ok i am no expert, and i have no intention to take part in any debate or proof anything, but i will try explain briefly based on my observations in case you do not agree for any extent:
speaker drivers are by far the most distorting component in the audio chain, and unfortunately, there still no established science to measure the objective sound quality.
speaker drivers come with all sorts of measurable distortions in high does such as THD, IMD, MTD etc, however these data are usually not disclosed (or even measured) by the manufacturers.
the non-ideal nature of sensitivity vs frequency (what you called FR) being just one of the measurable distortions, and i would rather describe it as SPL distortion.
luckily, SPL distortion is no big deal, as our clever speaker designers will ofcourse be able to compensate the dips and bumps with some crossover and enclosure design strategies.
then how about other distortions? no, nil, none, there is no fix at all. (actually, crossover caps/inductors/wiring, box shape, room etc do add another layer of distortion to the final sound, again, there is no fix)
thats why speakers with flat FR could sound fake, overly bright, dull, blurry, lifeless... while some with FR not as flat may simply sound great with crystal clear realism.
this is probably due to how our ear/brain works (for survival!), when we heard a sound, we dont really care whether its 83.1dB or 83.2dB, instead we compute instantly the distance, direction and most importantly, the represented information.
you can probably spot the sonic difference of a piano note coming out from the radio over a real live one, both played at the exact same key and SPL, with ease, right?
simply put, our ear/brain is much more sensitive to the timing and wave shape but not SPL, so speaker sound quality must only be judged by serious listening, not just FR.
ofcourse flat FR is still desired but its not everything, in practice the FR plot means nothing unless you are designing a speaker.
this thread is started since i often seen bold claims like "hey driver X which has ruler flat FR must sound better than the way more expensive driver Y".
which is obviously not true and appears to be a quite common misconception. hope this helps and please feel free to correct if anything wrong.
speaker drivers are by far the most distorting component in the audio chain, and unfortunately, there still no established science to measure the objective sound quality.
speaker drivers come with all sorts of measurable distortions in high does such as THD, IMD, MTD etc, however these data are usually not disclosed (or even measured) by the manufacturers.
the non-ideal nature of sensitivity vs frequency (what you called FR) being just one of the measurable distortions, and i would rather describe it as SPL distortion.
luckily, SPL distortion is no big deal, as our clever speaker designers will ofcourse be able to compensate the dips and bumps with some crossover and enclosure design strategies.
then how about other distortions? no, nil, none, there is no fix at all. (actually, crossover caps/inductors/wiring, box shape, room etc do add another layer of distortion to the final sound, again, there is no fix)
thats why speakers with flat FR could sound fake, overly bright, dull, blurry, lifeless... while some with FR not as flat may simply sound great with crystal clear realism.
this is probably due to how our ear/brain works (for survival!), when we heard a sound, we dont really care whether its 83.1dB or 83.2dB, instead we compute instantly the distance, direction and most importantly, the represented information.
you can probably spot the sonic difference of a piano note coming out from the radio over a real live one, both played at the exact same key and SPL, with ease, right?
simply put, our ear/brain is much more sensitive to the timing and wave shape but not SPL, so speaker sound quality must only be judged by serious listening, not just FR.
ofcourse flat FR is still desired but its not everything, in practice the FR plot means nothing unless you are designing a speaker.
this thread is started since i often seen bold claims like "hey driver X which has ruler flat FR must sound better than the way more expensive driver Y".
which is obviously not true and appears to be a quite common misconception. hope this helps and please feel free to correct if anything wrong.
Quote: "bold claims like "hey driver X which has ruler flat FR must sound better than the way more expensive driver Y".
which is obviously not true and appears to be a quite common misconception. hope this helps and please feel free to correct if etc etc."
And why is that a bold claim?
On axis is bnot the full story, but if a oudspearkr alrady fails at the on axis criterion, there is laittle help off axis
which is obviously not true and appears to be a quite common misconception. hope this helps and please feel free to correct if etc etc."
And why is that a bold claim?
On axis is bnot the full story, but if a oudspearkr alrady fails at the on axis criterion, there is laittle help off axis
Oops my previous post, that should read
On axis is not the full story, but a hard and fasr requirement for a good loudspeaker system. Then there is off axis& uniform directivity/the power response.
Read Floyd Toole.
Distortion is far less of an issue, although many here seem to be obsessed by it. Earl Geddes thinks otherwsie, based on research and not on beliefs
On axis is not the full story, but a hard and fasr requirement for a good loudspeaker system. Then there is off axis& uniform directivity/the power response.
Read Floyd Toole.
Distortion is far less of an issue, although many here seem to be obsessed by it. Earl Geddes thinks otherwsie, based on research and not on beliefs
You are not an expert, so the phrase "speaker drivers: flat frequency response does NOT mean good sound quality" and all the thoughts behind it can be attributed to a lack of education.
As does Toole for this also; see Sound Reproduction: Loudspeakers and Rooms (London: Focal, 2008) pp.451-453
I notice from published response curves that loudspeaker manufacturers (even high end ones) rarely design for a flat frequence response as this would result in the speakers lacking excitement in the retailer's demonstration room.
The most common deviation from a flat response is an elevated upper treble which gives showroom impact, but is ultimately tiring in a domestic situation.
The most common deviation from a flat response is an elevated upper treble which gives showroom impact, but is ultimately tiring in a domestic situation.
Fr is very important, but you make an excellent point. Measure multiple drivers of the same size, make an eq profile for each to make them all flat, and playing the same range of frequencies, play pink noise and they will all sound different to your ear. One of the factors is variation in off axis performance, but it is my belief that there are many factors at play that we have no idea how to measure. When you go to an audio show like rmaf, you might listen to 100 different speakers, most of which have been designed with acceptable fr, and while some sound similar, they each have a different sound. They vary tremendously. I think we like to believe that we understand and control all the factors, but remember the stratavarius, with all the science we have today, no one can control the factors adequately to surpass the efforts of someone hundreds of years ago. I see speaker designing in a somewhat similar way, perhaps it's 80% objective engineering and 20% black art. Just my 2 cents, I'm sure others will disagree. I would encourage you to get 4 or 5 full range drivers and try the experiment above.
Craig
Craig
IMO, it's a lot easier to design a speaker system using drivers with reasonably flat response than it otherwise would be. That said, the worst system I ever designed had remarkably flat response. Measured great on axis, sounded awful. There's much more involved in designing a system that sounds good in an actual room. I also found I voiced things slightly differently depending on the most used source, CDs or LPs.
I remember reading research from Bose shows that in natural sound that we hear most of the sound is actually the reflected sound. So much so the bose speaker had most of the drivers pointed to the back. Which is actually very logical if you think about it? Imagine a speaker, what is the solid angle a speaker makes with your ear at normal listening position. Or for that matter your ear makes to the speaker. In layman term, sound from the speaker after traveling 4 - 10 m . Where the area affected by the sound wavefront is probably something like a few 10 sqm. A few square cm enters your ear.
Off axis response probably play a more important role than what was originally thought.
Oon
Off axis response probably play a more important role than what was originally thought.
Oon
Why do you think that was? What was the speaker doing badly off axis, or was it a room problem - something else?
White noise and an idealised impulse both have a flat frequency response. They differ in phase response, though.
I'm starting to question whether speakers are really the "worst offender". I've found that in simulating an amplifier, substituting a more realistic LCR model in place of an 8 ohm load can instantly ruin the THD by a factor of 10 or more. It may be specific to the class-A topology that I've been investigating, but I'm starting to wonder about the legitimacy of those <0.001%THD claims in some designs.
It seems the methodology of "mixing and matching" modular components has built-in flaws. There's pressure to design generic amplifiers and generic speakers, while the typical use-case is the flat opposite. You plug them in once, and then use them together for many years until something breaks. OK, so it's fun to tinker, but there's a hidden price to pay, and IMO if people want better sound quality they should really be looking at "woofer amplifiers" and "tweeter amplifiers" that are fine-tuned to their woofers and tweeters.
As for distortion, here are some things I've noticed:
A high damping factor basically = the voice coil's ability to reflect vibrations, and that's a mixed bag. It is NOT a pure positive like some people make it out to be. OT1H, the system does a great job of reducing the amplitude of Helmholtz-like box modes, but OTOH those box modes can become much worse in character. If the energy has nowhere to go, it tends to ring for much longer with a higher Q.
When the speaker has to fight against unwanted high-SPL resonances, it draws more current from the amplifier, and the motor will produce more THD and IMD in response to that.
Even with OB, people often construct straight edged baffles with no apparent attempt to ease the transition between front and back. Shouldn't the edges be jagged, with tapered slots or padded holes? Without mechanical damping, I would expect peaks in the FR above the baffle step.
What's worse is that attempting to fix most cases of resonance with EQ seems like a game of carefully applying more poison and hoping it doesn't die.
In the same vein, I wonder what people's experiences are with controlling break-up modes? I used to think it was something inherent to the flawed nature of speakers, until I realised that maybe the amplifier wasn't doing its job properly? If the rubber surround is supposed to be soft and absorbent, why should the voice coil behave like a hard reflective surface? Shouldn't both sides have good absorption? Tailoring the damping factor for different frequency ranges is something for the designers to figure out, so they can enjoy tight bass and soft sweet treble at the same time.
I'm starting to question whether speakers are really the "worst offender". I've found that in simulating an amplifier, substituting a more realistic LCR model in place of an 8 ohm load can instantly ruin the THD by a factor of 10 or more. It may be specific to the class-A topology that I've been investigating, but I'm starting to wonder about the legitimacy of those <0.001%THD claims in some designs.
It seems the methodology of "mixing and matching" modular components has built-in flaws. There's pressure to design generic amplifiers and generic speakers, while the typical use-case is the flat opposite. You plug them in once, and then use them together for many years until something breaks. OK, so it's fun to tinker, but there's a hidden price to pay, and IMO if people want better sound quality they should really be looking at "woofer amplifiers" and "tweeter amplifiers" that are fine-tuned to their woofers and tweeters.
As for distortion, here are some things I've noticed:
A high damping factor basically = the voice coil's ability to reflect vibrations, and that's a mixed bag. It is NOT a pure positive like some people make it out to be. OT1H, the system does a great job of reducing the amplitude of Helmholtz-like box modes, but OTOH those box modes can become much worse in character. If the energy has nowhere to go, it tends to ring for much longer with a higher Q.
When the speaker has to fight against unwanted high-SPL resonances, it draws more current from the amplifier, and the motor will produce more THD and IMD in response to that.
Even with OB, people often construct straight edged baffles with no apparent attempt to ease the transition between front and back. Shouldn't the edges be jagged, with tapered slots or padded holes? Without mechanical damping, I would expect peaks in the FR above the baffle step.
What's worse is that attempting to fix most cases of resonance with EQ seems like a game of carefully applying more poison and hoping it doesn't die.
In the same vein, I wonder what people's experiences are with controlling break-up modes? I used to think it was something inherent to the flawed nature of speakers, until I realised that maybe the amplifier wasn't doing its job properly? If the rubber surround is supposed to be soft and absorbent, why should the voice coil behave like a hard reflective surface? Shouldn't both sides have good absorption? Tailoring the damping factor for different frequency ranges is something for the designers to figure out, so they can enjoy tight bass and soft sweet treble at the same time.
Another set of perspectives...
Any multi-way speaker system is a viciously complex creation with millions of potential variables... This discussion could be applied to any part of the sound recording - reproduction chain. Amplifiers and microphones come to mind....
And, performance of any of these components can be measured in a million different ways too..... The goal posts are in CONSTANT motion!
Designing a near state of the art speaker is not that difficult, unless you make it so!
Paying attention to basic up front priorities and goals like tonal balance. Selecting drivers that work well together with a reasonable native frequency response profile in the band width they are to be used.
Not forcing drivers to do things they were never intended to do.
Realistic performance expectations for a given build cost.
What is amazing is how good most low to moderate cost drivers are when used correctly. Most designs do not come close to extracting driver performance potential.
Yes, a reasonably flat on axis response is a nice goal, but not at the cost of say, power response. Or appropriate power handling. (Think of full range headphone drivers)
You can provide so - so drivers to a competent designer and by and large they can, within reason, make them sing. The converse is true. The best drivers in the world being miss used or woefully underutilized can and usually do sound, well, terrible!
BTW, in most residential environments, on axis at a typical listening distance, a slight 1-2 db dip in the 2.5Khz to 4.5 Khz region and slight 1-3 db roll off from say 12.5 Khz - 20 Khz tend to work well. So, not really flat. Not all will agree with this. There are no absolutes. End of rant!
Any multi-way speaker system is a viciously complex creation with millions of potential variables... This discussion could be applied to any part of the sound recording - reproduction chain. Amplifiers and microphones come to mind....
And, performance of any of these components can be measured in a million different ways too..... The goal posts are in CONSTANT motion!
Designing a near state of the art speaker is not that difficult, unless you make it so!
Paying attention to basic up front priorities and goals like tonal balance. Selecting drivers that work well together with a reasonable native frequency response profile in the band width they are to be used.
Not forcing drivers to do things they were never intended to do.
Realistic performance expectations for a given build cost.
What is amazing is how good most low to moderate cost drivers are when used correctly. Most designs do not come close to extracting driver performance potential.
Yes, a reasonably flat on axis response is a nice goal, but not at the cost of say, power response. Or appropriate power handling. (Think of full range headphone drivers)
You can provide so - so drivers to a competent designer and by and large they can, within reason, make them sing. The converse is true. The best drivers in the world being miss used or woefully underutilized can and usually do sound, well, terrible!
BTW, in most residential environments, on axis at a typical listening distance, a slight 1-2 db dip in the 2.5Khz to 4.5 Khz region and slight 1-3 db roll off from say 12.5 Khz - 20 Khz tend to work well. So, not really flat. Not all will agree with this. There are no absolutes. End of rant!
I don't think it was doing so badly off-axis, but it had high listener fatigue. It's been a long time but maybe I put a crossover in a band where I shouldn't have, or maybe I just like a bit of rolloff. The best speakers I've built were 3-way, using 4th order active crossovers and an amp for each driver, something hard to replicate with passive crossovers.
thanks for the many replies
surprisingly it seems that many (if not most) of you guys are using FR as the only and ultimate way to measure objective sound quality, sorry no way for me, flat FR is just a basic requirement and not the whole story.
speaker design books has little to do here (its main goal is just SPL correction), and i would rather prefer "The Audio Measurement Handbook" as discussed here:
Is there an objective way to measure sound quality? Audio community often cite uneven frequency in highs, mids and lows as poor audio quality, but how is that perceptually negative to someone who listens to music? - Quora
surprisingly it seems that many (if not most) of you guys are using FR as the only and ultimate way to measure objective sound quality, sorry no way for me, flat FR is just a basic requirement and not the whole story.
speaker design books has little to do here (its main goal is just SPL correction), and i would rather prefer "The Audio Measurement Handbook" as discussed here:
Is there an objective way to measure sound quality? Audio community often cite uneven frequency in highs, mids and lows as poor audio quality, but how is that perceptually negative to someone who listens to music? - Quora
- Status
- Not open for further replies.