I believe what we can agree on is that no speaker is perfect and striving for that desired perfection always poses choices in compromises necessary to produce a functional design.
You can get all the distortion components and non linearities down to such a minimum that the character of the speaker becomes almost inert. That in theory is desirable in terms of the speaker not adding anything to the music. The point this "near perfect" situation becomes noticeably imperfect when the speaker is driven harder, ie reproducing dynamic and complex musical events, shows the speaker's character, which can become highly objectionable depending on the listener's preference and hearing skill level. I've heard many speakers claiming to be low distortion designs, but at higher SPLs (not to the point of being overdriven) sound very off putting depending on how graceful the drivers handle these transients. Most 2 way designs with crossover points in the critical 1 - 3k region sound rough and harsh to my ears when driven harder, even if the drivers measure very well and great care was taken with the crossover design. I fight very hard to stay away from that critical midrange with all my designs because of my experiences.
There are some things in audio you just can't quantify, even with complex measurements. Everyone's hearing is different, which makes a speaker more subjective than a laboratory grade instrument. Many good speaker designs exist but very few exceptional ones. I've become a horn guy these days, mainly because they can sound more realistic, exciting and true to life (when executed properly) than almost any other direct radiating system. Horn loaded systems usually measure less transient accurate than other types of designs, but they sound more correct to the ear (at least to my own).
For my personal designs I listen to the raw driver playing music in free air and in a baffle first (even tweeters and CDs), then listen to sine wave sweeps and static sine waves from all angles of the driver, and finally look at what frequency range its audibly usable in. After that process I look at the measured data to quantify my findings and make the crossover work in sims, then in real life. I realize this is an unorthodox way to design a speaker, but I've come to this conclusion through my own experiences and it works for me. The human ear has to be part of the selection process, not just measurements and raw data.
You can get all the distortion components and non linearities down to such a minimum that the character of the speaker becomes almost inert. That in theory is desirable in terms of the speaker not adding anything to the music. The point this "near perfect" situation becomes noticeably imperfect when the speaker is driven harder, ie reproducing dynamic and complex musical events, shows the speaker's character, which can become highly objectionable depending on the listener's preference and hearing skill level. I've heard many speakers claiming to be low distortion designs, but at higher SPLs (not to the point of being overdriven) sound very off putting depending on how graceful the drivers handle these transients. Most 2 way designs with crossover points in the critical 1 - 3k region sound rough and harsh to my ears when driven harder, even if the drivers measure very well and great care was taken with the crossover design. I fight very hard to stay away from that critical midrange with all my designs because of my experiences.
There are some things in audio you just can't quantify, even with complex measurements. Everyone's hearing is different, which makes a speaker more subjective than a laboratory grade instrument. Many good speaker designs exist but very few exceptional ones. I've become a horn guy these days, mainly because they can sound more realistic, exciting and true to life (when executed properly) than almost any other direct radiating system. Horn loaded systems usually measure less transient accurate than other types of designs, but they sound more correct to the ear (at least to my own).
For my personal designs I listen to the raw driver playing music in free air and in a baffle first (even tweeters and CDs), then listen to sine wave sweeps and static sine waves from all angles of the driver, and finally look at what frequency range its audibly usable in. After that process I look at the measured data to quantify my findings and make the crossover work in sims, then in real life. I realize this is an unorthodox way to design a speaker, but I've come to this conclusion through my own experiences and it works for me. The human ear has to be part of the selection process, not just measurements and raw data.
I believe that even if you notch a driver flat, it will still ring.
I think a tone an octave or 2 below the driver's +6db 3.5khz resonance (notched to flat), will cause the resonance to ring (over time) ......
But you are right, the driver's spike at 3.5khz maybe +8db when the driver is running at 94db versus 80db and only +5db, i don't know.
I think there was a mention on the full range forum that a driver was resonating up near 10khz, but it also created bad distortion at 1/2 and 1/3 of the
Phase wrap is also an issue around the crossover point, that is an issue to many.
I prefer 6db time/phase aligned, purely subjective.
But freq response, dispersion, cost, room penetration, etc usually trumps a sharp triangular step response.
Without direct comparison, I imagine a 12db crossover would just sound "different" versus a 6db, unless you had say an actual guitar playing next to it to compare which sounds more real.............
Transients can usually be pick out as "sharper" such as wood block or keys jingling on a 6db time/phase aligned......
But music ? 12' away in your not perfect room ?
not sure...............
This video opened my eyes. I was dead set on 6db time/phase, now, not so much.
Dynamics are fun, so is detail, so now I use 2 different speakers, switching depending on music, mood, distance....
A 6.5" 2-way cannot do what a 12" 2-way can do, and vice versa.
I think it was Dunlavey that said there are good sounding speakers, bad sounding speakers, and accurate speakers.
What I define subjectively as good (or better) is not necessarily the most accurate.
I think a tone an octave or 2 below the driver's +6db 3.5khz resonance (notched to flat), will cause the resonance to ring (over time) ......
But you are right, the driver's spike at 3.5khz maybe +8db when the driver is running at 94db versus 80db and only +5db, i don't know.
I think there was a mention on the full range forum that a driver was resonating up near 10khz, but it also created bad distortion at 1/2 and 1/3 of the
Phase wrap is also an issue around the crossover point, that is an issue to many.
I prefer 6db time/phase aligned, purely subjective.
But freq response, dispersion, cost, room penetration, etc usually trumps a sharp triangular step response.
Without direct comparison, I imagine a 12db crossover would just sound "different" versus a 6db, unless you had say an actual guitar playing next to it to compare which sounds more real.............
Transients can usually be pick out as "sharper" such as wood block or keys jingling on a 6db time/phase aligned......
But music ? 12' away in your not perfect room ?
not sure...............
This video opened my eyes. I was dead set on 6db time/phase, now, not so much.
Dynamics are fun, so is detail, so now I use 2 different speakers, switching depending on music, mood, distance....
A 6.5" 2-way cannot do what a 12" 2-way can do, and vice versa.
I think it was Dunlavey that said there are good sounding speakers, bad sounding speakers, and accurate speakers.
What I define subjectively as good (or better) is not necessarily the most accurate.
Last edited:
There are people who believe you can hammer a speaker's lumpy FR into submission with DSP and have it sound like a more expensive speaker that has naturally linear FR and power response. That's simply not the case. Otherwise you could take just about any mediocre driver combo and have it sound like a higher end setup just with DSP correction.
Drivers with high Q resonance modes (hard composite, stiff paper or metal cones) will have a tendency to ring at specific frequencies even when they're flattened with EQ and other forms of processing. Just because you operate the driver outside of these resonance modes doesn't mean the driver won't ring audibly, plus the ringing can be masked by music harmonic content.
The amount of energy required to excite specific cone, dustcap or cavity resonance(s) isn't linear. The reactive behavior isn't linear in output amplitude either. Driver resonance modes can be excited by other complex waveforms in music (harmonics) not directly related to the discrete individual frequencies. These breakup modes often interact with other complex signals in music, sometimes octaves up or down in range and also with IMD artifacts that aren't present in the music signal itself. Measurement data won't necessarily reveal this behavior, which is why you can't always judge a driver by measurement results alone, even looking at the CSD results.
These are reasons why I like to audition raw drivers by playing sine waves and familiar music through them to listen to the character of audio that comes directly out of them in free air, open baffle or in a sealed test box. The driver will tell you alot about itself this way.
Midbass drivers will reveal their airflow restrictions and issues when the magnet, spider and VC venting is exposed using the above test, especially with higher level low frequency sine wave sweeps. It will show when and how gracefully the driver runs out of linear excursion. You'd be surprised how bad many of the more expensive LF drivers behave audibly due to poor VC venting. In this regard, a high BL offset can be obvious, observing the motor assembly offsetting the centering of the cone at higher MF drive levels (cone suck in/out).
I also audition HF drivers in a similar discrete method with music and sine waves, but you obviously need to be very careful using fullrange signals on tweeters or other bandwidth restricted drivers. I highly recommend a quality film cap in series just to be safe and be very careful with volume levels.
Baffle diffraction induced FR irregularities can in theory be EQed out of a speaker's FR, but the audible effects and sound signature can't be. That is one turd that can't be polished. Neither can the combing and off axis FR dips be fixed this way. Time domain issues can be resolved with FIR filtering and individual driver delay, but you can be stuck with pre-ringing which has the potential to be audible in many situations.
I'm not convinced added horizontal reflections (through minimal in room sound absorption) are a desirable way to enhance the L/R stereo image. Even if you're OK with an artificial sounding image, it will impact the speaker's CSD in a bad way, even moreso when playing music louder, which just extends the in room energy decay time and further smears intelligibility of individual instruments and vocals with increasing playback levels. This also applies to HF drivers radiating rearwards. Unless you prefer an artificially diffuse and vague spatial image, it won't recreate accurate sounding width or depth missing from the recording. If the room is treated properly and speakers are able to hold even horizontal and (relatively narrow) vertical directivity well into the lower midrange, you will end up with a great sounding system that can play all sorts of music with equal satisfaction. This of course means in room FR is linear, THD and IMD are low enough and sufficient SPL headroom is available.
Drivers with high Q resonance modes (hard composite, stiff paper or metal cones) will have a tendency to ring at specific frequencies even when they're flattened with EQ and other forms of processing. Just because you operate the driver outside of these resonance modes doesn't mean the driver won't ring audibly, plus the ringing can be masked by music harmonic content.
The amount of energy required to excite specific cone, dustcap or cavity resonance(s) isn't linear. The reactive behavior isn't linear in output amplitude either. Driver resonance modes can be excited by other complex waveforms in music (harmonics) not directly related to the discrete individual frequencies. These breakup modes often interact with other complex signals in music, sometimes octaves up or down in range and also with IMD artifacts that aren't present in the music signal itself. Measurement data won't necessarily reveal this behavior, which is why you can't always judge a driver by measurement results alone, even looking at the CSD results.
These are reasons why I like to audition raw drivers by playing sine waves and familiar music through them to listen to the character of audio that comes directly out of them in free air, open baffle or in a sealed test box. The driver will tell you alot about itself this way.
Midbass drivers will reveal their airflow restrictions and issues when the magnet, spider and VC venting is exposed using the above test, especially with higher level low frequency sine wave sweeps. It will show when and how gracefully the driver runs out of linear excursion. You'd be surprised how bad many of the more expensive LF drivers behave audibly due to poor VC venting. In this regard, a high BL offset can be obvious, observing the motor assembly offsetting the centering of the cone at higher MF drive levels (cone suck in/out).
I also audition HF drivers in a similar discrete method with music and sine waves, but you obviously need to be very careful using fullrange signals on tweeters or other bandwidth restricted drivers. I highly recommend a quality film cap in series just to be safe and be very careful with volume levels.
Baffle diffraction induced FR irregularities can in theory be EQed out of a speaker's FR, but the audible effects and sound signature can't be. That is one turd that can't be polished. Neither can the combing and off axis FR dips be fixed this way. Time domain issues can be resolved with FIR filtering and individual driver delay, but you can be stuck with pre-ringing which has the potential to be audible in many situations.
I'm not convinced added horizontal reflections (through minimal in room sound absorption) are a desirable way to enhance the L/R stereo image. Even if you're OK with an artificial sounding image, it will impact the speaker's CSD in a bad way, even moreso when playing music louder, which just extends the in room energy decay time and further smears intelligibility of individual instruments and vocals with increasing playback levels. This also applies to HF drivers radiating rearwards. Unless you prefer an artificially diffuse and vague spatial image, it won't recreate accurate sounding width or depth missing from the recording. If the room is treated properly and speakers are able to hold even horizontal and (relatively narrow) vertical directivity well into the lower midrange, you will end up with a great sounding system that can play all sorts of music with equal satisfaction. This of course means in room FR is linear, THD and IMD are low enough and sufficient SPL headroom is available.
How can we EQ out baffle diffraction (other than on a single axis)? From what I've seen it creates an uneven response on or off axis. EQ one and the other will get worse. Once you remove the diffraction ripple (for instance with a smooth shaped baffle or proper mouth termination on a horn) the on and off axis curves will start to track each other (look similar/the same shape). To me this means one cannot EQ baffle diffraction and one should deal with it in another way. I guess that was the reason this thread was started. EQ-ing away ripple induced by diffraction would work in 2D only. At a single point in space. Not very useful.
To me, the above depends on timing. In a large enough room the reflections (now arriving later) may not be that harmful and even pleasing. Take an open baffle speaker, where the nulls avoid early side wall reflections, but that back wave is going to come back. It there is a proper time difference, it can sound very pleasing. Believe it or not, it can actually enhance the front stage if the timing is right. Due to the larger distance that reflection has travelled, it's SPL level will be lower too and it will be largely decorrelated from the direct sound if it reflects off of distant walls and furniture (making it more diffuse). Same goes for that HF driver firing backwards, timing (and level) will dictate how pleasing the end result will or can be. A room void of reflections will show the flaws of Stereo. Ideally we'd have some level of reflections (timed right) to enhance the Stereo illusion.
We almost always have reflections though, even in a reasonably large room it will be tough to avoid the vertical reflections anyway.
Have you guys seen how tight the vertical pattern needs to be to really avoid floor and ceiling reflections? And do we even want to avoid all of them? Some level of reflections is quite welcome and we will still be able to listen "trough" it easily with our brain trained to do so. I take it when someone talks to you in a room you can still follow what he/she said... There's bound to be reflections during that conversation.
Take away all the reflections and a lot of things change. Stereo cross talk will become much more obvious (*), you'd hear the balance change with little head movements because the nulls caused by cross talk (right speaker heard by left ear and vice versa) will shift. Slightly off axis the balance might even be better, as in the exact sweet spot both ears have these dips at exactly the same spots while off axis each ear gets them at slightly different places. Luckily reflection present in a normal room hide these anomalies. There are bound to be reflections with most type of speakers in a normal room. Even with horns we'd still have floor and ceiling reflections, though at lower levels than with most direct radiators.
A measurement at the ear position in a very well treated room (not my room, think above studio grade room treatment):
In the exact sweet spot, you'd be getting something close to this curve at both ears, if the room is void of reflections. The green trace is without reflections, the other two colors show a Hass Kicker (orange) and Haas Kicker with ambient channels (red). Original link.
This room is painstakingly engineered to remove all direct reflections, including floor and ceiling. It doesn't even look like a normal room. It was build for one purpose only: as listening room. Here's a link with a description of that room, with a bit of luck there's still some links to the measurements he made inside that room at the listening position. I had never seen anything like it (except semi anechoic measurements but never at the listening spot).
The FR of one channel measured at a time, both channels shown:
So be careful what you wish for... you might be surprised what happens if you do get it...
Even though I quoted @profiguy a couple of times, I wrote this reply as a more informal information piece. A part that get's almost no thought on this forum, even though it does have a huge influence on what we perceive. What does our room do for us? Can it enhance our listening pleasure? I can recommend reading the many papers of Griesinger for what a room can do (even if most of his work is about larger spaces, the information he gives still is an eye opener).
(*) = if one listens much closer to the speakers, like in nearfield, the direct sound will hide the effects of this cross talk effect. there probably will be more shading from the head too.
Last edited:
DSP can fix some anomalies, but it cannot fix a lot of stuff… diffraction still exists even if you EQ the response flat. Flat does not fix it.
dave
This is totally false. There is nothing that the ear can hear that cannot be measured. The key is to know how the measurements correlate to subjective impression. That said, one can only draw these correlations based on subjects average response. There is no accounting for obscure subjective opinions.
But I guess that personal subjective impression "works for you." For me, measurements work far far better.
Diffraction cannot be fixed globally, even though it is linear. It's because it is multipath and hence uncorrelated to the original signal.
Mr Ged, I have a lot of respect for your work. I have done some of my own research in the area of FR correction with "difficult" drivers that have very tough to tame peaks and sound a lot like a tuning fork. Many subwoofer drivers with stiff cones behave like this, as you know.
I compare this tendency for ringing to acoustic feedback ie. a mic feeding back. The tipping point for feedback isn't always linear and often hair trigger in behavior. Once it starts you have to completely remove even the slightly out of band audio to stop the ringing. While not the exact same type of feedback mechanism, it functions in the same manner of excitability. No, I do not have graphs or formulas to explain this, just personal carefully isolated and controlled observations.
I have heard many speakers with cone materials and compositions that tend have a tonality of always having specific notes coming through in the background at times with louder playback levels. My ears can tell something is offputting and displeasing. Many aluminum and composite cone drivers have this edgy sound and there is no amount of external correction that will remove this, unless you suck out anything in that band with a very sharp digital filter, which is highly audible even with FIR filtering.
Some things can't be measured that are readily audible but not "in your face". i have to respectfully disagree with you that everything can be quantified in measurements. One could argue that many of these so called perceptions I speak of are psychosomatic in nature, but I attempt to remove as much subjective thought process as possible when doing my own evaluating and listening tests. Its hard to do, but not impossible.
Again, not trying to argue here for the heck of it. Just presenting an opposing view which can be applicable to the subject matter, even if you disagree. Human ears are hard to fool in many ways, but easy in others.
Attachments
Again, a very bad comparison. Feedback is not linear, but it requires an open loop gain, which is absent in the loudspeaker.
To me it is impossible.
We will never come to terms when you base everything on your subjective opinion. I almost never accept those - for obvious reasons. Without data people can claim anything, Like stollen elections.
Many years ago, I listened to a lot of small ensemble/vocal classical concerts (Early Music up to Bach) performed inside a Berkeley church, sitting as close to the musicians as possible -- usually front row, ocassionally seats added on stage. Hence very much "live sound". However, the most "beautiful" overall sound would be heard on the balcony, though missing a lot of faint/sharp details.
Thanks for starting this round of discussion.
I don't think its fair to label my findings strictly as opinions. I haven't done that with any of your statements, mainly out of respect. I would at least expect to receive such courtesy back from you. No, I don't have a doctorate title but that in itself shouldn't automatically qualify your statements as being flawless and absolute or discredit my statements or arguments.
There is quite a bit of factual observation in my alleged "subjective opinions". My ears (as many others) are quite capable of hearing specific fine details in audio reproduction and are able to quantify them to a certain extent (even comparatively), given the listening environment is sufficiently controlled, unbiased and conducted in a blind fashion.
A skilled conductor can pick out a specific out of tune instrument from an entire orchestra playing at the same time. You'd be hard pressed to find a piece of equipment or software algorithm which can do the same. VW has been working on a similar project that can locate and pinpoint faulty machinery from the noise signature alone ie a failing bearing in a large piece of machinery. A human with skilled hearing could easily perform the same task.
Obviously by nature, human ears aren't calibrated instruments, but they definitely are very capable of picking out all sorts of anomalies, some of which certain test equipment alone can't do (either due to sampling errors, microphone distortion, self noise, lack of resolution, lack of software to decode the data, etc).
Test equipment is.consistent, but not perfect and it can't always quantify or explain every tiny bit of discrepancy in audio quality perceived by human hearing. People listen with their ears, not through test equipment. The human brain is more capable than you give it credit. I actually took the time and listened to your interview with Erin. I figured you at least deserve to be heard. I don't agree with everything you said, but I did agree with most of it.
We no longer live in the stone age or use witch craft to solve or explain scientific problems (at least I don't, despite your possible opinion of me). Measurements and computer analysis greatly aid in the design of highly accurate and great sounding loudspeakers as you know, but measurements alone shouldn't have the final say in predicting how natural and transparent a speaker will sound to human ears.
Not everything in life can be explained or quantified with current science and technology. Otherwise we wouldn't have to deal with cancer or other diseases causing us to die at a premature age. We have animals that can smell minute quantities of cancer cells. No piece of test equipment can do that. This philosophy can be extended to many other areas ie. self driving cars and robots doing brain surgery. The human involvement factor is still critical and very capable as an analytical tool, including the fine perception of audio reproduction.
I respect your work as a speaker designer but I don't appreciate the way you tend to talk down to others.
My apologies to the OP for going off topic here, but I have to speak up when I feel its needed.
Profiguy, I agree with what you said. When I see someone that wants to be right and make other people feel bad I ask, what are they wanting to get? What do they win? Will the world be a better place because of what they said? Bottom line, what is the payoff to them? This is a hobby forum and we come here to communicate with like-minded people, kindred spirits, for enjoyment. The EGO should be kept under control. I agree with you that not everything that we hear can be measured. And for the people that say it can be measured, how do they know that they are measuring the right thing? Maybe the thing they are measuring is not what is causing the sound difference people hear. I have never found a real reason why cables can sound so different when they measure the same.
Dr, Wayne Dyer said “ if you have the choice to be right or to be kind, be kind”.
Dr, Wayne Dyer said “ if you have the choice to be right or to be kind, be kind”.
Last edited by a moderator:
I agree that any speaker in question that one is interesting in acquiring will benefit that person from being as "flat" as possible, because then there is a starting point that in theory isn't colored. After than, a given listener may want to EQ the speaker system is some way that he / she finds more pleasing, and the shape of their hearing system and neural aspects will be a factor. The ELC's are subjective and they can be different from person to person. And they vary by amplitude of course.
And we can never be sure, on that basis, how someone else actually (precisely) hears a piece of music, especially as it gets more complex. What one can determine is what he / she likes. And someone might like brighter mids and highs, while others like that those muted and prefer tons of bass. So reaching for the "tone controls" isn't necessarily a bad thing. But as one person listens to the accentuated highs, that might send another screaming out of the room
. Not too mention, the balance of tonality can also vary across recording's. And while the sound engineer has lots of fancy equipment, ultimately he /she also sits back and listens to the final product and says that's right.The three things that I personally always want to avoid is excess sibilants in the high end -- especially "s" on vocals, blariness in the midrange, and boominess (or resonance) in the bass. Any of these three can cause listening fatigue. Some speakers may be criticized for sounding too bland, but bland IMO can be listened to for longer periods of time.