AVS Forum banner
1 - 20 of 21 Posts

·
Registered
Joined
·
1,655 Posts
Discussion Starter · #1 ·
It occurs to me that there is a gap. There is a great amount of discussion between people who understand crossovers well; about the advantages and disadvantages to phase coherence and lobing for Butterworth vs Linkwitz, and odd-order vs even-order.


But there doesn't seem to be much information out there targeted at a basic understanding of crossover points for the guy who just wants to know what speaker to buy and whether to override the Audessy-chosen settings.


I assume that most of us understand what a crossover does. Without getting into the details (involving low and high pass filters and the like), the basic function of a crossover, what makes it a crossover, is to divide sound. One signal comes in, two signals go out, one containing high frequency sound, the other containing low frequency sound.


Some things I don't think are as well understood.

Crossovers are Gradual

It appears a common misconception that a crossover acts like a hard barrier... that a crossover point of (for example) 80Hz means that 100% of the 79Hz sound comes out one side and 100% of the 81Hz sound comes out the other. This misconception seems to cause a lot of confusion.


This is less than ideal, but will hopefully give an idea:



The red line is the "sum" line... that is to say, it represents the entire signal coming out of both speakers.


The dotted line is the low frequency channel. The dashed line is the high-frequency channel. The chart itself is arbitrary. The crossover point (1) can represent any frequency.


As you can see, at the crossover point, sound is coming out of both speakers equally. As we move away from the crossover point, one speaker gets more an more quiet while the other gets louder until the sound is entirely to one speaker.


How rapidly this happens is called the "slope"... which is measured in "decibels per octave".


Many people have not really considered how an octave ties in to a frequency. An octave can be measured as a doubling of frequency. To the ear, an octave is linear (every 8 keys on a piano is an octave), but to the frequency it is logarithmic.


That means that 20Hz-40Hz is one octave. 5Khz-10khz is also one octave.


The higher the decibel-per-octave, the steeper the curve.


Now as I mentioned before: the crossover is the point at which both speakers are receiving equal power. If the sum is flat, that means that the crossover point is -6db or so.


For setting subwoofers in an AVR: the most common crossover slope seems to be a very gentle 12db-octave (also called "2nd order"). Let's do the math for an 80Hz crossover point.


At 80Hz, the sub will be at -3db and falling at a rate of -12db per octave. At 160Hz, the sub will be a -15db, and so essentially silent. Working backwards, we find that the sub actually started getting less than 100% signal at a half-octave earlier (about 50hz).


Looking at the other side: the speaker starts loosing about a half-octave higher (100 Hz or so), crosses -3db at 80Hz, and at 40Hz is -15db.


This is why it is important that a subwoofer perform above the crossover point, and a speaker perform below the crossover point... because they are asked to. Ideally, both systems would perform 1 octave from the crossover (to both double and half the crossover frequency), to ensure that any distortion was inaudible. But this is not always possible, meaning there are trade-offs to consider.

What to give up

If we set the crossover too high for the subwoofer, then we get high-frequency (likely >160HZ) breakup.


If we set the crossover too low, we get low-frequency drop off.


Generally low-frequency drop-off is preferred (and easier for calibration to correct for) than high-frequency breakup. So when we err on crossover, we generally want to err "too low".


Another problem with letting the crossover point rise is one of localization. Our ability to tell the direction from which a sound comes (left vs right for example) is a function of our distance from the sound and the frequency of the sound. The lower the frequency, the less we can tell direction.


This is why a single sub, sitting on the floor, can handle all the low-frequency: but we want 2-9 speakers in our stereo to handle high frequencies.


If we choose to set the crossover point high, we risk that the subwoofer starts putting out localizable sound. Noise starts sounding like it's coming "from the sub" (which it is).


It is for this reason that, when in doubt, we try to set lower (80Hz) rather than higher (120Hz)... but as with anything: play with the settings to see which works best for you.
 

·
Registered
Joined
·
1,655 Posts
Discussion Starter · #2 ·
I didn't get around to but want to mention that a 12db-per-octave / 2nd order crossover is not the only kind in use (though it appears to be what the bulk of AVRs use for LFE).


There are crossovers that use steeper slopes (though none are infinite). A 4th-order crossover uses a 24db-per-octave slope. A 1st order would be 6bd-per-octave


I personally have not heard of crossovers above 4th order.

The importance of a smooth roll-off

Since a speaker does put out sound past the crossover point, and since that speaker may not offer a flat frequency response throughout the entirely audible range: what else should we look at?


As we move farther and farther from the crossover: the ability of the speaker fading out to affect overall sound decreases. Where we start to see a non-flat response: the ideal change is a "smooth roll-off". This is a steady decrease in output as we continue down (or up) the frequency response graph.


The reason is simple. Sudden, sharp changes, especially increases, are more audible than gradual reductions. The volume at a given frequency might not be completely correct, but it is "smooth" and, unless we are very familiar with what we are listening to, we are not likely to notice small but even distortions.



Here is the Frequency response chart of a bookshelf speaker. Notice on the left side of the graph how the reduction in performance makes a nice even curve with no spikes or dips. That would be a smooth roll-off.


We'd like to see the subwoofer do the same in upper-frequencies.
 

·
Registered
Joined
·
1,396 Posts
First of all, 12db/octave would be a 2nd order butterworth slope. 1st order would be 6db/octave.


Although what you're saying has merit to it, most subwoofers line-outs on most receivers are low passed with an 24db/octave, not a 12db/octave slope. At worst some are low-passed 3rd order, or 18db/octave. The THX spec states a 24db/octave slope which should have barely audible output above 120hz even at reference level.


and while the implication that a vented bookshelf tuned around 80hz may indeed experience a 24db/octave rolloff below tuning, which combined with the 12db/octave high pass creates a 6th order high pass slope is true, a bit of EQ boost around 65-75hz for the subwoofer usually cleans this up for a flat response. Being an active crossover, EQ boost is more acceptable, and most subs are more headroom capable than the bookshelves they are matched with.


So the idea you were suggesting in another thread, that a bookshelf which falls off dramatically around 80 hz needs to be crossed at 140-160hz is flawed because it assumes a passive 2nd order crossover as with a 3-way speaker. In reality the ability of an AVR to smooth out this crossover makes it much more negligible. As long as the bookshelf can reach down to 80hz as a -3db point, the blend can be made well. Maybe not seamless, but negligible. For perfection, I'd recommend a speaker which is -3db down at 35hz or so, high passed at 60hz 2nd order, and a subwoofer which is -3db at 14hz and 200hz with no cone breakup to be low passed at 60hz as well. And I wouldn't rely on an AVR crossover to do that. For most people's purposes, the AVR crossover suffices.
 

·
Registered
Joined
·
1,655 Posts
Discussion Starter · #5 ·

Quote:
Originally Posted by Eternal Velocity /forum/post/19520301


First of all, 12db/octave would be a 2nd order butterworth slope. 1st order would be 6db/octave.

Thanks for the correction... this post sadly borders the edge of my knowledge and so such mistakes are likely.


If I can edit, I will correct the OP.

Quote:
Although what you're saying has merit to it, most subwoofers line-outs on most receivers are low passed with an 24db/octave, not a 12db/octave slope. At worst some are low-passed 3rd order, or 18db/octave. The THX spec states a 24db/octave slope which should have barely audible output above 120hz even at reference level.

Do you have a reference on the non-THX slopes? What I could find said 12db.

Quote:
So the idea you were suggesting in another thread, that a bookshelf which falls off dramatically around 80 hz needs to be crossed at 140-160hz is flawed because it assumes a passive 2nd order crossover as with a 3-way speaker. In reality the ability of an AVR to smooth out this crossover makes it much more negligible. As long as the bookshelf can reach down to 80hz as a -3db point, the blend can be made well. Maybe not seamless, but negligible. For perfection, I'd recommend a speaker which is -3db down at 35hz or so, high passed at 60hz 2nd order, and a subwoofer which is -3db at 14hz and 200hz with no cone breakup to be low passed at 60hz as well. And I wouldn't rely on an AVR crossover to do that. For most people's purposes, the AVR crossover suffices.

Obviously -24db slope would change the needed response up: but how do you figure that the response below crossover is irrelevant?


A speaker -3 at 80 but (say) -12 at 70 wouldn't be a problem?
 

·
Registered
Joined
·
1,396 Posts

Quote:
Originally Posted by JerryLove /forum/post/19520471


Obviously -24db slope would change the needed response up: but how do you figure that the response below crossover is irrelevant?


A speaker -3 at 80 but (say) -12 at 70 wouldn't be a problem?

The response below crossover isn't irrelevant, however level matching the upper output of the subwoofer with EQ (IE Audessey) will result in a smooth overall frequency response. The goal is the overall frequency response, not the individual frequency response of the bookshelf or the sub to be flat.


Perhaps not as seemless a blend as a simple two-way 2nd order slope, but not a vast issue either.


I presume you've got frequency response data of a sub and bookshelf crossed over via AVR which suggests otherwise? Or is this conjecture? (yes, an irritating word to be accused of at times, but I'm just curious).



Now regarding non-THX receivers with a steep crossover slope, I can't claim to know how all of them work. I do know emotiva's UMC allows for variation in the slope, where 12db and 24db may be selected.


I also read somewhere that marantz/denon/onkyo at the very least use THX slopes for all their receivers, regardless of THX certification. "read somewhere" doesn't mean much, I agree. But it makes sense regardless. I'd be interested in seeing LFE-out measurements though, if you've got them.
 

·
Registered
Joined
·
1,655 Posts
Discussion Starter · #7 ·

Quote:
Originally Posted by Eternal Velocity /forum/post/19521156


The response below crossover isn't irrelevant, however level matching the upper output of the subwoofer with EQ (IE Audessey) will result in a smooth overall frequency response. The goal is the overall frequency response, not the individual frequency response of the bookshelf or the sub to be flat.

Level matching would be insufficient given roll-off. If frequency response is not consistent across the frequency range in question, then level matching at one frequency will create non-matched levels at another frequency.


The overall effect of that match failure will be logarithmic... and so if you match at crossover, and if what we are discussing is roll-off (as opposed to some other abhorrent response) the effect on real sound will be less than the effect on the FR chart.

Quote:
Perhaps not as seemless a blend as a simple two-way 2nd order slope, but not a vast issue either.

Qualifying how important a given amount of non-flat response at a given frequency is would be beyond my knowledge at this moment (and more than a bit subjective).

Quote:
I presume you've got frequency response data of a sub and bookshelf crossed over via AVR which suggests otherwise? Or is this conjecture? (yes, an irritating word to be accused of at times, but I'm just curious).

I suspect that I have the same amount of data on this that you do.

Quote:
Now regarding non-THX receivers with a steep crossover slope, I can't claim to know how all of them work. I do know emotiva's UMC allows for variation in the slope, where 12db and 24db may be selected.


I also read somewhere that marantz/denon/onkyo at the very least use THX slopes for all their receivers, regardless of THX certification. "read somewhere" doesn't mean much, I agree. But it makes sense regardless. I'd be interested in seeing LFE-out measurements though, if you've got them.

I beginning to want to take them. I'll have to play around with the 4.2 system (the only one using LFE from an AVR) and see.


Obviously: If the slope is steeper than I've asserted the result will be that the important frequency response will decrease. If we care about anything above -18db (a bit more than 1%), then we move from "1 octave from crossover" to "1/2 octave from crossover".
 

·
Registered
Joined
·
1,396 Posts

Quote:
Originally Posted by JerryLove /forum/post/19523594


Level matching would be insufficient given roll-off. If frequency response is not consistent across the frequency range in question, then level matching at one frequency will create non-matched levels at another frequency.

*Level matching via EQ


IE a 3db parametric boost around 75db


Another thing to keep in mind is that getting bass flat in-room is difficult as is. You could have the most impressive theoretical setup, but from 300hz down, you'd still have plenty of peaks and valleys. Any presumed suckout near the crossover region may in fact be not perceivably different from the modes and nodes of in-room interaction.
 

·
Registered
Joined
·
1,655 Posts
Discussion Starter · #10 ·
Quote:
Originally Posted by Eternal Velocity
*Level matching via EQ


IE a 3db parametric boost around 75db
Sorry I missed that part: my bad.


While not a perfect solution: EQ has indeed helped such issues a great deal. It's also why roll-off being controlled is so important (I added info on that to post 2)

Quote:
Another thing to keep in mind is that getting bass flat in-room is difficult as is. You could have the most impressive theoretical setup, but from 300hz down, you'd still have plenty of peaks and valleys. Any presumed suckout near the crossover region may in fact be not perceivably different from the modes and nodes of in-room interaction.
Yes. Unfortunately the room will have a *lot* of involvement in these frequencies. But I cannot deal with every issue simultaneously. Since it's simpler to measure a speaker than a room: I start there.


I do think that there is room for debate as to the quality of a given measurement. Is -10db at 1/2-octave below crossover (assuming smooth roll-off) something to be concerned about or something to fix with EQ? What about -20? (asked rhetorically, not seeking an answer)


So certainly some values matter and some do not. We generally agree +/- 3db is irrelevant, and that -20db is inaudible (I assume we do). So I start there and then begin accepting compromise. Perhaps, with some discussion and data, we can find something less stringent we agree will work in all normal situations. If so, I'll amend my thresholds accordingly.
 

·
Banned
Joined
·
11,550 Posts
Guys - let's use an arbritary SPL level of 100 dB from both the mains and the subwoofer as we approach the crossover point of 80 Hz. At 80 Hz, each of the subwoofer and main will be producing 97 dB (- 3dB from the reference), which will have a net yield of 100 dB at the crossover point.


Now let's think about what happens at 60 and 100 Hz. By 60 Hz, the subwoofer is delivering at a full 100 dB, and by 100 Hz, the mains are at 100 dB.


Let's look at a speaker which is ruler flat to 30 Hz. At 60 Hz, it will be down appx. 9 dB from the electronic crossover. This speaker is producing 91 dB at 60 Hz.


Compare this to a speaker which rolls off at 24 dB per octave with a -3 dB point at 24 dB per octave. From 72 to 60 Hz, we are looking at appx. 1/3 octave. This means this speaker will be down appx. 11 dB more than will the ruler flat speaker, totalling being down 20 dB. This speaker is producing 80 dB at 60 Hz.


At 60 Hz, the subwoofer is delivering 100 dB on its own.


If we add 100 + 91 dB for speaker "A", we get a total of 100.5 dB at 60 Hz.


If we add 100 + 80 dB for speaker "B", we get a total of 100 dB at 60 Hz.


Is anyone going to say with a straight face that this is an important difference? A case could be made that we want 100 dB here (meaning 60 Hz), not 100.5, but that really is splitting hairs.


In the vast majority of systems, the real problem is having a subwoofer which is linear up to and through the crossover point.


The Klipsch Klipsch KL-525-THX is rated down to 80 Hz at - 3 dB.


Mark Seaton's Sparks are rated to 70 Hz.


JTR's Triple 8 is rated to 70 Hz at - 3 dB.


Both Klipsch and Seaton make great subwoofers which are linear up to and through the crossover of 80 Hz. They do this because both companies understand that what a subwoofer does from 40-80 Hz is as important to the listening experience as what it does from 20-40 Hz.


JTR offers methods that assure a smooth transition to subwoofers at well.
 

·
Registered
Joined
·
1,396 Posts

Quote:
Originally Posted by craigsub /forum/post/19524644


Guys - let's use an arbritary SPL level of 100 dB from both the mains and the subwoofer as we approach the crossover point of 80 Hz. At 80 Hz, each of the subwoofer and main will be producing 97 dB (- 3dB from the reference), which will have a net yield of 100 dB at the crossover point.

Wait, wasn't the topic of discussion that the natural rolloff of the main would cascade with the digital crossover at the crossover point... thus that the main would be -6db down at the crossover point?
 

·
Banned
Joined
·
11,550 Posts

Quote:
Originally Posted by Eternal Velocity /forum/post/19524949


Wait, wasn't the topic of discussion that the natural rolloff of the main would cascade with the digital crossover at the crossover point... thus that the main would be -6db down at the crossover point?

If the - 3dB down point of the speaker is at 80 Hz, yes. The SHO-10 - 3 dB point is 72 Hz. It is flat at 80 Hz.
 

·
Registered
Joined
·
1,655 Posts
Discussion Starter · #14 ·

Quote:
Originally Posted by craigsub /forum/post/19524644


Guys - let's use an arbritary SPL level of 100 dB from both the mains and the subwoofer as we approach the crossover point of 80 Hz. At 80 Hz, each of the subwoofer and main will be producing 97 dB (- 3dB from the reference), which will have a net yield of 100 dB at the crossover point.


Now let's think about what happens at 60 and 100 Hz. By 60 Hz, the subwoofer is delivering at a full 100 dB, and by 100 Hz, the mains are at 100 dB.

But at 60Hz the mains are still outputting audible sounds. It's the same with the sub at 100.


Let's look at a speaker which is ruler flat to 30 Hz. At 60 Hz, it will be down appx. 9 dB from the electronic crossover. This speaker is producing 91 dB at 60 Hz.


Compare this to a speaker which rolls off at 24 dB per octave with a -3 dB point at 24 dB per octave. From 72 to 60 Hz, we are looking at appx. 1/3 octave. This means this speaker will be down appx. 11 dB more than will the ruler flat speaker, totalling being down 20 dB. This speaker is producing 80 dB at 60 Hz.

Quote:
If we add 100 + 91 dB for speaker "A", we get a total of 100.5 dB at 60 Hz.


If we add 100 + 80 dB for speaker "B", we get a total of 100 dB at 60 Hz.


Is anyone going to say with a straight face that this is an important difference? A case could be made that we want 100 dB here (meaning 60 Hz), not 100.5, but that really is splitting hairs.


In the vast majority of systems, the real problem is having a subwoofer which is linear up to and through the crossover point.

So as I understand your position: Half-an-octave (about) past the crossover: the contribution of the waning speaker (assuming it's simply rolling off too quiet) is not perceptible?


What would you set as the minimum acceptable change?
 

·
Banned
Joined
·
11,550 Posts

Quote:
Originally Posted by JerryLove /forum/post/19525014


But at 60Hz the mains are still outputting audible sounds. It's the same with the sub at 100.


Let's look at a speaker which is ruler flat to 30 Hz. At 60 Hz, it will be down appx. 9 dB from the electronic crossover. This speaker is producing 91 dB at 60 Hz.


Compare this to a speaker which rolls off at 24 dB per octave with a -3 dB point at 24 dB per octave. From 72 to 60 Hz, we are looking at appx. 1/3 octave. This means this speaker will be down appx. 11 dB more than will the ruler flat speaker, totalling being down 20 dB. This speaker is producing 80 dB at 60 Hz.


So as I understand your position: Half-an-octave (about) past the crossover: the contribution of the waning speaker (assuming it's simply rolling off too quiet) is not perceptible?


What would you set as the minimum acceptable change?

It is not my position. It is merely the math involved.
 

·
Registered
Joined
·
1,655 Posts
Discussion Starter · #16 ·

Quote:
Originally Posted by craigsub /forum/post/19525036


It is not my position. It is merely the math involved.

So then I *did* accurately sum up your post?


What would you set as the minimum acceptable change?


If we set the threshold at "overall sound at -3db) then either speaker could go dead silent (-infinity db) at the crossover as long as the other was 0db from reference at the crossover.


To use your example. If at 80Hz our sub is putting out 97db (100 reference -3db) and our bookshelf is putting out 0db (-100 from reference) we should be just fine because the sum is 97db and so +/- 3db from reference.


Is that correct math?
 

·
Banned
Joined
·
11,550 Posts

Quote:
Originally Posted by JerryLove /forum/post/19525073


So then I *did* accurately sum up your post?


What would you set as the minimum acceptable change?


If we set the threshold at "overall sound at -3db) then either speaker could go dead silent (-infinity db) at the crossover as long as the other was 0db from reference at the crossover.


To use your example. If at 80Hz our sub is putting out 97db (100 reference -3db) and our bookshelf is putting out 0db (-100 from reference) we should be just fine because the sum is 97db and so +/- 3db from reference.


Is that correct math?

It is the correct math at 80 Hz, but the system is going to sound pretty weak above 80 Hz. If you don't believe this, try it sometime.
 

·
Registered
Joined
·
1,655 Posts
Discussion Starter · #18 ·

Quote:
Originally Posted by craigsub /forum/post/19525137


It is the correct math at 80 Hz, but the system is going to sound pretty weak above 80 Hz. If you don't believe this, try it sometime.

So then the math is right but doesn't tell the whole story. What is the rest of the story? What's the magic number and how did we arrive there?
 

·
Registered
Joined
·
26,476 Posts

Quote:
Originally Posted by Eternal Velocity /forum/post/19520301


So the idea you were suggesting in another thread, that a bookshelf which falls off dramatically around 80 hz needs to be crossed at 140-160hz is flawed because it assumes a passive 2nd order crossover as with a 3-way speaker. In reality the ability of an AVR to smooth out this crossover makes it much more negligible. As long as the bookshelf can reach down to 80hz as a -3db point, the blend can be made well. Maybe not seamless, but negligible. For perfection, I'd recommend a speaker which is -3db down at 35hz or so, high passed at 60hz 2nd order, and a subwoofer which is -3db at 14hz and 200hz with no cone breakup to be low passed at 60hz as well. And I wouldn't rely on an AVR crossover to do that. For most people's purposes, the AVR crossover suffices.

Why is this discussion in two threads??


Anyways, I will repost my plot here to explain why we should XO speakers closer to an octave higher then their F3 point when that F3 is too close to the Sub XO point.




Sharp knee is just bad in design theory. The 50Hz choice shows a much better slope when using a 80Hz XO common in most AVRs. Remember all three designs are flat to 80Hz on their own but we have to show the slope when the Filter from the AVR is applied.


Of course as you posted we can EQ anything and we can EQ out the problem but that is only a "FIX" and when discussion speaker XO theory we really should be sticking to the proper fundamentals involved otherwise we won't have a good reference starting point.


Imo we should get the XO between the sub and the mains done properly and lets not worry about EQing out problems we do not need to create.


Its one of the reasons I just added bass bins that run from 30Hz to 200Hz to my 2-way waveguides that did run down to 60Hz.


In the end the only thing that matters is understanding and maybe living with the compromises of any product we buy.
 
1 - 20 of 21 Posts
Top