gedlee said:
I don't think it is a fair statement to say that "none of that is applicable".
The examples in the article were using MLSSA from DRA Labs, which has become something of a de facto standard. Now it may be that other measurement systems don't suffer from these problems, but look at the problems that are inherent with the MLSSA system.
When it is gated at 6 msec, there is insufficient resolution to see anything useful below about 2 kHz.
What's even stranger are the artifacts created by the gating. When the window is 217 msec, the resonances behave as one would expect, delaying with time. But as the gating time is reduced then something bizarre happens. The "resonances" *change frequency with time*.
This is something that is very difficult to create with an actual physical device. In the real world, it would sound like a "chirp" that increased in frequency as the amplitude decayed. Not something that I've ever heard...
I haven't read the article for a while, so maybe he addresses the reasons for these artifacts. I would guess that it has to do with whatever window is chosen for the gating operation. And unfortunately, I'm sure that there is no possible window that won't create artifacts. Different gating windows will just produce different artifacts.
It is possible that there is some other measurement system besides MLSSA that won't have these artifacts with short gating windows, but I kind of doubt it. Remember, there's no such thing as a free lunch. And I don't think there is some new kind of math that is better than the math used in MLSSA....
Hi Charles
TDS has outstanding noise immunity, much better than MLS for example although it takes a little longer.
One finds that the response of most speakers does change some, until you are some distance away.
For what I do, measuring speakers (usually for the purpose of gathering data for crossover development) I use 2 meters most of the time.
Even with the TDS system, at 2 meters, it is desirable to use a tower to reduce the magnitude of anything other than the direct sound as much as possible.
Here is a shot of a speaker on a tower with a mic at 3 meters.
In this case, 20 feet high is sufficient.
Measuring a subwoofer on the other hand, essentially has to be done in halfspace ground plane, away from buildings to avoid reflections and if the subwoofer is large, then to do it at a larger distance than one meter, which reduces the error the physical volume causes. At work for larger woofer we use 10 meters which is –20dB from 1 meter and then drive the system at what would be 100Watts (+20dB from 1 Watt) into that nominal impedance.
Best,
Tom Danley
TDS has outstanding noise immunity, much better than MLS for example although it takes a little longer.
One finds that the response of most speakers does change some, until you are some distance away.
For what I do, measuring speakers (usually for the purpose of gathering data for crossover development) I use 2 meters most of the time.
Even with the TDS system, at 2 meters, it is desirable to use a tower to reduce the magnitude of anything other than the direct sound as much as possible.
Here is a shot of a speaker on a tower with a mic at 3 meters.
In this case, 20 feet high is sufficient.
Measuring a subwoofer on the other hand, essentially has to be done in halfspace ground plane, away from buildings to avoid reflections and if the subwoofer is large, then to do it at a larger distance than one meter, which reduces the error the physical volume causes. At work for larger woofer we use 10 meters which is –20dB from 1 meter and then drive the system at what would be 100Watts (+20dB from 1 Watt) into that nominal impedance.
Best,
Tom Danley
Attachments
Tom Danley said:
Yes, when I was designing speakers (in a former lifetime) I used the Techron TEF 12 system.
Tom Danley said:
Yes, I used many different techniques to obtain as large of a time window without reflections as possible. Unfortunately, outdoor measurements can be very problematic due to temperature variations. I suppose if you live in Southern California it might not be too bad. But many suspension components are very sensitive to temperature, especially many of the commonly used ferrofluids. When I first used Dynaudio tweeters with ferrofluid 20 years ago I found that only a few degrees F could make a big difference in the T/S parameters.
Indoors it is difficult to find a large enough space. Our building had one section with a 20' ceiling. Putting the DUT and mic in the center allows roughly a 20 msec window.
But as you can see from the article in Stereophile, this is still only giving good results down to perhaps 500 or 1000 Hz.
Tom Danley said:
Yes, the ground plane technique is discussed in part two of the linked article in Stereophile. But I still have not found a way to measure between 100 Hz and 1000 Hz that I trust. Perhaps a good anechoic chamber would work, but I don't have access to anything like that.
I think that this is actually one of the biggest problems with speaker design. Almost nobody can make accurate measurements in this region that can account for cabinet diffraction effects. So everyone ends up more or less guessing -- which is one reason why different designs sound so different.
Charles Hansen said:
I don't think that this is true because cabinet diffraction is very short in time, certainly much less than 6 ms.
The problems that you allude to are not measurement system dependent but just natural limitations of time windowing. But your 2 kHz. limit for 6 ms. is way off the mark. Its more like 200 Hz. for normal Q's and maybe 800 Hz. for very high Q's. The simulated signals in the Stereophile article were unreasonably high Q's. That was my point.
gedlee said:
Well, the problem is the transition from 2-pi space at high frequencies to 4-pi space at lower frequencies (and then back to 2-pi space again at even lower frequencies when the floor comes into play, although at that point we are well into the area where room modes become a big factor and all bets are off).
The amateurs have taken to calling this "the baffle step", but whatever you call it, it clearly affects the frequency response of the speaker in a very critical region.
Most people who are aware of this phenomenon use a full 6 dB compensation in the crossover. But there are others who say that the diffraction that causes the "step" in the frequency response is delayed in time and that the ear is more sensitive to the first arrival. Sometimes they will split the difference and use a 3 dB compensation in the crossover.
At any rate, this effect (and how it is handled - or not) by the speaker designer will make a large difference in the perceived tonal balance of the speaker.
gedlee said:
And I think your point is misguided.
If you look at every measurement done on every speaker in Stereophile's tests, they *all* show artifacts exactly like those show in these simulated tests. So I would assert that the simulations are very realistic indeed. And if you read the footnotes to the article, you would see that he didn't pull the parameters for his simulations out of thin air. He based them upon research that was done on real loudspeakers earlier by a researcher in the UK.
I would say that 6 msec gives very good data down to 2 kHz or so. The data down to 1 kHz is usable, but requires a fair amount of mental interpretation. Below that is basically just garbage.
I think that unless one expects high Q resonances between points limited by resolution, one can still use data down to 200Hz for design purposes with a 6ms gate.
soongsc said:
Thinking is good.
However, you might also try reading -- you could learn something.
From the linked article:
"By limiting the time window of a measurement, you also limit its frequency resolution. The relationship is a simple one: if you apply a time window of 6 milliseconds (0.006s—not unusual for measurements conducted in a typical domestic room), then the frequency resolution of the measurement will be limited to 1/0.006 = 167Hz."
So if you think that a frequency resolution of 167 Hz is useful for making measurements all the way down to 200 Hz, be my guest. But that is not a useful approach in my experience. (And I've been using quasi-anechoic measurements for over 20 years.)
If you try using a gated sine sweep, you will find that the results are pretty must similar. If you have high Q resonances anywhere within a 167Hz range, the driver you are using must be pretty bad in the first place. I am not used to using such ill quality drivers. Additionally, if you use only one method and never compared, I guess it is reasonable to worry. But in reality, do you have any data that actually shows it influenced your design decisions? Or did you just make judgement as you reason above?Charles Hansen said:
I have used quasi-anechoic chamber, and the results really do not make enough difference from a normal room of similar size. Room size determine how you set the gate window.
Charles Hansen said:
Its the first time I hear that some of us amateurs relax the full BSC compensation for psychoacoustic time perception reasons. You obviously got it from somewhere but I don't know how can this stand.
Me for one I always do it for Hi-Fi speakers since I don't remember anymore, for one reason. And that reason is room gain.
I find the linked
study By John Kreskovsky, quite right.
salas said:
I don't understand you people.
You don't bother to even read the source that *you* cite yourself!!!
Here are some quotes from *your* link:
"It is often said that while full baffle step compensation yields flat on axis response such system often sound bass heavy."
"The last figure on this page shows the power response when only a 3dB correction to the baffle step is applied."
Next time try opening your eyes instead of your mouth.
soongsc said:
It has to be. There is no way around it. The math is the math. And the point is that it is extremely difficult to achieve any kind of accurate measurements between 100 Hz and 1000 Hz.
Try reading the articles I linked before. Then read them again. Then again. Until you finally understand them.
Then you will realize that 99% of all published measurements are virtually worthless in this frequency range. There are only two ways to get accurate measurements down to 100 Hz:
a) Use a large anechoic chamber.
b) Measure outdoors with the speaker and microphone suspended at least 50 feet (or better yet 100 feet) in the air.
Option (b) is clearly impractical. And option (a) is only possible for a literal handful of facilities in the entire world.
So if you think about it, you will realize that you actually have very little idea of how your speaker responds in this range. If you don't know the actual response, you cannot design the crossover properly. You may have a target, but you have no idea what correction needs to be applied to reach the target.
Bass heaviness is possible, especialy if the bass enclosure is not properly damped.Charles Hansen said:
Charles Hansen said:
Next time try opening your eyes instead of your mouth and see that you have to read next page also. See about correlation in a stereo pair.
Here is one quote from my link in RED letters:
'' Additionally, we must also consider the interaction of both speakers in a stereo configuration. This is addressed on the next page.''
salas said:
Again, you are arguing against yourself!
First you say that the "baffle step correction" must always be 6 dB and that you have never even heard it suggested that anything less should be used.
Then you post a link that says the exact opposite of what you say.
And now you post *another* link that says the exact opposite of what you say.
Here is the quote from the "next page" about "correlation in a stereo pair":
"Clearly, 3dB baffle step correction would yield flat power response
down to 50 Hz and below that the baffle step correction can be reduced or eliminated."
What part of this reference don't you understand???
The room mode that are shown in the stereophile test data look funny. Since there is no information on the actual room they measured in, it make no sense to be to spend time talking about those articles. Most anechoic chambers are designed for down to 100Hz, whatever that means in actual levels...Charles Hansen said:
It seems like you don't actualy have your own data that shows different windowing will change your design decisions.🙂
To Charles:
What is that you don't understand? That 3 dB is the final better choice for a stereo pair in a room?
I did not post another link, you just flipped page as it is recommended in the one and only link I posted.
You understood:
''Again, you are arguing against yourself!
First you say that the "baffle step correction" must always be 6 dB and that you have never even heard it suggested that anything less should be used.''
I wrote:
''Its the first time I hear that some of us amateurs relax the full BSC compensation for psychoacoustic time perception reasons. You obviously got it from somewhere but I don't know how can this stand.
Me for one I always do it for Hi-Fi speakers since I don't remember anymore, for one reason. And that reason is room gain.''
''Relaxing'' as ''lessening''. Less!
See what I said? I always use less than full. I wasn't arguing against 3 dB. I was arguing for the reason. Psychoacoustics VS plain room gain.
I wonder how you kept on reading a link that points to the opposite that you understood from me in a hurry, without just wondering that you may ask for a clarification from me. An oxymoron should raise a question mark not an out of hand dismissal.
What is that you don't understand? That 3 dB is the final better choice for a stereo pair in a room?
I did not post another link, you just flipped page as it is recommended in the one and only link I posted.
You understood:
''Again, you are arguing against yourself!
First you say that the "baffle step correction" must always be 6 dB and that you have never even heard it suggested that anything less should be used.''
I wrote:
''Its the first time I hear that some of us amateurs relax the full BSC compensation for psychoacoustic time perception reasons. You obviously got it from somewhere but I don't know how can this stand.
Me for one I always do it for Hi-Fi speakers since I don't remember anymore, for one reason. And that reason is room gain.''
''Relaxing'' as ''lessening''. Less!
See what I said? I always use less than full. I wasn't arguing against 3 dB. I was arguing for the reason. Psychoacoustics VS plain room gain.
I wonder how you kept on reading a link that points to the opposite that you understood from me in a hurry, without just wondering that you may ask for a clarification from me. An oxymoron should raise a question mark not an out of hand dismissal.
salas said:
OK Salas, now I understand what you are saying. Please accept my apologies as I misunderstood your initial post to mean that one should *always* use a full 6 dB "baffle step correction".
Now I see that we both agree that the "baffle step correction" can be less than 6 dB. Your point was that the reason to reduce it is because of room gain instead of any other factor. This is something of a sidetrack from my original topic, but since you brought it up we can look at this more closely.
Kreskovsky's studies are largely focused on certain parameters and are made using specific assumptions. And that may be the complete and whole story. And it may not be. (In fact, I am sure that it is not.)
The point I was originally making is that it is extremely difficult to make accurate measurements of a real loudspeaker in this frequency range. That is why Kreskovsky is forced to use *models* in his exposition. And even he points out some of the shortcomings of trying to make predictions solely through the use of his models:
"there are other factors to consider. Vibration of wall dissipates power at low frequency and low frequency energy can easily be transmitted through most structures."
"Other factors must also be considered though. For example are still faced with room pressurization and room modes which will further complicate the issue. It is no wonder that clean low frequency reproduction in a confined space is so difficult to obtain."
My point is that it would be great if we could *actually measure* the loudspeaker's response instead of using a simplified model that can never fully describe the behavior of the actual system. But that is extremely difficult. And, back to my original point, impossible if one is using a 6 msec time window.
soongsc said:
I am not talking about the averaged in-room response curves that Stereophile sometimes publishes. I am talking about the CSD (waterfall) plots that they publish. If you look at these, you will see the exact same problems that Keith Howard described in his article I previously linked.
The point here is that the assumptions that KH made when modeling his data were valid ones, that correlate well with actual measurements made of actual speakers when using a 6 msec time window.
And if you take the time to read the article, you will see that the data gathered with a 6 msec time window is almost useless below 1 kHz. And furthermore, it is extremely difficult to achieve better measurements without the use of facilities which are far beyond the reach of all but a handful of large research institutions (eg, large anechoic chambers).
Charles Hansen said:
Of course. Accepted.
I also always measure. Each room has different room gain for different positions. Different nodes, different construction and losses, different LF alignments for each speaker system.
John just got the main trend in a simple short text using sims right. That is why I linked his study. It is -mainly- real.
When we DIY we know the system, the positioning and the room. We can use non gated for the nodal region. And adjust for our custom situation. Nice one!
When designing for commercial, its a complex guessing game if not having access to a huge anechoic room or measuring a speaker hanging from a crane at the quiet vast Sahara.😀
Thank you for accepting my apology so quickly!
But in the nodal region, the response varies with time. It takes time for the standing waves to develop and create nodes and anti-nodes. And I don't think anyone knows for sure how the ear/brain interprets the initial energy versus the energy integrated over time.
If the ear/brain integrated *fully* over time, then we could just use pink noise and graphic equalizers to get perfect bass in any listening room with any speaker (at least at one listening position). But it's not that simple.
salas said:
But in the nodal region, the response varies with time. It takes time for the standing waves to develop and create nodes and anti-nodes. And I don't think anyone knows for sure how the ear/brain interprets the initial energy versus the energy integrated over time.
If the ear/brain integrated *fully* over time, then we could just use pink noise and graphic equalizers to get perfect bass in any listening room with any speaker (at least at one listening position). But it's not that simple.
- Status
- Not open for further replies.