• 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!

Ideal speaker for reproducing voice

OP
A

ajd578

Member
Joined
Sep 27, 2022
Messages
22
Likes
4
I agree, not in an A/B comparison anyway. But able to fool a listener into thinking there is a live, unamplified singer in the room, without direct comparison? Absolutely.
Has such an experiment been conducted? If not, how would it be?

What about this: record a set of N individuals speaking/singing P passages Q times, for N*P*Q recordings. Set up a room such that the loudspeaker/individual are positioned at the same height, ~15 degrees left/right of the listener, or vice versa. The speaker plays one of the recordings, the real individual performs the same passage live, and the blinded listener has to report which was which. Repeat for each N*P*Q recording (and n listeners), randomizing the placement and order of the sound sources on each trial.
 

theREALdotnet

Major Contributor
Joined
Mar 11, 2022
Messages
1,202
Likes
2,078
Has such an experiment been conducted? If not, how would it be?

What about this: record a set of N individuals speaking/singing P passages Q times, for N*P*Q recordings. Set up a room such that the loudspeaker/individual are positioned at the same height, ~15 degrees left/right of the listener, or vice versa. The speaker plays one of the recordings, the real individual performs the same passage live, and the blinded listener has to report which was which. Repeat for each N*P*Q recording (and n listeners), randomizing the placement and order of the sound sources on each trial.

Sounds like a plan. And while we have the facilities and test subjects, let’s investigate different scenarios wrt speaker placement and room treatment. Also, the impact of mono vs stereo on perceived realism.
 
OP
A

ajd578

Member
Joined
Sep 27, 2022
Messages
22
Likes
4
Sounds like a plan. And while we have the facilities and test subjects, let’s investigate different scenarios wrt speaker placement and room treatment. Also, the impact of mono vs stereo on perceived realism.
Definitely easiest to start with mono, but room treatment and speaker placement could be interesting. If you did it in a perfectly anechoic room, then I think this would be a test of what direct response characteristics are needed (frequency, phase response, and dynamic range only, I think). If reflections are allowed, then the perceived realism may depend on proper radiation pattern. I think?
 

theREALdotnet

Major Contributor
Joined
Mar 11, 2022
Messages
1,202
Likes
2,078
Definitely easiest to start with mono, but room treatment and speaker placement could be interesting. If you did it in a perfectly anechoic room, then I think this would be a test of what direct response characteristics are needed (frequency, phase response, and dynamic range only, I think). If reflections are allowed, then the perceived realism may depend on proper radiation pattern. I think?

Yes, but I think the tests should be conducted with a realistic listening environment in mind, i.e. a home setup (perhaps not a typically one but rather a well thought-out one). I don’t think pitching a singer against a loudspeaker in an anechoic chamber will reveal much in the way of realism. Then again, who knows? But it would make the test impractically expensive, you’d have to fit all the test subjects into the anechoic room somehow. Which would make it far less anechoic.
 
OP
A

ajd578

Member
Joined
Sep 27, 2022
Messages
22
Likes
4
Yes, but I think the tests should be conducted with a realistic listening environment in mind, i.e. a home setup (perhaps not a typically one but rather a well thought-out one). I don’t think pitching a singer against a loudspeaker in an anechoic chamber will reveal much in the way of realism. Then again, who knows? But it would make the test impractically expensive, you’d have to fit all the test subjects into the anechoic room somehow. Which would make it far less anechoic.
I think you would at least need to make the recordings under anechoic conditions, or else they would capture the acoustics of the recording space, serving as a cue for the listener.
 

kongwee

Major Contributor
Joined
Jan 22, 2022
Messages
1,024
Likes
276
Just to add, the vocal in song are highly processed, you wouldn't know the actual timbre of that singer. You can see plenty how it done in youtube.
 
OP
A

ajd578

Member
Joined
Sep 27, 2022
Messages
22
Likes
4
Just to add, the vocal in song are highly processed, you wouldn't know the actual timbre of that singer. You can see plenty how it done in youtube.
Yeah, the recordings would need to be raw.
 

CREMA

Member
Joined
Jan 24, 2020
Messages
21
Likes
57
Location
South Korea
Why is that? A natural voice have wide directivity.
I was mistaken listening environment for a speech, or a for a monitoring. If he's only playing the raw vocals, it'll be far from the studio sound recommendations.

It will depend on the listening space and listenining distance, but a speaker with a wide directivity will definitely be good.
 
Last edited:

CREMA

Member
Joined
Jan 24, 2020
Messages
21
Likes
57
Location
South Korea
I don't see why reflections should be low - are you saying you need a specific room to achieve this? A person sounds like a person regardless of the room they are in.
I misunderstood the meaning of the post. If you only play the raw vocals, it would be better to keep the reflections rich and wet.

This is because the voice sounds terrible when you play the raw vocal source in a dry room.
 
OP
A

ajd578

Member
Joined
Sep 27, 2022
Messages
22
Likes
4
I misunderstood the meaning of the post. If you only play the raw vocals, it would be better to keep the reflections rich and wet.

This is because the voice sounds terrible when you play the raw vocal source in a dry room.
Thanks, but to be clear, the goal isn't for the voice to sound good - just real.
 

Jim Shaw

Addicted to Fun and Learning
Forum Donor
Joined
Mar 16, 2021
Messages
616
Likes
1,160
Location
North central USA
I've read that no speaker can be perfect - every speaker is an optimization to a specific goal, with necessary compromises.

So if the goal was to have a speaker through which playback of a vocalist *alone* was indistinguishable from a live, unamplified singer, to a blinded listener, what measurement characteristics would that speaker have, and how would it be best approximated with real components?
It's a damn good question. The human voice typically uses less than half the spectrum we need for much music. And it typically requires only a part of the dynamic range we tend to think we need. But there are some major issues with reproducing it, to wit:

Voice recordings are often badly distorted and falsely embellished when the microphone is very near the mouth. There are effects of proximity -- the false emphasis on certain frequencies and bands, and wide variations in transient reproduction.

This closeness has some (or a lot) of the result that what is recorded is as if the singer/speaker were 3 inches from your ear, and not heard from several feet away, minimum, as you would usually hear them live without amplification. Sort of like having a singer sing into your ear while you wear hearing protection to keep the volume below pain. The typical mic does strange and wonderful things to an up-close singer. The result is unreal.
[It's why classical concert vocal soloists are generally miked at least 3-4 feet away, while pop concert singers are often holding a mic at 1-3 inches.]

Then there's Autotune. In classical concert recordings, we never use Autotune on singers. At pop concerts and studio pop recordings, we always use Autotune. -Always. When a singer is singing 3" from your ear, pitch becomes important. Autotune actually becomes part of the vocal character of the singer.

Then, there are a ton of producer-friendly studio tricks available on pop music recordings: Compression (attack, hold, decay), limiting, reverb, EQ, Pultecing, and a sea of other tricks up the producer's sleeves. If they succeed, the singer sounds 'better' than life to most, but at the cost of non-reality.

And, there's commonly a 'singing booth' used in pop recording. It offers its own brand of jams and jellies to the bread and butter of the voice.

I understand your wishes. But with the vagaries of voice recording, it's a tough errand to find a solution. I'll be interested in others' comments, especially those who practice the art of vocal recording.

Sidebar: I once worked for a studio owner/producer/recordist who typically recorded female vocalists with a U87 (male vocalists with maybe a 44B or other). Especially with the U87, he could 'hear' the slightest off-axis position of the mic. I would regularly get told to " go in there and rotate the vocal mic 3 degrees to her left" or such. Sure enough, the vocal character would change noticeably. Such is the perception of a musician. With such as the 44B, his concern was more about the backside of the mic.
iu
iu
 
Last edited:

Philbo King

Addicted to Fun and Learning
Joined
May 30, 2022
Messages
669
Likes
877
This is like the medieval arguments about how many angels can fit on the head of a pin. And just as useful.
 

dfuller

Major Contributor
Joined
Apr 26, 2020
Messages
3,414
Likes
5,259
That seems like a fairly limited specification. In essence, Amir’s ideal speaker has a flat frequency response and smooth directivity all around. Nothing else matters. Compression, phase, time alignment, stereo performance, pair matching and other parameters are not evaluated. It’s because these things are not mentioned (or are dismissed) in the scripture of St Toole of Harman.
Man basically every speaker has massive phase rotation around crossovers. It really doesn't matter, ultimately. Stereo performance is influenced heavily by mono. Pair matching is important, yes, but impractical to test. Compression should be measured but isn't.
 

Purité Audio

Master Contributor
Industry Insider
Barrowmaster
Forum Donor
Joined
Feb 29, 2016
Messages
9,192
Likes
12,489
Location
London
Contemporary active designs are phase coherent at the crossover and Erin I believe tests for compression.
Keith
 

Jim Shaw

Addicted to Fun and Learning
Forum Donor
Joined
Mar 16, 2021
Messages
616
Likes
1,160
Location
North central USA
Compression should be measured but isn't.
Compression testing is missing in action and greatly lamented. I suppose this test parameter died to speed things up in the test.
But if you are trying to fill something much bigger than a 10' x 12' room, compression matters a lot.
 

dfuller

Major Contributor
Joined
Apr 26, 2020
Messages
3,414
Likes
5,259
Contemporary active designs are phase coherent at the crossover and Erin I believe tests for compression.
Keith
Erin does test for compression. I'm absolutely not sold on linear phase crossovers being necessary or even useful.
 

preload

Major Contributor
Forum Donor
Joined
May 19, 2020
Messages
1,559
Likes
1,704
Location
California
I've read that no speaker can be perfect - every speaker is an optimization to a specific goal, with necessary compromises.

So if the goal was to have a speaker through which playback of a vocalist *alone* was indistinguishable from a live, unamplified singer, to a blinded listener, what measurement characteristics would that speaker have, and how would it be best approximated with real components?
Interesting exercise. It's partly a trick question, because it would also depend on the recording conditions. For starters, it would probably matter a great deal more that the total system frequency response from mic->playback was perfect than the individial frequency response of the loudspeaker.
 

dfuller

Major Contributor
Joined
Apr 26, 2020
Messages
3,414
Likes
5,259
I've read that no speaker can be perfect - every speaker is an optimization to a specific goal, with necessary compromises.
This is true.
So if the goal was to have a speaker through which playback of a vocalist *alone* was indistinguishable from a live, unamplified singer, to a blinded listener, what measurement characteristics would that speaker have, and how would it be best approximated with real components?
You'd want very flat, very low distortion, very even midrange dispersion. You don't need tons of LF extension - F3 of ~80hz is probably good enough.
 

Jim Shaw

Addicted to Fun and Learning
Forum Donor
Joined
Mar 16, 2021
Messages
616
Likes
1,160
Location
North central USA
Contemporary active designs are phase coherent at the crossover and Erin I believe tests for compression.
Keith
Erin is currently off the air.
Apparently, Amir doesn't consider compression worthy of testing and reporting. But relatively few speakers currently sold can do even 96 dB at 1 meter without a bunch of compression and distortion. (A symphony orchestra can readily achieve 115-120 dB at the front row.)
 
Top Bottom