Flat Response

[DDF]

Re: Flat Responce

3Khz? The human voice is fundamentally from about 80Hz-800Hz .. I don't believe this business about 1-3Khz. Listen to a sine wave at those frequencies and tell me you've heard someone speak that high :eek: ... The harmonics in my voice probably don't get above 320Hz, besides things like glottal noises and fricatives.

Vocal intelligibility is based - at least in english - in vowels; a dipthong, for instance, is not going to include high harmonics like the "ess" or "t" noise, which are the things that get amplified in the 1-3Khz range.

100 yrs of scientific research in the area proves otherwise. Thems the facts. I used to work professionally in telecom audio. The intelligibility curves are dead on, boost 2 to 5 khz to enhance it. There are many papers available on the Internet on the topic.

Its why standard phone voice bandwidth extends to 4 kHz. Extended bandwidth plays out to 7 khz. The added 3 kHz even makes the channel sound louder (3 to 6 dB) than the 4 kHz bandwidth, with the same channel gain. All arrived at through piles of DBT testing (some of it ours).

[mzisserson]

Re: Flat Responce

I am surprised this has not come up. Here is how our ears precieve loudness at frequency.

http://en.wikipedia.org/wiki/Fletcher%E2%80%93Munson_curves

I do not believe (and it has been proven by many commercial designs) That flat anechoic response does not equal good sound, or perceived sound.

I DO agree that flat power response is critical.

I believe that lobing should be controlled.

I also believe that it is critical to consider things like speaker-to-room transition and HRTF's when designing.

I do not think a flat speaker sounds good, it is a one-up thing. It will only sound good, in one specific application set up absolutely perfectly. As with any speaker, you move it just a millimeter, and the "flat" is destroyed. It is unrealistic for normal speakers, even in the high-end world.

To me, it is far more important to get lobing, timbre, voicing, bass response (considering room placement), diffraction and step right. If you do these things right, you will have a listenable speaker with a great presentation in any environment.

[UGP]

Re: Flat Responce

as far as the 3k voice thing goes all I know(well not actually all) is that when I mix and if you do too much de-essing as 3-4k voices will sound muffled, dull, pushed back... I'm sure you've listened to a woof/mid combo with out a tweeter and noticed this, even with just speaking voice's.

[dlr]

Re: Flat Responce

Reductio ad absurdum this early?
C'mon Pete, let's let it play out a little longer before pulling that card eh?

So how did your system sound in the bathroom?
Did you listen to it in the refrigerator also? ;)

p.s. Remember, this is another "A" flat response thread

Do you guys really intend to make every thread a pi**ing match? Is it possible to enter into it as a simple debate of ideas rather than trying to force someone else to submit (and we know neither side is willing to concede anything anyway)?

We are at a site where not every person desires the same kind of system nor does everyone here have the means for "a good system (that) needs no room treatment". Some are even going to want a simple 2-way, imagine that. Not going to be a system that needs no room or placement consideration.

Debate ideas, not personalities. It doesn't really matter who "pulled the card" first, all of us know that's where it's always headed. This sort of tiresome posting is making this place a lot less interesting and certainly less friendly that it used to be. I've already noticed a drop in participation by some I consider to have been extremely helpful over the years.

dlr

[billfitzmaurice]

Re: Flat Responce

100 yrs of scientific research in the area proves otherwise. Thems the facts. I used to work professionally in telecom audio. The intelligibility curves are dead on, boost 2 to 5 khz to enhance it. There are many papers available on the Internet on the topic.

Its why standard phone voice bandwidth extends to 4 kHz. Extended bandwidth plays out to 7 khz. The added 3 kHz even makes the channel sound louder (3 to 6 dB) than the 4 kHz bandwidth, with the same channel gain. All arrived at through piles of DBT testing (some of it ours).

+1. For an easy A/B comparison listen to FM radio, with a 15kHz bandwidth, then switch to AM, with a 5kHz bandwidth (transmission as high as 10kHz is allowed, but receivers tend to be low-passed at 5kHz to filter out noise). It's not that difficult to program a receiver with an AM and FM station broadcasting the same syndicated program, and if it's a talk radio show playing all the better. But if that's too much trouble just listen to FM and turn the treble EQ all the way down. Most receivers use shelving EQ with a +5kHz corner frequency, and the difference of flat versus rolled off is quite noticeable.

As to the benefit of flat response, media is mixed on speakers with relatively flat response, so if you want to hear what the engineer heard then speakers with relatively flat response are a necessity. But as already mentioned so is a room that doesn't have severe response anomalies.

[MSaturn]

Re: Flat Responce

I'm talking about fundamentals, here. You can boost the 'presence' of anything by making it bright. If you cut all but 800Hz+ for vocals, you'd have a shrill mess .. cut everything above, it'd still be perfectly intelligible.

Is that thing seriously claiming nearly 5000Hz? Jesus. Just because it's on the internet doesn't make it right. I haven't produced a note above 1000Hz in my life, and that was back when my falsetto worked right.

Breath noise perhaps?

My argument here is pure vocal frequency, not the combined effect of the voice, mouth, tongue, etc. - I've no doubt those sounds are high in frequency.

[DDF]

Re: Flat Responce

I'm talking about fundamentals, here. You can boost the 'presence' of anything by making it bright. If you cut all but 800Hz+ for vocals, you'd have a shrill mess .. cut everything above, it'd still be perfectly intelligible.

Is that thing seriously claiming nearly 5000Hz? Jesus. Just because it's on the internet doesn't make it right. I haven't produced a note above 1000Hz in my life, and that was back when my falsetto worked right.

Breath noise perhaps?

My argument here is pure vocal frequency, not the combined effect of the voice, mouth, tongue, etc. - I've no doubt those sounds are high in frequency.

This is incorrect. Voice intelligibility is greatly enhanced by extending the channel bandwidth to 4 Khz, and further enhanced by extending it to 7 kHz, and adding additional pre-emphasis.

This isn't a matter of opinion, its one of fact as backed up by extensive testing over many many years, first in AT&T labs in the 1930s.

There are lots of studies. I think you're getting up on volwels vs constanants (fricatives etc). Standard intelligibility tests look at missed constantants. For example, on the phone everyone thinks I said my last name was Dal Sarra, not Dal Farra, due to the poor channel bandwidth.

The page I linked is dead on. There are many many other references to voice intelligibility available. Its not "true because its on the internet". Its true because it was tested repeatedly over and over for years with consistent results. Our team executed DBT with over 200 participants in rigorously controlled test conitions, to study this in detail. I was there involved with it at the time.

No point in arguing this, its fact.

But again teh point is that boosting 2 to 5 kHz can improve intelligibility. Its in all the sound reinforcement texts (I have 5 of them I checked). Best to try it for yourself before stating everyone else with hundred of years collective experience in the area is wrong.

[ligs]

Re: Flat Responce

Like Villchur and Allison wanted.... ;)

http://www.classicspeakerpages.net/l...d_in_home_lis/

In the above article it certainly shows the destructive influence of the thick molding of AR3a on its FR:(

I am surprised nobody mentioned this, sort like a modern version of the above.

Listening Evaluation Regarding to frequency responses.
Nov 1, 2005 12:00 PM, By Rick Kamlet http://svconline.com/mag/avinstall_l...ng_evaluation/

Listener preferences correspond not with a single technical measurement, such as on-axis frequency response or total sound power, but rather to a family of curves. Each of these curves reveals something that the other curves might not. These curves represent:

On-axis frequency response
The response averaged over a specific off-axis listening window
The early reflected sound response
A predicted “in-room” response (more typically useful for home speakers)
The sound power (the total of sound emitted in all directions)
The directivity indices related to the sound power and to early reflections.

[LouC]

Re: Flat Responce

DDF: You pushed this in the right direction. These discussions come up, and there's a tendency to treat speakers like they are the only part of the system. Room acoustics, personal hearing ability, the electronic signal chain, the engineering of the original recording, and taste all enter into the final listening experience.

Speaker design will only carry you part of the way. If cap quality, cables and interconnects are supposed to be discernible, then why not carry this all the way back through to the amps, cd/dvd player, the recording engineer, the acoustics of the recording studio, the altitude & humidity, ...

At best one could "optimize" a design for a single piece of music in a single listening environment. So every design is a compromise.

Personally, I find a bit of BBC dip and slightly depressed top end works for me and most of the people I've built for. Imaging and seamless driver integration are most frequently what people notice.

Pete: If you lived in the Southwest, you'd find a lot of homes that are acoustically "big bathrooms". Masonry walls, tile floors on concrete slabs, single story, flat roof. "Live" doesn't adequately describe it.

[MSaturn]

Re: Flat Responce

This at least confirms that most vocal energy (vowels) lie below 3Khz.

I have a different perspective on this because I'm a singer, probably. I don't care so much about consonants. Note comments on 'realism'.

[DDF]

Re: Flat Responce

In the above article it certainly shows the destructive influence of the thick molding of AR3a on its FR:(

I am surprised nobody mentioned this, sort like a modern version of the above.

Listening Evaluation Regarding to frequency responses.
Nov 1, 2005 12:00 PM, By Rick Kamlet http://svconline.com/mag/avinstall_l...ng_evaluation/

Listener preferences correspond not with a single technical measurement, such as on-axis frequency response or total sound power, but rather to a family of curves. Each of these curves reveals something that the other curves might not. These curves represent:

On-axis frequency response
The response averaged over a specific off-axis listening window
The early reflected sound response
A predicted “in-room” response (more typically useful for home speakers)
The sound power (the total of sound emitted in all directions)
The directivity indices related to the sound power and to early reflections.

I was reading something in another forum, and the poster dug into the detailed outcomes of Toole's famous preference tests. What he found was that a couple of the most highly prefered speakers had poor off axis response and a couple with extremely good off axis response were poorly rated for preference.

The poster indicated these were less outliers than thought, and that thecorrelation of good sound = good off axis is less strong than the current common wisdom, based on Toole's own results with drove this common wisdom in teh first place. This puts some solid caveats on the ubiquitous goal of smooth power response.

I haven't had the time to go back yet and review the papers for these results but they are like big gongs going off, demanding attention to see if they're true.

Lou, I completely agree. My personal approach is to not design speakers optimized to only sound amazing with amazing recordings. I also design them to minimize annoyance with poor recordings. Frequency response balance plays into this and invariably involves some deviation from flat response. You could argue that the response is less ideal for stellar recordings in a perfect room, but OTOH, the response is far better on average and for less than perfect rooms. So, non flat response can increase the average good sound from recording to recording, but also reduce the statistical distribution (there are fewer poor listening experiences with poor recordings but also teh very best listening experiences are a hair less high).

Jumping a bit into the Sean Olive driven discussion, my personal philosophy is also to not make my living space a slave to my speakers. Believe it or not, there's more to life than audio equipment. :D They have to sound good in a real multi-use room that look sgood to me, so this also invariably results in some trade offs from flat, for many many (if not all) designs.

Dave, I agree, these discussions shouldn't have to turn pedantic.

Dave

[DDF]

Re: Flat Responce

This at least confirms that most vocal energy (vowels) lie below 3Khz.

I have a different perspective on this because I'm a singer, probably. I don't care so much about consonants. Note comments on 'realism'.

Sure, I agree. There's an ITU-T standard for spoken voice speech energy (P.76 if I remeber correctly), that captures long term averaged spectral power and short term distribution.

The interesting paradox is how so much of the intelligibility data resides in so little of the energy.

The whole Bandwidth=intelligibility connection is why low bit rate voice coders for wireless will throw away bits and create distortion, vs the easier solution of cutting bandwidth. Intelligibility is job one in that aplication, quality job 2.

[m.a.c.]

Re: Flat Responce

I wonder how a system would sound in a tiled bathroom?

These would probably sound excellent provided you are seated in the sweet spot.

[billfitzmaurice]

Re: Flat Responce

This at least confirms that most vocal energy (vowels) lie below 3Khz.

I have a different perspective on this because I'm a singer, probably. I don't care so much about consonants. Note comments on 'realism'.

If you're a singer you should be aware more than most as to the importance of harmonic content, and the relative insignificance of the fundamentals. It's what gives every instrument its voicing and character. Without harmonics you wouldn't be able to tell a trumpet from a viola, or for that matter Barbra Streisand from Bea Arthur. :eek:
The entire point behind THD as a significant measurement is that getting the harmonic content right is the key to accurate sound reproduction. If fundamentals were even nearly as significant there would be a TFD spec as well.

[Hearing Specialist]

Re: Flat Responce

OK, here we go. Lets cover some speech stuff that seems to be up for debate here....

1. The normal average ear canal resonance for an average adult ear is 2700hz.
2. When talking about "fundamentals" a person's height and physical size determines their resonance frequency of their voice. The taller and thinner the lower their voice is pitched. The heavier and shorter the higher that person's voice will be pitched.
3. All English speech is between 250 cycles and 8K fundamentals and harmonic. The only way to measure is with real ear measurements done in a clinical setting. Not with our mics used to calibrate our receivers and systems. It has to be a calibrated real-ear measurement done using probe tube mics placed close to the eardrum (tympanic membrane) laying in the ear canal itself. Any variable other than just mentioned makes the measurement invalid.
4. Everyone speaks vowel (below 1K) sounds and consonant sounds (1K-8K) which would mean that a fundamental consonant sound could be said between 1k to 8k.
5. "TH" is at 6K at 25db HL, "S" is a hair below 6K at 28db HL, "F" is a hair over 4K at 23db HL, "K" is a hair under 3K at 38db HL, "P" is at 1K at 40db HL, "H" is at 1500hz at 40db HL, and "G" is at 2K at 40db HL, "A" is at 750hz at 40db HL, "I" 500hz at 40db HL, "O" is at 500hz at 45db HL, i'll stop with just those. Point is we will speak fundamentals at various frequencies.
6. There are 3 kinds of DB: db HL (hearing level), db IL (intensity level), and SPL. To convert the above mentioned letters to SPL add 20 to those above mentioned numbers.