45 responses Add your response
I've got a better question: who cares?
Seriously. The sarcasm is an unintended but usefully jarring side effect. But seriously: Who cares? Take any speaker that has at any time ever been considered the best, or a classic, or whatever. Any greatly admired speaker. Can you think of one of them that the big reason it was so revered was its ruler flat response?
I mean aside from the followers of speakers like Dunlavy who are happy to have the sound stepped on all over the place just so long as its flat. The vast majority however vote with their wallets for a whole lot of other sonic attributes. Of which perfectly flat response seems to be pretty far down the list.
So its less "why can't" than "why don't". And the answer is: Because hardly anyone cares all that much. Not because it doesn't matter. Because its just one of a whole lot of things that matter.
A powerful DSP unit can probably get you pretty close at the microphone location, but to do it right, you'd need to exclude reflections.
When I first had access to good test equipment, as an enthusiastic amateur, I built a speaker that was about +/- 1.25 dB over most of the spectrum (room size limited how low I could get good data). I remember vividly tweaking the crossover to get closer and closer to "flat", and that as I did so, the speaker sounded worse and worse! I persevered, having faith that everything would sound right when I finally got to the promised land. Well, it didn't happen. When I was done, it was a truly dreadful sounding speaker.
I won't bore you with my quest for understanding that experience and where it led me, but let's just say that "flat" would not be my goal for a high-end home audio loudspeaker.
@erik_squires makes an interesting observation: "...in the room it all goes to hell."
In my opinion there are really TWO frequency responses that a designer needs to get "right": The frequency response of the first-arrival sound, and the frequency response of the reflections. By way of example, live unamplified instruments get both of these "right", and the result sounds pretty good.
"It was probably not measured properly and not really flat. Why else would it sound dreadful?"
Actually, you are right! I was only measuring the on-axis response, and ignoring the off-axis response, because back then I didn’t know any better!
But here is another critical piece of the puzzle: "Flat" sounds "thin and bright" to most people.
And a gently downward-sloping curve sounds "flat" to most people.
There’s an excellent discussion of the subject in this review of the Dutch & Dutch 8c loudspeaker, scroll down about 1/4 of the way:
The author draws on an extensive controlled blind study conducted by Sean Olive and Floyd Toole, and says,
"Most participants in the study preferred a frequency response from 20 Hz with a straight line to -10 dB at 20 kHz. A measured “flat” in-room frequency response is not the preferred target, as it sounds too thin or lacking bass."
If I was designing a studio monitor the goal posts would be in a different place, and "flat" measured response might very well make sense, depending on what type of monitor it was (tracking, mixing, or mastering).
Agreed, other aspects are also vital to good reproduction of music and sound. Seems there are questions about why frequency response accuracy is important.
Very simple. The aim of loudspeakers is to accurately reproduce sound (as recorded with the best of modern technology). For that, accurate frequency response must be a goal. Or am I mistaken?
Would you please describe your current speaker syste, as well as any widely available commercial speaker you stan?I don't have a current system. None that meet my standards can be found for a reasonable cost or perhaps even at any cost.
The last speaker I had were called Green mountain audio chroma. However, they suffer from severe colorations which make them sound very cold. Additionally these speakers have zero baffle step compensation, in addition to poor bass response. There's no evidence they are time coherent either, only the manufacturers claim.
Thanks to all for input.
My question is simple.
Why is it obviously technically difficult for speaker designers to provide precisely level frequency response. @kenjit. I accept some listeners want to adjust the sound for different recordings. Old fashioned tone controls and modern equalizers provide that.
I don’t think that it is that hard to design a flat anechoic response, which is the way that speakers are typically measured. But it may not be desirable.
Tell me, do you listen to your speakers out of doors, where there is no room gain at low frequencies and no reflections? Or perhaps your listening room is an anechoic chamber, such as the one at the NRC. In which case, you would be in luck with your flat response speaker.
This is another instance of where the measurements are taken in such a way as to be uncorrelated with the practical use of the product, IMHO. Who listens to their speakers in an anechoic environment?
Most of the rest of us stiffs listen in conventional room with bass reinforcement, side wall, ceiling and floor reflections. If you gave me that perfect speaker it certainly wouldn’t measure very well once it was in my room. I am confident of that.
@ptss - The simple answer is $$$. Mr. Dunlavy personally HAND TUNED every one of his speakers to achieve this tight tolerance. This tuning took into account the cross-over components AND the drivers in each cabinet. Doing custom tuning is not practical for anything other than a boutique enterprise. Why is it important to have tight tolerances? If you have a speaker with +/- 3db tolerance, you can have a 6db variance between frequencies coming out of your speaker. Is this important for realistic playback? You decide. Also, if speakers were made with +/- .5 db and music sounded funky in your room, you would know it's not the speakers causing the problem.
I realize the simplest crossover better the sound. Frequency response starts with flat responce drivers. Next the cabinet is, especially the front baffle, designed to help naturally flatten drivers. Finally proper crossover points and the least amount of notch filters possible. The more crossover to flatten response can certainly make a flat responce sound horrible. Keep it simple, simple, simple.
I do want to point out that components are a lot better than they were in the 1980s.
Drivers are, by and large, more consistent, and cheap 1% film caps are readily available.
It's a lot less important to match drivers and caps than it was then, and it is a lot less effort. This is a good thing especially when you consider having replacement parts on hand. Imagine if a driver in a hard to match speaker goes out. You'd have to send both speakers back to the factory to have them matched and possibly have the crossover updated. What a giant PITA!
I don’t think that is uniformly correct at all. Snell kept a record of the fr of every driver, by serial number. If you lost a driver they could just send you a match. I am speaking here of the Peter Snell days, when drivers, and crossovers were very tightly matched. No idea what happened in the Kevin Voecks era and the dark days thereafter.
I agree with the others that it really isnt that important. I think the main point is and should be what does a manufacturer have to do to get his speaker to measure flat. In my experience you lose more designing a speaker (with this as its main goal) that measures flat than you gain when this goal is achieved. But I was never a Dunlavy fan so what the hell do I know.
Audiokinesis is dead right on this one. My system is adjusted to be down 9 db at 20 kHz at 0 db. I would like at add one other variable to Audiokinesis's point and that is volume. As Fletcher and Munson noted the frequency response of our ears changes with volume. As volume increases we become more sensitive to low bass and treble. Or conversely as volume drops we become less sensitive to bass and treble.
This was the theory behind Loudness controls which boosted treble and bass for low volume listening. But, the curves were fixed and only accurate at one volume. Most of us never used them. We just turned the volume up until things sounded balanced. People gravitate to speakers that have a response curve that suites their taste at the volume they usually listen at, in the room they listen in. In this context 1 or 2 db here and there can mean quite a lot. I impulse tested a friend's Watt/Puppys several years back in the near field and sure enough both speakers had a 2 db depression centered at 3500 Hz. This is exactly where you put a notch filter if you want to cut out sibilance in female voices and is responsible for the "smoothness" that people found so attractive in these speakers.
As Audiokinesis mentioned, Flat sounds awful. The problem is that what sounds good is a moving target. Tone controls and loudness filters are severely compromised. They just are not flexible enough. The only way to deal with this is with digital frequency response control or room control or whatever you want to call it. And not just bass, full range. The speakers are impulse tested from the listening position. A frequency response curve is generated for each loudspeaker along with a correction filter that brings the response of the system to flat making both speakers exactly the same which is vitally important for imaging. You can view all these curves on your PC. Next you can overlay a response curve of your choice and design. You want your speakers to sound like Wilson Watt/Puppys? Just put a 2 db depression at 3500 Hz and you are good to go. If you are very fond of your speaker's house curve you can overlay that and you will have exactly the same sound you started with except with better imaging. You can have a series of curves for various purposes. The very best units have dynamic loudness control. The unit automatically shifts loudness curves with volume. The end result of this is that the music has exactly the same tonal balance regardless of volume which is very spooky when you first hear it.
FWIW Revel Salon specify +/- 1/2 dB from about 29 to 18,000. They are also the longest lasting reference level speaker that has remained unchanged. I find that interesting when most manufacturers seem to make changes simply to make a change. I think the Salon’s uniquely sculpted face also leaves little room for improvement in dispersion. Also interesting as only B&W seem to also have a significant design to eliminate dispersion
Ptss, you can make speakers pretty close to flat when tested in an anechoic chamber. There will be some unit to unit variation. The problem is that once you put them in a real room their response curve changes. The Japanese tried to design loudspeakers this way and they were awful. It is impossible for a speaker manufacturer to predict let alone compensate for the acoustic environment their speakers will be placed in. If you want to be assured of flat frequency response get headphones.
How do they, the various speaker manufacturers, measure frequency response? Do they use the same anechoic chamber? Their own anechoic chamber? Is there a standard for anechoic chambers? Or some other kind of specially constructed room? One wonders if they construct a room that produces a flat frequency response for a particular speaker, if you see what I mean.
Why is it hard to get flat frequency response, on axis, in an anechoic chamber? What is it about the drivers that needs “technical” improvement? We’ve put men on the moon, extreme precision required; what’s the big deal with loudspeakers. With the amounts charged for the “spensive” models — I expect better; more accuracy.
My question is simple.Duke answered your question above.
How about because human ears are nowhere near as sensitive to amplitude as they are tone.This is true and I find it amusing that in amplifiers the constant voltage characteristic is considered so important, when in the end, the ear actually places a greater value on the tonality caused by distortion than it does on tonality created by actual FR errors. Clearly a constant voltage characteristic (double power as impedance is halved) isn't that important in the overall scheme of things- but getting the distortion signature right is. That is why the tubes/transistors debate rages on ad nauseam.
Sorry to have to disagree Atmasphere. I was hoping it might be folks with your technical acumen that might provide real insight |Well as you saw, Duke did it- so it can be done and isn't that crazy difficult. Just turns out that isn't how you want to do it. What sort of insight beyond that are you looking for?
The goal of perfectly flat response is simply a goal unto itself and has limited relevance or value in terms of creating technically excellent speakers. I don't understand the obsession with this particular measurement. Lots of speakers have been designed to perfectly embrace on particular measurement, be it phase, impulse response, flat response, some kind of directionality, and they all fail to live up to the claims a perfect measurement is said achieve.
The problem as Duke pointed out is that speakers are used in rooms- and rooms have an acoustic signature as well as reflections- the latter playing an enormous role in how we perceive sound.
If one is being pragmatic, its probably better to work *with* human hearing perceptual rules rather than against them.
Saying reproducing the frequencies of music - any less accurately than “Spot On” -is simply to accept the failure of today’s technology. That’s been a common problem with mankind in general and is why we laid breakthroughs that eventually move us forward to a new paradigm.
Its interesting to me that responses to this thread have virtually all been to say why frequency response is unimportant.
“Groupthinking acceptance of the status quo”?
i suggest that in a very few years; when dead flat frequency response is common; the invaluable contribution it makes to reproducing sound will be considered a “no brainer”.
Just my 2 cents worth.
I can absolutely promise you that dead flat frequency response will never, ever be a common thing. If flat frequency response is ever a thing in a real room, it will be the product of active signal processing, not some advancement in speaker technology.
You're the one showing up here boldly claiming that the ideal speaker has dead flat anechoic frequency response. Where's your research and proof that's a desirable characteristic? I don't think you have any to justify your claim. You don't have any research or proof of some "group think" mentality either. It looks like your just throwing stupid claims around with no proof at all. I'm certain you have nowhere near the body of knowledge Ralph does and you speak with an unwarranted arrogance that your feeble understanding doesn't justify. When you gather your evidence and research that proves flat frequency response is so very important, cone back and educate us all on the wrongs of our ways. Until then, try to avoid slinging baseless accusations and claims at people who know vastly more than you.
@ptss wrote: "responses to this thread have virtually all been to say why frequency response is unimportant. "
Imo frequency response is the most important thing, but imo "frequency response" MUST include the off-axis behavior of the speaker because most of the sound that you hear in a non-anechoic, non-nearfield setup starts out as off-axis sound. One place where you and I disagree is on WHAT that frequency response should look like.
"No one has responded with any indication they know why it’s not practicable to make speakers with “flat” frequency response."
It can be done but imo it's an uninformed mis-allocation of funds and effort for home audio. In my opinion (and in the opinions of Floyd Toole, Sean Olive, Earl Geddes, and most other speaker designers) "flat" is not the goal because it doesn't sound as good as a gently downward-sloping curve.
If you want "flat", you might want to thoroughly check out Neumann studio monitors. I did back when they were Klein & Hummel, and elected not to become a dealer. But they are a wonderful tool for recording studios.
Its interesting to me that responses to this thread have virtually all been to say why frequency response is unimportant.
No one has responded with any indication they know why it’s not practicable to make speakers with “flat” frequency response
whether its important or not is irrelevant. The reason its not possible is because it would take too many iterations and man hours to optimise the response to be flat. We therefore settle for flatness within a specified limit such as +- 3db.
And as duke suggests , flatness at one point is pointless. But then flatness at every point in space is even harder to achieve. DSP will only optimise one point in space. Typical rectangular shaped cabinets are not the best for flat response. But spherical speakers are impractical to make so we have to accept the peaks and dips you get with a non spherical box.