New Record Day - Binaural Shootout (Revel, Klipsch, Spatial, Q Acoustics)

[richard12511]

Loaded all samples into foobar, and applied replaygain to the files. That fixed the loudness difference issue, but since tonal balance is faaaar off on #1, it is not possible to get this to sound equally loud no matter how you do it.

Have not heard of replaygain. Is it free? I was trying to manually turn up and down the volume, bit didn't know exactly how much to do.

For me 1 actually sounds the best(closest to the reference), 3 is second. Can't decide between 4 and 5, but leaning 4, then 2.

When I was watching the video through the first time, I thought 1 and 4 sounded the best(slight edge to 4 for the spaciousness. 2 and 5 were the only two I thought sounded awful. 5 had the same spaciousness as 4 but seemed like it was either emphasizing the bass and treble, or the mids were way off. Sounded like I was listening to 2 separate recordings playing at the same time. 2 was not as bad as 5 tonally, but much worse spatially. Seems are interpretations are very different, which is likely why the only useful thing to do is compare against the reference.

 

[Kvalsvoll]

Have not heard of replaygain. Is it free? I was trying to manually turn up and down the volume, bit didn't know exactly how much to do.

For me 1 actually sounds the best(closest to the reference), 3 is second. Can't decide between 4 and 5, but leaning 4, then 2.

When I was watching the video through the first time, I thought 1 and 4 sounded the best(slight edge to 4 for the spaciousness. 2 and 5 were the only two I thought sounded awful. 5 had the same spaciousness as 4 but seemed like it was either emphasizing the bass and treble, or the mids were way off. Sounded like I was listening to 2 separate recordings playing at the same time. 2 was not as bad as 5 tonally, but much worse spatially. Seems are interpretations are very different, which is likely why the only useful thing to do is compare against the reference.

The software is called foobar2000, install, open tracks, right-click, select Replaygain - Scan per track gain, then Apply replaygain to each track. When this is done, you can close foobar and play the files in your preferred player.

This makes it much easier to compare, when you can switch instantly between original and each speaker.

 

[ctrl]

This is not so simple. How the binaural recording represents the experience heard in-room depends on more factors than you can see in a simple frequency response. The response of this type of recording is very sensitive to sound field directivity.

Today's dummy heads are designed for diffuse field equalization. Headphones optimized to that target AFAIK are very rare (too bright sound with non-binaural recordings?). However, Sean Olive has shown that headphones with a target curve slightly different from the diffuse field equalization curve give better results.
Isn't that how Sean Olive developed his target curve? Binaural recording of the highest rated speaker in a listening room, then adjusting the headphone target curve until the sound matched that of the speaker?
After several iterations of the headphone target curve, one (untrained and trained listeners) can now be fairly certain that when the headphones are tuned to the Harman target curve, binaural dummy head recordings will sound close to what is heard in the recording room.

#1 is dynamically compressed, sounds soft. like transients are muted.

As others have said, this is due to the mismatched levels of the compared speakers. Listen to the binaural recordings of speakers I linked in the last post (here at lowbeats.de).
Compare the sound levels and dynamics of the Grimm LS1 and the Tannoy Canterbury. The Grimm would be around [email protected] sensitivity when built passively. The Tannoy is listed at [email protected], which is about the same level as the Spatial X5.
This is what it sounds like when the sound pressure levels are set fairly.

How would this be affected by listening to this on speakers? That's what I've been using for all of these comparisons.

Then you have people like me, who insist on listening to this on speakers.

Neumann claims that recordings with the dummy head can also be heard through loudspeakers. But I would doubt that, because actually the signal would have to be convolved with an inverse HRTF to be suitable for playback over loudspeakers.

When I listen through speakers, the "gramophone/megaphone sound" is clearly boosted with the Spatial X5 (much less pronounced than with the Klipsch, of course). The effect is less pronounced over closed headphones and least pronounced over in-ear.

 

[MrPeabody]

When adjusting the volume on the short samples by turning the volume control, the Revel and the QAcoustics (slightly less good) would be okay for me, the Spatial X5 doesn't come close tonally to the original. With the M3, I miss the midrange, it sounds "thin". The Klipsch sounds as hollow as an empty poured out can of Pepsi.

My impressions as well. I thought that the Revel and the QAcoustics were the only two that sounded very similar to the original, and while the difference was slight, I thought that the Revel sounded a little bit better. This didn't seem right so I listened a while longer and decided that the QAcoustics possibly sounded a bit more like the original although I still liked the Revel better. With both of the Spatial speakers, the bass didn't sound tonally correct to me. I kept hearing a low-level background noise that didn't seem to belong there. I couldn't quite put my finger on it at first, but after listening a while longer, to both the MP3 files and the Youtube video, I decided that it is probably a combination of uneven response and some amount of distortion. The uneven response is possibly attributable to a room effect, and given that it is a dipole, I think it likely that the strong reflection from the wall behind the speakers is introducing coloration. In any case I did not care for the sound of the Spatial speakers. They weren't bad speakers, just not in the same league as the Revel or the QAcoustics. As for the Klipsch, what can I say ... fish out of water? I was very surprised at how colored the sound was. Not the least bit pleasant to listen to. Even though I'm not convinced that the Revel is the most accurate, it is still the one that I would buy.

As richard12511 has pointed out, the level matching was not. Way too much variation in level from one speaker to the next, and in the MP3 files, the original is at a much higher level than the others.

Something else I noticed, which I have noticed before, is that after this kind of back-and-forth test for more than a couple minutes, something happens to my hearing. This has happened to me before on various occasions when auditioning speakers. At first I hear obvious differences and I have a clear sense of the differences and of which one sounds more natural to me. Then after a few minutes the differences seem to sort of fade, and after about fifteen minutes I wouldn't even be able to tell you which speaker I'm listening to.

 

[amirm]

Ron is using the Neumann Binaural recording head:

As a result, this lacks the torso portion that a proper HATS for sound evaluation would have. Specifically, reflections from the shoulder to the ear are not represented.

 

[MrPeabody]

I let my hearing have a rest then took another listen. I have a clearer understanding the differences. The QAcoustics is easily the one that sounds closest to the original. To match the level of the Revel to the original you have to turn the Revel up a bit, and once you do that, to where the overall level is closely matched to the original, it is apparent that the Revel applies an emphasis to the bass. Not unpleasant, but not accurate. And on the second listen, the deficiencies of the two Spatial speakers, and especially the Klipsch, were more apparent than previously. Both Spatial speakers sound very similar, with the same coloration. There is coloration in the bass, and a clear emphasis at higher frequency that gives the piano an unnatural sound, which some people might describe as "forward". And as for the Klipsch, well, it doesn't belong with this group of speakers. Even the two Spatial speakers outclass it by a wide margin. My thought about which of these speakers I would most want to buy has changed from the Revel to the QAcoustics.

 

[MrPeabody]

Ron is using the Neumann Binaural recording head:

As a result, this lacks the torso portion that a proper HATS for sound evaluation would have. Specifically, reflections from the shoulder to the ear are not represented.

Also, it looks like it is asleep, and there is no way for it to eat. Sorry Amir, I couldn't help it.

 

[amirm]

Also, it looks like it is asleep, and there is no way for it to eat. Sorry Amir, I couldn't help it.

It is meditating as to figure out which speaker sounds better to it.....

 

[Kvalsvoll]

Here is the masvis analysis, it shows tonal differences that correlates well with listening observations - #1 being more tonal correct because the 700hz peak is lower, #5 being overall more tonal correct because the excessive bass booOOooom is gone:

Due to the booOOOoom, level is not matched on #1 across mid-higher frequencies. Differences in measured DR is caused by phase shift which inevitably occurs when you play back on speakers in a room, in combination with use of limiters to push levels loUUDDD in the production.

Enjoy the graph.

 

[ctrl]

I let my hearing have a rest then took another listen. I have a clearer understanding the differences. The QAcoustics is easily the one that sounds closest to the original...

You really got into it!
I have to agree with @Kvalsvoll and you about the Revel, it seems like the Revel excites a room mode at 65Hz - because the measurements I know of show no abnormalities in that frequency range.

If we now had reliable measurements from all speakers, we could try whether the listener's impressions can be justified with the measurements.

 

[Kvalsvoll]

If we now had reliable measurements from all speakers, we could try whether the listener's impressions can be justified with the measurements.

In-room measurements (with full data) can tell something. But there are problems with this type of recording, and using speakers to listen, which the binaural was not made for.

A recorder better suited for speakers could give better results. I suspect much of the differences in tonality in midrange is caused by the binaural recorder.

When I listen ot these samples, I find my observations also vary when I use different systems for listening. The system currently in Room2 seems to work much better than the media room, much easier to hear how the sound changes.

I really would like this to work. To be able to hear what other people's systems sound like, people far away. And other speakers. But we know it can never be, at least never more accurate than our own system that we use to listen.

 

[Daverz]

Ron is using the Neumann Binaural recording head:

As a result, this lacks the torso portion that a proper HATS for sound evaluation would have. Specifically, reflections from the shoulder to the ear are not represented.

How did you guys miss the obvious Harlan Ellison reference? Not SF fans, I guess.

 

[TheInquiring]

I wonder how would quality planars like Quad ESL's or upper tier Magnepan's, for example, fare in this test...

 

[Propheticus]

Couldn't help it.

 

[ctrl]

As long as the binaural speaker recording is only ever compared to the original source, the comparison is pretty fair - this applies regardless of the headphones used, since only differences between the original and the recording are taken into account. So, one can rate how much the loudspeaker deviates from a "neutral reproduction".

On second thought, I wrote a lot of crap there.

With binaural recordings, the HRTF is included in the recording and is thus optimized specifically for headphone listening - even if the KU100 dummy head lacks the upper body to generate an "optimal" HRTF.

Here, the already often shown, graph about the composition of the HRTF. There you can also see the reason why the missing upper body hardly makes a difference (compared to the other influences) in the dummy head recordings:


The original source is in most cases not optimized for listening over headphones, but for stereo listening over loudspeakers and was mastered accordingly over studio monitors.

That means via headphones alone, the timbre of the original source (if it is not a binaural recording) just cannot be compared to the binaural recordings, because the HRTF is not included. Rather, the orignal source would have to be heard through neutral speakers in order to compare the timbre.

 

[Propheticus]

What about processing the original file through a HRTF function (convolution?) and comparing that with the spatial recording of these speakers?
Would that not tell the difference (timbre) the speakers add to the source?

 

[ctrl]

What about processing the original file through a HRTF function (convolution?) and comparing that with the spatial recording of these speakers?
Would that not tell the difference (timbre) the speakers add to the source?

Yep. That would solve the problem. In this case, the HRTF of the KU 100 dummy head would have to be used.
As far as I know, Neumann supplies additional software, perhaps there is already the possibility to perform such a convolution.

 

[Jim Matthews]

It is meditating as to figure out which speaker sounds better to it.....

It's dreaming of a vacation.

 

[Propheticus]

It's dreaming of a vacation.

 

[McFly]

I know it’s sighted and probably rubbish, but through akg371’s late at night I was genuinely surprised how much I could hear that klipsch cabinet humming, and how much i liked the Q 500’s. So much I started looking around for dealers to go have a listen. Then my wallet shrivelled up when I saw the price. Nice looking though makes the revels look a bit pedestrian.