Revel Salon2 vs Genelec 8351B - Blind Test Preparations

[KSTR]

IME, one cannot ABX speakers like this, using different positions. The different position gives more than enough cues to easily identify each speaker. Even with two same speakers. One is easily recognized as the upper speaker and the other is the lower, simple as that. Please try that first (stack the Genelecs on top of each other) and only if you completely fail to ABX them you might have a chance to test for actual differences of the speakers proper.
There is a reason why all professional blind speaker tests use a (semi-)automatic shuffler to load the exact same playback position for every speaker.

 

[echopraxia]

IME, one cannot ABX speakers like this, using different positions. The different position gives more than enough cues to easily identify each speaker. Even with two same speakers. One is easily recognized as the upper speaker and the other is the lower, simple as that. Please try that first (stack the Genelecs on top of each other) and only if you completely fail to ABX them you might have a chance to test for actual differences of the speakers proper.
There is a reason why all professional blind speaker tests use a (semi-)automatic shuffler to load the exact same playback position for every speaker.

I tried, but I cannot tell the difference in height. This is at a 20 foot listening distance, and the tweeter height difference here is maybe a foot or two when stacking.

Have you tried distinguishing height differences like this? I find it vastly more difficult than horizontal differences, which are quite obvious.

Building a robot capable of shuffling a speaker as heavy as the Salon2’s is just not practical for me, so an alternate needs to be chosen. This vertical stack seems the best candidate so far.

Anyway, I will never be able to match the positions perfectly even if I could use a robot, because this is comparing a 3-way coaxial standmount with a 4-way non-coaxial floorstander. I suspect Revel’s robot does not normalize height differences between towers (though I don’t know this), because if you did you would have to choose a reference axis, and this wouldn’t really even make sense for towers.

For example, how would Revel make the Salon2’s tweeter as low as the F208? You just can’t do this unless you either raise the F208 way up (not how it was designed to be used), or dig a hole and sink the Salon2 into it (not how it was designed to be used). They were designed to be floor standing. So I would be surprised and skeptical of any test procedure that tries to mess with that.

 

[Beave]

The importance of doing speaker comparisons 'blind' is partially (mostly? totally?) negated when:
1. You're only doing two speakers
2. You know ahead of time what those two speakers are
3. You'll probably have little to no difficulty immediately determining which speaker is which, both by sound differences and by location differences.

All of that kinda makes the 'blind' part of the comparison a moot point.

You might as well just have a 'shootout,' without pretensions of doing it blinded.

I'd put them side by side, with tweeters near each other.

Level match with pink noise, bandlimited if possible to something like 200Hz to 2kHz or 4kHz, and measured as precisely as you can (you can probably get within a dB or so).

 

[echopraxia]

The importance of doing speaker comparisons 'blind' is partially (mostly? totally?) negated when:
1. You're only doing two speakers
2. You know ahead of time what those two speakers are
3. You'll probably have little to no difficulty immediately determining which speaker is which, both by sound differences and by location differences.

All of that kinda makes the 'blind' part of the comparison a moot point.

You might as well just have a 'shootout,' without pretensions of doing it blinded.

I'd put them side by side, with tweeters near each other.

Level match with pink noise, bandlimited if possible to something like 200Hz to 2kHz or 4kHz, and measured as precisely as you can (you can probably get within a dB or so).

As you can see from the measurements, they’re both pretty flat. No major colorations. Once I EQ bass, they should both be flat 10hz to 20khz, +-5db.

So I don’t think this will be a case of “no difficulty determining which is playing”, but I guess we will find out.

I don’t think knowing that there are two speakers will pollute the blind part of the test. It may magnify perceived differences, but ABX vs AB testing should at least somewhat mitigate that.

 

[bigjacko]

I have one questions about whether to eq above 200Hz or not. If both 8351 and salon 2 had same sweep condition but the result shows 8351 was flat and salon 2 was not flat, why not eq salon 2 to flat? I think if 8351 is flat, the room did not create unevenness, so the unevenness should came from salon 2 itself. At off angles the frequency response changes making perfect eq impossible, but I think we can still try average out the difference.

 

[Beave]

Yeah, they're both "pretty flat," but I bet they're still pretty easily distinguishable. On-axis isn't identical and neither is off-axis. A broad hump of 1dB in one or a broad dip of just 1dB in the other is enough to make them sound different. "Major colorations" aren't needed to make differences audible, though "major colorations" in one or the other (or both) sure make it easier to distinguish two speakers.

 

[echopraxia]

Yeah, they're both "pretty flat," but I bet they're still pretty easily distinguishable. On-axis isn't identical and neither is off-axis. A broad hump of 1dB in one or a broad dip of just 1dB in the other is enough to make them sound different. "Major colorations" aren't needed to make differences audible, though "major colorations" in one or the other (or both) sure make it easier to distinguish two speakers.

Even if they are clearly distinguishable, I still think it’s best keeping it as blind as possible. Because at least we won’t know which one is playing, and this will prevent brand/appearance/size/price or other biases from entering the picture and distorting impressions.

Also, if what you’re saying is true (+-1db making blind tests pointless), then all blind tests are useless because no two speakers are identical within 1db across all frequencies.

 

[KSTR]

@echopraxia , Yes, we tested for this, even under semi-anechoic conditions a few degrees of vertical source change are readily audible with wide-band signal (pink noise). With added floor and ceiling reflexion the comb filtering then gives really strong cues even for minor offsets.
Acoustical axis is where the upper midrange / lower treble is (1kHz....4kHz).

 

[echopraxia]

I have one questions about whether to eq above 200Hz or not. If both 8351 and salon 2 had same sweep condition but the result shows 8351 was flat and salon 2 was not flat, why not eq salon 2 to flat? I think if 8351 is flat, the room did not create unevenness, so the unevenness should came from salon 2 itself. At off angles the frequency response changes making perfect eq impossible, but I think we can still try average out the difference.

It’s tempting to try, but it’s hard to get right and probably would raise more criticisms of the test validity than it would solve.

For example, as I move the mic around the room, the Genelec remains impressively flat no matter what I do. The Salon2 frequency response fluctuates around a lot more, but this makes sense since it’s not a coaxial and so there will be more interference pattern ripples throughout the room caused by the distance between the drivers. However, I think it’s more of an open question as to how audible this is since our perception tends to do a good job of averaging this out in a way a fixed position microphone does not.

I could probably EQ the Revels reliably only if I did so based on CEA2034 or ideally anechoic or Klippel measurements. But since I am not equipped to do this, it’s probably best (and arguably most fair) leaving it alone aside from EQ’ing the bass response only, which is where all the major room issues appears anyway.

P.S. Tomorrow, I will try doing a many-position averaged measurement for these, which may be more useful to compare frequency response.

 

[echopraxia]

@echopraxia , Yes, we tested for this, even under semi-anechoic conditions a few degrees of vertical source change are readily audible with wide-band signal (pink noise). With added floor and ceiling reflexion the comb filtering then gives really strong cues even for minor offsets.
Acoustical axis is where the upper midrange / lower treble is (1kHz....4kHz).

Then does this invalidate all of Revel’s testing of floorstanding towers of differing height?

 

[Senior NEET Engineer]

The Genelec can be placed horizontally to bring the tweeters closer together. It may change the sound a bit.

 

[echopraxia]

The Genelec can be placed horizontally to bring the tweeters closer together. It may change the sound a bit.

Yeah I considered this. It’s tough to decide which is better, because Genelec’s published measurements show slightly better horizontal off axis response than vertical, though of course both are excellent as it is a coaxial design. But they are not 100% identical. So I’m not sure which is best — to bring the tweeters closer via horizontal orientation, or to leave vertical which most people will use and which has slightly better horizontal off axis smoothness.

Definitely open to thoughts/opinions here.

 

[Senior NEET Engineer]

Yeah I considered this. It’s tough to decide which is better, because Genelec’s published measurements show slightly better horizontal off axis response than vertical, though of course both are excellent as it is a coaxial design. But they are not 100% identical. So I’m not sure which is best — to bring the tweeters closer via horizontal orientation, or to leave vertical which most people will use and which has slightly better horizontal off axis smoothness.

Definitely open to thoughts/opinions here.

I wouldn't say vertical orientation is better. Horizontal will still be smooth. You'll get less ceiling reflections which don't contribute to spatiousness, and more sidewall reflections that do.

With far field listening, there is a lot more reflected sound. The size of the recording gets masked by the size of the room. Strong early reflections can help to get back some of the spaciousness.

I think one of the studies in Sound Reproduction showed that smoothness should not necessarily be prioritized over width in home listening scenario.

 

[echopraxia]

I wouldn't say vertical orientation is better. Horizontal will still be smooth. You'll get less ceiling reflections which don't contribute to spatiousness, and more sidewall reflections that do.

With far field listening, there is a lot more reflected sound. The size of the recording gets masked by the size of the room. Strong early reflections can help to get back some of the spaciousness.

I think one of the studies in Sound Reproduction showed that smoothness should not necessarily be prioritized over width in home listening scenario.

From these specs, it’s hard for me to see which would be wider: https://www.genelec.com/8351b#section-technical-specifications

From this, what orientation do you think would be best for this test? “Portrait” (default), or “landscape”? (To use slightly different terms to prevent confusion of vertical vs horizontal responses.)

 

[muad]

Hey echo,

Thanks for doing this shootout!

One thing to consider with the speaker stacked so closely above each other is the baffle step. The frequencies that would normally radiate up and over the top are now being reflected forward. It's essentially going to change the vertical response of both speakers. Side by side with some space between them may give you a better result.

I would guess the genelec would fare better in this regard as its designed to have stands or something below it. The revel definitely wasn't designed to have a flat surface above it.

 

[Senior NEET Engineer]

From these specs, it’s hard for me to see which would be wider: https://www.genelec.com/8351b#section-technical-specifications

From this, what orientation do you think would be best for this test? “Portrait” (default), or “landscape”? (To use slightly different terms to prevent confusion of vertical vs horizontal responses.)

For some reason, the specs don't agree with Amir measurements of 8341A. At 1000hz 70 degrees, there is 10 dB difference in Amir's measurement and none in Genelec's.

I don't think there is any definitive answer for which way to go. I just wouldn't rule out the landscape orientation, considering it also brings the tweeters closer.

 

[echopraxia]

For some reason, the specs don't agree with Amir measurements of 8341A. At 1000hz 70 degrees, there is 10 dB difference in Amir's measurement and none in Genelec's.

I don't think there is any definitive answer for which way to go. I just wouldn't rule out the landscape orientation, considering it also brings the tweeters closer.

I also worry about interfering with the baffle shape mentioned by @muad .

Maybe I should just use stands. That certainly would be more realistic. But my worry is how to solve the very audible horizontal position shift, since the Salon2’s are too heavy to constantly shuffle around horizontally in the room.

One thought is that I could place two Genelecs on each side of the Salon2, but only wire up one of them. Then, when the listener hears the relative horizontal shift, at least it won’t be obvious which of the speakers are being heard as long as the listener was not told which of the Genelecs is plugged in — since hearing a shift to the right (for example) will tell you nothing about which speaker you are now listening to. At a listening distance of 20ft I don’t think it will be easy to pinpoint in absolute terms which speaker is playing.

 

[q3cpma]

I would test in stereo, otherwise the Salon's wide dispersion will be dramatically advantageous.

 

[echopraxia]

I could also try a symmetric stereo setup with speakers configured: AB ———— BA.

This gives an unfair soundstage difference to the outer speakers, though I could make the Genelecs the outer ones to compensate for their narrower dispersion, which I think is still somewhat fair.

However as mentioned above, if I run a stereo comparison, I can't easily test with subs integrated, since my miniDSP SHD has no more than 4 programmable output channels.

 

[q3cpma]

I could also try a symmetric stereo setup with speakers configured: AB ———— BA.

This gives an unfair soundstage difference to the outer speakers, though I could make the Genelecs the outer ones to compensate for their narrower dispersion, which I think is still somewhat fair.

However as mentioned above, if I run a stereo comparison, I can't easily test with subs integrated, since my miniDSP SHD has no more than 4 programmable output channels.

You'd have to switch manually, but I'm afraid that both stereo and a sub are needed to make this useful: there's no "halfway useful" here. Not to mention that power handling differences may be heard too soon in mono.