Live vs. Recorded - can you hear a difference?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
During the XFI Premium Hifi Show (in the Netherlands) my company Dutch & Dutch and the audiophile music label TRPTK will do a test that as far as I know has never been done this thoroughly before.

3NcRTH0.jpg



We are going to do a comparative listening test, with music pre-recorded in an anechoic chamber and the same music played live in our demo room. We are going to listen to voice, a snare drum and cello.

High Fidelity basically is about reproducing a sound with the greatest level of fidelity. A direct comparison to live music perhaps is the hardest test imaginable for both the recording and playback system. I think it's very exciting!

When an instrument is played live in a space, you hear the instrument's sound plus a special sauce added to it, which is the interaction of the instrument with the acoustics of space. Thus, with our demo it is the aim to mimic that sound - the combination of a live instrument plus the sound of the room.

We can not do that with "normal" recordings, because those always include acoustics of the space in which the instrument was recorded. Should we use such a recording, you would get acoustics on top of acoustics. To record the sound of the instrument, we made recordings in the anechoic chamber of the Delft University of Technology.

The idea is that if we play this through an accurate playback system (Dutch & Dutch 8c), the sound in theory should match with live. Of course there are some caveats, the biggest of which probably is directivity.


What do you think? Can we 'fool' people and have them believe they are listening to live, while it's actually reproduction they're listening to?

Will it work better with one instrument than the other?

Short video of recording in the anechoic chamber
 
Last edited:

ra7

Member
Joined 2009
Paid Member
When you say "played live in our demo room" you mean played by musicians, correct? I think that has been done before, where musicians were playing in a room, and then they stop and act as if they are playing, but really it is the speakers, and the audience does not notice. It would cool to understand what happens in your demo.

You could also do another test, which is similar to one of Tom Danley's ideas. Once you have recorded the musician in the anechoic chamber, you can replay that recording through your speakers in the anechoic chamber and record the playback from your speakers. Then you can share these two recordings and ask people to identify which is which.

Please keep us posted on how it goes.
 
You could also do another test, which is similar to one of Tom Danley's ideas. Once you have recorded the musician in the anechoic chamber, you can replay that recording through your speakers in the anechoic chamber and record the playback from your speakers. Then you can share these two recordings and ask people to identify which is which.
But only people familiar with the sound of real instruments live should take part otherwise it's a pointless exercise, and it would be better if it was done in a normal sounding room
 
Won't sound the same. Escpecially a cello. An instrument is a very directionally dependant source. Let's say it emits sound off the back that has very little high freqs. This sound will reflect off the walls and thicken up the lower freqs. At the listeners position. Your recording won't have these because the anechoic room has absorbed them all.
 
Moderator
Joined 2011
To record the sound of the instrument, we have taken recordings in the anechoic chamber
of the Delft University of Technology. Can we 'fool' people and have them believe they are
listening to live, while it's actually reproduction they're listening to?

Acoustic Research did many such demonstrations in the 60s. The music was recorded out of doors,
instead of in an anechoic chamber. http://www.audiocircle.com/image.php?id=32179
 
What do you think?

HRTF might be a challenge. A centrally located single source like a cello will interact with a listener's head/ears differently than a spaced pair of speakers presenting a phantom mono image. The cello is also large enough to potentially present height information. A better starting point might be a point source like a flute, snare, triangle etc. played back through a single speaker before moving on to stereo.
 
Won't sound the same. Escpecially a cello. An instrument is a very directionally dependant source. Let's say it emits sound off the back that has very little high freqs. This sound will reflect off the walls and thicken up the lower freqs. At the listeners position. Your recording won't have these because the anechoic room has absorbed them all.

^this

the proposed experiment is flawed
 
Do the test so that the source angle between top and bottom of the sound source are the same. So the speaker will be close and the cellist far away.

Otherwise the most basic difference will the size of the thing the sound is coming from - easy to spot from a close listener in the average show demo room. With a snare, the dynamics will come from a point source innitially so not so much of a problem and the drum small anyway.
 
Administrator
Joined 2004
Paid Member
We did some demos like this in France back in the 1980s. We did not do the recordings in an anechoic chamber, but in well treated rooms or small theaters. Much of the success or failure depended on the mic technique, the hyper-cardioid mics being the most successful. With no acoustics in the recording venue, you may have more leeway in the mic placement and pattern. Playback level is important for a successful imitation.

Good luck and let us know the results! :up:

Unlike most here, I think you will do really well and have a high chance of fooling people.
 
How do I get a copy of your recordings?

Are you planning to use stereo or mono? Mono should be easier. Less dynamic instruments should be easier to match. The closer the match between the directivity of the instrument and the speaker the better.

It's hard to predict because I haven't heard any recordings done in an anechoic chamber.

I do have a friend with a home studio who ended up having a conversation with a recording, because he had accidentally left the mic on earlier in the day and captured spoken voice. When he played it back he thought the person he had recorded was talking to him. He was distracted at the time though, and not questioning what he was hearing.

You'll probably have better success if your demo room has significant acoustic treatments. Pictures of the room to see the size and treatments would be nice.

How do you plan to determine whether or not people are fooled? Document their guesses and compare to reality like ABX?
 
To me it looks like keyser is recording the cellist with a single mic. Then this recording is played back with the speaker and recorded again. I guess that both recordings will sound similar, but that is because of the anechoic room.

This test has no relevance to practical sound reproduction, a sine sweep would tell the same.

Dr. Tapio Lokki and his team are doing some really interesting research.
Check also www.davidgriesinger.com

Recording and mixing acoustic music is art in itself The Decca Sound: Secrets Of The Engineers – The Polymath Perspective
 
Last edited:
First of all one has to know what one is doing. What is the stereo sound? How does this illusion work?

We need a theory.

For example David Moulton has a theory of how speakers in stereo work:

loudspeakers themselves are perceived in stereo as early reflections of a sound whose direct version we missed. So we're listening to the first of a set of early reflections

Sounds right BUT OTOH the first of a set of early reflections is not what a stereo mic records. So there is fundamental inconsistence between what the mic records and how the loudspeakers are perceived.

Anyone has a better theory?
 
Thanks for all the questions and tips!

Let me start with some general clarifications.

The room in which we'll be demoing has dimensions of (l*w*h) 8.6 * 6.2 * 3.1 meter. It has clipped carpet on the floor, a suspended ceiling system you often see in office buildings, there are drapes by the windows.

We recorded the musicians in the anechoic chamber and we'll be comparing the recordings played back over an accurate system (Dutch & Dutch 8c), with the same musicians playing live in the room.

We'll be using an LCR setup, with primarily mono feeds from a single microphone to the speakers.


You could also do another test, which is similar to one of Tom Danley's ideas. Once you have recorded the musician in the anechoic chamber, you can replay that recording through your speakers in the anechoic chamber and record the playback from your speakers. Then you can share these two recordings and ask people to identify which is which.

Please keep us posted on how it goes.

That's actually a great idea. We didn't do those recordings now unfortunately, but I'll make some recordings next time I'm there. Given how flat and smooth the 8c's measure, I think it will be very difficult to hear any difference between an original and a recorded version. Perhaps after several loops? I imagine the cumulative noise will give it away at some point.

Dynamics, phase, directivity, acoustics.

Too many variables.

I'm not so worried about dynamics and phase. The speakers used have wide dynamic range and exemplary phase response. I'm not really worried about the acoustics either, because they're the same for the live instrument and the reproduced sound.

I'm mostly worried about directivity.

Especially the cello has a very complex radiation pattern, so there is no one microphone position that fully captures the character of the instrument. That imperfect recording is then played back over a speakers with a very even dispersion pattern, which is then directly compared to a live cello. I suspect the combination of an imperfect recording and non-matching directivity will make the cello the easiest one to recognize for being not live.

The spoken voice and the snare are probably easier to pull off.

Won't sound the same. Escpecially a cello. An instrument is a very directionally dependant source. Let's say it emits sound off the back that has very little high freqs. This sound will reflect off the walls and thicken up the lower freqs. At the listeners position. Your recording won't have these because the anechoic room has absorbed them all.

I'm afraid you're 100% right. We did make a couple of recording with a panel right behind the musician though, to capture a virtual front-wall reflection in the playback space. That's because the directional speakers themselves don't excite a front-wall reflection, while the live musicians do.

HRTF might be a challenge. A centrally located single source like a cello will interact with a listener's head/ears differently than a spaced pair of speakers presenting a phantom mono image. The cello is also large enough to potentially present height information. A better starting point might be a point source like a flute, snare, triangle etc. played back through a single speaker before moving on to stereo.

You're right. For that reason we'll mainly do mono playback over a single speaker. Good point about the height information. I'm not sure how to work around that.

We also recorded the sound of the rattling of keys on a chain by the way. We checked it on the monitoring system and the recording comes eerily close to live!

^this

the proposed experiment is flawed

Yes it is. We're trying to do the impossible, and do it as good as we can!

I remember a "work coach" saying, "aspirations are things you aspire to"

Thanks Scott :D .

We did some demos like this in France back in the 1980s. We did not do the recordings in an anechoic chamber, but in well treated rooms or small theaters. Much of the success or failure depended on the mic technique, the hyper-cardioid mics being the most successful. With no acoustics in the recording venue, you may have more leeway in the mic placement and pattern.

Nice! Did you do it for a hifi show as well?

Playback level is important for a successful imitation.

We've done some exploratory tests with the recordings and I think you're absolutely right. I was amazed at how critical it is to get the level exactly right for especially the spoken voice. It's really a matter of tenths of a dB, for it to sound really natural.

Good luck and let us know the results! :up:

Unlike most here, I think you will do really well and have a high chance of fooling people.

Thanks Pano, I will. I really hope we will be able to fool the people with some of the sounds. That would be the most interesting and entertaining outcome for everyone involved.

What speaker? Surly not the crappy cone and dome things in the photo, right? It might work with an MBL.

Unfortunately it's the crappy speakers in the photo. To be honest, I don't really think they're crappy (I actually think they're among the most accurate you'll find), but you're probably right in that an omnipole would be a better speaker for this particular job. That's because an omni has a dispersion pattern that's closer to that of the snare, the keys on a chain and the cello. To accurately reproduce spoken voice, a conventional 2-way box the size of a human head would probably be best.

We're trying to create the illusion of "here and now". The speakers we're using are better equipped to create an illusion of "there and then", which is an even more difficult one to pull off.

How do I get a copy of your recordings?

I'll make the recordings available for download after the test.

How do you plan to determine whether or not people are fooled? Document their guesses and compare to reality like ABX?

I'm planning do indeed do a blind test. Arnold Krueger pointed me to this material: Scrim (material - Wikipedia)

The idea is to have people write down the exact moment in time there's a switch from live to reproduced, or vice versa. Those who get it right will for instance win an album from the music label we're doing this with, TRPTK (triptych).
 

Attachments

  • 8c aanzichten.jpg
    8c aanzichten.jpg
    406.8 KB · Views: 353
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.