No. I don't have the software. I am a mac user and have never found suitable software that runs under mac OS.
I take the speakers out on the lawn on a quiet day, away from reflective surfaces. This is the closest to free space I have available. With the selected x-over filter enabled, I match the mid and tweeter outputs separately within 0.5 dB, the highest resolution available on my RTA, on the listening axis, at the listening distance and height. I reverse the phase of tweeter or mid and observe the mic feed on the RTA. It should be a single vertical line but in practice there is a central line with rapid fall off to either side of the x-over frequency. With a perfect null, the RTA line falls off the scale in a dramatic way. I slowly adjust the magnitude and delay through the DCX until the notch is at its deepest. In practice, I can get at least –40 dB quite easily. This is limited by the 0.1 dB gain resolution and 2 mm delay resolution of the DCX. I reverse the phase of mid or tweeter and they are now aligned. Identical procedure for the mid-woofer alignment.
Interesting approach. Thinking about this briefly, it seems that getting deepest null requires forcing crossover point to center of RTA band.
For mid/tweeter at >2kHz alignment their is no need for outdoors with gated IR technique. Microphone on listening axis from midpoint of drivers at 6x or so driver spacing to centers with small windowing. Phases of mid and tweeter may be checked for slopes and intercept, revealing whole wave, half wave, and polarity issues. As such repeatability of notch depth more than -60dB often possible when using DCX2496.
I've also done this with short broadband burst, an effectively self gated signal.
Very nice build!
Regards,
Andrew
For mid/tweeter at >2kHz alignment their is no need for outdoors with gated IR technique. Microphone on listening axis from midpoint of drivers at 6x or so driver spacing to centers with small windowing. Phases of mid and tweeter may be checked for slopes and intercept, revealing whole wave, half wave, and polarity issues. As such repeatability of notch depth more than -60dB often possible when using DCX2496.
I've also done this with short broadband burst, an effectively self gated signal.
Very nice build!
Regards,
Andrew
Bon
After using the 4th order Bessel alignment for the last month, I have to say that it sounds way, way better than the 4th order LR or Butterworth filters. Playing a good recording at a low level, you can put your ear right up to the tweeter and hear the difference. Way more musical and pure sounding. I don't know if it's because of the basic nature of the Bessel alignment with it's relative lack of oscillation, or if it was just implemented better than the others by the Behringer team.
After using the 4th order Bessel alignment for the last month, I have to say that it sounds way, way better than the 4th order LR or Butterworth filters. Playing a good recording at a low level, you can put your ear right up to the tweeter and hear the difference. Way more musical and pure sounding. I don't know if it's because of the basic nature of the Bessel alignment with it's relative lack of oscillation, or if it was just implemented better than the others by the Behringer team.
bessel filters have the best phase response and transient response. Their main drawback is the initial shallow rolloff meaning a wider range over which the drivers are still contributing. If however the drivers can tolerate this then all the better 
The attached image is from the analog devices paper on filter design. It is for analog filters not digital but should still be relevant. available here http://www.analog.com/library/analogdialogue/archives/43-09/EDCh 8 filter.pdf , the image is from page 8.25
After reading lots about the pros and cons of the various crossovers, I went with an accoustic 4th order bessel passive crossover for my MTM's
Tony.
The attached image is from the analog devices paper on filter design. It is for analog filters not digital but should still be relevant. available here http://www.analog.com/library/analogdialogue/archives/43-09/EDCh 8 filter.pdf , the image is from page 8.25
After reading lots about the pros and cons of the various crossovers, I went with an accoustic 4th order bessel passive crossover for my MTM's
Tony.
Attachments
If you look at the photo of my RTA output in post #140, you will see the 1/3 octave bands at 80, 100, 125, 160 Hz. It is convenient to select one of these frequencies as x-over to maximise the discrimination of the nulling procedure. I forgot to to mention that the signal applied to both drivers is a sine at the crossover frequency. With careful adjustment of the driver gains in 0.1 dB increments and delay in 2 mm increments, the precision of the nulling process is perfectly adequate for excellent time alignment. Someday I might invest in software to do this in room, but it only has to be done once outdoors for my purposes. BTW, the DCX2496 has an automatic time alignment based on pseudo-random binary sequences. I do not use this facility, preferring the old-school approach.
I selected drivers that were capable of performing comfortably at least 2 octaves beyond the chosen crossover frequencies. I also checked the behaviour of the unequalised drivers with wide-band pink noise.bessel filters have the best phase response and transient response. Their main drawback is the initial shallow rolloff meaning a wider range over which the drivers are still contributing. If however the drivers can tolerate this then all the better
The attached image is from the analog devices paper on filter design. It is for analog filters not digital but should still be relevant. available here http://www.analog.com/library/analogdialogue/archives/43-09/EDCh 8 filter.pdf , the image is from page 8.25
After reading lots about the pros and cons of the various crossovers, I went with an accoustic 4th order bessel passive crossover for my MTM's
Tony.
It is really heartening to have technical support and actual listening experience about the attributes of the Bessel x-over alignment. It has been my preferred x-over filter for over a decade now. The behringer guys really did a good service by including it in the DCX2496 at a very affordable cost. It is a very effective solution to excellent driver integration at the cost of 6-channels of power amplification. Although power amps are quite readily available, it is still an extreme solution for many.
Bon,
Great build.
I am a Mac user, and use Smaart, which has IR and phase screens, etc.,.
I use its Automatic Delay Locator to get close, but the method you use for time alignment is still the most precise.
Since your RTA reads on ISO centers, doing the nulling process at those frequencies is most accurate, in practice once you have set the time alignment, moving a bit off the ISO center either way would have little effect as far as phase alignment since your drivers are all so well behaved on either side of the crossover point.
How many dB down at the nominal crossover frequency is the Behringer 4th order Bessel ?
Art
DCX2496 BES24 crossover 1.00kHz filters captured with swept sine via digital input:
Zoom of summed filters:
Yes, appears Behringer did OK by this one. It's been modified so phase of LR and HP match very well, and dip in response of sum isn't too bad. So this filter looks very much like DCX2496 LR24 filter, but doesn't sum perfectly flat.
Regards,
Andrew
Zoom of summed filters:
Yes, appears Behringer did OK by this one. It's been modified so phase of LR and HP match very well, and dip in response of sum isn't too bad. So this filter looks very much like DCX2496 LR24 filter, but doesn't sum perfectly flat.
Regards,
Andrew
They must have the frequencies spread apart pretty far to align the phase (ballpark looks like 1520/3700 mid/tweet .... 100/250 woofer/mid). I was trying to simulate what they were doing on some other makers software here ... could duplicate the response but phase was still 45* out. No amount of TA could align the phase and keep the FR Bon was showing (with the frequencies linked).
Adding (4) filters should get the on axis response near ideal ... a guy would need to add one for the tweeter @2.5k +2gain Q=0.7 ... two for the mid, one @2.5K +2gain Q=0.7 and one @160hz +2 gain Q=0.7 ... last on on the woofer @160hz +2 gain Q=0.7. This might get Bon close.
Adding (4) filters should get the on axis response near ideal ... a guy would need to add one for the tweeter @2.5k +2gain Q=0.7 ... two for the mid, one @2.5K +2gain Q=0.7 and one @160hz +2 gain Q=0.7 ... last on on the woofer @160hz +2 gain Q=0.7. This might get Bon close.
Last edited:
Good observation. This is how I do it. The behringer allows filtering, gain adjustment and delay of the input signal. I apply the required filters with chosen gain and Q to the input feed of each channel prior to crossover. This saves on two filters per channel over doing the same thing on the individual drivers feeds. My 10 year old DCX (v 1.14) has just enough processing power to manage it all. I have a newer version (v 1.17) but have not yet compared to see whether the code efficiency has improved the memory capacity.
That's why I asked about the Bessel 24 response at the crossover point of the specific crossover the OP was using, digital implementation is often quite different in different software/hardware, sometimes even from the same manufacturer.
I use the DBX Driverack PA, it is polarity reversed from input to output on all three bands.
Radically different settings would not correct the polarity inversion of the mid band (only) on the DBX DriverackPa +.
Neither the DriverackPA or the + version offer polarity reverse on inputs or outputs...
If a speaker measures good in phase and frequency response, it usually is good, but unfortunately different filters on different units may be needed to get the same measured response.
Art
I understand what you're saying Art. The bessel seems to have a different interpretation from unit to unit. Probably due to differing philosophies/priorities pertaining to polar response/phase linearity/magnitude flatness.
I like bessel networks in the LF bands ... in the MF/HF transitions it moves the voices too far forward for my tastes. Otherwise, overall imaging is excellent.
I like bessel networks in the LF bands ... in the MF/HF transitions it moves the voices too far forward for my tastes. Otherwise, overall imaging is excellent.
The DCX2496 allows phase reversal of any of the outputs and even phase advance by 5 degree increments up to 180 degrees. I never could understand the price difference between the DriverackPA and the DCX2496, especially since the DCX is 96 kHz sampling rate and the DBX is 48 kHz. Maybe the build quality of the DBX is higher. However, my first DCX did developed a noise issue after 6 months. The warranty replacement has been rock solid over 10 years.
Going by Wintermutes post #145, you probably are hearing the Bessels cleaner impulse response and lack of ringing. It seems that you can have excellent impulse response (Bessel) or excellent FR flatness (L-R) but not both simultaneously. There must be some uncertainty principle working. I come down on the side of the impulse response since the FR dip is easy to fix with a shallow bandpass filter.
Bon,
Just found out about this from an AVS thread; very impressive design and execution!
I do though want to clarify the mechanism of constrained layer damping.
The above statements indicate what may be a misunderstanding.
First, it's not clear what you mean by cork not shearing, as it has a very low elastic and shear moduli.
Second, the crucial parameter of the mid-layer compound is viscoelasticity, where "visco" refers to high internal damping which is required to dissipate energy with shear strain; a bouncy low-loss rubber with the same elasticity as the polyurethane, which fortunately for us is very lossy.
Just found out about this from an AVS thread; very impressive design and execution!
I do though want to clarify the mechanism of constrained layer damping.
The above statements indicate what may be a misunderstanding.
First, it's not clear what you mean by cork not shearing, as it has a very low elastic and shear moduli.
Second, the crucial parameter of the mid-layer compound is viscoelasticity, where "visco" refers to high internal damping which is required to dissipate energy with shear strain; a bouncy low-loss rubber with the same elasticity as the polyurethane, which fortunately for us is very lossy.
hi
thanks for your explication of your technics for time alignment, I will try to do the same with my speakers!
amazing job!
I'm not sure if I understand perfectly this though:
'' With the sine applied to both drivers, reverse the phase of one driver and adjust the phase and delay until the mic reading bottoms out. I can easily get a 30-40 dB null. With a little practice this can be improved.''
what does it mean?
thanks for your explication of your technics for time alignment, I will try to do the same with my speakers!
amazing job!
I'm not sure if I understand perfectly this though:
'' With the sine applied to both drivers, reverse the phase of one driver and adjust the phase and delay until the mic reading bottoms out. I can easily get a 30-40 dB null. With a little practice this can be improved.''
what does it mean?
Last edited:
This procedure is for an active set up.
Required equipment:Active crossover (Behringer DCX2496), Pink noise source (HFN Test disc), sine generator, RTA (Behringer 8024), Measuring mic (Behringer ECM8000), mic stand.
Set up a measuring mic at the listening position aligned with the tweeter or mid-way between tweeter and mid axis. At the selected mid-tweeter crossover frequency, play a sine signal with the tweeter only driven at moderate SPL. Play the same signal through the mid and adjust the SPL level to be identical at the listening position. Now reverse the phase of ONE ONLY of the mid or tweeter and play the same sine through both drivers. Slowly adjust the delay of the tweeter while observing the combined mic output on an RTA. You should see a strong vertical bar at the crossover frequency. As the tweeter delay is adjusted, the vertical level bar typically starts to drop. If it drops off the bottom of the screen, increase the pre-amp gain to raise the level. (Wear ear protection!!). At some point the delay will minimise the signal. The two drivers are now in perfect anti-phase. Reversing the out-of-phase driver will now have both drivers in perfect phase alignment. If you swap the phase of one driver back, the mic output will jump back up. The difference between the out of phase and the in phase is the measure of the alignment accuracy. Anything beyond 40 dB difference is adequate.
Without changing the mic position, repeat the same procedure with the woofer-mid at the lower crossover frequency. If the mid is delayed relative to the woofer then this delay must eventually be also added to the tweeter delay. With all the drivers connected in positive phase, they are now perfectly in phase at the crossover frequencies. I prefer the 24dB/octave Bessel alignment. It has the most gradual frequency dependent phase shift compared to L-R, Butterworth, Chebyshev. The out of band roll-off is more gradual, so I choose drivers that are well behaved outside their normal operational range (mostly Scanspeak). Finally, with all drivers operating in phase, I play pink noise (HFN Test CD, track 74) and adjust levels for a smooth overall response mon The RTA. Bessel alignment is not-flat summing so there may be a shallow dip through the crossover region. This can be corrected by low Q bandpass filters at each crossover frequency.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.