your experience about crossover slope

tropico

Member
2010-04-18 3:09 am
Hello All

I had tweaked my crossover constantly for the last almost two month.
and found it is so sensitive to sound quality.

I cannot help but wonder how you guys tweak crossover for a better refined sound. Is any common rule based on your experience?

What I look is toward a general principle..to squash the best sound from driver/box.....at least know I was in the right direction toward a better sound.


For example in a common two way design (say 6.5 woofer + 1 dome tweeter), assuming the crossover had effectively tackle the breakup freq by some kind of notch, then what crossover slope would you like?

The steeper slope has narrower cover freq between woofer and tweeter, but with more phase change. In the contrast, the gradual slope has wider cover freq and with less phase change. (as in jpg). What is better?
 

Attachments

  • cross.jpg
    cross.jpg
    91 KB · Views: 331

AllenB

Moderator
Paid Member
2008-10-18 11:31 am
The breakup region is more than just a peak, it is an unevenness at different angles. Try to get the peak down, say -24dB and keep it smooth. This way the two drivers should be reasonably matched in directivity and the variations/sensitivity you are noticing should lessen.
 

eriksquires

Member
2013-05-10 4:11 pm
Hi Tropico,

As you probably know, the electronic filter slope isn't as important as the combined electro-acoustical slope of a filter + driver. And that's not as important as the combination of the two drivers working harmoniously.

Now as for answering your real question, I'm going to have to pardon myself for feeling qualified to answer you, but I'll express my views with the understanding that I'm not very experienced and there's no reason for you to listen to me.

First, it's nearly impossible to get perfect LR or Butterworth or Chebyshev or any other slope. You have to be receptive to how the drivers behave and how they interact with each other. You might want an LR 4 alignment, but find that it's nearly impossible to get the right phase and amplitude and therefore compromise. So, I suggest keep an open mind, don't force a speaker into a filter shape that it's not cooperating with. Gravesen mentions running into this at times.

My current monitors are asymmetrical, with about a 2 1/2 order high pass and 1 1/2 order low pass if you measure order by db/octave. The principle I use for the phase matching is that they should work like Fred Astaire and Ginger Rogers. Like the famous dancing couple, they need to come together and then split apart, but not violently. The longer they linger together the better.

I know this will probably drive some people nuts, and I'm sure I'll learn more, but right now I'm just not that interested in close miked measurements for the overall crossover FR. I use close mike testing only to evaluate a driver and examine it's distortion so I can get a good idea of it's useful range.

The FR data I import into XSim are from 3' distances and a mix of gating and ungated measurements with the loudspeaker in place. I know there will be all sorts of confounding effects from the room but I prefer to leave those gremlins in and ignore them in exchange for the accurate bass to mid-range response. I know what the comb filtering and floor bounce notches are, I can ignore them and focus on the overall bass shape knowing it's accurate to the listening location.

Next, again going against the grain, I prefer speakers of controlled directivity, not wide or omni. This gives me the most clarity at my listening location with modest room treatment, as opposed to heavier room treatment that would be needed with wider dispersion speakers. This is something you can control somewhat by choosing the appropriate crossover locations.

Lastly, I don't use a flat FR. I use one I like, which is similar to one I got from Madisound a while ago. A mild boost around 100 - 600 Hz, descending to 1 kHz, where there's a knee, then very gently descending 2-3 db by 20kHz. Find an FR that works for you and your environments. I think many have given up on a flat FR, but there are variations everywhere. The famous East Coast Sound vs. West Coast Sound are examples of these ongoing arguments. The best part about being in DIY of course is you don't have to accept anything but your own taste. Make what you want. Make it do exactly what you want, and be happy. :) But finding that out may take investigating on your part.

I hope you find some of this useful or perhaps stimulating of your own ideas.

Best,


Erik
 
Last edited:
First, it's nearly impossible to get perfect LR or Butterworth or Chebyshev or any other slope.

Oh, believe me, it isn`t hard at all. Real question would be what benefits would this bring you.

So, I suggest keep an open mind, don't force a speaker into a filter shape that it's not cooperating with.

Once installed, the drive units would scream loudly how you should crossover them :) If you ignore their suggestion, you would likely design a poor crossover. They`d show you the path.


I know this will probably drive some people nuts, and I'm sure I'll learn more, but right now I'm just not that interested in close miked measurements for the overall crossover FR.

It won`t drive anyone nuts. Speaker design is all about personal view and preferences. But in general this is a pretty serious mistake you`re doing. Nearfield can tell you a lot about your whole design, not just the driver itself. Without it, you may be unable to isolate other problems related to the baffle and the loudspeaker interaction with the room, for example. Moreover, unless you have access to an anechoic or can do semi-anechoic measurements in a very large enclosed space with high roofs, you won`t understand what`s happening below the typical gating frequency ( 250-350Hz ) and a lot of music stuff is in there.

I know there will be all sorts of confounding effects from the room but I prefer to leave those gremlins in and ignore them in exchange for the accurate bass to mid-range response. I know what the comb filtering and floor bounce notches are, I can ignore them and focus on the overall bass shape knowing it's accurate to the listening location.

Those effects also affect phase and cause dips and peaks in the power response you can`t know in terms of origin. And you get power response in this exact room, not on-axis response. While this approach could allow you to fine tune your speaker to your own room, it will not make it a good speaker until you design it to be such and then add room energy ( and identify what treatment it needs ).

Next, again going against the grain, I prefer speakers of controlled directivity, not wide or omni. This gives me the most clarity at my listening location with modest room treatment, as opposed to heavier room treatment that would be needed with wider dispersion speakers. This is something you can control somewhat by choosing the appropriate crossover locations.

Not sure why against the grain :) We see more and more such speakers today. If i had to do a new speaker for my pretty small living room, I would go the same way - using as large drivers as possible and then shifting to waveguides. I think someone did post a thread here on diyaudio on the use of very large drivers and using their beaming to create a c.dir. speaker for smaller rooms. It would be a different animal if you crossover above the directivity knee of a drive unit.

Lastly, I don't use a flat FR. I use one I like

And this is a good example of how colouration can be introduced. The speaker is tested with a "flat" signal, be it pink/sine/mls or whatever you want. This way it would be able to replicate the recording. if you make this speaker non-flat - you get colouration and that usually leads some people to say a specific system has more detail or they heard things they never heard before. Well, while this can be attributed to resolution as well, in most cases ( especially with full range drivers ) this is colouration, a product of ragged frequency response.
 
Last edited:

eriksquires

Member
2013-05-10 4:11 pm
I'm not saying no to nearfield. I'm saying that when I design a crossover it's not my primary source of "truth". I'll keep nearfield measurements in mind, such as distortion components, but what I import into XSim is the 3' distance measurements. That's what I design the crossover slopes, baffle step and driver amplitiudes to.

As for coloration, sorry, disagree again. A measured flat speaker sounds terrible to me. The issue isn't just about the speaker's FR, but about the FR in the room with reflections. The amount of energy your ear receives from a flat speaker isn't flat. Adjusting that is quite positive. But again, I'm designing for a speaker in a specific location, which means it would probably measure less than flat in an anechoic environment. That's fine with me. I'm not alone here in using a speaker with a target FR other than flat. Others smarter than me have elucidated on the need to use a curve other than flat. I often see a simpler curve going gently down from 20 to 20k as being touted as more ideal. It's also a question of how mixing rooms are. I don't think they are flat either. Someone refered to it as the B&K curve, but I have no idea if that's correct.

You are more than welcome to use a flat FR though, that's the beauty of DIY. I suggest that if you can experiment with a DSP before commiting it's really the best way to judge the final result. Try flat, try the B&K, try mine. See what you like, because unless you are a scientist designing for the pleasure of other instruments, that's all that should matter to your wallet is what you like.

As for usability below 300 Hz, I have no such problems.

Best,


Erik
 
Last edited:

tropico

Member
2010-04-18 3:09 am
Hi, Eriksquires , Many thanks the reply

Sorry I still get confusion.

I often see a simpler curve going gently down from 20 to 20k as being touted as more ideal. It's also a question of how mixing rooms are. I don't think they are flat eithe

Does it mean u measured at 3 meter distance and find a gradully down from 20 to 20khz is best for sound?

This is where I true get confusion. When I put microphone at roughly 3 meter (listening position), I find flat FR sound too bright and harsh. when I put microphone closed to speaker (mine is MTM). the height of microphone would have large variation to FR and seem more unpredictable to me. So I back to 3 meter measurement when fine tuning my crossover. Did I miss anything?

gating and ungated measurements with the loudspeaker
What is that? How to to do that (mine is Omnimic Measurement System)?

I use for the phase matching is that they should work like Fred Astaire and Ginger Rogers. Like the famous dancing couple, they need to come together and then split apart, but not violently. The longer they linger together the better.

Does this mean you let phase change more gently at the crossover cover freq? This seem imply less slope and wider cover range.

Best,

tropico
 

krivium

Member
Paid Member
2009-10-13 2:43 am
but what I import into XSim is the 3' distance measurements.

3' refer to feet, so approx 0.9/1m.

B&K curve:

http://i47.tinypic.com/rrt8yd.jpg

the height of microphone would have large variation to FR and seem more unpredictable to me.

This typical behavior of mtm: one of the goal of this design is too maintain a restricted vertical coverage. But this is frequency dependent: the lower frequency you go the less you should notice difference as the 2 woofer tends to radiate omni (and that once their center to center distance is 1/4 fc they 'behave as one driver', both driver are considered 'coupled'). here is a 3 d rendering of polar response comparison between an mt and a mtm:

http://hc.nextgeneration.free.fr/electroacoustique/images/comparaison MTM et MT 1.jpg
 
Last edited:

AllenB

Moderator
Paid Member
2008-10-18 11:31 am
Does this mean you let phase change more gently at the crossover cover freq?
No, this isn't a bad thing but it's not what makes a good crossover. Combining the response from the box, to the driver layout, and splitting it with filters paying attention to keeping the drivers working within the range where they behave predictably and consistently is most important.
 

eriksquires

Member
2013-05-10 4:11 pm
No, this isn't a bad thing but it's not what makes a good crossover. Combining the response from the box, to the driver layout, and splitting it with filters paying attention to keeping the drivers working within the range where they behave predictably and consistently is most important.

Great quote, Allen! Kind of what I'm trying to say. The point about phase matching like I do is of course, a personal opinion, which you don't have to agree with. :)
 

eriksquires

Member
2013-05-10 4:11 pm
Does it mean u measured at 3 meter distance and find a gradully down from 20 to 20khz is best for sound?

3' = 3 feet, or approximately 1m. I'm not going to say 3' is better than 3.2' or 2.8' so exactly how you translate this into metric is not as important as the principle.

This is where I true get confusion. When I put microphone at roughly 3 meter (listening position), I find flat FR sound too bright and harsh. when I put microphone closed to speaker (mine is MTM). the height of microphone would have large variation to FR and seem more unpredictable to me. So I back to 3 meter measurement when fine tuning my crossover. Did I miss anything?

Not sure. Make sure you measure your speaker on the same axis as you listen. For instance, I try to listen with the midpoint of my tweeters (long AMT's) pointing at my ears.

Of course, I don't listen at 3', I listen around 9 feet (~ 3 meters), but I measure with the tweeter on the same axis at 3'

What is that? How to to do that (mine is Omnimic Measurement System)?

That's what I use. It's the default FR measurement mode. See the radio button that says "Blended?" that's what I mean.

Does this mean you let phase change more gently at the crossover cover freq? This seem imply less slope and wider cover range.

The target slope/filter alignment is not what I meant to focus on. I meant, regardless of whether you use 1st or 4th order, I like to align the phase of the drivers in more than just one place. Let me include a graph.
attachment.php

Notice that the phase curves (dashed) approaches gently from 2k and doesn't really separate until 3.5kHz.

To answer another point made elsewhere, drivers don't need or care about ideal Butterworth curves. We care about this in purely electronic filters, but in passive crossovers as some one else said, we care about amplitude and phase of the finished product. In electronics using LR4 or Butterworth, etc. is great because it helps us match up a low pass and high pass section that we know will be perfect, or at least very close to our desired result.

In speakers it's much harder to get all of this to happen. Within the realm of possible priorities during crossover design:


  • Ideal frequency response (whether flat or not is up to you)
  • Ideal phase matching at crossover points
  • Time alignment (physically moving the drivers so that the arrival time is perfect)
  • Minimal lobing / interference from driver to driver
  • Time alignment
  • Ideal matching of driver curves to theoretical filter
  • Power handling
  • Constant directivity
Of all of these I would put the "ideal matching to a theoretical filter" last.

If you look at the chart above, you can see that the amplitudes don't match a standard curve, but look at the finished curves (note change in F scale):

attachment.php


Now, maybe you don't like my curve, or my speakers. That's OK! The point is I gave up "ideal filter curves" for everything else I wanted. :)

To be honest, in my process I often use an ideal LR2 (Linkwitz-Riley 2nd order) curve as a target when I'm starting to match up the drivers. However I throw it under the bus pretty quickly depending on how the phase matching is working, or not.

To Mario's point, yes you can wrangle (wrestle) any driver into a close-to-ideal 2nd order Butterworth for instance, however wrestling it and the matching driver across all the important paramters AllenB and I have mentioned is another story.

Gaining Absolute Truth
Tropico, one of the best ways to gain experience about this is to use XSim.

First, use a couple of "ideal" drivers. If you use an empty driver it's an ideal 8 ohm driver at 70 db or something like that. Try putting together different filter alingments. Like, 1st and second order. Use the Circuit blocks. That is really easy. Now take your woofer and add 1.3" of distance. Try again.

After this, go to Parts-Express and find some Dayton drivers you might like to build a 2 way from. Don't worry, this experiment will be free. PE provides the FRD and impedance data for all the Dayton drivers. Import them, use the same distance, and re-do.

By the time you are done going through this I think you'll have a better appreciation for what is hard and what is easy. :)

I admit I have not read the good doctor D'Apolitto's work on speaker measurements. It's on it's way. You should also grab a copy of it for a thorough understanding. Perhaps when I am done with his book I'll find out that doing nearfield and relying on that data is in fact faster. So, my point is, this is my best advice to you. After I become more well read this may change too. :)



Best,


Erik
 

Attachments

  • Phase_Matching_Closeup.jpg
    Phase_Matching_Closeup.jpg
    80.2 KB · Views: 500
  • Predicted_FR.jpg
    Predicted_FR.jpg
    125.5 KB · Views: 495
Last edited:

eriksquires

Member
2013-05-10 4:11 pm
Tropico,

I also want to say some designers have very strong opinions about crossover slopes. For instance, some try very hard to reduce the entire crossover to 1 capacitor. Others insist that only first order crossovers sound good. Whatever the type of filter slope/Q some one out there insists it's "Best". We've deviated from your question somewhat. One DIY'er does everything to avoid resistors. There are also other type of crossover designs where the phase matching is not as discussed here. Others will have to comment about the pro's and cons. I've never heard them nor read enough to make an informed opinion.

The biggest benefits of high order filter alignments is that there's less off-axis combing. For instance, for center channel or D'Apollito aligned MTM speakers 4th order is really the best idea. But as an article from Gravesen discusses, it's just not always possible. I think it was one of his Center Channel builds.

It's also a different story if you can time-align your drivers. That is, move your tweeter back so your woofer and tweeter have 0 acoustical distance. Try this with XSim, you'll see how this changes the ease of crossover design.

Best,


Erik
 
Last edited:

AllenB

Moderator
Paid Member
2008-10-18 11:31 am
The FR data I import into XSim are from 3' distances and a mix of gating and ungated measurements with the loudspeaker in place. I know there will be all sorts of confounding effects from the room but I prefer to leave those gremlins in and ignore them in exchange for the accurate bass to mid-range response. I know what the comb filtering and floor bounce notches are, I can ignore them and focus on the overall bass shape knowing it's accurate to the listening location.
At middle and higher frequencies the room is easy to gate and doing so is ordinarily desirable. The opposite could be said about the baffle and its contribution will be measured inaccurately up close, so I agree.

Bass will be affected by the room to the point that none of this makes much sense anyway.
Ideal phase matching at crossover points
Broadly speaking this only happens in one ~plane. A common assumption would be that this should cross the listening axis but I dont see this as strictly necessary or even correct.
 

eriksquires

Member
2013-05-10 4:11 pm
All truth! However I really liked the end results. It's impossible to tell where one driver begins and one ends.

At middle and higher frequencies the room is easy to gate and doing so is ordinarily desirable. The opposite could be said about the baffle and its contribution will be measured inaccurately up close, so I agree.

Bass will be affected by the room to the point that none of this makes much sense anyway.

Broadly speaking this only happens in one ~plane. A common assumption would be that this should cross the listening axis but I dont see this as strictly necessary or even correct.
 
I use the Harsch XO slopes for all my speakers to get quasi transient perfect performance and almost flat phase. It uses a steep 4th order BW low pass on woofer which is nice as that prevent cone breakup from leaking through. It uses a second order Bessel high pass on tweeter which may be challenging for some tweeters. So if requires a tweeter with a lower fs and smooth flat wide response. A good 3in or 2.5in full range driver can work well here.

http://www.diyaudio.com/forums/multi-way/277691-s-harsch-xo.html

The near transient perfect response sounds very realistic with percussion.

Recently, Bushmeister implemented the Harsch with his mini point source horn. More here:

http://www.diyaudio.com/forums/mult...lti-way-point-source-horn-98.html#post4615797

Xo is at 500Hz here is phase:
530875d1455378296-bookshelf-multi-way-point-source-horn-phase.jpg