Binaural blind comparison test of 4 loudspeakers - II

[ctrl]

As far as I know one would need headphones that have binaural target curves or room acoustics that show a neutral linear music power vs frequency (thus enabling using headphones with Harman target).

But yes, obviously if the binaurally recorded speaker has a drop that is 1) faithfully recorded by the microphones but 2) compensated by a peak on the headphones used for comparison... any analysis is off the table.

I don't know if I understand you correctly, there are several types of binaural recordings (but beware, my knowledge on this subject is limited).

One contains a complete HRTF (for example KEMAR dummy head) and cannot be listened to with "normal" headphones if not further processed, because the "normal" headphones themselves reproduce the frequency response of a HRTF. So you would hear the HRTF twice, so to speak - once contained in the binaural recording and once generated by the headphones to simulate the human outer ear.
Such a recording requires specially equalized headphones.

But if you record with normal stereo microphones and only take care of the physical separation of the microphones (by a head model or a wooden plate), there is no (or no complete) HRTF in the recordings, but for example only the delay that our head causes.

Or the dummy head with the included microphones is itself calibrated again to a "flat" playback curve, allowing the binaural recording to be listened to with headphones or loudspeakers, but with headphones the spatial information contained in the binaural recording is reproduced more realistically - this is what the Neumann KU100 dummy head does.

Here is a comparison of two different dummy heads and HRTF of subjects. Of course, the tiny recording microphone could not be placed directly in front of the eardrum of the test subjects, therefore their HRTF is always "erroneous", since the ear canal resonance is partially missing.
You can nicely see that the KU100 is calibrated for almost flat playback, this prevents the HRTF from being applied a second time when listening with headphones.

Source: A Perceptual Evaluation of Individual and Non-Individual HRTFs: A Case Study of the SADIE II Database

 

[PierreV]

I don't know if I understand you correctly, there are several types of binaural recordings (but beware, my knowledge on this subject is limited).

to recap, I purchased a cheap set of binaural microphones, Roland CS10-EM and used my own head.
Edit: I think @Blumlein 88 suggested potentially better microphones but they weren't available locally.

 

[ctrl]

to recap, I purchased a cheap set of binaural microphones, Roland CS10-EM and used my own head.

Okay, that means the microphones were just next to your head and so there is no full HRTF in the recording, but spatial information due to the separation of the microphones.
The recordings used here for the poll were made in a similar way.

 

[ctrl]

Like: "clip1-> original - speaker A- speaker B- speaker C - Speaker D" in one file. and the same for clip 2 and 3. Or make separate files for each speaker and clip. The shorter the clip the better and the faster one can switch the easier the comparison.

You could be right.

I did not find any reasonable way to compare them so I did not vote.

Every vote counts and every vote not cast is a vote for the manufacturer you don't like

If the four Short_Samples are in the audio player, with a little practice you can jump back and forth between the samples quite quickly.

 

[RHO]

Every vote counts and every vote not cast is a vote for the manufacturer you don't like

I only own DIY speakers.
Still, eventually I did cast a vote.
It took me some time to hit the progress bar in the right spot to get a good comparison but I managed to pick one.

 

[Thomas_A]

Okay, that means the microphones were just next to your head and so there is no full HRTF in the recording, but spatial information due to the separation of the microphones.
The recordings used here for the poll were made in a similar way.

The question is then how the specific room curve is. A boosted bass and a drop in HF a la Harman room curve? Adding a headphone with a Harman target curve will then also add the room ”twice”, once in the recording itself and secondly during playback. It must be some recommendation of the headphone target response. Free or diffuse field?

 

[ctrl]

The question is then how the specific room curve is. A boosted bass and a drop in HF a la Harman room curve?Adding a headphone with a Harman target curve will then also add the room ”twice”, once in the recording itself and secondly during playback.

The Harman (update: headphone) target curve is, from the trajectory, primarily a special HRTF, only the bass boost by the "non-experts" in the listening evaluations, could be a problem. But most headphones do not have this bass boost.

A binaural recording with a pair of diffuse-field equalized microphones (attached to the head) should thus provide a frequency response similar to that of the KU100 (with some interference through head and torso).

And yes, a room that is too damped, for example, would be heard on the binaural recording.

 

[LTig]

I don't know that it is a regular problem with binaural, but in this case I also hear a narrowing of the stage.

Of course we do, because the original is not a binaural recording and hence sounds wider in a headphone without crosstalk.

 

[LTig]

With the 1st and 3rd clip it was not too difficult to select the speaker which was the least wrong regarding tonality.

Spoiler

All speakers seemed to suffer from a loss of mid and upper mid frequencies. Two sounded hollow as well, and two had problems with resonances in the lower bass in so far as they were not existent in the original.

The speaker I finally chose had a smaller loss of mids and did not show resonances in the low bass (bass drum slam). The latter was my main criterium because the differences were gross (my hearing is probably not good enough to judge higher frequencies reliably). I think it's a smaller bookshelf which does not excite those low frequencies in the first place.

I understand that others chose different exactly because of more low bass. If the task had been to chose for preference I would have chosen the favorite because I like low bass. That one could be a floorstander or a bigger standmount with active equalization to get a lower f3 in the bass

I had more problems with the piano clip, probably because its frequency range is smaller, especially in the lows.

 

[LTig]

@ctrl : Please do first reveal the result of the poll and let's discuss what people think the speakers are, before you reveal them some time later.

 

[Thomas_A]

The Harman target curve is, from the trajectory, primarily a special HRTF, only the bass boost by the "non-experts" in the listening evaluations, could be a problem. But most headphones do not have this bass boost.

A binaural recording with a pair of diffuse-field equalized microphones (attached to the head) should thus provide a frequency response similar to that of the KU100 (with some interference through head and torso).

And yes, a room that is too damped, for example, would be heard on the binaural recording.

Hm but my understanding that even headphones with flat bass (e.g. Etymotic) has a decreasing HF response. Adding the typical room response of a domestic room (both due to speaker directivity not being flat 360° and the absorption in the room at higher frequencies) will typically lead to a darker sound, unless the headphones are calibrated for completely flat. Below my room response vs. "Harman" trained listeners. If I do a recording with microphones sitting close to my ears, I get a raw file that is too dark when listened through a pair of headphones. Even adjusting the low bass will not do - I have to boost the 1-10 kHz range by 4-5 dB.

 

[Blumlein 88]

Hm but my understanding that even headphones with flat bass (e.g. Etymotic) has a decreasing HF response. Adding the typical room response of a domestic room (both due to speaker directivity not being flat 360° and the absorption in the room at higher frequencies) will typically lead to a darker sound, unless the headphones are calibrated for completely flat. Below my room response vs. "Harman" trained listeners. If I do a recording with microphones sitting close to my ears, I get a raw file that is too dark when listened through a pair of headphones. Even adjusting the low bass will not do - I have to boost the 1-10 kHz range by 4-5 dB.

View attachment 159953

Even for binaural, above 1 khz there is a diffuse sound field with much more reflected vs direct sound. In person our hearing manages to filter out much of the early reflections so the ratio of reflected vs direct sound we hear is much less tilted toward reflected.

So if you boost your recording above 1 khz it can make the balance sound closer to what you hear in person, but you've also boosted some reflected energy which you'll hear over headphones. The end result still not being the same sound as you perceive with your ears.

 

[ctrl]

If I do a recording with microphones sitting close to my ears, I get a raw file that is too dark when listened through a pair of headphones. Even adjusting the low bass will not do - I have to boost the 1-10 kHz range by 4-5 dB.

Do your microphones have a flat diffuse field frequency response?
If not, you would have to compensate accordingly. Could this be the reason for the required treble boost?

The calibrated diffuse field frequency response of a KU100 is very linear, I could imagine that normal microphones cannot realize this.

 

[Thomas_A]

Do your microphones have a flat diffuse field frequency response?
If not, you would have to compensate accordingly. Could this be the reason for the required treble boost?

The calibrated diffuse field frequency response of a KU100 is very linear, I could imagine that normal microphones cannot realize this.
View attachment 159955

The microphones are the Line Audio Om1.

Line Audio Design - Made in Sweden

Line Audio Design

www.lineaudio.se

 

[PierreV]

Even for binaural, above 1 khz there is a diffuse sound field with much more reflected vs direct sound. In person our hearing manages to filter out much of the early reflections so the ratio of reflected vs direct sound we hear is much less tilted toward reflected.

If I understand you well, that means that in these binaural recordings, when we hear a FR that sounds closer to what we are used to hear with our real-life speakers, we are probably picking speakers that we would find too bright if we listened to them in person?

 

[ctrl]

The microphones are the Line Audio Om1.

Pretty impressive specs, almost too good - better than a Class 1 mic.
The frequency response shown seems to me to be the free field frequency response, I can't see the diffuse field frequency response anywhere.

The difference between free and diffuse field frequency response (calibration files) looks like this for my microphone (normalized to free field frequency response)

 

[Blumlein 88]

If I understand you well, that means that in these binaural recordings, when we hear a FR that sounds closer to what we are used to hear with our real-life speakers, we are probably picking speakers that we would find too bright if we listened to them in person?

I would think that is quite possible yes.

 

[ctrl]

I was asked to give you some time to speculate what kind of speakers it could be.
So go ahead! I will name the speaker models in about nine hours.

Like, for probably everyone, no speaker was perfect. I basically have a stalemate between LS A and LS D - sometimes one sounded better, sometimes the other.
I tend to LS D after I evaluated the Short_Samples, but since I know all the models, I am not unbiased.

 

[thewas]

Since the voting is closed, here are my listening impressions:

Speaker A sounds quite balanced on all tracks.

Speaker B doesn't sound bad either, has the most low bass which is a bit too much without room correction.

Speaker C is the most coloured in the mids (muffled and boomy) and the only one that doesn't sound good to me in the third track.

Speaker D sounds pretty similar to A.

In terms of sound quality, the third track is the best, but unfortunately, as is usually the case with such audiophile recordings, there were no single male of female vocals used which imho are best at picking up midrange colourations.

It wasn't an easy choice, I just decided on A based on the third song (D being very close), with room EQ I might have taken B instead.

Thank you @ctrl for another interesting and fun experiment, looking forward to the revealing of the loudspeakers.

 

[Blumlein 88]

I once again chose the 2nd most favorite speaker. Listening over phones it was down to A or D for me. Of the other two C was just out. B received more consideration from me, but in the end A and D seemed to be more even and covered more bandwidth.

D has the wider more spacious stage, and seemed the choice at times. A was a narrow stage and some less air to it by a little bit. Otherwise it seemed to get the most inner detail and the meat of the midrange down closer to the reference original track. Without the original track to compare I might have settled on D, but I felt it was less accurate across the midrange than A. The difference wasn't large between those.