A good tweeter to match 5" SB Acoustics Magnesium Woofers

[dlr]

Even a small waveguide like you see in Seas DXT, Morel CAT378, or Wavecor TW030WA11/12 can make a difference to that directivity pattern around the crossover frequency, with added benefit of avoiding the baffle diffraction and improving low end distortion.

That's what I hoped for with the DXT, but it didn't prove to be nearly as impervious to diffraction as one would think.

dlr

[dlr]

I don't disagree with most of what you've said, but there is a lot related to taste that enters into it and there is no one answer to all driver combinations. I'm probably much closer to your position on this, but I don't do first/third, though I have experimented some and found that BW1 can be a good option with the right drivers. My preference has mostly been for LR2 and depending on the tweeter used, introduce a slight droop from around 1-3KHz. Long ago we had a discussion back at Madisound I think related to this. Jeff was involved in that and I think that his analysis was probably correct and it's related to your thoughts here. It wasn't about any XO peaking at the tweeter low end or not, but with the use of LR2 primarily. That relates to power response which of course is primarily about polar He pointed out that LR2 provides a broader power response dip. I usually add a small dip in 1-3Khz area that I think recesses the vocalists just a bit for "soundstage" if you want to call it that. It's just more pleasant and natural sounding to me, I hate a forward presentation.

You've probably done more real research into the off-axis and room influence than most here, so I really can't go into that detail, but in general I think that my preferences probably align with yours just from my own experiments.

WRT to the low vs high Fc, I do take issue with one part of that. The driver combination really controls which way to go. My current main system is a 3-way dipole, with the tweeter (monopole) front hemisphere polar response similar to that of a full dipole. My goal was to get a front hemisphere horizontal response operating primarily in dipole mode with the tweeter integration matching and with a smooth off-axis integration, no tweeter bloom. That is a problem and I agree with you on that. I went about it in an entirely different manner and could not have done this easily in the past.

First I fully treated the baffle for tweeter diffraction. That's always first after creating a box/baffle (baffle only here, no box) as I'm sure you know. That was key, but the other factor was the Ultimate Equalizer (UE) from Bohdan (SoundEasy creator). I had started using SoundEasy earlier to audition designs before building a crossover. It was my primary use as I still measure with LAUD. This provided a unique ability to audition literally dozens of designs in a single evening. What I realized made this unique is that if you use accurate driver SPL measurements (no impedance required) and set any design target on-axis, then no matter what XO you target, the on-axis will always be that target. The DSP in it guarantees it. From this point on I knew that I could completely dispense with on-axis measurements! When I realized this and confirmed it for multiple cases, I started measuring horizontal polars.

My system includes a 5" midrange and 1" woofer. Exceptional midrange, the single most important driver IMO, and a good, but not exceptional tweeter, one that I expected to have better front directionality, the DXT. But I found that it's not nearly as directional as I expected, so any tweeter would probably be fairly close in polar response.

Here's my experience with M/T crossovers on this pair which I would expect to work on similar small midwoofer 2-ways. I fully expected to go with about 2.5KHz LR2. I thought that higher Fc with broad overlap would provide the best polar response integration. I quickly found that to be entirely the opposite. There were significant off-axis integration problems. I did not have to consider any tweeter transfer functions because the UE created whatever filter was required to achieve the on-axis target. I spent several hours measuring dozens of designs off-axis. I went out to 60 degrees initially, but decided that only 45 degrees was necessary because the changes were consistent with other angles. In the end I designed the system XO strictly at 45 degrees.

High Fc with broad overlap had the most off-axis irregularities. The final XO that I now use was LR8 @1200Hz. Everything higher Fc and/or lower order was progressively worse, which completely surprised me, but it was the measurements that proved it to me. I'm sure that the tweeter transfer function to achieve that must require peaking and the equivalent of a Linkwitz Transform to extend it flat, but that's what I use and I've never heard any stress in the system at my typical listening levels or even short term testing. The tweeter might fail due to heat stress if I ran it too long, but I don't stress my ears for any length of time anyway. But I couldn't find any other crossover that didn't have off-axis flare and it was higher with high Fc. I never did consider trying a higher Fc, then altering it for tweeter flare. I wanted the best on-axis direct response and good polars.

Measurements are without the woofer, due to the 250Hz Fc and my basement with a low ceiling, so limited time response. Besides, at 250Hz, integration isn't as difficult for W/M. I was interested primarily in the M/T. Ignore anything below 500Hz. The woofer would have filled that in.

These are direct first, then normalized to the on-axis in the other. Keep in mind that this is LR8 @1200Hz.




These next two are also limited to 500-5000Hz to cover the mid/tweeter integration.



dlr

p.s. I could not have designed it in this manor if there was any significant tweeter diffraction influencing axes differently. It was the controlled diffraction that gave good on-axis response that was maintained in the off-axis as well and provided consistent results no matter the crossover used.

[dcibel]

I do the same thing, although I sometimes refer to it as a "BBC dip". Flat on-axis on so many of my design attempts sounded poor, when I finally disciplined myself to take off-axis measurements, I figured out why.

IMO that is essentially what the "BBC dip" was correcting, as well back in the day when it was being used you would often see higher distortion in this frequency range due to the midwoofer breakup, contributing as well to the "energy" present in these frequencies, so the effect was two-fold.

The tweeter dispersion "mushroom" became very clear to me on my first serious design - an 8" 2-way. Flat response on axis sounded a bit too shouty to me, added a couple dB drop just above the crossover frequency to compensate for the wide tweeter directivity here made a huge difference.

I am still surprised that you don't see more tweeters with waveguide faceplates available, even in very high end tweeters they still are built on a flat plate. Even a small waveguide like you see in Seas DXT, Morel CAT378, or Wavecor TW030WA11/12 can make a difference to that directivity pattern around the crossover frequency, with added benefit of avoiding the baffle diffraction and improving low end distortion. The testing available online at AudioXpress for the Wavecor mentioned are a perfect example of the diffraction benefit. I am also of the opinion that a waveguide tweeter making for a more "constant directivity" speaker has a much greater chance of creating the same sound tone in different rooms, since there is greater forward facing energy, less off-axis energy, so the room being "live" or "dead" has a lesser effect on the overall sound at the listener.

[hongrn]

xmax, you should really post the build on this forum, especially the crossover work. I don't know about the rest of you, but I need to learn from the best.

[johnnyrichards]

Touching on this again, above is the reason I sometimes favour the "first to third" order crossover function (basically a very overdamped third) that I've posted about numerous times. Also why I favour a slight on axis system dip near tweeter roll off knee (recall being berated here on this forum at length for that one). Exact opposite of the speaker under discussion here.

FWIW, I'm not the only one with the opinion that bump at the low end of the tweeter usually (again I haven't heard this speaker) sounds overly aggressive in that range once room contribution is factored in . George Short from North Creek etc (as I've posted quite a few times) came to the same conclusion as I about overdamped 3rds and used those and "gfiandy" over at diyaudio is a commercial loudspeaker system designer (speakers for Arcam and others) and also advocates for the on axis dip, only achievable with a soft high pass knee:
http://www.diyaudio.com/forums/multi...ml#post5306291

I do the same thing, although I sometimes refer to it as a "BBC dip". Flat on-axis on so many of my design attempts sounded poor, when I finally disciplined myself to take off-axis measurements, I figured out why.

[xmax]

Maybe It's not crappy and I'm a crazy person, this will be the only way to tell for sure.

Also perhaps I should be volume matching since that is what will be happening essentially.
Then with the mic a few inches from the tweeter and nothing else within 3ft you can say
it's a reflection problem...

[dcibel]

Why would you build such a crappy speaker?

[xmax]

And to be clear I am aware he would have tweaked the filter for the addition of the resistor,
but perhaps not enough. I have all the parts to build this speaker and intend on doing it
although I was hoping I did not need to go to these extremes for you guys to get it.

[xmax]

Some will never learn. When I get done with these Monitors I am doing for the Grammy award winning recording
engineer for "Rock Record of the Year" (Grammys) last year. I will take some more time with this silly test. I have mounds
of test equipment I have not dug out yet.

[DDF]

I can't comment on the sound of the speaker so my comments here are removed from that and based only on my experiences with my designs. My experience on how to deal with blending drivers of differing dispersions has been the exact opposite. I found pushing the tweeter lower and peaking it with a high Q crossover increases the 'tweeter bloom" and makes the design sound harsh in a small lively room. I listen in this type of environment allot and have many experiments and simulations/measurements followed by years of listening to come to that conclusion. IME, its better to use a low Q on the tweeter and on the woofer and have them overlap over a greater range. This assumes small speakers, drivers closely spaced. Off axis comes out smoother, design sounds less harsh in play. in a large room with late reflections, there's much less value to that, IME. FWIW

[ATTACH=CONFIG]n1359898[/ATTACH]

Touching on this again, above is the reason I sometimes favour the "first to third" order crossover function (basically a very overdamped third) that I've posted about numerous times. Also why I favour a slight on axis system dip near tweeter roll off knee (recall being berated here on this forum at length for that one). Exact opposite of the speaker under discussion here.

FWIW, I'm not the only one with the opinion that bump at the low end of the tweeter usually (again I haven't heard this speaker) sounds overly aggressive in that range once room contribution is factored in . George Short from North Creek etc (as I've posted quite a few times) came to the same conclusion as I about overdamped 3rds and used those and "gfiandy" over at diyaudio is a commercial loudspeaker system designer (speakers for Arcam and others) and also advocates for the on axis dip, only achievable with a soft high pass knee:

http://www.diyaudio.com/forums/multi-way/316528-diy-vs-brand-speaker-26.html#post5306291

[DDF]

In fact, that low crossover Fc Jeff used is to be commended since as a bookshelf the listener may be quite a bit off-axis, so a low crossover Fc of a small, low power bookshelf would be highly recommended. For dispersion characteristics. By someone who takes all factors into account, that is.

First, thank you to the original poster for not fretting about the thread hijack!

I can't comment on the sound of the speaker so my comments here are removed from that and based only on my experiences with my designs. My experience on how to deal with blending drivers of differing dispersions has been the exact opposite. I found pushing the tweeter lower and peaking it with a high Q crossover increases the 'tweeter bloom" and makes the design sound harsh in a small lively room. I listen in this type of environment allot and have many experiments and simulations/measurements followed by years of listening to come to that conclusion. IME, its better to use a low Q on the tweeter and on the woofer and have them overlap over a greater range. This assumes small speakers, drivers closely spaced. Off axis comes out smoother, design sounds less harsh in play. in a large room with late reflections, there's much less value to that, IME. FWIW

The following might be a bit more of general interest. During the design phase of a puny (3.3L) system ("The Napoleons", I posted design at DIYAudio at one time; what's posted there isn't the final design, need to find time to get that documented) using W4-1720 & D26NC55, I simulated effect of changing resistors at the input and output of the high pass. Reason was I had the xover built externally and wanted to try some quick voicing changes using alligator clippers and shunt jumpers. The impact on response shape is shown attached, again FWIW (I found it interesting). I can't recall the baseline (these are relative curves, not absolute) but it does show the differences.

[dlr]

Food for thought. I modified WinPCD to export the individual section transfer functions properly, I had the wrong complex data values prior to this, but most users probably aren't aware that the driver equalized response and circuit transfer functions can be exported to file. I also modified WinGraph to allow data below zero, since the transfer functions export as normalized values. Now the filter transfer functions can be viewed in a single graph. I'll be putting update files at my site, but these are minor changes, so it's not necessary to get the update unless you want to export the transfer function.

Two graphs. The first one is the original comparison made without/with the 6.2 ohm resistor after the tweeter network for a second order electrical. The second one is doing the same for another filter that's third order electrical as I used in my old 3-way for years (design at my site). The results of just sticking a resistor in blindly without re-optimizing the crossover to take it into account (as one would normally do) shows quite a different result.


Second order electrical network

\
Third order electrical network


dlr

[dcibel]

As you can see the third harmonic is higher at 3K (the most sensitive area of our hearing)
with the resistor only the B&K mic has low enough distortion to see that. If you look
at the Via wave you can see the how low I like the 3rd harmonic. The test conditions should not have
anyone concerned about reflections. The other 2 mics are the UMM-6 and the
USB OMNI.



Maybe you guys measure speakers for different reasons than I do.
Although looking at it again the third harmonic is so low I don't think it's an issue.
I will return to this test at another date in the mean time think for yourselves and please don't drink the kool-aid.

Yeah I don't like kool-aid either - only the finest snake squeezin's for this guy!

Looks like you have a problem around 1kHz without the resistor, that pesky Fs poking through again haha ;)

3rd harmonic higher at 3kHz? We must be looking at different charts. One thing is clear, and the point that you've poorly made noting "resistor distortion" is that if you normalize the amplitude through the rest of the pass band, the tweeter with the resistor will play louder at 3kHz due to the change in transfer function it imposes. Distortion vs amplitude remains unchanged, unfortunately your measurements have not convinced me otherwise. We however already knew that and had it sorted pages back in this thread, the issue was blaming the resistor for distortion, which is doesn't inject to the circuit, it only changes the transfer function which we of course expect and intend it to do. You tried it and didn't like it - that's fine but let's not write off the resistor, follow some sort of scientific method to prove correlation and not make assumptions based on coincidence. Maybe you have had problems with Fs in the past due to non-ferrofluid tweeter with high Q resonance where the filter crossover couldn't compensate for - this will of course be represented in the transfer function and a common issue when using a calculator to determine a crossover instead of real-world data. Let's get one thing straight though - harmonic distortion caused by Fs is not a thing.

DDF makes a good point about the dispersion, if we correct for baffle diffraction at a single axis, we are likely over-compensating as the diffraction effects change a fair bit off-axis. It is important to consider a wide angle of radiation for diffraction and directivity when making design decisions, not just a single axis unless the intended installation is an anechoic chamber.

So we can look at the crossover design another way - what if we increase the crossover frequency to 4kHz+ to avoid all this pesky distortion stressing the tweeter? Well there are cons to that as well. The woofer then has to play higher in frequency where its directivity narrows (aka beaming) and we run into frequency non-linearities and is most cases distortion as the midwoofer encounters cone breakup in the high frequencies as well. It is always a trade-off, one has to make decisions based on all the data present to make the best of what we have.

[dlr]

He has no idea how a to gate an impulse response measurement. I'd disregard all his driver measurements, distortion plots, and anything related to speaker building. Go learn the basics first before spewing nonsense here.

The screen capture he posted of the measurement system has settings, Shape: "Rectangular" "Auto Window". It's essentially a time slice with no window setting to prevent FFT irregularities due to no tapering of the window and the start/stop times are left to the system to select if I read that correctly. I would suggest that two successive measurements with no changes whatsoever may result in two significantly different distortion plots. SPL results even with a proper tapered window will have variation, especially near the low end due to insufficient impulse length.

I have to say again that what I see does not give me confidence in the reliability of the distortion test results, separate from the other arguments made.

dlr

[hongrn]

Looking at all the vitriolic comments lobbed at this forum by this guy, we can safely come to a few conclusions: 1) His crossovers are done using online calculators. 2) He has no idea how a to gate an impulse response measurement. I'd disregard all his driver measurements, distortion plots, and anything related to speaker building. Go learn the basics first before spewing nonsense here.