LR2 experiment with delay network for 8945P/810921 *PIC*

[jkim]

Although I read Roman's comparison of LR2 vs LR4, I wanted to do my own experiment to see any audible effects of different slopes and phase rotations of LR2 and LR4 xovers. As Roman stated, the biggest difference should be in their power response. But intuitively, the LR2 should provide more seamless integration of drivers due to its larger overlap. And how about their different levels of phase distortion (180 degrees versus 360 degrees out of phase)?

My experiement won't be like Roman's A/B test since I can't compare both versions at the same time---I can't afford two sets of drivers now But I heard my LR4 version for long enough (several months) with my favorite CD collection. So, I hope I can hear the difference, if any, by changing the xover back and forth (so, not a quick A/B test).

Since I use a flat baffle, a delay network is necessary for LR2 implementation. So, I used a first order ladder delay topology in the tweeter section just as Zaph did for his ZD5 design. In fact, I tried to pull asymmetric LR2 slopes (just like I do for LR4) without a delay network, but only realized that it's not possible with this driver combination---it either requires too shallow a slope for the tweeter or a nearly 1st order slope for the woofer, which is not good due to the woofer's beaming.

Surprisingly, I came up with an almost identical network topology, for both woofer and tweeter sections, to Zaph's (see the image below).

One thing I learned from using this first order ladder delay network is that although there is a room for adjusting the amount of delay, it's not possible to obtain linear levels of delay for the signals at all frequencies. Good phase alignment can be obtained around the xover frequency, but the delay is not enough at higher frequencies (e.g., above 6 kHz). So, it's not possible to obtain exactly the same phase tracking as that of LR2 with drivers that are physically time-aligned on a slanted/stepped baffle. Zaph's ZD5 also suffers from this---I confirmed this by replicating his design and also simulating LR2 with time aligned drivers. But the phase misalignment occurs far enough from the xover point. So, the effect won't be noticed in practice.

I ordered all xover parts today. More fun to come!

-jAy


Edit (as of 2-17-2009): The above simulation was for a preliminary crossover version. The actual crossover I ended up using is a bit different with a little higher xover point.

[jkim]

LR4 is popular because of its easy implementation. Without much experience, however, I personally believe that LR2 should be better than LR4 for some reasons. I know their power/polar responses are different, but let me give other reasons for my belief.

First, LR2's larger response overlap between drivers should provide more seamless integration of drivers (e.g., no abrupt transition of system harmonic distortions around the xover frequency). For example, take a look at Zaph's SR71 system harmonic distortion profile:

<A HREF="http://www.zaphaudio.com/SR71-measured-HD.gif">http://www.zaphaudio.com/SR71-measured-HD.gif</A>

This is a LR4 design. Notice a sudden fall of the 3rd order harmonic around 1.7 kHz and a fall of the 5th order harmonic above 1 kHz. This is due to the tweeter's much lower odd-order distortions than the woofer's around the xover point, and a narrow overlap between the drivers. This sudden transition can be an audible cue for where the tweeter takes up its duty. Now, see Zaph's ZD5 system harmonic distortion:

<A HREF="http://www.zaphaudio.com/ZD5-measured-HD-system.gif">http://www.zaphaudio.com/ZD5-measured-HD-system.gif</A>

This is a LR2 design. Compared to SR71's plot, it has a smoother transition w.r.t. system harmonic distortion over a wide frequency range.

Second, LR2's 180 degrees out of phase property (versus LR4's 360 degrees out of phase) should give a less amount of mismatch between woofer and tweeter in the time domain. Their difference can be shown in step responses and square wave reproduction. But will this be audible in normal music listening? I'm not sure.

I'm currently putting more weight on the first reason. But investigation of these issues will require more complicated experiments, which I can't perform now---a digital xover with DSP phase control may be handy for this kind of experiment.

Anyway, I'll see which I'll prefer between LR2 and LR4 versions of my 8945P/810921 2-way. But even if I prefer LR2 to LR4, I won't be able to figure out exactly why I do.

-jAy

[romanbednarek]

> LR4 is popular because of its easy
> implementation. Without much experience,
> however, I personally believe that LR2
> should be better than LR4 for some reasons.
> I know their power/polar responses are
> different, but let me give other reasons for
> my belief.

> First, LR2's larger response overlap between
> drivers should provide more seamless
> integration of drivers (e.g., no abrupt
> transition of system harmonic distortions
> around the xover frequency). For example,
> take a look at Zaph's SR71 system harmonic
> distortion profile:

>
> <A HREF="http://www.zaphaudio.com/SR71-measured-HD.gif">http://www.zaphaudio.com/SR71-measured-HD.gif</A>
> This is a LR4 design. Notice a sudden fall
> of the 3rd order harmonic around 1.7 kHz and
> a fall of the 5th order harmonic above 1
> kHz. This is due to the tweeter's much lower
> odd-order distortions than the woofer's
> around the xover point, and a narrow overlap
> between the drivers. This sudden transition
> can be an audible cue for where the tweeter
> takes up its duty. Now, see Zaph's ZD5
> system harmonic distortion:

>
> <A HREF="http://www.zaphaudio.com/ZD5-measured-HD-system.gif">http://www.zaphaudio.com/ZD5-measured-HD-system.gif</A>
> This is a LR2 design. Compared to SR71's
> plot, it has a smoother transition w.r.t.
> system harmonic distortion over a wide
> frequency range.

> Second, LR2's 180 degrees out of phase
> property (versus LR4's 360 degrees out of
> phase) should give a less amount of mismatch
> between woofer and tweeter in the time
> domain. Their difference can be shown in
> step responses and square wave reproduction.
> But will this be audible in normal music
> listening? I'm not sure.

> I'm currently putting more weight on the
> first reason. But investigation of these
> issues will require more complicated
> experiments, which I can't perform now---a
> digital xover with DSP phase control may be
> handy for this kind of experiment.

> Anyway, I'll see which I'll prefer between
> LR2 and LR4 versions of my 8945P/810921
> 2-way. But even if I prefer LR2 to LR4, I
> won't be able to figure out exactly why I
> do.

> -jAy

I'm interested in hearing your findings after your listening comparison but I have some comments regarding your current theories.

I don't agree that it is important to have distortion levels blend between drivers. Distortion is bad and if steeper slopes can reduce the amount of distortion overall this is a good thing. Tweeters can handle more power with steeper slopes.

Group delay and phase shift is another story and is one of those areas that is debatable. In theory a transient perfect design should do a better job of reproducing the original signal. However, it is believed that our ears interpret things differently and can't discern small delays across a frequency band. I always thought that this might have an impact on imaging and soundstage characteristics but when considering the small amounts of delays with higher order filters I'm not so sure our ears can tell the difference. If you are really concerned about group delay having an impact on the sound, look at the effects lower in frequency where the corresponding time delays are on a much larger scale (however the wavelengths are so long that our ears have trouble localizing anything).

I am still convinced that the main difference between LR2 and LR4 designs is in the power response. With a wider overlap between drivers there is a wider range where the drivers can be out of phase and not combine constructively. In this case a smoother transition may be obtainable with shallower slopes. A lot also depends on the dispersion characteristics of the driver as well as diffraction interactions. I'm convinced that some drivers/enclosures are better suited for one type over the other just based on the way they blend when considering the off axis performance.

When I did my comparison I actually tried two different 2nd order systems. The first one didn't use a ladder delay circuit but required that I tilt the cabinet back a few degrees to get the proper phase alignment. There was little difference between this version and the ladder delay version. I just found it easier to compare the ladder delay version with the 4th order because I was able to instantaneously switch between crossovers (with a relay switchbox) without having to move the speakers.

[Jeff B.]

> I'm interested in hearing your findings
> after your listening comparison but I have
> some comments regarding your current
> theories.

> I don't agree that it is important to have
> distortion levels blend between drivers.
> Distortion is bad and if steeper slopes can
> reduce the amount of distortion overall this
> is a good thing. Tweeters can handle more
> power with steeper slopes.

> Group delay and phase shift is another story
> and is one of those areas that is debatable.
> In theory a transient perfect design should
> do a better job of reproducing the original
> signal. However, it is believed that our
> ears interpret things differently and can't
> discern small delays across a frequency
> band. I always thought that this might have
> an impact on imaging and soundstage
> characteristics but when considering the
> small amounts of delays with higher order
> filters I'm not so sure our ears can tell
> the difference. If you are really concerned
> about group delay having an impact on the
> sound, look at the effects lower in
> frequency where the corresponding time
> delays are on a much larger scale (however
> the wavelengths are so long that our ears
> have trouble localizing anything).

> I am still convinced that the main
> difference between LR2 and LR4 designs is in
> the power response. With a wider overlap
> between drivers there is a wider range where
> the drivers can be out of phase and not
> combine constructively. In this case a
> smoother transition may be obtainable with
> shallower slopes. A lot also depends on the
> dispersion characteristics of the driver as
> well as diffraction interactions. I'm
> convinced that some drivers/enclosures are
> better suited for one type over the other
> just based on the way they blend when
> considering the off axis performance.

> When I did my comparison I actually tried
> two different 2nd order systems. The first
> one didn't use a ladder delay circuit but
> required that I tilt the cabinet back a few
> degrees to get the proper phase alignment.
> There was little difference between this
> version and the ladder delay version. I just
> found it easier to compare the ladder delay
> version with the 4th order because I was
> able to instantaneously switch between
> crossovers (with a relay switchbox) without
> having to move the speakers.

My vote is for the power response as well, but I believe I may the one who brought that up a few years ago ;- )

[jkim]

> I'm interested in hearing your findings
> after your listening comparison but I have
> some comments regarding your current
> theories.

Thanks for your comments, Roman.

> I don't agree that it is important to have
> distortion levels blend between drivers.
> Distortion is bad and if steeper slopes can
> reduce the amount of distortion overall this
> is a good thing. Tweeters can handle more
> power with steeper slopes.

Try to think a little deeper, Roman. A driver's (or a system's) nonlinear distortion profile is an important key to its unique tonality (e.g., paper versus metal cones). A sudden change in distortions over a frequency range is certainly unnatural. I do understand low distortions are always better. But my point is not about this. It's about smoother vs abrupt transition in distortion profiles. I believe that this can't be easily explained away until we perform a listening test in controlled conditions.

> Group delay and phase shift is another story
> and is one of those areas that is debatable.
> In theory a transient perfect design should
> do a better job of reproducing the original
> signal. However, it is believed that our
> ears interpret things differently and can't
> discern small delays across a frequency
> band. I always thought that this might have
> an impact on imaging and soundstage
> characteristics but when considering the
> small amounts of delays with higher order
> filters I'm not so sure our ears can tell
> the difference. If you are really concerned
> about group delay having an impact on the
> sound, look at the effects lower in
> frequency where the corresponding time
> delays are on a much larger scale (however
> the wavelengths are so long that our ears
> have trouble localizing anything).

As I said, I'm not sure on this issue. Like you, I don't believe such a short time delay can have an audible effect. But then why do some people (e.g., Thiel and John k) make so much effort to build a T-P design? Is it simply because they are obsessed with time coherence that has no audible effect? I don't know. One thing I'm sure of as a cognitive psychologist is that there are a lot of things yet to be known about human perception. AFAK, this narrow topic of human auditory perception of delayed signals hasn't been fully investigated.

> I am still convinced that the main
> difference between LR2 and LR4 designs is in
> the power response.

I agree. For example, LR2 should have narrower vertical lobes than LR4, over a wide frequency range.

-jAy

[Jed K.]

> Try to think a little deeper, Roman.
> -jAy

No offense, but just because someone disagrees with you (I also agree with Roman's take on the system's harmonic distortion), it doesn't mean they haven't thought about their response "deeply."

I'll go on by saying your test is completely flawed if you don't bother doing a direct A/B comparison and MEASURE the before and after effects of harmonic distortion, FR, and off axis response before AND after applying different slopes to each design.

On the other hand, if you want to be subjective, I can accept that. It will be interesting to hear about your listening impressions.

[jkim]

> No offense, but just because someone
> disagrees with you (I also agree with
> Roman's take on the system's harmonic
> distortion), it doesn't mean they haven't
> thought about their response
> "deeply."

Not offended at all. Don't be too sensitive here, Jed. I just wanted to make sure he fully understood my point. That's all. As I said, the effect of unsmooth driver transition w.r.t harmonic distortions is not an issue that can be easily disagreed or agreed upon without a rigorous experiment. I clearly said that it's my belief. I'm not 100% sure just as you're not 100% sure.

> I'll go on by saying your test is completely
> flawed if you don't bother doing a direct
> A/B comparison and MEASURE the before and
> after effects of harmonic distortion, FR,
> and off axis response before AND after
> applying different slopes to each design.

Did you miss this?

"Anyway, I'll see which I'll prefer between LR2 and LR4 versions of my 8945P/810921 2-way. But even if I prefer LR2 to LR4, I won't be able to figure out exactly why I do."

-jAy

[Jed K.]

> Did you miss this?

> "Anyway, I'll see which I'll prefer
> between LR2 and LR4 versions of my
> 8945P/810921 2-way. But even if I prefer LR2
> to LR4, I won't be able to figure out
> exactly why I do."

> -jAy

Did you miss this?--"On the other hand, if you want to be subjective, I can accept that. It will be interesting to hear about your listening impressions. Jed"

When you use someone else's research based on objective data, and I'll go on by saying, research I admire because of its scientific methods; it seems silly to imply other's opinions are not valid when you are taking a subjective approach to your own theory.

This is not an argument, just an observation.

Back to speaker building.

Jed

[jkim]

> Did you miss this?--"On the other hand,
> if you want to be subjective, I can accept
> that. It will be interesting to hear about
> your listening impressions. Jed"

Nope, either. But then I don't understand why you said such a thing (re objective testing) in your reply if your intention is not to offend me.

> When you use someone else's research based
> on objective data, and I'll go on by saying,
> research I admire because of its scientific
> methods; it seems silly to imply other's
> opinions are not valid when you are taking a
> subjective approach to your own theory.

I didn't simply disagree with Roman. I just wanted to point out that the problem is not simple as it appears but needs empirical validation. I do admit that my words "try to think deeper" can be misunderstood. But I don't understand why you respond this way. What we do here, anyway, is casual forum talks. We exchange our thoughts, beliefs, and theories. Go back to my original post. Even if you measure everything you can, most of these issues can't be resolved unless we perform well designed, behavioral experiements with human listeners. Even so, it may need time for similar experiments to be replicated by others.

-jAy

[jkrutke]

That system will probably sound great at low levels. At higher levels you will have a power handling issue with the tweeter. Take a look at your tweeter transfer function - it's too shallow. If anything, it's a good experiment to let you know what an overdriven tweeter sounds like. A LR2 with low Qts tweeters at a low Fc will always be problematic.

Don't be too quick to pass judgement on LR2 based on your results with this one experiment. While I do prefer LR2 myself, I really didn't know until I heard 10 different well-designed LR2 systems.

As to why I like LR2 better, I'm with Jeff B on this one. I suspect it's mostly power response, with phase wrap being a secondary, unproven possibility. Most drivers are just meant for LR4 however, and as such you could say that LR4 is "better". (I hate just using the word "better", it's too vague) Better, as in LR4 is simply more likely to work. The stars really have to align to pull off a well-designed LR2.




(Originally posted by: jkrutke)

[jkim]

Thanks for chiming in, John.

Yes, the issue of tweeter power handling. I understand what you say. Let me see what'll happen. In fact, I compared this system's tweeter rolloff to many existing LR2 and LR4 design's tweeter rolloffs that use low xover points, including your ZD5, BAMTM, my 2-ways, etc. My tentative conclusion is that it'll be okay even at a moderately high level. Don't be misled by the xover frequency in my plot. It seems to be 2.1 to 2.2 kHz but it's low because of the woofer's dip around the frequency. The actual tweeter rolloff is only a little bit shifted towards the lower side than your ZD5's XT25 rolloff. I know the 810921's natural rolloff starts early due to low Qts. That's why you're concerned. I compared its filtered rolloff to some of low Fc LR4 design's rolloffs (e.g., my 2-ways and your BAMTM). Actually, this LR2 rolloff has lower amplitude than LR4 responses in the 800 Hz to 2 kHz range, which should be critical to the tweeter's power handling. Below 500-600 Hz, this LR2 rolloff has significantly higher amplitude than low Fc LR4's. This observation applies to your ZD5, too.

I know you once said you subjectively prefer LR2. But why? As you "suspect," it could be mostly power response. I just wanted to think of a couple of other possible measurable factors. As I said, however, we can't be sure which is the main cause, until we perform well-designed, scientific experiments with listeners.

-jAy

[romanbednarek]

> My vote is for the power response as well,
> but I believe I may the one who brought that
> up a few years ago ;- )

Yeah, we discussed this one many moons ago. I think I had come to the power response conclusion after listening to the two versions but your plots made it more evident (and they're still up on my site, thanks again).

[romanbednarek]

> Thanks for your comments, Roman.

> Try to think a little deeper, Roman. A
> driver's (or a system's) nonlinear
> distortion profile is an important key to
> its unique tonality (e.g., paper versus
> metal cones). A sudden change in distortions
> over a frequency range is certainly
> unnatural. I do understand low distortions
> are always better. But my point is not about
> this. It's about smoother vs abrupt
> transition in distortion profiles. I believe
> that this can't be easily explained away
> until we perform a listening test in
> controlled conditions.

I can understand your point to a degree but a LOT is going to depend on the drivers used and steeper slopes might not always present an abrupt transition. If you are worried about abrupt transitions just look at the harmonic distortion peaks that are present with most metal cone drivers when the resonance is excited by a subharmonic. In this case a steeper sloped crossover has a better chance of removing these non-linear distortion peaks out of the picture. The other point is whether, subjectively, a non-even non-linear distortion profile actually does sound worse than one that has more distortion but is more even across the bandwidth. This is just an assumption you have made based on a theory and it may be difficult to prove or disprove this theory due to the abundance of other variables present in the analysis. My opinion is still that my ears would prefer less distortion but I haven't had the chance to prove this or not.

[jkim]

Provided Link: My Speaker Building Pages


> I can understand your point to a degree but
> a LOT is going to depend on the drivers used
> and steeper slopes might not always present
> an abrupt transition. If you are worried
> about abrupt transitions just look at the
> harmonic distortion peaks that are present
> with most metal cone drivers when the
> resonance is excited by a subharmonic. In
> this case a steeper sloped crossover has a
> better chance of removing these non-linear
> distortion peaks out of the picture. The
> other point is whether, subjectively, a
> non-even non-linear distortion profile
> actually does sound worse than one that has
> more distortion but is more even across the
> bandwidth. This is just an assumption you
> have made based on a theory and it may be
> difficult to prove or disprove this theory
> due to the abundance of other variables
> present in the analysis. My opinion is still
> that my ears would prefer less distortion
> but I haven't had the chance to prove this
> or not.

Thanks for another consideration, Roman. Oh, BTW, if my "try to think a little deeper" words may have offended you even a tiny bit, as Jed pointed out yesterday, please accept my apology.

Actually, I decided not to say more on this point after I saw people's responses. Except you, it seems that other people simply think they understand what I mean. Often times, it takes frequent communications for people to understand what each other really means. This happens even in review processes of academic journal articles according to my experience.

Since you replied, I'll try a bit more. Let me ask a question. Why do we want to have a flat frequency response on our speaker system? It's a problem of balance. We want our speakers to reproduce accurate balance of sound amplitude over a frequency band. If a frequency range is emphasized over another, it should be obtrusive and audible.

Now, do you remember that Zaph once talked about the effect of a driver's distortion profile upon choosing a proper amount of BSC? For instance, accroding to his experience, a midwoofer that has a rising distortion profile (e.g., Seas P18RNX and ER18RNX) requires a larger amount of BSC than the one that doesn't (e.g., Adire Extremis). I still agree with him. But why? Because this is another problem of balance. A driver's harmonic distortion profile is simply another amplitude curve, just like the fundemental tone's frequency response, that can affect a system's tonal balance.

If you have an experience of finding a good balance between midwoofer and tweeter's SPLs in a 2-way system or balance between bass and midrange units' SPLs in a 3-way system (certainly you have much), you must have noticed that abrupt SPL imbalance close to the xover point (like a brick-wall step function) is more obtrusive than a gradual imbalance far from the xover frequency (like a logistic transfer function).

Likewise, suppose that we use a bass and a midrange drivers that have very different distortion characteristics. For instance, the bass driver is a metal cone woofer with consistently higher 3rd order harmonics than 2nd order ones over a wide frequency range, but the midrange is a paper cone with consistenly higher 2nd order harmonics than 3rd order ones. If we use low order acoustic filters in this case, their difference in distortion profile will be mixed well over a wide frequency range, but if we use higher order filters, the abrupt change in the system harmonic distortion profile may be obtrusive. I strongly suspect that this is a situation in which people feel that their drivers are not integrated well even if the system has a nice, flat frequency response.

According to my experience of doing this audio hobby (not just DIY), I remember I sometimes heard "tweeter" instead of the "speaker" although the speaker should have a reasonably flat FR. And we often see people say this kind of experience. Why? Just because of imbalance in the FR of fundamental tones? I strongly suspect that this is the situation that may be explained by my theory. If two drivers have very different distortion profiles (i.e., poor driver match), the problem won't be hidden by any type of xover. But I believe using low order filters with a wide driver overlap can mitigate the problem.

There may be other factors that may cause this. But I think my theory is certainly "plausible." What do you think?

-jAy

[jkim]

Provided Link: My Speaker Building Pages


Assuming that two systems' on-axis, anechoic responses are matched with each other, LR2 and LR4 versions' different power responses will definitely affect our perceived "in-room" loudness, over a certain frequency range around the xover point, at the listening position. Also, room interference may be smaller for LR2 than for LR4 due to the LR2's narrower vertical lobing.

Although I can't control room interference, I can control the sytems' tonal balance. Since I don't have measurement equipment, I'll simply try to match two system's tonal balance as closely as possible. So, this will be a different test from Roman's. Actually, even if I had measurement equipment, I would do the test this way since I'm interested in LR2 vs LR4's subjective sound quality rather than the effect of their different, already-evident-on-paper power responses. That is, even if I could match their on-axis frequency responses by measurement, I wouldn't do that but instead I would match their "perceived, in-room" tonal balance as closely as possible. In this sense, the goal of my test is a bit different from Roman's.

BTW, I've decided to use a switch between xovers for a quick A/B test.

-jAy

[romanbednarek]

> Assuming that two systems' on-axis, anechoic
> responses are matched with each other, LR2
> and LR4 versions' different power responses
> will definitely affect our perceived
> "in-room" loudness, over a certain
> frequency range around the xover point, at
> the listening position. Also, room
> interference may be smaller for LR2 than for
> LR4 due to the LR2's narrower vertical
> lobing.

> Although I can't control room interference,
> I can control the sytems' tonal balance.
> Since I don't have measurement equipment,
> I'll simply try to match two system's tonal
> balance as closely as possible. So, this
> will be a different test from Roman's.
> Actually, even if I had measurement
> equipment, I would do the test this way
> since I'm interested in LR2 vs LR4's
> subjective sound quality rather than the
> effect of their different,
> already-evident-on-paper power responses.
> That is, even if I could match their on-axis
> frequency responses by measurement, I
> wouldn't do that but instead I would match
> their "perceived, in-room" tonal
> balance as closely as possible. In this
> sense, the goal of my test is a bit
> different from Roman's.

> BTW, I've decided to use a switch between
> xovers for a quick A/B test.

> -jAy

I suspect that in your attempt to make both crossovers similar that if you were to measure each version that the on-axis responses wouldn't be match but instead would include a bit of emphasis to compensate for the differences in power response. I was really surprised at how similar I was able to get the on-axis responses of my LR2 and LR4 versions when I tuned them.

I've learned that when performing experiments it is best to remove as many variables as possible because it can be easy to draw conclusions that may be influence by other factors that weren't under consideration. It still doesn't hurt to experiment though and the A/B switch makes it easier to compare without being hindered by acoustic memory issues.

[romanbednarek]

> Thanks for another consideration, Roman. Oh,
> BTW, if my "try to think a little
> deeper" words may have offended you
> even a tiny bit, as Jed pointed out
> yesterday, please accept my apology.

> Actually, I decided not to say more on this
> point after I saw people's responses. Except
> you, it seems that other people simply think
> they understand what I mean. Often times, it
> takes frequent communications for people to
> understand what each other really means.
> This happens even in review processes of
> academic journal articles according to my
> experience.

> Since you replied, I'll try a bit more. Let
> me ask a question. Why do we want to have a
> flat frequency response on our speaker
> system? It's a problem of balance. We want
> our speakers to reproduce accurate balance
> of sound amplitude over a frequency band. If
> a frequency range is emphasized over
> another, it should be obtrusive and audible.

> Now, do you remember that Zaph once talked
> about the effect of a driver's distortion
> profile upon choosing a proper amount of
> BSC? For instance, accroding to his
> experience, a midwoofer that has a rising
> distortion profile (e.g., Seas P18RNX and
> ER18RNX) requires a larger amount of BSC
> than the one that doesn't (e.g., Adire
> Extremis). I still agree with him. But why?
> Because this is another problem of balance.
> A driver's harmonic distortion profile is
> simply another amplitude curve, just like
> the fundemental tone's frequency response,
> that can affect a system's tonal balance.

> If you have an experience of finding a good
> balance between midwoofer and tweeter's SPLs
> in a 2-way system or balance between bass
> and midrange units' SPLs in a 3-way system
> (certainly you have much), you must have
> noticed that abrupt SPL imbalance close to
> the xover point (like a brick-wall step
> function) is more obtrusive than a gradual
> imbalance far from the xover frequency (like
> a logistic transfer function).

> Likewise, suppose that we use a bass and a
> midrange drivers that have very different
> distortion characteristics. For instance,
> the bass driver is a metal cone woofer with
> consistently higher 3rd order harmonics than
> 2nd order ones over a wide frequency range,
> but the midrange is a paper cone with
> consistenly higher 2nd order harmonics than
> 3rd order ones. If we use low order acoustic
> filters in this case, their difference in
> distortion profile will be mixed well over a
> wide frequency range, but if we use higher
> order filters, the abrupt change in the
> system harmonic distortion profile may be
> obtrusive. I strongly suspect that this is a
> situation in which people feel that their
> drivers are not integrated well even if the
> system has a nice, flat frequency response.

> According to my experience of doing this
> audio hobby (not just DIY), I remember I
> sometimes heard "tweeter" instead
> of the "speaker" although the
> speaker should have a reasonably flat FR.
> And we often see people say this kind of
> experience. Why? Just because of imbalance
> in the FR of fundamental tones? I strongly
> suspect that this is the situation that may
> be explained by my theory. If two drivers
> have very different distortion profiles
> (i.e., poor driver match), the problem won't
> be hidden by any type of xover. But I
> believe using low order filters with a wide
> driver overlap can mitigate the problem.

> There may be other factors that may cause
> this. But I think my theory is certainly
> "plausible." What do you think?

> -jAy

Don't worry about offending me, I'm pretty passive and it takes quite a bit to ruffle my feathers.

Your points about smoothing out the transition between drivers in terms of distortion has some merit. However, one thing that I've learned over the years that distortion in one frequency range will sound different than distortion in another frequency range. For example, I don't mind odd order harmonic distortion from metal cone drivers in the bass region because I don't find it to be harsh and it actually improves the imaging by superimposing localizable higher frequency tones over the fundamental.

Another thing to consider is if higher high-order distortion harmonics from a lower frequency driver fall in the same range as low-order distortion harmonics from the higher frequency driver, does the abundance of harmonic tones focused in one area cause issues.

I think that your theory makes the most sense when considering drivers which are vastly different in distortion profiles and a quick transition between the drivers can change the texture of the tones within that range.

I still think that the fundamental thing to consider is the fact that distortion sounds different depending on which range it is in. I don't like odd order distortion in midrange drivers because the harmonics fall into a region where my ears are very sensitive and the nature of odd harmonics make them easier to detect. However, with a tweeter or bass driver, odd order harmonics aren't as offensive to me due to the region those harmonics exist.

This topic is pretty complex and I don't think that there is a single right answer due to the variety of drivers and distortion profiles. Even my comments above are a bit weak and stray a bit from the original topic but I just thought of something that seems more relevant that I'll add next.

Consider the case of a midbass driver with high odd order distortion and a tweeter with very low distortion in general or even relatively higher even order distortion. Would I prefer to use a larger amount of overlap so that the odd and even order distortion blend across that region or would it sound better to use a steeper slopes so that the amount of, what I call offensive, odd order distortion is less even though the transition is more abrupt? I guess what I keep coming back to is that there are so many different situations that it is difficult to generalize a solution for all of them and with the situation just presented I think that I would prefer the steeper slopes just to keep the odd order distortion down regardless of the transition.

[jkim]

> Don't worry about offending me, I'm pretty
> passive and it takes quite a bit to ruffle
> my feathers.

> Your points about smoothing out the
> transition between drivers in terms of
> distortion has some merit. However, one
> thing that I've learned over the years that
> distortion in one frequency range will sound
> different than distortion in another
> frequency range. For example, I don't mind
> odd order harmonic distortion from metal
> cone drivers in the bass region because I
> don't find it to be harsh and it actually
> improves the imaging by superimposing
> localizable higher frequency tones over the
> fundamental.

> Another thing to consider is if higher
> high-order distortion harmonics from a lower
> frequency driver fall in the same range as
> low-order distortion harmonics from the
> higher frequency driver, does the abundance
> of harmonic tones focused in one area cause
> issues.

> I think that your theory makes the most
> sense when considering drivers which are
> vastly different in distortion profiles and
> a quick transition between the drivers can
> change the texture of the tones within that
> range.

> I still think that the fundamental thing to
> consider is the fact that distortion sounds
> different depending on which range it is in.
> I don't like odd order distortion in
> midrange drivers because the harmonics fall
> into a region where my ears are very
> sensitive and the nature of odd harmonics
> make them easier to detect. However, with a
> tweeter or bass driver, odd order harmonics
> aren't as offensive to me due to the region
> those harmonics exist.

> This topic is pretty complex and I don't
> think that there is a single right answer
> due to the variety of drivers and distortion
> profiles. Even my comments above are a bit
> weak and stray a bit from the original topic
> but I just thought of something that seems
> more relevant that I'll add next.

> Consider the case of a midbass driver with
> high odd order distortion and a tweeter with
> very low distortion in general or even
> relatively higher even order distortion.
> Would I prefer to use a larger amount of
> overlap so that the odd and even order
> distortion blend across that region or would
> it sound better to use a steeper slopes so
> that the amount of, what I call offensive,
> odd order distortion is less even though the
> transition is more abrupt? I guess what I
> keep coming back to is that there are so
> many different situations that it is
> difficult to generalize a solution for all
> of them and with the situation just
> presented I think that I would prefer the
> steeper slopes just to keep the odd order
> distortion down regardless of the
> transition.

Thanks for the reply, Roman. This discussion is very interesting. As for the last example you came up with, I think it really needs another experiment to see which one we will prefer. I remember that, to the statement that the Seas Excel metal cone should be used with a very low Fc steep filter, Dennis Murphy once replied like "Is it really so?" The level of odd-order distortion will definitely be lowered by using a low Fc, steep filter. But is it really critical to real listening experience? Perhaps, the issue may be reduced to a matter of personal preference. Some people may like lower distortion sound regardless of transition being smooth or not, and others may prefer sound quality of smooth transition at the cost of a little lower distortion from a driver.

-jAy

[jkim]

> I suspect that in your attempt to make both
> crossovers similar that if you were to
> measure each version that the on-axis
> responses wouldn't be match but instead
> would include a bit of emphasis to
> compensate for the differences in power
> response. I was really surprised at how
> similar I was able to get the on-axis
> responses of my LR2 and LR4 versions when I
> tuned them.

Actually, I think I can see the effect you said above quite accurately even in my simulations. Since I use exactly the same in-box modeled responses between the two crossover designs, absolute measured accuracy for a single design has little importance. Simulation will give me quite accurate comparison between them in this case.

-jAy

[brianpowers27]

THe following article (Frequency Filtering, in practice) does a good job of explaining the problem associated with higher order filters, in the time domain.

It is possible that what we end up hearing in higher order filters is the ringing associated with the filters response. I would like to learn a little bit more about the time based filter response of the lr4.