Jump to content

Settings for tri-amping an '82 Belle with an active digital crossover

Chris A

Recommended Posts

Good clean sound is the main reason to tri-amp a three-way (or bi-amp a two-way system), but another benefit is the ability to easily switch off individual drivers when analyzing the sound system. These pictures show the pink noise response of my K-horns at the main listening position, using 48 db/octave Linkwitz-Riley filter slopes. Looks like I need to adjust the PEQ to make the speakers sound even better!

Left to right pictures show, #1 sub only, #2 sub + woofer, #3 sub + squawker, #4 sub + woofer + tweeter, #5 sub + tweeter.






Link to comment
Share on other sites

You might also try listening to 24 dB/octave L-R filters centered at the same frequencies in order to hear if they might sound a bit smoother in the crossover regions. The reasons have to do with relative phase of the two types of filters electronically, but you may enjoy the higher slope L-R filters more in order to minimize the effective crossover bands. YMMV.


Edited by Chris A
Link to comment
Share on other sites

Thanks for the suggestion, I will try the different crossover slopes, and adjust the PEQ.

My preferences are influenced by listening to live music a couple nights each week, usually hearing 102db SPL peaks through fixed installation, in-house calibrated, curved line-array speakers like at the right side of this dark picture. There is a consistency from venue to venue that leads me to believe this is good sound, since hundreds and/or thousands of paying audience members seem happy with the sound and I usually think it sounds "real". I am not a sound pro, just another music fan, and I appreciate the wisdom of the Klipsch community helping to improve my home audio system.


Link to comment
Share on other sites

My preferences are influenced by listening to live music a couple nights each week, usually hearing 102db SPL peaks through fixed installation, in-house calibrated, curved line-array speakers like at the right side of this dark picture. There is a consistency from venue to venue that leads me to believe this is good sound, since hundreds and/or thousands of paying audience members seem happy with the sound and I usually think it sounds "real". I am not a sound pro, just another music fan, and I appreciate the wisdom of the Klipsch community helping to improve my home audio system.

My response is: that sound IS real for that venue and genre. It's only artificial if someone records it and plays it back on something other than the original setup in the original venue. And "artificial" or "reproduced" can be either good or otherwise.

Here is another interesting insight (at least, IMHO) into this subject: https://community.klipsch.com/index.php?/topic/148991-david-byrne-how-architecture-helped-music-evolve/

Edited by Chris A
Link to comment
Share on other sites

Here are new pictures of my outputs running 24 db/octave Linkwitz-Riley filters, compared to the 48 db/oct. pictured in my above post. The midrange to treble blends well this way, but the K-horn woofer to midrange at 48 db/oct. seems cleaner, more precise at 95+ db SPL, so I will continue to try both ways on the bass crossover.

Left to right pictures show, #1 full spectrum, #2 sub + woofer, #3 sub + squawker, #4 sub + woofer + tweeter, #5 sub + tweeter






Link to comment
Share on other sites

Chris, I hope this isn't too off-topic; however, since I noticed that Khornukopia has a Behringer DCX-2496 I wanted to cross-reference to a thread when Rudy81 had Klipschorns and used cross-over settings developed be Le Cleac'h when he had a DCX-2496.

Yes I have the Behringer DCX-2496

Here is a link to the page where the quotes below can be found and hope you may find it useful.


Le Cleac'h Expanded Soundstaging and 3D-Imaging


Rod: As you have gathered, I am no expert on this subject and am anxious to learn as much as possible. I am familiar with the issues you cite regarding the DCX2496 but I must say I have also found several opposing opinions citing it's great performance for the price. I am sure there are 'better' quality products out there, but at what cost? I can only compare the results I am getting with the DC-ONE, the DCX2496 and my old setup with ALK passives and Parasound HALO amps.

I am still in the early stages of evaluating performance since I am just now tweaking the crossover. The only time I ever turned the system on and thought "this isn't going to work" was when I used the DC-ONE with the HALO amps because I could not attenuate the sensitivity and had a lot of hiss and when I used the Crown D75 due to popping noises. Otherwise, I have not had any bad reactions to what I have heard on a superficial basis.

I find the DCX very quiet and it's performance very acceptable....as I said, I'm no expert.

I will point out this site to you, which has some reviews and opinions on the DCX as well as an intriguing way to set up the delays to enhance imaging and soundstage. I have not tried this yet, but am about to try their technique.

General page: http://freerider.dyndns.org/

Details on improving crossover setup: http://freerider.dyndns.org/anlage/LeCleach.htm

I would venture to say that some of the improvements I hear come not from the DCX, but form individually amping the drivers. The benefits stand out, although it is a pain to turn on 3 amps everytime, but I'm working on a power sequencer to automate the task.

The more I study this subject, the more I begin to think that we worry way too much on such things as interconnects, power cords etc. Even in a simple setup think about how many electronic pieces are in the way of the source signal from laser or hard drive head to speaker driver! That's a lot of chinese parts of which we have no quality control unless you are buying some very expensive gear. The bottom line is becoming 'how does it sound'.

I have endeavored to keep a running 'diary' of my experience for the next poor explorer that wishes to try this. Well, I'm off to try this improved filter design thing mentioned above.

OMG, I don't know where Mr. Le Cleac'h came up with his crossover calculations, but holy cow this sounds good!!!

Anyone who has the DCX needs to try this thing (don't take my word for it). Prior to trying this 3D combination, I had the folllowing crossover setup: L-R filter at 24dB slope all the way around, crossed at 499Hz and 6000Hz. The delays were mid 3.99 ms and 4.9 ms. Sounded good, imaging was good, but I kept struggling with a slight offset to the right (because my speakers were not exactly distant from my center point). That is, the center image was slightly righ of my center position.

After running the aforementioned DCX auto setup, making the speakers equidistant from the center (this time for real), and using the dowloaded spreadsheets here is what I came up with for my listening position and my Khorns:

Butterworth 3rd order filter, crossover points at 434Hz for woofer, 569Hz-5223Hz for mid, and 6843Hz for the tweeter. Mid polarity inverted with delay of 8.83ms. Tweeter was normal polarity with delay of 9.86ms.

I did quite a bit of A/B comparrisons with these two crossover settings and with all types of music. The Butterworth solution as indicated above sounds WAY better.

The web site is right, it opens the soundstage up and imaging is pinpoint accurate. Mind you, I have no idea what, why or how, I just followed the directions and found that it is worth the effort. This kind of flexiblity makes me sure I will never go back to passives for my mains!

There is no way I would have stumbled upon this setting anytime before winning the powerball lotto. I wonder how this would work for a two way system?

I'm sure somebody is going to throw the BS flag, but before you do, please try this technique. Certainly better than anything I would have come up with no matter how many REW readings I took. I ran the excel spreadsheet that simulates the settings and here are the results for the L-R and the Butterworth settings.

Link to comment
Share on other sites

Looks like Le Cleac'h Thomas Hartwig did some experimenting with not only the crossover filter types and midrange/tweeter delays, but also moving the center frequencies outward of each side of the crossover filter, all while looking at a square wave output on a oscilloscope. This is one way to get there.

I would expect that any time that you can get the phases more correct in the crossover regions, there will be an overall improvement in the resulting sound, but when you get the delays set properly between midrange and tweeter, the soundstage and "3D imaging" will suddenly appear to explode onto the scene.

Edited by Chris A
Link to comment
Share on other sites

I just made some quick adjustments using the numbers from Rudy81 and it sounds better already. I am heading out for the weekend, so will have to wait until next week to fine tune, but I am impressed with the immediate improvement. Moving the crossover frequencies apart was something I wanted to do while looking at the RTAs of my individual drivers, but I did not know if that was acceptable until reading Rudy's description and it makes good sense. Thanks to Rudy81. Thanks to Fjd.

Link to comment
Share on other sites

All the good folks of the forum with great speaker systems live far away, so my only audio quality comparisons are sound stages like this view from my front row seat the other night. Listening to a variety of music at home now, I am very pleased with the sound of my tri-amped, time aligned speakers.


Link to comment
Share on other sites

Here are the settings that I'm currently using on an ElectroVoice Dx38 active crossover for a center Belle (1982) with a Beyma CP25 tweeter. This was such an upgrade in sound quality that I'm wondering whether my prior K77 stock tweeter inside the Belle top hat might have some issue with its performance. However, my philosophy on this subject is "don't look back" since I never really liked the admittedly harsh sound of the K77 tweeter (i.e., a re-branded EV T35 tweeter).


EDIT (26 October, 2015):  Here is the link to the "JuBelle" settings using a K-510 horn and K-69-A driver instead of the Belle's stock horn. 


The Beyma allows the crossover frequency to be dropped to 2.5 KHz (from 3.5-5 KHz for the K77--a problem area for the stock Belle tweeter-midrange combination that really doesn't have a good crossover point). These settings break the Belle's frequency response into the following three bands: 65-600 Hz (a decade for the bass bin), 600-2500 Hz (2 octaves for the midrange), and 2500-20,000 Hz (about a decade for the tweeter). This choice was deliberate in order to minimize the bottom end of the K500 midrange and its top end, which has never been very pleasing to the ears. The current settings are like night-and-day to the K77 version of the Belle triamped.


I bypassed the Belle's "AB" passive crossover and connected to each driver using the AB crossover terminal blocks to each amplifier channel.  The Beyma CP25 tweeter sits on top of the top hat's sound absorption panel, centered on the top hat with its horn mouth slightly protruding in front in order to level the tweeter horn's axis with the Belle's front face.


The increase in sound quality over a stock passive-crossover Belle is very significant, IMO.


Belle Klipsch with Beyma CP25 tweeter - EV Dx38 Tri-Amp Settings

v0.2 (9 August 2014 - Complete overhaul with delays corrected using TrueRTA)


IN1: Master EQ1, Type Bypass
IN1: Master EQ2, Type Bypass
IN1: Master EQ3, Type Bypass
IN1: Master EQ4, Type Bypass
IN1: Master EQ5, Type Bypass
IN1: Master Delay -- Delay: 1.9ms, Unit: ms

IN2: Master EQ1, Type: Bypass
IN2: Master EQ2, Type: Bypass
IN2: Master EQ3, Type: Bypass
IN2: Master EQ4, Type: Bypass
IN2: Master EQ5, Type: Bypass
IN2:Master Delay -- Delay: 1.9ms, Unit: ms
IN1+In2: Master Delay -- Delay: 1.9ms, Unit: ms


Out1:Routing, Source IN1, Label LO
Out1: CHANNEL EQ1,Type: PEQ, f: 180Hz, Q: 2.0, Gain: -12dB
Out1: CHANNEL EQ2,Type: PEQ, f: 192Hz, Q: 1.4, Gain: -5dB
Out1: CHANNEL EQ3,Type: PEQ, f: 420Hz, Q: 1.4, Gain: -3dB
Out1: CHANNEL EQ4,Type: PEQ, f: 300Hz, Q: 6.0, Gain: +6dB
Out1: HIPASS XOVER, Type: thru
Out1: LOPASS XOVER, Type: Linkwitz 24 dB, f: 560Hz, Pol: norm
Out1: CHANNEL DELAY, Delay: 0.0 us, Unit: us
Out1: COMPRESSOR, Thrsh: 0dB, Rat:1/2.0, Attack: 0ms, Rels:100ms
Out1: LIMITER, Thrsh: 0dB, Release: 100ms
Out1: Level: 0dB, Polarity: Normal


Out2: Routing, Source IN1, Label MID
Out2: CHANNEL EQ1,Type: Bypass
Out2: CHANNEL EQ2,Type: Bypass
Out2: CHANNEL EQ3,Type: Bypass
Out2: CHANNEL EQ4,Type: Bypass
Out2::HIPASS XOVER, Type: Linkwitz 24 dB, f: 560Hz, Pol: norm
Out2: LOPASS XOVER, Type: Linkwitz 24 dB, f: 2.50KHz, Pol: norm
Out2: CHANNEL DELAY, Delay: 1125 us, Unit: us
Out2: COMPRESSOR, Thrsh: 0dB, Rat:1/2.0, Attack: 0ms, Rels:100ms
Out2: LIMITER, Thrsh: 0dB, Release: 100ms
Out2: Level: -12.0dB, Polarity: Normal


Out3: Routing, Source IN1, Label HI
Out3: CHANNEL EQ1,Type: PEQ, f: 12.4KHz, Q: 3, Gain: -11dB
Out3: CHANNEL EQ2,Type: PEQ, f: 5.0KHz, Q: 0.7, Gain: -8dB
Out3: CHANNEL EQ3,Type: Bypass
Out3: CHANNEL EQ4,Type: Bypass
Out3: HIPASS XOVER, Type: Linkwitz 24 dB, f: 2.5KHz, Pol: norm
Out3: LOPASS XOVER, Type: thru
Out3: CHANNEL DELAY, Delay: 2083 us, Unit: us|
Out3: COMPRESSOR, Thrsh: 0dB, Rat:1/2.0, Attack: 0ms, Rels:100ms
Out3: LIMITER, Thrsh: 0dB, Release: 100ms
Out3: Level: +3.0dB, Polarity: Normal


(Channel 4 not used for this application - center channel only)

* EV recommends using the same settings for its low pass/high pass filters.


belle+beyma cp25 tweeter 1-48th octave smoothing.jpg

Link to comment
Share on other sites

I corrected a microphone calibration issue that made the top end of prior PEQ and high shelf settings too hot by a couple of dB. I also started over on all the filters and managed to eliminate all midrange filters and reduce by more than half the number of tweeter filters. I also found a new set of filters for the bass bin that correct for known Belle FR issues. Only one high-"Q" filter is used to partially correct a narrow 300 Hz notch in FR found there.


I also found, downloaded, and used TrueRTA (the free version) to check the channel delays using a square wave input and TrueRTA's oscilloscope view. I plan to spend more time in this area in the future. But I can say that time delay mismatches within the two crossovers' bands while using the Dx38 active crossover to dial on-and-off the correct delay times is more than just a little audible - it will almost drive you out of the room when the time delays are incorrect, but "refocuses" when the time delays are corrected.


After listening to the current settings, the results are extremely close to the Jubs and the Cornwall surrounds, especially while using pink noise recalibration of channels, which tends to reveal a lot about the "voicing" of each speaker (a hint and lesson learned...). The current settings are now very pleasing for even typically strident music, such as cymbals, hammered electric guitar and electric bass lines, and high-sibilance vocals--which was one of two clues that I had a microphone calibration issue--the other being the FR of the Jubs on each side that I knew to be anechoically flat from Roy's settings.



Link to comment
Share on other sites

When reading comments about how an amp or speakers sound, I often wonder "compared to what"?

Here is a picture of some speakers I listened to the other night. The small private audience was allowed to stand right in front of the stage during the music video taping of a famous country music band, and the music quality sounded fantastic.

Listening to my speakers the next day is not an A-B comparison, but when I play my tri-amp K-horns they sound good and "live" to me.


Link to comment
Share on other sites

Updated the midrange delay very slightly after testing the two crossover points (560 and 2500 Hz) much more extensively. Now the frequency response and phase at the two crossover bands is smooth and without discontinuities.




belle bass bin-midrange crossover ampl and phase 16 aug 2014.jpg


belle midrange-tweeter crossover ampl and phase 16 aug 2014.jpg




Link to comment
Share on other sites

  • 2 weeks later...

Chris A,

On the specs for your Belle, there is a 1.9 ms delay applied to the input. What is the purpose of an input delay?

Also on the Belle specs, the output delay time is in microseconds (Unit: us), I hope it is okay that I use milliseconds (Unit: ms) across the board, just to keep it simple.

On the graphs in your latest post, the phase display has a sawtooth shape. How do I read that line?

Link to comment
Share on other sites

The 1.9ms delay is the minimum input delay of the unit.  I showed all input screens, even for those functions/filters that I didn't use.  The millisecond/microsecond unit selection is also on the input screen - so it's really up to you which units to use.


The phase should change without encountering any large discontinuities--smoothly changing as shown above.  The plot above of phase/frequency has no filtering in order to show this clearly even though you also see higher frequency noise on the FR and phase traces. It just so happens that the phase passes through 180 degrees near the center crossover point at the cursor position in the bottom plot.  I produced two new phase plots using REW on the same data but unwrapping the phase, below:

belle bass-midrange phase unwrapped 9 aug 2014.jpg


belle midrange-tweeter phase unwrapped 9 aug 2014.jpg

Link to comment
Share on other sites

Do you select the crossover point because of this phase, or do you adjust the phase based on the x-o frequency. Whatever the answer, is it an automatic process or do I need to study this subject carefully and make adjustments? I have REW and an ECM8000 microphone, but don't have much time right now to experiment with my already great sounding system. Thanks

Link to comment
Share on other sites

I selected the crossover point looking at the FR of each horn/driver on-axis, then calculated the relative time delays based on the math length differences of the bass bin --> midrange horns and midrange --> tweeter horns. 


I initialized the delays of the midrange and tweeter channels based on those path length differences. 


I then used REW up-sweeps to measure the phase of the bass bin-midrange crossover, and midrange-tweeter.  I adjusted midrange delay until I got a smooth phase in the crossover region with the bass bin, 


Once I got these two horns aligned, I measured the crossover band between the midrange and tweeter, updating the tweeter channel delay until I got a smooth phase curve.


For a starting point, I'd recommend the delays that Greg Oshiro documented in his Khorn tri-amping thread, assuming that you have stock Khorns.



Edited by Chris A
Link to comment
Share on other sites

Looking at the sawtooth pattern a few posts earlier, my reaction was this is getting complicated, but your latest explanation and graph make it easy to understand now. It is like the piece of a puzzle that allows me to see the big picture.


The flexibility and precision of the digital active crossover allows the woofer, midrange and tweeter to perform their best as individual units, and better together as a loudspeaker system. 

Link to comment
Share on other sites



You've been busy. If you are in an experimental mood, try this:

  1. Measurement mic at the listening position.
  2. A large pile of fuzzy stuff (technical term) on the floor to kill the floor reflections from all 3 speakers to the measurement mic. Use REW impulse response to verify that the floor reflection is sufficiently attenuated.
  3. Use REW upsweeps to plot frequency response (magnitude *and* phase) with the longest time window you can to exclude significant room reflections.
  4. Measure one Jub first and set the window parameters. Use the same settings to measure all 3 speakers.
  5. Make the magnitude and phase responses from all three speakers match as best you can. This will be difficult at high frequencies. One inch of distance difference at ~13.5Khz is 360 degrees
  6. I expect additional input delay will be needed on the Belle to compensate for the (I assume) shorter distance to the Belle relative to the Jubs.
  7. Allpass filters may have to be added to the Jubs to make the phase response match that of the Belle around the Belle MF/HF crossover frequency.
  8. Adjust level of the Belle relative to Jubs to taste by ear.

I've always wanted to try this, but I have no center channel.



Link to comment
Share on other sites

  • 3 months later...

Thanks Greg--you gave me about three months worth of work that I've just managed to complete/conquer. ;)  


  1. I found that taking measurements at the listening position were infused greatly with room acoustics "additions" and off-axis effects, so I place my microphone at 1 metre directly in front of each loudspeaker with the microphone slightly below the midrange horn and the axis of the microphone pointed at the tweeter.  Now I get FR and phase plots that make sense (actually the difference was like night and day for the phase - which I can use now for setting the delays, and it works quite well now, whereas the phase measurements really didn't work before from the listening position).  I check everything out from the listening position after the one metre measurements and adjustments are made.  This method works pretty well.
  2. I also found that using the one metre measurements dramatically cuts down on the amount of material I have to lay down on the floor - now I use one 2' by 2' pad between the loudspeaker and the microphone, and that seems to kill the floor bounce.  I also found that I can time gate the measurements after the fact at multiple time gates to see more of what is happening. 
  3. I can use the full 500 ms gate and still see what I need to see when measuring at one metre, without having to reduce the time gate.
  4. I use the same settings for all loudspeakers, moving the microphone to presurveyed one metre measurement positions for the sake of repeatability.
  5. Matching the FR and phase responses is the most challenging and time consuming task, but the payoff is very great.  I still cannot believe the difference that the time spent on doing this task has made (and could never be done via any automated means, I found).  I have the two surround loudspeakers (Cornwalls) yet to EQ/compensate for phase/delay between drivers (not channel delays, which have been compensated using REW), but I don't believe that the results will be as dramatic as from the front three loudspeakers, since most surround material uses the loudspeaker channels as "echo channels" [since completed, and the results were much greater than I would've imagined]. I do however have a few recordings that use the surround channels as real hard-cut music channels, and I'm looking forward to getting the surrounds EQed/compensation properly.
  6. I've used REW phase response measurements to compensate for delays for each driver within each loudspeaker, and used Audyssey to set the 5.1 channel delays using their measurement microphone at my listening position(s) - since I'm using three different active crossover DSP setups, all having different overall delay.  I have to add 21 feet to the delay that Audyssey wants to use for my TH subs (TH-SPUD clones with a 21 foot horn path length) since the precursor TH pulse seems to trip the Audyssey software into thinking that the precursor is the real pulse.  Since I cross over at 40 Hz to the subs in each corner, it really isn't terribly sensitive to phase adjustments.  I can still see the phase in REW for the Jub bass bin-TH sub crossover, and I use that measurement to set the delays on the Jub bass bin channel.
  7. I haven't had to use all-pass filters at all.  Am I missing something?
  8. I use REW and the calibrated microphone to set the channel gains from the listening position - since I can read 1/10th dB overall quite easily, which is better than I can hear or read using a handheld.  Getting the FR/phase matched from channel to channel in step #5 helped a lot here.

The center channel has been the reason why I've kept after this, since it never seemed to sound truly "seamless" before.  Now it does--and I can walk across the entire width of the room without being able to sense a "sweet spot" - which is a spectacular achievement, IMHO.


Thanks Greg--I needed something to do... :blink:


front three spks - one metre + sub on jubs.png


The green trace is the 3-way JuBelle (K-510/K-69-A midrange and Beyma CP25 tweeter) and the other two traces are the Jubs+subs, all taken at one metre on-axis (each). The dips in response are non-minimum-phase cancellations from the microphone measurement positions.




Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

  • Create New...