The myth about comb filtering in columns is busted

[dickwest]

Has this been posted? An interesting and long history of different types of speaker systems culminating in a line array of what appears to be the $0.69 drivers many of you used. In addition, the add-on page to this person's web site poses a position on the apparent fallacy of comb filtering effects in line arrays. Interesting . . .

<A HREF="http://www.roger-russell.com/columns/columns.htm#single">http://www.roger-russell.com/columns/columns.htm#single</A>

[rudyjakubin]

http://img.photobucket.com/albums/v641/rjakubin/Speaker-1.jpg


> Has this been posted?
>
> <A HREF="http://www.roger-russell.com/columns/columns.htm#single">http://www.roger-russell.com/columns/columns.htm#single</A>

Yep, 'bout a couple days ago. Nice marketing speak.
My first ever project was the line array in this pic and 2 years later it's still my reference standard I compare all of my daring endeavors to unseat the 'king'...
Sheeesh! I keep knocking my head on 1 or 2 dozen different designs and ask myself 'Why'...

[marlboro]

It generated a huge series of responses not all of which were even ordered harmonics of the orginal.

One of the better ones was by Jim Griffin at:

<A HREF="http://www.pesupport.com/cgi-bin/config.pl?read=299451">http://www.pesupport.com/cgi-bin/config.pl?read=299451</A>

Since Russell is now no longer is employed by McIntosh, I am not so sure its completely a marketing event. But h***, I don't know the man, and most surely Jim Griffin may.

Marlboro
(formerly called Pinky: got tired of a weasily alias that doesn't actually describe me very well)

[dickwest]

Thanks. Sorry about the redundancy. My computer glasses are in the shop getting new lenses so it has been difficult keeping up with all the excellent posts here.

> Yep, 'bout a couple days ago. Nice marketing
> speak.
> My first ever project was the line array in
> this pic and 2 years later it's still my
> reference standard I compare all of my
> daring endeavors to unseat the 'king'...
> Sheeesh! I keep knocking my head on 1 or 2
> dozen different designs and ask myself
> 'Why'...

[aarond]

Cool link.

Anyone have any idea of what drivers he is using? 16.5k for 2 cabs, 50 drivers and an EQ?

Does anyone else think that if the drivers could be sourced (or go w/ another full ranger) that 1/3 octave EQ on each channel would give better performance than a non adjustable unit( you could correct for room response)?

Might be fun trying to duplicate for much less $$. I'm sure they sound good but they cost way too much for me. Enter in the DIY factor and I would be these might almost seem reasonable.

Aaron

Aaron D

[cjd]

Not sure what that myth *is* - perhaps that it doesn't exist or something?

The physics of it, however, remains what it is. :)

C

[preiter]

His argument is that comb filtering also happens extensively with traditional speakers due to room reflections. The reason we don't hear it is that our brains filter it out. So you shouldn't be concerned with comb filtering in line arrays because our brains will filter that out, too.

[jimgriffin]

I’m not going to agree as Roger Russell suggests that comb filtering (or comb lines) are a myth but let me describe the effect in more detail. In my opinion Roger goes too far in claiming that comb lines aren’t an issue for line arrays.
I’ll talk first about the horizontal plane comb lining and later address the more important vertical plane.

Line arrays and point source speakers will have essentially the same comb lining or filtering characteristics in their horizontal planes so what Russell’s horizontal axis comments are essentially valid. We have very acute ability to discern sounds in the horizontal plane and it is mostly due to head shadowing of our ears. So sounds that arrive from different directions (different times) are easier to discriminate. Furthermore, horizontal axis reflections from room related surfaces (mainly sidewalls) are handled by the Haas or precedence effect in which the ear/brain system plays a role in the first 35 milliseconds or so. The direct or first arrival sound trumps the secondary or reflected sounds for either the line source or point source.

The situation in the vertical plane is different compared to the horizontal plane. With a line array you have significantly different path lengths between the drivers to the listener's ears from those at the end of the array vs. those in the center of the array. The center to center separation between drivers comes into play in the vertical plane. But in the vertical plane our ability to localize sound is significantly less as ears/brain cues (no head shadowing) are more limited. The Haas or precedence effect will also help to integrate sounds in the vertical plane. You can conclude that comb lines are less discernable in the vertical plane than they are in the horizontal plane.

Near field arrays radiate their energy some differently vs. point sources so their vertical radiation characteristics will differ vs. shorter line arrays. In the near field sound energy radiates outward from the source parallel to the floor and ceiling planes. You can stand or squat in front of near field array and the sound will follow you as you move up or down in front of the array. Hence, there is little spreading of the sound so few reflections occur from the ceiling and floor. Essentially, you hear sound from the wavefront as you listen and move up and down. The specific conditions for near field radiation are detailed in my white paper. Those conditions include a specific height of the array and maintaining a center to center separation between drivers less than one wavelength at the highest frequency of operation.

A few comments on Roger Russell’s full range driver line array are that the vertical spacing between the 3.5” diameter drivers indicates that comb lining will start happening above 3874 Hz (one wavelength spacing). The reality is that the amplitude response of his array will be uneven and start to decline above that frequency. Now you can use an equalizer as Russell does to help flatten out the amplitude so that the overall response will be close to acceptable. He does have decreased sensitivity of our ears working for him as you move into the 10-20 kHz region per the Fletcher-Munson curves. Plus as we age our higher frequency hearing starts to diminish so for us older guys this may help cover sound issues in this upper range. My comment on Russell’s array is that I suspect that you will hear ‘phasey’ sound as you move off axis and around the room for music with higher frequency content.

For what it is worth the Darren Kuzma’s 32 driver line array in the Parts Express Project Showcase demonstrates what one would have with drivers that are spaced 2.125” c-t-c. Note how Darren’s array shows significant decline of the amplitude response vs. frequency (see the unequalized and equalized plots in his report). An array with 3.5” spacing will have even lower frequency onset of amplitude response roll off.

Now let us take Russell’s no comb lining myth to the limit via using larger drivers or move the drivers even further apart in the array. I suspect that you will quickly hear even more frequency response shortcomings regardless of any equalization. It is not a design approach what I would advise one to pursue.

[jimgriffin]

> Cool link.

> Anyone have any idea of what drivers he is
> using? 16.5k for 2 cabs, 50 drivers and an
> EQ?

> Does anyone else think that if the drivers
> could be sourced (or go w/ another full
> ranger) that 1/3 octave EQ on each channel
> would give better performance than a non
> adjustable unit( you could correct for room
> response)?

> Might be fun trying to duplicate for much
> less $$. I'm sure they sound good but they
> cost way too much for me. Enter in the DIY
> factor and I would be these might almost
> seem reasonable.

I suspect that this array is very similar to his June 2006 Audioxpress article. He is using a Vifa near full range driver in the article.

Ask yourself whether you would like to be listening to a speaker which has a decent sized woofer vs. a heavily EQ'ed 3.5" driver to produce the low frequencies? Now ask yourself if you wish to hear your highs from that same EQ'ed 3.5" driver vs. a good quality planar or ribbon?

My point do assume that this is the ultimate speaker by any means.

[jimgriffin]

> My point do assume that this is the ultimate
> speaker by any means.

Oops. I meant to say: ... do not assume ....

[jerrald]

> I’m not going to agree as Roger Russell
> suggests that comb filtering (or comb lines)
> are a myth but let me describe the effect in
> more detail. In my opinion Roger goes too
> far in claiming that comb lines aren’t an
> issue for line arrays.
> I’ll talk first about the horizontal plane
> comb lining and later address the more
> important vertical plane.

> Line arrays and point source speakers will
> have essentially the same comb lining or
> filtering characteristics in their
> horizontal planes so what Russell’s
> horizontal axis comments are essentially
> valid. We have very acute ability to discern
> sounds in the horizontal plane and it is
> mostly due to head shadowing of our ears. So
> sounds that arrive from different directions
> (different times) are easier to
> discriminate. Furthermore, horizontal axis
> reflections from room related surfaces
> (mainly sidewalls) are handled by the Haas
> or precedence effect in which the ear/brain
> system plays a role in the first 35
> milliseconds or so. The direct or first
> arrival sound trumps the secondary or
> reflected sounds for either the line source
> or point source.

> The situation in the vertical plane is
> different compared to the horizontal plane.
> With a line array you have significantly
> different path lengths between the drivers
> to the listener's ears from those at the end
> of the array vs. those in the center of the
> array. The center to center separation
> between drivers comes into play in the
> vertical plane. But in the vertical plane
> our ability to localize sound is
> significantly less as ears/brain cues (no
> head shadowing) are more limited. The Haas
> or precedence effect will also help to
> integrate sounds in the vertical plane. You
> can conclude that comb lines are less
> discernable in the vertical plane than they
> are in the horizontal plane.

> Near field arrays radiate their energy some
> differently vs. point sources so their
> vertical radiation characteristics will
> differ vs. shorter line arrays. In the near
> field sound energy radiates outward from the
> source parallel to the floor and ceiling
> planes. You can stand or squat in front of
> near field array and the sound will follow
> you as you move up or down in front of the
> array. Hence, there is little spreading of
> the sound so few reflections occur from the
> ceiling and floor. Essentially, you hear
> sound from the wavefront as you listen and
> move up and down. The specific conditions
> for near field radiation are detailed in my
> white paper. Those conditions include a
> specific height of the array and maintaining
> a center to center separation between
> drivers less than one wavelength at the
> highest frequency of operation.

> A few comments on Roger Russell’s full range
> driver line array are that the vertical
> spacing between the 3.5” diameter drivers
> indicates that comb lining will start
> happening above 3874 Hz (one wavelength
> spacing). The reality is that the amplitude
> response of his array will be uneven and
> start to decline above that frequency. Now
> you can use an equalizer as Russell does to
> help flatten out the amplitude so that the
> overall response will be close to
> acceptable. He does have decreased
> sensitivity of our ears working for him as
> you move into the 10-20 kHz region per the
> Fletcher-Munson curves. Plus as we age our
> higher frequency hearing starts to diminish
> so for us older guys this may help cover
> sound issues in this upper range. My comment
> on Russell’s array is that I suspect that
> you will hear ‘phasey’ sound as you move off
> axis and around the room for music with
> higher frequency content.

> For what it is worth the Darren Kuzma’s 32
> driver line array in the Parts Express
> Project Showcase demonstrates what one would
> have with drivers that are spaced 2.125”
> c-t-c. Note how Darren’s array shows
> significant decline of the amplitude
> response vs. frequency (see the unequalized
> and equalized plots in his report). An array
> with 3.5” spacing will have even lower
> frequency onset of amplitude response roll
> off.

> Now let us take Russell’s no comb lining
> myth to the limit via using larger drivers
> or move the drivers even further apart in
> the array. I suspect that you will quickly
> hear even more frequency response
> shortcomings regardless of any equalization.
> It is not a design approach what I would
> advise one to pursue.
When designing line arrays using drivers of X size just keep in mind the formula for wavelenght: f=13560/diameter.
This will tell you when frequency response starts to decrease 3dB/octave. Using this formula will determine the pro's and con's to using fullrange drivers or tweeters of unusal diameters.
So for the use of 3.5" drivers:
f=13560/3.5
f=3874Hz (as Jim pointed out)
And if you're wondering about 2.125" drivers:
f=13560/2.125
f=6381Hz
Nothing beats a two-way line array.

[Andy_G]

What I would like to see as PROOF that comb filtering doesn't exist is several full sweep frequency measurements made at slightly different levels (say 1/2" differences for about a foot or height change) in front of the speaker, and at different distances from the speaker.

Now if they all came out the same then I would be reasonably convinced. There is no way I will be convinced that combing doesn't occur just by viewing one measurement !

[thylantyr]

Lets summarize.

He made some line arrays for McIntosh that were
sweet in the dinosaur era, left McIntosh in 1992
and in year 2000 tests 'a theory' using 69 drivers
in a line source and then published an article in
audioXpress magazine in November 2005.

Then, he spend a few more dollars and bought
"twenty five high-technology 3-1/2” drivers"
and got better results and published another
article in 2006.

Meanwhile, other companies have offered
much better line array designs albiet, not cheap,
but they exist, then Jim Griffin/Rick Craig
popularized the cyberspace DIY line array many
years ago offering high quality designs and
many folks have built project that are far superior to what has been posted by Roger -- and
his word is golden in year 2006 ? ROFL.

[bogie]

Build something. Listen to it. Decide if Bose makes something better. If so, scrap it, and try again.

[chrism]

> Oops. I meant to say: ... do not assume ....

Jim is correct. A planar ribbon is a better solution to multiple tweeters or a line of heavily eq'ed smaller drivers. I brought my quasi line arrays last year at DIY Dayton. These consist of 9 buyout 299-432 Vifa 4 1/2"ers with 4 Neo 6 tweeters mounted in a WWWTTTTWWWWWW arrangement. If you are not looking for output below 70 Hz without heavy EQ'ing, forget it. If you put them right up against the wall, they seem to do OK in the lower registers but do not produce any real bass. I do not hear any comb filtering from the 4 tweeters but what I do hear is a lot of lobing due to the arrangement. My crossover is at 4500 hz 3rd order. I have resorted to using the LFE output from my HT amp to run 2 RS270's for the low end. If you stay seated, the sound is awesome, if you stand, you can hear the weakness of the configuration.

[marlboro]

Yes, Chris. With my Neo's I have a line in each speaker of 20 of them. They have their flanges cut so that their c-to-c distance is about .8. This cuts the plastic almost into the metal of the magnet casing.

I believe that if you use them with out their flanges cut, and with a cross that high, you will probably experience what you do. But then the lower you cross them, the more of them you will have to have, in my opinion.

I also have them so that horizontally they are at no time more than 4 inches c-to-c from the 3.5 in mid woofers to avoid horizontal lobing.

Its definitely an issue that in my estimation with cone and dome drivers can only be handled by using really small ones. I believe that one uses domes instead of planars in an array they have to be either 1/2 inch domes(good luck finding one) or they have to be cut tight like I have to do with the neos.

I may cross them to a horn woofer from Bill Fitzmaurice, though currently they are connected to my 12 inch DVC woofers still in their ported box.

kind regards,

Marlboro

[marlboro]

[Pete Schumacher ®]

> Yes, Chris. With my Neo's I have a line in
> each speaker of 20 of them. They have their
> flanges cut so that their c-to-c distance is
> about .8. This cuts the plastic almost into
> the metal of the magnet casing.

> I believe that if you use them with out
> their flanges cut, and with a cross that
> high, you will probably experience what you
> do. But then the lower you cross them, the
> more of them you will have to have, in my
> opinion.

> I also have them so that horizontally they
> are at no time more than 4 inches c-to-c
> from the 3.5 in mid woofers to avoid
> horizontal lobing.

One small correction . . . you can't avoid horizontal lobing, you can only affect the size of the null and where it occurs. Close spacing AND a lower crossover point will improve the horizontal pattern. A steeper crossover slope will also help, but as long as you have two sources (line or point) that are placed side-to-side playing the same frequency, there will be certain angles off axis that produce a null. This is purely a function of the time delay produced by being closer to one driver than the other as you move off axis.

> Its definitely an issue that in my
> estimation with cone and dome drivers can
> only be handled by using really small ones.
> I believe that one uses domes instead of
> planars in an array they have to be either
> 1/2 inch domes(good luck finding one) or
> they have to be cut tight like I have to do
> with the neos.

> I may cross them to a horn woofer from Bill
> Fitzmaurice, though currently they are
> connected to my 12 inch DVC woofers still in
> their ported box.

> kind regards,

> Marlboro

[killerkowalczyk]

> Yep, 'bout a couple days ago. Nice marketing
> speak.
> My first ever project was the line array in
> this pic and 2 years later it's still my
> reference standard I compare all of my
> daring endeavors to unseat the 'king'...
> Sheeesh! I keep knocking my head on 1 or 2
> dozen different designs and ask myself
> 'Why'...

What would you rate the efficiency on those line arrays you build? 97/98 db@1w/1m ???

TIA

Killer

[marlboro]

OK, Pete, then....

Three followup Questions:

a. How steep a crossover slope would be necessary to move this horizontal lobing into inaudibility?

b. At what crossover point does the crossover slope become less helpful, or is there some other issue here than a simple relationship between the crossover slope and the frequency?

c. And, are there types of music where this type of musical lobing would be more noticeable or less noticable than other types?

Marlboro

> One small correction . . . you can't avoid
> horizontal lobing, you can only affect the
> size of the null and where it occurs. Close
> spacing AND a lower crossover point will
> improve the horizontal pattern. A steeper
> crossover slope will also help, but as long
> as you have two sources (line or point) that
> are placed side-to-side playing the same
> frequency, there will be certain angles off
> axis that produce a null. This is purely a
> function of the time delay produced by being
> closer to one driver than the other as you
> move off axis.

[Pete Schumacher ®]

> OK, Pete, then....

> Three followup Questions:

> a. How steep a crossover slope would be
> necessary to move this horizontal lobing
> into inaudibility?

Infinite slope would eliminate the null since no frequencies would overlap. Anything less will produce a null, but the steeper the slope, the narrower the range of frequencies that are nulled.

> b. At what crossover point does the
> crossover slope become less helpful, or is
> there some other issue here than a simple
> relationship between the crossover slope and
> the frequency?

As you go lower in frequency, the cancellation will cease to be a problem. Once you have a separation of less than a quarter wavelength, you've pretty much made both drivers look like a single source so there would be no null at all.

> c. And, are there types of music where this
> type of musical lobing would be more
> noticeable or less noticable than other
> types?

I don't really think the music type would matter since the whole spectrum is used by most music types. And again, you'll only notice this as you move left to right of the on-axis location.

> Marlboro