remove people cannot differentiate slight differences.
So we're right back to the issue of whether all of this has any value to most of the posters on this DIY audio forum. I maintain it does not, and that people should take your advice cautiously.
I would say that Jon didn't make enough restrictions😉 Running bass drivers so high is not normal, yet..
Yep. The correction files came with my cheap mics were very questionable, and I ended up making my own custom correction files. It seems that your frequency matching accuracy is not too different from mine, but I was comparing 2 way vs 3 way at that time, I remember that it was really hard to match FR @crossover points.
A wide variety of music was present at this test.
Rock, jazz, classical, electronic, pop, you name it. All with enough material so the 360hz-7200hz bandwith would be tested in all sorts of ways. That's sure not an angle you can attack the test's integrity. But thanks for trying. 😉
Now, about the unfiltered thing, like explained previously the 360hz-7200hz was not an option, it was mandatory. Anything else would have made the drivers exceeds their limits, which defeats the purpose entirely.
As for the usefulness of the test, for DIYers, I think, on the contrary, it gives a very useful guideline in regards of the midrange selection in a 3-way speaker projet.
It shows very clearly that a midrange should be selected for his output (SPL) first, related to the bandwith needed, and the sub/woofer + tweeters own mechanical limits.
So basically, if you're using a subwoofer than cannot be properly used above 120hz, you cannot use a compression driver as a midrange.
Also, if you're using a small ribbon tweeter that is limited to 3khz+, you know you can't use most of the bigger cones as a midrange.
Also, if you're looking for 95db in-room, it's entirely different than a 110db project, which will restrict a lot the midrange's selection.
So, again, FR is king.
And SPL factor just follows the FR...
Last edited:
classicalfan, obviously, didnt think your music selection was the limiting factor. stop playing dumb.
As a concrete exemple of the ''usefulness'' of my test:
Before, I was convinced that 4-way was the optimal configuration.
Now, I'm not even convinced that 3-way is necessary, to achieve an optimal hifi experience... I sure can live with an EQd fullrange if the SPL needed is low.
And i'm working at this very moment on a high-performance 2-way that gives me plenty of satisfaction.
I'm not seing the midrange bandwith as critical as before.
I'm not thinking anymore that something esoteric is hidden behind the measures, such as micro-dynamic details or a magic tone or sound signature or that these frequencies are more important to the human ears or any of the usual audiophile mumbo-jumbo...
Now, it's basically SPL, FR and far behind power response, which is also important but sure not as critical as SPL and FR.
Before, I was convinced that 4-way was the optimal configuration.
Now, I'm not even convinced that 3-way is necessary, to achieve an optimal hifi experience... I sure can live with an EQd fullrange if the SPL needed is low.
And i'm working at this very moment on a high-performance 2-way that gives me plenty of satisfaction.
I'm not seing the midrange bandwith as critical as before.
I'm not thinking anymore that something esoteric is hidden behind the measures, such as micro-dynamic details or a magic tone or sound signature or that these frequencies are more important to the human ears or any of the usual audiophile mumbo-jumbo...
Now, it's basically SPL, FR and far behind power response, which is also important but sure not as critical as SPL and FR.
Last edited:
classicfan, as many others, are trying to find bugs under every single rocks.
Don't get me wrong: I would love to find a fundamental flaw in this test, or any other similar blind test, some sort of ''eureka'' moment that would make me (us) realise that we are indeed able to hear what we think we can hear. That this is not hallucinations, that psycho-acoustics are for others...
But unfortunately, the cold truth is: we are humans.
The Dayton RS225 is not really a ''bass driver'' though... 😛
had to boost the 3khz dip, but it was not like I was trying to make sing a 18'' subwoofer....
https://www.parts-express.com/pedocs/specs/295-356--rs225-8-spec-sheet.pdf
bytheway, the power response of the RS225....
4khz = 20db differential, even at 30 degrees...
still, even by swinging our heads left and right on the chair, we were unable to spot it. BUT, below 3khz the power response was much more manageable, of course, so on the whole 360-7200hz blended result, it was indistinguishable.
Most probably, we wouldnt have the same result with a a driver that beams 20 or 30db differential from 1khz, or on a smaller angle....
4khz = 20db differential, even at 30 degrees...
still, even by swinging our heads left and right on the chair, we were unable to spot it. BUT, below 3khz the power response was much more manageable, of course, so on the whole 360-7200hz blended result, it was indistinguishable.
Most probably, we wouldnt have the same result with a a driver that beams 20 or 30db differential from 1khz, or on a smaller angle....
You are not going to accept that listening to music over a limited bandwidth is not a useful test so what's the point of arguing?
- Status
- Not open for further replies.