I've read books and articles about loudspeakers, drivers and crossover and I have listened to perhaps all types of loudspeakers (dynamic, horn etc) as well as driver technology: different cone materials and so forth. I have learned about the crossover and its function and there are many school of thoughts regarding speaker filtering, how much and how little it should correct.
There is one thing that stand out: Everyone is creating an interpretation - Beyond very few objective measurements like Equal Loudness, exactly how we interpret sound entering our ear is largely uncharted. We also have psycho-acoustics which perhaps play a larger role on what we enjoy, aka what sound good or not. And since loudspeakers to a large degree is open to interpretation, lets bypass what we personally like and look at an actual objective topic.
- TOPIC -
As a general rule, it is said that a loudspeaker should have a ruler flat frequency response. Except drivers who measure extremely flat, all drivers have peaks and valleys over the range of its accepted frequency range. Unlike amplifiers which can measure flat from perhaps 5 Hz to 100 kHz or more, speaker drivers do not. And it is the flat frequency response concept that I want to ask about.
I have attached a picture of the synthetic crossover network I've been working on. Some quick facts:
Drivers; 3/4 Dome Tweeter, 5.25" Midrange, 2x 6.5" Woofers.
Crossover points: 200 and 2800 Hz
Sensitivity: 89 dB / 2.83V
Filter type: 2nd order.
The picture show a filter which maintain a +/- 2.5 dB response between 77 Hz and 8 kHz. At most, +/- 3 dB variation between 76 Hz to 25 kHz.
So this is a manufacture response and the question is as follow: What do you accept as a dB response change over the useful frequency response ?
I see many loudspeakers that accept 5, 6 even 7 dB "window", so am I pushing it to far with my 2.5 dB variation ? I guess the reason for asking is that as you "compress" the freq.respons, the C, R and L values go up which is equal to higher cost... so what is the motivations here ?
Oneminde
There is one thing that stand out: Everyone is creating an interpretation - Beyond very few objective measurements like Equal Loudness, exactly how we interpret sound entering our ear is largely uncharted. We also have psycho-acoustics which perhaps play a larger role on what we enjoy, aka what sound good or not. And since loudspeakers to a large degree is open to interpretation, lets bypass what we personally like and look at an actual objective topic.
- TOPIC -
As a general rule, it is said that a loudspeaker should have a ruler flat frequency response. Except drivers who measure extremely flat, all drivers have peaks and valleys over the range of its accepted frequency range. Unlike amplifiers which can measure flat from perhaps 5 Hz to 100 kHz or more, speaker drivers do not. And it is the flat frequency response concept that I want to ask about.
I have attached a picture of the synthetic crossover network I've been working on. Some quick facts:
Drivers; 3/4 Dome Tweeter, 5.25" Midrange, 2x 6.5" Woofers.
Crossover points: 200 and 2800 Hz
Sensitivity: 89 dB / 2.83V
Filter type: 2nd order.
The picture show a filter which maintain a +/- 2.5 dB response between 77 Hz and 8 kHz. At most, +/- 3 dB variation between 76 Hz to 25 kHz.
So this is a manufacture response and the question is as follow: What do you accept as a dB response change over the useful frequency response ?
I see many loudspeakers that accept 5, 6 even 7 dB "window", so am I pushing it to far with my 2.5 dB variation ? I guess the reason for asking is that as you "compress" the freq.respons, the C, R and L values go up which is equal to higher cost... so what is the motivations here ?
Oneminde
Last edited:
You are right, this is listening with my eyes, that is why i stressed the point that its an objective topic with regards to the ruler flat rule. Because if absolute neutral is only desirable in theory AND you have not listened to the tonality of the drivers/loudspeakers - then what "rule" is applied when designing the filter and the +/- dB window.
I can see you deserve a good answer. Unfortunately, that is "you can't get there from here" or at least not too accurately.
What you need to do is play music in your room and at your chair and adjust the EQ until you like the sound. While that may sound like establishing your "house curve" it inherently also takes into account your room acoustics.
Then you assess the output needed to achieve that (which has to include much more than simple near-field on-axis FR). Obviously you aren't talking about upstream EQ but the sound that is actually reaching your chair.
To return to the way you presented the question initially, it really is more a matter of tolerances ("and"*) per octave or per relevant unit of bandwidth and, in turn, how that translates into the tolerance (or smoothing) of measurements. But applied to your house curve, not some eye-candy FR.
Like a pretzel to compare that outlook to the T/S sim approach.
B.
*
Just-noticeable difference - Wikipedia
Last edited:
Hi Onemind,
There shall always be an interpretation how a source will/should sound.
From a technical point of view a loudspeaker is the usually the worse thing in the signal line regarding how well it can follow/produce it`s incoming audio signal.
Measuring your loudspeaker with flat frequency response and zero acoustic phase is the goal to go for. With respect to all kind of measurement systems, i am only aware of one system that can reveal that acoustic phase and reflect the radiation impedance felt by the loudspeaker. This is the TEF (TDS) system devoloped under licence (Heyser) by Techron and later Goldline.
What most of us don`t realize is that we can hear how a source is radiating thru the air. Once measured (with above system) and get it technical perfect, listening with your ears you will never forget how a source have to radiate as spherical (wave) source. (following the input).
gr. Marcel
There shall always be an interpretation how a source will/should sound.
From a technical point of view a loudspeaker is the usually the worse thing in the signal line regarding how well it can follow/produce it`s incoming audio signal.
Measuring your loudspeaker with flat frequency response and zero acoustic phase is the goal to go for. With respect to all kind of measurement systems, i am only aware of one system that can reveal that acoustic phase and reflect the radiation impedance felt by the loudspeaker. This is the TEF (TDS) system devoloped under licence (Heyser) by Techron and later Goldline.
What most of us don`t realize is that we can hear how a source is radiating thru the air. Once measured (with above system) and get it technical perfect, listening with your ears you will never forget how a source have to radiate as spherical (wave) source. (following the input).
gr. Marcel
One problem here, if you do the "in room filter construction", that loudspeaker will only work in that room and the spot it sits in - I don't like the filter to behave in this way, that is why active system exist. So the loudspeaker must be room neutral and any issue in the room must be addressed accordingly.
Tolerances per octave, interesting... but that is related to order slope aka 1st, 2nd etc and that is not the issue here. The dB fluctuation is. Lets say the midrange is 89 dB sensitive, this then is the point you can go up and down from, lets say 3 dB or +/- 1.5 dB up/down so that you average the 3 dB since 3dB is accepted as that range which is least detectable, meaning less than 3 dB and you can't notice a change.
Am I to believe that you have to listen for flatness or desirable perceived flatness rather than making that the goal in software like I did with the response I showed in the picture ?
Tolerances per octave, interesting... but that is related to order slope aka 1st, 2nd etc and that is not the issue here. The dB fluctuation is. Lets say the midrange is 89 dB sensitive, this then is the point you can go up and down from, lets say 3 dB or +/- 1.5 dB up/down so that you average the 3 dB since 3dB is accepted as that range which is least detectable, meaning less than 3 dB and you can't notice a change.
Am I to believe that you have to listen for flatness or desirable perceived flatness rather than making that the goal in software like I did with the response I showed in the picture ?
Was it Archimedes who said "give me a long enough lever and I can move the world".
Aside from the other reasons for thinking that talking about phase is not helpful, Marcello's conception is analogous to Archimedes, "let me stick the perfect sound in your ears and you'll be in Carnegie Hall".
B.
That is the reason I ask. So objectively we can say that the loudspeaker should be flat, like I have done in the current filter. But in reality, that filter might not work because I want it to sound different compared to the ruler flat rule ?
To be honest, if that is the case, then it is impossible to make a filter in software in advanced, yet, people actually design filters in advance and they have a an accepted dB up and down between 20 - 20 000 Hz which they apply - but NO ONE is talking about what acceptable is... grrrr
Oneminde, I think you are familiar with our forum friend, Joachim Gerhard:
Always good to read the experts.
One further aspect I notice when I get it wrong is overall tonal balance. Bass is a bit of a moveable feast, depending on room and placement, but I start to appreciate what Troels Gravesen does with his LCR presence notches around 500-1500Hz on problematic small speakers. A couple of dB less energy on vocals seems to make for better flatness and a natural rendition.
DIY-Loudspeakers
A last aspect that doesn't show up on the frequency response is power response. LR2 and LR4 phase aligned fare much worse on this than 90 degree BW3, and you can hear it along with worse dispersion at crossover.
SpeakerBuilding.com - Interview with Joachim Gerhard of Audio Physic, Page 1
Always good to read the experts.
One further aspect I notice when I get it wrong is overall tonal balance. Bass is a bit of a moveable feast, depending on room and placement, but I start to appreciate what Troels Gravesen does with his LCR presence notches around 500-1500Hz on problematic small speakers. A couple of dB less energy on vocals seems to make for better flatness and a natural rendition.
DIY-Loudspeakers
A last aspect that doesn't show up on the frequency response is power response. LR2 and LR4 phase aligned fare much worse on this than 90 degree BW3, and you can hear it along with worse dispersion at crossover.
I am actually returning with some knowledge. Design of filter is best done when driver response is measured when it is mounted in the cabinet with damping material and baffle / geometric shape diffraction revealed. Yes I know about J. Gerhard and thank you for posting the quote. I am currently using +/- 3dB with less as the goal
Once more, the "its an interpretation" or personal taste thing. This makes loudspeakers very easy and very difficult to design - but forward we move non the less for there is no other option.
There's another fly in the ointment, and a very large one. Loudspeakers don't have "a" frequency response like, say, an amplifier does. An amplifier has just one pair of speaker terminals, and you measure its frequency response from its one input pair of input nodes to its one pair of output nodes.
But a speaker has an infinite number of output points -- everywhere in 3D space out in front of it. And everywhere you put the microphone you're going to get a different frequency response curve. All places you might put the mic might not be equally relevant, but the one place designers or manufacturers choose to put the mic is certainly not the only one that matters. Off-axis response into different directions matters a LOT -- my usual (maybe too often expressed here) suggestion is to just put a large board (or maybe a 33rmp record cover) between your ears and a speaker, and notice how little - hardly at all - the volume you hear is reduced. You will get some sound from the speaker no matter what direction it left the speaker from. To make things even less tied-with-a-bow, each of those frequency responses are going to change depending on things in the room (walls, furniture, ceiling) and their size and placements relative to both you and the mic. Worse, still, the time frame you measure from will also get you a different measurement (first 5msec, first 25msec, first 250msec) and of course the amount of smoothing, too, since when the time lengths get long there is no way you're going to stay within 2dB without smoothing.
My point is, thinking that if you find a place where it all measures within X decibels in some room and some placement, and plopping yourself in that orientation from the speaker will make it all great is very optimistic. It doesn't even remotely define what the speaker is going to sound like.
But a speaker has an infinite number of output points -- everywhere in 3D space out in front of it. And everywhere you put the microphone you're going to get a different frequency response curve. All places you might put the mic might not be equally relevant, but the one place designers or manufacturers choose to put the mic is certainly not the only one that matters. Off-axis response into different directions matters a LOT -- my usual (maybe too often expressed here) suggestion is to just put a large board (or maybe a 33rmp record cover) between your ears and a speaker, and notice how little - hardly at all - the volume you hear is reduced. You will get some sound from the speaker no matter what direction it left the speaker from. To make things even less tied-with-a-bow, each of those frequency responses are going to change depending on things in the room (walls, furniture, ceiling) and their size and placements relative to both you and the mic. Worse, still, the time frame you measure from will also get you a different measurement (first 5msec, first 25msec, first 250msec) and of course the amount of smoothing, too, since when the time lengths get long there is no way you're going to stay within 2dB without smoothing.
My point is, thinking that if you find a place where it all measures within X decibels in some room and some placement, and plopping yourself in that orientation from the speaker will make it all great is very optimistic. It doesn't even remotely define what the speaker is going to sound like.
Last edited:
I've been playing around with VituixCAD this week, but sort of realizing that if I want to have a fighting chance to know in advance if things will work, I have to jump over to Soundeasy, LEAP or LspCAD and include the cabinet and simulate everything on a professional level .... OR, build, measure and design the filter in post. The later is cheaper since you don't need to purchase software's but can quickly become more expensive since your are more or less blind, unless you follow a proven concept which basically is copying someone ells's work.
Thanks for adding that to my knowledge bank and I realize that a room is equally if not more important, especially the low end response of a room.
What I am getting at is not does issues or personal taste, it is making sure that - if this is the goal ofc - the loudspeaker is as neutral as possible in order to not add or take away anything from the source ... ideally yes. The real world is different. Everything depend on the quality of the source, the DAC, the amplifier, placement of speaker and so forth. But as any loudspeaker engineer knows, unless it is a very personal loudspeaker, he or she cannot control the environment or the source material (some call it source program, don't know why).
Lets say you designed a loudspeaker that performs extraordinary with classic, acoustical and Jazz music but not so good with rock. The one who listen to rock might be very upset with the loudspeaker ... perhaps because it is voiced in a specific way and is no longer neutral. This is very common and yes, cone material, size, shape of cabinet and so forth matter. But, that does not qualify for not aiming at ruler flat response and let the cone material do its thing. After all, cone material are chosen to be paired with the idea of what the loudspeaker should accomplish. I know first hand what paper, metal and ceramic drivers accomplish and they are aimed at different type of listening styles - all of them have pro and con's attached.
Loudspeakers is an art form and no one can say which one is the best ... that is what I am learning
I like it as flat as i can make it (taking directivity in account). The nature of deviations must be considered when deciding weather to cope with them or not. If a deviation is a diffraction issue, like dip higher in frequency, then it usually doesn't need treatment because it will disappear when going slightly off axis. If it is a peak, we are quite sensitive to them so you might want to address it - depending on how persistent it is. I consider +/-1dB good and +/-0.75dB or less great above the modal region of the listening room - i'm speaking about anechoic on listening window, not on one given axis. I'll throw some examples of what i think are good loudspeakers regarding frequency response.
https://www.soundstage.com/images/s...urements/kef_blade_two/fr_listeningwindow.gif
https://www.soundstage.com/images/s...ements/kef_reference_1/fr_listeningwindow.gif
https://www.soundstage.com/images/stories/loudspeakermeasurements/magico_s5/fr_listeningwindow.gif
https://www.soundstage.com/images/s...urements/vivid_giya_g2/fr_listeningwindow.gif
http://www.audioheritage.org/vbulletin/attachment.php?attachmentid=62448&d=1402913922
All of the these measurements are anechoic and are taken +/-15 degrees horizontal and vertical and averaged to get the listening window measurement. That is close to what you will hear. Now, there are debates of wether it should be +/-30 degrees horizontal or +/-15 degrees horizontal. I'm all in for +/-30 degrees or more but the slight drop from lower to higher frequencies is to be expected in that case.
Why anechoic above the modal region ? Well if anechoic listening window measurement is flat, then any problem you might encounter in you room is position or room related and can be treated in the right manner depending on frequency where the problem is. If measured listening window deviations are +/-1.5dB or +/-2dB i'd consider that very average/badly engineered loudspeaker.
https://www.soundstage.com/images/s...urements/kef_blade_two/fr_listeningwindow.gif
https://www.soundstage.com/images/s...ements/kef_reference_1/fr_listeningwindow.gif
https://www.soundstage.com/images/stories/loudspeakermeasurements/magico_s5/fr_listeningwindow.gif
https://www.soundstage.com/images/s...urements/vivid_giya_g2/fr_listeningwindow.gif
http://www.audioheritage.org/vbulletin/attachment.php?attachmentid=62448&d=1402913922
All of the these measurements are anechoic and are taken +/-15 degrees horizontal and vertical and averaged to get the listening window measurement. That is close to what you will hear. Now, there are debates of wether it should be +/-30 degrees horizontal or +/-15 degrees horizontal. I'm all in for +/-30 degrees or more but the slight drop from lower to higher frequencies is to be expected in that case.
Why anechoic above the modal region ? Well if anechoic listening window measurement is flat, then any problem you might encounter in you room is position or room related and can be treated in the right manner depending on frequency where the problem is. If measured listening window deviations are +/-1.5dB or +/-2dB i'd consider that very average/badly engineered loudspeaker.
Last edited:
That is an interesting take on loudspeakers. First, the Magico S5 measure really well, almost textbook. How it actually sound is a different story. The Be tweeter definitely adds to the neutral response.
Lets talk listening window. The +/- is up and down from the famous 1kHz sensitivity measurment yes ? so lets use the +/- 3dB which many use, this then would imply that the response as a whole can deviate within a 6dB window - 3 bellow and 3 above from the 1kHz sensitivity point. If so, then I agree on your statement. A 3 dB change is accepted as something we can detect and it is fairly safe to stay within this range, which result in a maximum of +/- 1.5 dB. More than this and you will start to detect shifts in output.
But it isn't as simple. Unless you are a toddler, our ears are unbalanced. They are frequency SPL unbalanced. The left and right ear hear frequency's with different sensitivity and as we grow older, the worse it gets. I can't complain too much since I am fairly balanced and detect stuff very high up in range.
To end on a positive note, I have learned a lot this day. Especially that I need to move to a better software that can give me as much data as possible regarding simulation so that I can see what the cabinet add or subtract.
Something to think about. Assuming you're ok with the idea of evolution, what do you think would be more beneficial to develop sensitivity of, for survival of a species? Would a fine sense of the exact spectrum shape of a sound be a big deal to optimize for, or would it be as sense for the directions and closeness of sounds? Ultra-tuned response shape is B.S., you might be able to tell a difference in a quick A/B, but it's not going to magically make it sound like real sounds in life. Given a stereo source, the radiation pattern will be way more significant.
@Oneminde
By listening window i was talking about angles at which the frequency response is measured. All of the measurements above are done from on axis up to 15 degrees off axis in increments. That resulted in a number of frequency response curves that when averaged, gave the result i posted. Frequency response is not the only parameter that determins how a loudspeaker will sound - although it is generally considered one of the most important - so you can't know how Magico sounds based only on that. If you are reffering to Fletcher-Munson curve when saying that our ears are "frequency SPL unbalanced", i don't see what that's got to do with this topic. If ones ears differ in sensitivity or there is a hearing loss there are hearing devices that can help.
Back on topic: i was saying that frequency response needs to be as flat as it can and have as uniform directivity that is physically achievable to stand a chance of being called neutral and to be able to reproduce the recording with as little changes possible.
By listening window i was talking about angles at which the frequency response is measured. All of the measurements above are done from on axis up to 15 degrees off axis in increments. That resulted in a number of frequency response curves that when averaged, gave the result i posted. Frequency response is not the only parameter that determins how a loudspeaker will sound - although it is generally considered one of the most important - so you can't know how Magico sounds based only on that. If you are reffering to Fletcher-Munson curve when saying that our ears are "frequency SPL unbalanced", i don't see what that's got to do with this topic. If ones ears differ in sensitivity or there is a hearing loss there are hearing devices that can help.
Back on topic: i was saying that frequency response needs to be as flat as it can and have as uniform directivity that is physically achievable to stand a chance of being called neutral and to be able to reproduce the recording with as little changes possible.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.