They dont disappear. They are just less obvious in stereo.
Last edited:
-
WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Revel Salon2 vs Genelec 8351B - Blind Test Preparations
- Thread starter echopraxia
- Start date
Probably.
But thus far I've only seen mono vs stereo data for four speakers: Rega, KEF, Quad, and whatever speaker is being EQ'd (or not) in Amir's post #273.
I'm not really interested in who wins and who loses. I'm much more interested in WHY the scores change. And I don't think it's necessarily because "flaws are less noticeable in stereo" because for TWO of the four EQ Models in Amir's post, the stereo score was LOWER than the mono score.
There are particular pieces of information which I'm on the lookout for, and from time to time something comes along which might be relevant, and then I start poking around. I want to know how the sausage is made.
Last edited:
I love your curiosity, @Duke!
As for the blind test, I can't wait to see the result. I think there's too much energy being spent on finding flaws in the experiment than necessary.
EQ and/or bass management in the lowest frequencies makes sense because low frequencies is room dependent and not universally applicable.
Some rooms will benefit from a lack of shitstorm in the deepest frequencies while others will benefit from plenty of oomph. All rooms will benefit from EQ and/or bass management regardless of speaker, so why not just eliminate the biggest factor of subjective sound quality that has the least relevance to speaker quality itself by simply even out the difference with room EQ?
In my mind there's still another factor that will differentiate the speakers when it comes to bass, and that's the mid/upper bass oomph. A speaker with bass drivers spread out in height and/or close to the ground will have less cancellation in the important area between 100-300 hz and will likely still be experienced as fuller/bigger/warmer even if the frequencies below 100 hz are equalized.
Personal experience here is that this particular area is highly important for sound quality but not really suitable for heavy EQ due to our increased sensitivity for room decay in that area. It's also a concern that at those frequencies we'll see a bigger difference over small movements of the microphone than below 100 hz, so any EQ will possibly make the sound (and decay) worse at other places where it still can be audible for all intents and purposes.
My guess is that Revel will win because of "bigger sound" and "soundstage releasing the speakers" and think this might be traced back to the lesser effects of cancellation (or ill-effects of EQ to match the response below 200 hz) in the 100-300 hz range if explored rigorously in the future.
As for the blind test, I can't wait to see the result. I think there's too much energy being spent on finding flaws in the experiment than necessary.
EQ and/or bass management in the lowest frequencies makes sense because low frequencies is room dependent and not universally applicable.
Some rooms will benefit from a lack of shitstorm in the deepest frequencies while others will benefit from plenty of oomph. All rooms will benefit from EQ and/or bass management regardless of speaker, so why not just eliminate the biggest factor of subjective sound quality that has the least relevance to speaker quality itself by simply even out the difference with room EQ?
In my mind there's still another factor that will differentiate the speakers when it comes to bass, and that's the mid/upper bass oomph. A speaker with bass drivers spread out in height and/or close to the ground will have less cancellation in the important area between 100-300 hz and will likely still be experienced as fuller/bigger/warmer even if the frequencies below 100 hz are equalized.
Personal experience here is that this particular area is highly important for sound quality but not really suitable for heavy EQ due to our increased sensitivity for room decay in that area. It's also a concern that at those frequencies we'll see a bigger difference over small movements of the microphone than below 100 hz, so any EQ will possibly make the sound (and decay) worse at other places where it still can be audible for all intents and purposes.
My guess is that Revel will win because of "bigger sound" and "soundstage releasing the speakers" and think this might be traced back to the lesser effects of cancellation (or ill-effects of EQ to match the response below 200 hz) in the 100-300 hz range if explored rigorously in the future.
- Joined
- Nov 29, 2017
- Messages
- 555
- Likes
- 1,621
I have not followed this thread thoroughly, and don't really want to get involved in whatever speaker-world conflict is occurring here as there are plenty well enough of those in headphones. However, a quick scan revealed that this plot is coming up a lot
And people keep saying things like "imagine the deviations", and since I had a moment spare and access to WebPlotDigitizer and excel, I figured I'd throw them into excel, make some box plots, and possibly slightly reduce the amount of imagination in the world (my primary goal in life).
All data was scraped using WebPlotDigitizer so take the digits right of the period with some grains of salt.
All data
Mono only
Stereo only
Sound quality only
Spatial quality only
And here's the excel sheet. I didn't get terribly creative since I had about 20 minutes spare, but if I have even slightly decreased the amount of imagination in the world, I can sleep easier tonight.
Edit: The box plots were created with literally whatever the Excel 365 defaults are, other than adding median lines. I make no promises, commentary, assurances, or statements regarding the accuracy or usefulness of the box plots for the task at hand.
And people keep saying things like "imagine the deviations", and since I had a moment spare and access to WebPlotDigitizer and excel, I figured I'd throw them into excel, make some box plots, and possibly slightly reduce the amount of imagination in the world (my primary goal in life).
All data was scraped using WebPlotDigitizer so take the digits right of the period with some grains of salt.
All data
Mono only
Stereo only
Sound quality only
Spatial quality only
And here's the excel sheet. I didn't get terribly creative since I had about 20 minutes spare, but if I have even slightly decreased the amount of imagination in the world, I can sleep easier tonight.
Edit: The box plots were created with literally whatever the Excel 365 defaults are, other than adding median lines. I make no promises, commentary, assurances, or statements regarding the accuracy or usefulness of the box plots for the task at hand.
That's pretty cool! Shows deviations are larger in mono if I read it rightly, implying (perhaps) that more differences can be determined in a mono test than in stereo. I'd expect stereo adds interactions between speakers, room, etc. that make it harder to distinguish differences among the speakers themselves.
- Joined
- Jan 23, 2020
- Messages
- 4,337
- Likes
- 6,709
I agree completely with those saying that the blind listening tests need not be perfect to be of use. No blind listening test will ever be perfect, but even if it's not perfect, it can still be useful. I'm sure that all of the blinds I've hosted would be torn to shreds if I posted them here, but(at least for me), they've still been incredibly enlightening, and I'm so glad that I did them. Had I posted about them here, I may have been discouraged enough to stop doing them after the first one, which is a shame . We should be encouraging blind tests, not discouraging them. Even if they're not perfect, they're better than similarly performed sighted tests.
My next blind will probably bee a Revel M105 vs Genelec 8030c, with the JBL 308p used as the third speaker to add context(recommendation of Toole). I'll post about it here, and I'm sure there will be many complaints, and many people saying it's useless, but I'm of the opinion that it's still better than nothing.
My next blind will probably bee a Revel M105 vs Genelec 8030c, with the JBL 308p used as the third speaker to add context(recommendation of Toole). I'll post about it here, and I'm sure there will be many complaints, and many people saying it's useless, but I'm of the opinion that it's still better than nothing.
@Duke have you read Floyd's book "sound reproduction" ?
It covers a lot more than a few graphs shown in this thread. Chapters 3.4, 7.4.2 etc
The commentary around the data will probably answer a lot of your questions.
Last edited:
Precisely what Toole says.
In summary, while it is undeniable that good stereo and multichannel recordings are more entertaining than monophonic versions, it is an inconvenient truth that monophonic listening provides circumstances in which the strongest perceptual differentiation of sound quality occurs .
When multiple loudspeakers are simultaneously active, the uncorrelated spatial sounds in the recordings modify the ability of listeners to judge timbral quality.
Last edited:
- Joined
- Jun 5, 2020
- Messages
- 4,809
- Likes
- 3,749
What's the best way to test in mono? I've tried downmixing from stereo but it does not sound good. One-channel only does not sound good, either. Maybe I need true mono recordings.
Downmixing to mono should work fine unless there are phase anomalies between sounds mixed across left and right. Such recordings do exist and effects can cause this, sort of issue, but generally it should be ok.
How have you mixed to mono?
Yes. And thank you for asking in a non-confrontational way.
Something that I question is shown in Figure 7.11, which is labelled Figure 8.8 in post #246. I lifted the figure from that post to make following this one a bit less tedious:
I didn't go into this before because I didn't want you and I to get into an unproductive tangential argument about me looking for inconsequential tiny faults with Toole's work. So let me say in advance that I HOPE you can accept this as me expressing a difference of opinion.
Note that the walls behind the screen are covered by medium-weight drapes, which would absorb short wavelengths. The caption contains what I believe to be erroneous information about the Quads: "This would have the greatest effect on the bidirectional dipole loudspeaker, but the product was already equipped with absorbent pads on the rear half of the enclosure to attenuate the output above about 500 Hz."
In the 3rd edition of his book (I have both editions) Toole identifies the Quads as the "ESL 63". The ESL 63 does not have absorbent pads on the rear half of the enclosure. The original Quads did, but the ESL 63's that I owned for two years did not. If you'd like proof I'll post a link to a YouTube video showing the disassembly of an ESL 63.
So in my opinion the backwave of the Quad 63's had their high frequency energy attenuated by the drapes lining the walls behind the screen. And in my opinion that handicapped them. (Likewise imo the Martin Logans were handicapped in that and other ways in the speaker-shuffler room, but that's a more complicated topic.)
As a longtime dipole speaker owner and dealer, there are virtually no circumstances under which I would recommend light absorption of the backwave energy of a dipole speaker. So in my opinion the mono and stereo tests of the Quads did not adequately evaluate the contribution of their backwave. And imo that has implications for the subsequent discussion of loudspeaker directivity in that chapter, because the (now degraded) backwave constitutes one-half of the Quad's radiation pattern. To be more specific, unasked and unanswered (in that chapter) is the question of whether the backwave energy would have made a beneficial contribution had it been spectrally correct. Imo, ime, ymmv, etc.
Last edited:
I think it's 24 or 48db per octave on the miniDSP. Both the high pass and low pass filters set to 80hz. The rolloff does cross over a little, but it's not nearly as much as most diagrams I see of speaker crossovers.
Seems like the perfect handoff would be a subwoofer playing 80hz, and then the main speakers playing 80.0001hz and up.
watchnerd
Grand Contributor
Brickwall filters are not benign.
- Joined
- May 18, 2020
- Messages
- 1,285
- Likes
- 2,939
Interesting. So was it the ESL 57? Which would be ironic in this context, since the 57 was used and admired in its early years almost exclusively as a mono system. It was an affluent-middle-class purchase, likely to be used in not-too-small rooms with hard plaster walls, with a very specific listening chair. Very difficult for it to shine in a test of this nature, I guess.
As the crossover slope increases, so too does the phase angle between the highpass and lowpass drivers. I'm going to make some simplifying assumptions in the following description, but I think the relevant concepts will be accurately portrayed:
With a second order (12 dB per octave) crossover, the phase of the woofer (or subwoofer) is 180 degrees behind the phase of the tweeter (or main speakers, if we're talking about subwoofer crossovers). This is usually addressed by reversing the polarity of the signal going to the tweeter, but sometimes the tweeter is set back relative to the woofer, perhaps by slanting the front baffle. (A horn may do this inherently due to its depth.)
With a 4th order (24 dB per octave) crossover, the woofer is 360 degrees behind the tweeter, so they are back in-phase. Perfection?? Well, not really. Yes they are in-phase, but the woofer's output is arrives behind the tweeter's output by one wavelength! So the next time you read ad copy bragging about a speaker's "phase coherent" or "phase correct" crossover, recognize that in MOST cases, that crossover is NOT correct in the time domain.
Let's look at 8th order (48 dB per octave). Now the woofer lags the tweeter by 720 degrees, or two wavelengths. Do you see the direction this is going?
So, what do you think will happen with a super-steep brick wall filter?
^^This.^^
The ear has poor time-domain resolution at low frequencies so we can get away with more phase shift than you might think, but I'm pretty sure a brickwall filter indeed would not be benign.
That being said, I THINK it is possible to make corrections in the digital domain which are not possible in the analog domain, so it MIGHT be possible to make a benign brick wall filter. That is not my area of expertise.
Back to blending mains with subwoofers: The real world is more complicated than my examples above because the main speakers have their own inherent rolloff, with its accompanying phase shift, which are added to the rolloff and phase shift of the highpass filter, and which then interact with the rolloff and phase shift of the lowpass filter on the sub(s). Imo the secret weapon for fine-tuning the blend of all these rolloff slopes and phase shifts is having a continuously-variable phase control on your subwoofer amp(s).
So here is my suggestion for manually dialing-in the blend between subwoofer(s) and mains, applicable to most subwoofer amps that have a continuously-variable phase control:
1. Adjust the gain control on the subwoofer amp. This usually makes the most difference.
2. Adjust the low-pass filter frequency of the subwoofer amp. This usually makes the next most difference.
3. Adjust the phase control on the subwoofer amp. This usually makes the least difference but it still can make or break the combination.
4. Cycle back through several times to fine-tune.
* * * *
I guess there is some uncertainty as to whether or not it was the 57 or the 63. I've owned both.
When I had 57's, I "hotrodded" them by adding a tweeter, BEHIND the panel and facing up, to add some high frequency energy to the reverberant field. In my Biased Uncontrolled Loudspeaker Listening, Sighted, Hearing Imaginary Things (BULLSHIT) testing, it was judged to be a worthwhile improvement.
Last edited:
First let me emphasize the fact that is obvious to you but may not be obvious to some forum members - digital crossovers make the same mess of the phase as analog do. LR 24 will make equal mess of the phase if you make it with analog elements as if you do it digitally. Hence both need to be corrected via adequate FIR filters. So, without XO phase correction you can forget about getting nice step response, either with analog or digital XO.
Now, from what I know you can't phase correct brick wall XO but you certainly can do it with LR 192. This is how LR 192 high pass looks when applied at 80Hz:
As you correctly pointed out it would make a huge mess with the phase, but with appropriate correction phase can be perfectly restored to linear behaviour:
So, this works nicely and is reasonably steep.
However, this kind of crossover correction will work nicely between main and sub only if their phase was previously aligned. That same issue exists between main woofer and the tweeter but those 2 drivers are usually sitting on the same front panel of the speaker and they don't have different delay as they are passive drivers. That is not the case with sub and main so you have to take care of it, usually by delaying the SW to compensate for difference in distance between them toward LP and to delay in DSP circuitry which is often found in modern subs. Of course, if you have sub with passive XO which is further away from LP than mains you would need to delay the mains to get their phases to match around XO point.
Due to differences in LF roll-off you can't hope to get the phase equal down to 20Hz but that is not neccessary - you need to get the phase of sub and main to be as equal as possible only in the region around crossover to ensure correct summation. Width of that region depends on how steep crossover is.
It goes without saying that levels should also be matched so adjusting gain of the sub to match level with main(s) around XO point is equally (or even more!) crucial.
P.S. I wouldn't use the term "amp" here as the same rules apply to passive subs with passive XO so better to leave amp out of it
Last edited:
Can we stick to OP please ,new topics new threads .
Transporting old arguments from other threads by the same participants is incredibly dull and a bit insulting to OP .
Transporting old arguments from other threads by the same participants is incredibly dull and a bit insulting to OP .
OP
- Thread Starter
- #360
If I end up blind testing with subs (though I don’t know if we have a strong consensus either way yet on that), your advise and others would be helpful on how to correctly set it up.
Just as a sanity check, I would want to make sure I haven’t been doing it all wrong. When you mention adjusting sub level to match main, I’m a little unclear on what this means if there are large nulls and peaks — so what are you matching with what, for example?
I have usually followed a process of (1) adjust mains delay until REW measures they greatest response with the least nulls, (2) adjust sub level so that the lowest point is roughly around the SPL of the mains so there is EQ headroom, and (3) apply EQ corrections (Dirac Live via miniDSP SHD which as far as I know doesn’t seem to do the integration itself, just the EQ).
Maybe I’ve been doing it all wrong? Maybe I need to adjust sub level so the crossover slopes are more symmetric with each other rather than worrying about EQ headroom. But I also don’t know how it works if a huge boost is needed at some frequencies; does that risk clipping?
Similar threads
