With the B&C DE10-8 CD and the Eminence Beta 10A I am actively crossing over using a 24db linkwitz riley at 2.2khz.
Adding a 10uF ($14 here!) in series with the compression driver would be adding a first order 6db high pass at around 2khz that should be a big enough gap to not mess up the active crossover?
I think I will add a 6db lpad which will put a 8ohm resistor in parallel to the CD and a 4ohm in series. That should help both with hiss and smoothing the impedance curve.
col.
Adding a 10uF ($14 here!) in series with the compression driver would be adding a first order 6db high pass at around 2khz that should be a big enough gap to not mess up the active crossover?
I think I will add a 6db lpad which will put a 8ohm resistor in parallel to the CD and a 4ohm in series. That should help both with hiss and smoothing the impedance curve.
col.
gedlee said:
This only happens when the compression driver has a perfectly flat impedance curve which is never the case. In practice, the pole from the capacitor interacts with impedance peaks and dips leading to a more uneven frequency response, uneven group delay (and a more spoiled crossover).
With LR24 crossovers the capacitor should be chosen to produce minimal phase shift, like less than 20º, at least down to half the crossover frequency.
Eva said:
I said "to first order" meaning that there would be deviations from this ideal. The impedance peaks are at the lower end of the pass band and as such this still works well above that. And don't forget that the resistor across the driver flattens out the peaks a great deal. Hence, my statement was pretty accurate as it stands.
And for the record, I wouldn't worry about the "phase shift", except that, of course, everything needs to be choosen so that the crossover with real drivers works right. This cannot be done on paper - you must use the real drivers FR and phase data. Anything else is just a guess and likely to be a bad one.
In most cases the group delay of a compression driver mounted on a horn is efectively flattened by applying a parametric eq at the cutoff frequency to tame the main group delay peak (to reduce the Q of the last horn resonant mode and rear chamber resonance) and by applying a 2nd order allpass (phase shifter) to the midrange. Other parametrics may be used to compensate for group delay peaks and dips due to the few lower horn resonant modes. This scheme works very well and results in matched impulse responses and optimum on-axis summing (it also allows to use midrange horns down to the very cutoff point, even if they are undersized and peaky) but the capacitor will spoil it unless it's sized to introduce very little phase shift in the overlapping region.
Can someone confirm the following as a starting point?
I am using the B&C DE250 with the 10-inch DDS waveguide. I am crossing over actively around 1-1.2 KHz , 48 db/octave, LR using Behringer DCX2496.
Should I still use a 10 uF cap in series with the DE250 and a 10 ohm resistor in parallel? Then measure and Re-EQ the system?
TIA
I am using the B&C DE250 with the 10-inch DDS waveguide. I am crossing over actively around 1-1.2 KHz , 48 db/octave, LR using Behringer DCX2496.
Should I still use a 10 uF cap in series with the DE250 and a 10 ohm resistor in parallel? Then measure and Re-EQ the system?
TIA
Hello,
Could you, please, give us a pulse reponse of such phase equalized horn.
From my experience, I doubt a classical equalization will improve the group delay of a horn near its cutoff (as generally any equalization increases the group delay inside the interval of frequency inside which it is active).
There is another other way to perform electronically a phase equalization.
You'll find in the attached file an example of linearization of the group delay curve of one of my TAD TD2001 on a Le Cléac'h horn having a Fc = 320Hz. As you can see we can reduce by one octave or so the frequency above which the group delay is constant.
This example was obtained using a summation with a 2 resistors summator of 2 output channels of a DCX2496 digital crossover. I described the method in 2006 on the forum [son-qc] of which I am moderator.
http://fr.groups.yahoo.com/group/son-qc/message/21764
first channel programmation:
(no equalization, no polarity inversion)
high pass filter : Fc= 20Hz
low pass filter : Fc = 20kHz
polarity: normal
phase: 0°
delay: 550 millimeters
second channel programmation:
(no equalization, no polarity inversion)
high pass : Fc= 20Hz
low pass Linkwitz-Riley 48dB/octave Fc = 738Hz, inverted polarity
delay: 0mm
The obtained transfer function for the correction is the one of a high pass filter (Fc = 253Hz) that possess the property of delaying the high frequencies but not the low frequencies.
Best regard from Paris, France
Jean-Michel Le Cléac'h
Could you, please, give us a pulse reponse of such phase equalized horn.
From my experience, I doubt a classical equalization will improve the group delay of a horn near its cutoff (as generally any equalization increases the group delay inside the interval of frequency inside which it is active).
There is another other way to perform electronically a phase equalization.
You'll find in the attached file an example of linearization of the group delay curve of one of my TAD TD2001 on a Le Cléac'h horn having a Fc = 320Hz. As you can see we can reduce by one octave or so the frequency above which the group delay is constant.
This example was obtained using a summation with a 2 resistors summator of 2 output channels of a DCX2496 digital crossover. I described the method in 2006 on the forum [son-qc] of which I am moderator.
http://fr.groups.yahoo.com/group/son-qc/message/21764
first channel programmation:
(no equalization, no polarity inversion)
high pass filter : Fc= 20Hz
low pass filter : Fc = 20kHz
polarity: normal
phase: 0°
delay: 550 millimeters
second channel programmation:
(no equalization, no polarity inversion)
high pass : Fc= 20Hz
low pass Linkwitz-Riley 48dB/octave Fc = 738Hz, inverted polarity
delay: 0mm
The obtained transfer function for the correction is the one of a high pass filter (Fc = 253Hz) that possess the property of delaying the high frequencies but not the low frequencies.
Best regard from Paris, France
Jean-Michel Le Cléac'h
Eva said:
Attachments
I can't provide plots for the moment, I can only measure phase (of each driver and the sum) and I use to do most of these adjustments by ear (and comparing polarity flipping), but consider this:
- A parametric EQ with gain results in a group delay peak, which may be used to compensate a dip.
- A parametric EQ with attenuation results in a group delay dip, which may be used to cancel a peak (at the expense of attenuation, since you are really changing the effective Q of the lower cutoff, and you may match it to the group delay of a phase shifter).
Notice the dip in frequency response around 730 Hz, it's the responsible of your basic group delay correction.
Try one (or two, one "wide" and one "narrow") parametric(s) around 400Hz with suitable Q and attenuation and you will be able to flatten the group delay (once again, at the expense of rolling off the frequency response, but this may be corrected later with global equalization on the input signal).
You can use the waveguide down to the cutoff that way (if the compression driver can handle it).
- A parametric EQ with gain results in a group delay peak, which may be used to compensate a dip.
- A parametric EQ with attenuation results in a group delay dip, which may be used to cancel a peak (at the expense of attenuation, since you are really changing the effective Q of the lower cutoff, and you may match it to the group delay of a phase shifter).
Notice the dip in frequency response around 730 Hz, it's the responsible of your basic group delay correction.
Try one (or two, one "wide" and one "narrow") parametric(s) around 400Hz with suitable Q and attenuation and you will be able to flatten the group delay (once again, at the expense of rolling off the frequency response, but this may be corrected later with global equalization on the input signal).
You can use the waveguide down to the cutoff that way (if the compression driver can handle it).
Hello,
I did this kind of research quite extensively so I know what is doable or not.
So I'll wait for your pulse measurements and I'll study them very accurately.
(If a simple minimum phase equalization could make a time reversal it will be very interesting but this is science fiction...)
Best regards from Paris, France
Jean-Michel Le Cléac'h
I did this kind of research quite extensively so I know what is doable or not.
So I'll wait for your pulse measurements and I'll study them very accurately.
(If a simple minimum phase equalization could make a time reversal it will be very interesting but this is science fiction...)
Best regards from Paris, France
Jean-Michel Le Cléac'h
Eva said:
The lower cutoff of the horn is a minimum phase phenomena and may be altered with minimum phase filters to reduce Q. The group delay peak of a filter is directly related to its Q.
It's like linkwitz transform, where you change effective loudspeaker T/S parameters. It's not science fiction, it's just a generalized lack of creativity.
It's like linkwitz transform, where you change effective loudspeaker T/S parameters. It's not science fiction, it's just a generalized lack of creativity.
The resistor is optional. You can try it, or not. Either way will not hurt the driver.
My understanding is that even though you are using the DCX2496 for a crossover, you want to put a capacitor is series for extra protection. If this is the case, with the 1k kHz crossover point, then the 10uF cap is too small. It would work for about 2.5 kHz. For your 1 kHz crossover point, you need about a 27 uF cap.
Tom
Horizons, you can calculate it here:
http://ccs.exl.info/calc_cr.html#first
just use the high pass section.
col.
http://ccs.exl.info/calc_cr.html#first
just use the high pass section.
col.
hello Eva,
If you think that you can do it, so do it and I'll still wait for your results to analyse them.
I did since long time that kind of search and I know what will be the results ;-)
BTW : Linkwitz transform for an OB or equalisation of a closed enclosure are only effective inside a limited frequency range and always the high frequencies still arrive before the low frequencies (the group delay cannot tend toward zero at low frequency...).
Futhermore in those 2 cases the transfer fuction of the response is second order ...
Best regards from Paris, France
Jean-Michel Le Cléac'h
If you think that you can do it, so do it and I'll still wait for your results to analyse them.
I did since long time that kind of search and I know what will be the results ;-)
BTW : Linkwitz transform for an OB or equalisation of a closed enclosure are only effective inside a limited frequency range and always the high frequencies still arrive before the low frequencies (the group delay cannot tend toward zero at low frequency...).
Futhermore in those 2 cases the transfer fuction of the response is second order ...
Best regards from Paris, France
Jean-Michel Le Cléac'h
Eva said:
Jmmlc said:
There are a couple of full-horn working systems based on that group delay correction principle, but I don't have any software capable of measuring group delay. My current adjustment criteria is the quality of the null that you get across the crossover region when the wrong polarity is used and the quality of the summing with the right polarity, and impressive improvements result from group delay correction/matching.
Hello Eva,
If the crossover frequency is set quite far above the acoustical cut-off frequency of the horn then a partial electronic compensation of the reactance of the horn is not doubtful. But a complete compensation until frequencies very near of the acoustical cut-off of the horn, with zero group delay is a very doubtful claim for me.
The problem of reactance compensation by electronic means is quite the same problem as reactance annulling by acoustic means (dimensioning of a rear cavity...) most of the time the results are not at the height of the expectation.
Also you have to consider that most time distortion artefacts in horns like reflected energy from mouth to throat results in the pulse response as secondary delayed pulses from the main pulse. By the way theses artefacts are reflected in the response curve most people are thinking they are minimum phase but in reality they are not.
But we are here in the fog and it can only be eliminated by measurements. For me it is easy, using my own Matlab routine, to obtain the group delay curve from a pulse response. Send me the impulses response of the horn systems you mentionned and I'll do it.
Best regards from Paris, France
Jean-Michel Le Cléac'h
If the crossover frequency is set quite far above the acoustical cut-off frequency of the horn then a partial electronic compensation of the reactance of the horn is not doubtful. But a complete compensation until frequencies very near of the acoustical cut-off of the horn, with zero group delay is a very doubtful claim for me.
The problem of reactance compensation by electronic means is quite the same problem as reactance annulling by acoustic means (dimensioning of a rear cavity...) most of the time the results are not at the height of the expectation.
Also you have to consider that most time distortion artefacts in horns like reflected energy from mouth to throat results in the pulse response as secondary delayed pulses from the main pulse. By the way theses artefacts are reflected in the response curve most people are thinking they are minimum phase but in reality they are not.
But we are here in the fog and it can only be eliminated by measurements. For me it is easy, using my own Matlab routine, to obtain the group delay curve from a pulse response. Send me the impulses response of the horn systems you mentionned and I'll do it.
Best regards from Paris, France
Jean-Michel Le Cléac'h
Eva said:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.