Joachim,
I completely agree. I actually up to now have had repeat failures in making series crossovers work really well.
I suspect in this particular speaker it works well especially in the Bass/Midrange crossover because of the very different look-back impedance each driver sees compared to the first order parallel version.
So a very primitive crossover without ton's of impedance linearisation etc. gives excellent results, in ways a similarly simple parallel crossover does not.
And in this case the on-axis response of the series crossover is not hugely different from the off axis one, but the off axis one is indeed different and much better than before, in part likely due to the much better acoustic phase tracking of the drivers in the crossover region.
Subjectively they do quite well, unspectacular, spacious and dynamic. Nothing like a really good horn system or a high efficiency cardioid design, but very listenable and natural for such a simple and primitive design.
Ciao T
I completely agree. I actually up to now have had repeat failures in making series crossovers work really well.
I suspect in this particular speaker it works well especially in the Bass/Midrange crossover because of the very different look-back impedance each driver sees compared to the first order parallel version.
So a very primitive crossover without ton's of impedance linearisation etc. gives excellent results, in ways a similarly simple parallel crossover does not.
And in this case the on-axis response of the series crossover is not hugely different from the off axis one, but the off axis one is indeed different and much better than before, in part likely due to the much better acoustic phase tracking of the drivers in the crossover region.
Subjectively they do quite well, unspectacular, spacious and dynamic. Nothing like a really good horn system or a high efficiency cardioid design, but very listenable and natural for such a simple and primitive design.
Ciao T
I have not yet heard a horn that I'd like to own. Seems like lots of stored energy in horns. I do wonder whether digital equalizing can reduce it. Joachim, have you any experience in this area?
Hi,
I do not know what horns you have heard. Those that I consider "good" have very low stored energy, based on the various ways to measure it.
But agreed, many of the mainly PA horns have ton's of resonances and delayed sound...
EQ does not help that much IMHO, but using them quite narrowband does help. In the 80's we where using 5 to 6 way all horn PA Systems, mainly sticking to the fairly narrow "clear" frequency range of each horn.
What Horns do well is a kind of explosive but natural dynamic that I yet have to hear direct radiators match. And of course they do (if well designed and matched) do controlled and even DI down into the 10's of Hz (at least the huge walls we used for concert sound).
Ciao T
I do not know what horns you have heard. Those that I consider "good" have very low stored energy, based on the various ways to measure it.
But agreed, many of the mainly PA horns have ton's of resonances and delayed sound...
EQ does not help that much IMHO, but using them quite narrowband does help. In the 80's we where using 5 to 6 way all horn PA Systems, mainly sticking to the fairly narrow "clear" frequency range of each horn.
What Horns do well is a kind of explosive but natural dynamic that I yet have to hear direct radiators match. And of course they do (if well designed and matched) do controlled and even DI down into the 10's of Hz (at least the huge walls we used for concert sound).
Ciao T
True, the dynamics of good horns are very good, which is also what prompted me to study them and do tests. Basically, it seems the less compression, the less stored energy. They also need good drivers to drive them as well since the effeciency starts to reveal more problems upstream of the playback chain.
I will do some simulations on series against parallel xovers, also the vertical radiation pattern. Then you will see the difference.
Joachim,
I did tons of sims (using different tools including still Calsod and measured a lot using pink noise and 1/12th octave smoothing.
I find MLSA measurements tell me a lot of things I do not want to know and very little that I do - I remember an avid "meter reader" doing an MLSA measurement of my system and suggesting it was severely defficient in frequency response.
So, it being easy to do and having the original settings saved I obliged him and corrrected the frequency response of system until his ETF measurement system told us "it is good now".
After almost an hours tweaking to get there EVERYONE PRESENT agreed the new system sounded REALLY BAD.
So I went back to my at best 1/2 Octave band equalised (the digital EQ could actually do 1/6th octave, but I never went down to that level, Auto-Eq doing it sounded bad as well) original setup ane all present (including the Avid Meter Reader) agreed that it sounded great.
So another case of "measures perfect, sounds crap, so you measure the wrong things...".
Ciao T
PS, I also do MLSA measurements, strictly to get a close-up view in high resolution around the crossover region, but I'd NEVER, EVER attempt to EQ a system or judge sound quality of a system based on MLSA...
I did tons of sims (using different tools including still Calsod and measured a lot using pink noise and 1/12th octave smoothing.
I find MLSA measurements tell me a lot of things I do not want to know and very little that I do - I remember an avid "meter reader" doing an MLSA measurement of my system and suggesting it was severely defficient in frequency response.
So, it being easy to do and having the original settings saved I obliged him and corrrected the frequency response of system until his ETF measurement system told us "it is good now".
After almost an hours tweaking to get there EVERYONE PRESENT agreed the new system sounded REALLY BAD.
So I went back to my at best 1/2 Octave band equalised (the digital EQ could actually do 1/6th octave, but I never went down to that level, Auto-Eq doing it sounded bad as well) original setup ane all present (including the Avid Meter Reader) agreed that it sounded great.
So another case of "measures perfect, sounds crap, so you measure the wrong things...".
Ciao T
PS, I also do MLSA measurements, strictly to get a close-up view in high resolution around the crossover region, but I'd NEVER, EVER attempt to EQ a system or judge sound quality of a system based on MLSA...
I've been trying auto EQ based on old measurement data using Unlimate Equalizer. VERY interesting results. No measurements yet, but here are some current impressions.
1. The sound really become more coherent. Image depth and focus is better, and the recording room reflections seem more realistic. (the magnitude of difference is just a bit less than properly tuned interconnects.)
2. The details and the dynamics of vocals and instruments seem to have better resolution. I enjoy even more the singing and playing skill of the performers.
3. The sound is cleaner, and I tend to want to increase the volume because it needs to be at a certain level to get the right "feel". This was not the case before where I was happy with lower levels.
4. I still feel something is a bit lacking, don't know whether it's because the power amp is not enough or whether it's because I'm playing through two sound cards in series to get things to work.
Automatic EQing is not bad if it's done right, but I think there are a few things we should be aware of:
1. The speaker still needs to have as little diffraction artifacts as possible. To have the EQ handle treat the diffraction artifacts is not good.
2. Speaker EQ and room EQ should be handled separately.
3. MLS based measurements reveal the diffraction artifacts better, so if the speaker was not designed to minimize and spread diffraction out over a wide range, auto EQing to that will not be good.
But this digital EQing thing really has great potential.
1. The sound really become more coherent. Image depth and focus is better, and the recording room reflections seem more realistic. (the magnitude of difference is just a bit less than properly tuned interconnects.)
2. The details and the dynamics of vocals and instruments seem to have better resolution. I enjoy even more the singing and playing skill of the performers.
3. The sound is cleaner, and I tend to want to increase the volume because it needs to be at a certain level to get the right "feel". This was not the case before where I was happy with lower levels.
4. I still feel something is a bit lacking, don't know whether it's because the power amp is not enough or whether it's because I'm playing through two sound cards in series to get things to work.
Automatic EQing is not bad if it's done right, but I think there are a few things we should be aware of:
1. The speaker still needs to have as little diffraction artifacts as possible. To have the EQ handle treat the diffraction artifacts is not good.
2. Speaker EQ and room EQ should be handled separately.
3. MLS based measurements reveal the diffraction artifacts better, so if the speaker was not designed to minimize and spread diffraction out over a wide range, auto EQing to that will not be good.
But this digital EQing thing really has great potential.
Thorsten, where did you measure ? To make the response flat at the listening seat is not a good idea when you sit in the farfield. In most cases the sound will be very bright then. In the old times companies like Canton, Heco and Braun did that and the infamous "Taunus Sound " was the result. Back in the 80th and 90th i had to fight that prejudice in countries like England and USA. They did not want to listen to my speakers at first because "German speakers sound too bright". I had luck because my speakers where made flat in the freefield or nearfield however you may call it. I solved the sound problem with accurate placement back then and my seminars about setup got popular world wide. My method even made it into Martin Colloms book together with the Cardas and Wilson placement. Noverdays we have digital equalisers but even Lyngdorf makes the EQ more wideband and soft recently. It is simply not posible to turn a Klipschorn into a Quad. I had that ilusion too in the beginning of the 90th. Correct EQ needs the particular speaker and room taken into acount and the result is a "house curve" that may not work in other systems. I am for one are a bit underwhelmed by the capabilities of digital room EQ but it can work rarther well in an active speaker when all components are of that ilusive "audiophile quality".
Later today i will post two simulations on a parallel and series network in comparison. Xover is arbitrary set at 2kHz, 3rd order.
Later today i will post two simulations on a parallel and series network in comparison. Xover is arbitrary set at 2kHz, 3rd order.
So here come my two hypothetical two way speakers. I set the xover at 2kHz and use a third order topology. The result is on the tweeter axis at 1m. The woofer is put 150mm down and 32mm back to acount for the mechanical setup. I made the frequency response as flat as posibe and got a plus minus 1dB result. The values of the xover parts are of cause not textbook after optimisation. First comes the parallel xover.
Series Vertical polar:
Parallel Vertical polar:
Series on axis FR:
Parallel on axis FR:
Just trying to present the comparison simpler. Thanks for these Mr. Gerhard.
Dan
Parallel Vertical polar:
Series on axis FR:
Parallel on axis FR:
Just trying to present the comparison simpler. Thanks for these Mr. Gerhard.
Dan
Last edited:
Thanks Dan, that makes it easier to compare. My point is that it is nearly imposible to build a series and a parallel xover hat has the same on and off axis performance. The series xover comes with a nice phase response on axis but also lots of lobing. No wonder these two version will sound considerably different.
Hello,
I have been working on a pair of PHOENIX clones made from
Two of these
http://www.tymphany.com/files/resources/dvifa/PL18WO09-04.pdf
one of these.
http://www.tymphany.com/files/resources/dvifa/DX25SC05-04.pdf
and bi-amped with a DCX2496 Digital XO. Plus single sub.
Set up in near field these things get lost in the room. The sound stage is MUCH larger than with B2031P’s. I am going to leave these things set up for a few days they need an M di-pole woofer.
This is the first use of the DSP XO.
Any possibility of using DSP XO with your ZDL?
DT
All just for fun!
I have been working on a pair of PHOENIX clones made from
Two of these
http://www.tymphany.com/files/resources/dvifa/PL18WO09-04.pdf
one of these.
http://www.tymphany.com/files/resources/dvifa/DX25SC05-04.pdf
and bi-amped with a DCX2496 Digital XO. Plus single sub.
Set up in near field these things get lost in the room. The sound stage is MUCH larger than with B2031P’s. I am going to leave these things set up for a few days they need an M di-pole woofer.
This is the first use of the DSP XO.
Any possibility of using DSP XO with your ZDL?
DT
All just for fun!
Joachim,
This is not a speaker I would EVER attempt to make. I don't where to start as to what is wrong with it... 😛
Ciao T
This is not a speaker I would EVER attempt to make. I don't where to start as to what is wrong with it... 😛
Ciao T
Joachim,
I always use a range of measurements. Non of which are very traditional or produce nice curves for publication.
If I equalise the system (I do not currently) I do use "house-curve" overlays (plus others) to add to a completely "flat" in room curve. I actually wrote quite extensively on that in the early 2K's...
Ciao T
I always use a range of measurements. Non of which are very traditional or produce nice curves for publication.
If I equalise the system (I do not currently) I do use "house-curve" overlays (plus others) to add to a completely "flat" in room curve. I actually wrote quite extensively on that in the early 2K's...
Ciao T
DT, yes i like a good dipole too. Actually my main speaker, the MPL, is a dipole.
Thorsten, i agree, this example speaker i simulated has a lot of flaws and your 3 way with series network may work much better. I do not say that series will work less well then parallel in general. What i found though is that one simply can not compare this two methods directly. Some time back i compared a Sonics Argenta with parallel network to a Sonics Argenta with series network. I gave my best to make them both as linear as posible.
To my surprsie the series etwork ended up to use more parts although the reason to use it for many is the argument that you can safe on parts. I found that to be untrue but others could have more luck.
So you make a series of measurements to EQ a speaker plus room. I understand.
Thorsten, i agree, this example speaker i simulated has a lot of flaws and your 3 way with series network may work much better. I do not say that series will work less well then parallel in general. What i found though is that one simply can not compare this two methods directly. Some time back i compared a Sonics Argenta with parallel network to a Sonics Argenta with series network. I gave my best to make them both as linear as posible.
To my surprsie the series etwork ended up to use more parts although the reason to use it for many is the argument that you can safe on parts. I found that to be untrue but others could have more luck.
So you make a series of measurements to EQ a speaker plus room. I understand.
DT, i have a digital Behringer crossover and a digital Behringer equalizer so i could try to do digital crossover and EQ. At this moment i will do the ZDL in the analog domain though.
Joachim,
I have up to now not found Series Crossovers a good alternative, tehyn typically did worse than parallel options. It is just in this single case it worked out out in exactly the way those who promote low order series networks as cure it all claim.
I think it has to do with using a very wideband mid. In it's 3 litre irregular shaped ceramic enclosure, okay, in it's thick walled flower pot, the "midrange" offers around 100Hz - 15KHz usable bandwidth.
The choke in parallel to the midrange offers a low look-back impedance to the driver at low frequencies, the polar opposite of a first order parallel crossover where the lookback impedance at LF for the midrange rises. So it eliminates the problem with the LF Peak without extra components (LCR Trap). Equally the big (200uF film) cap in parallel with the woofer kills the rising inductance, as does the overall cap across the LF/MF section for midrange.
The Ribbon tweeter does not as suck benefit from having an inductor across it (it is rather self damping), but if it was a dome tweeter without ferro fluid we again could operate without LCR trap for the impedance peak.
Much of this makes only sense in the context of first order, minimalist networks. If the crossover is higher order or more complex the benefits seem to rapidly evaporate.
In this case the acoustic crossover between the woofer & midrange is a true first order for a wide overlap range, with 1/6th of the wavelength of the crossover point as driver spacing, so it works well. The MF/HF is asymmetric first order LPF and third order (acoustic) HPF, so the wideband midrange still contributes a lot of the signal above the nominal crossover point, for this some EQ is added to the Tweeter so there is actually a transtional slope which is first order above the nominal crossover point and third order below...
Anyway, enough boring you with details...
Yes, also to get a speaker X-Over optimised... For that I typically make near-field gated measurements in precise positions where I compare against the simulation and make sure I get the expected responses. I find this is more useful than the traditional 1m at tweeter high measurements, when combining these "extreme closeups" with far-field measurements.
Ciao T
I have up to now not found Series Crossovers a good alternative, tehyn typically did worse than parallel options. It is just in this single case it worked out out in exactly the way those who promote low order series networks as cure it all claim.
I think it has to do with using a very wideband mid. In it's 3 litre irregular shaped ceramic enclosure, okay, in it's thick walled flower pot, the "midrange" offers around 100Hz - 15KHz usable bandwidth.
The choke in parallel to the midrange offers a low look-back impedance to the driver at low frequencies, the polar opposite of a first order parallel crossover where the lookback impedance at LF for the midrange rises. So it eliminates the problem with the LF Peak without extra components (LCR Trap). Equally the big (200uF film) cap in parallel with the woofer kills the rising inductance, as does the overall cap across the LF/MF section for midrange.
The Ribbon tweeter does not as suck benefit from having an inductor across it (it is rather self damping), but if it was a dome tweeter without ferro fluid we again could operate without LCR trap for the impedance peak.
Much of this makes only sense in the context of first order, minimalist networks. If the crossover is higher order or more complex the benefits seem to rapidly evaporate.
In this case the acoustic crossover between the woofer & midrange is a true first order for a wide overlap range, with 1/6th of the wavelength of the crossover point as driver spacing, so it works well. The MF/HF is asymmetric first order LPF and third order (acoustic) HPF, so the wideband midrange still contributes a lot of the signal above the nominal crossover point, for this some EQ is added to the Tweeter so there is actually a transtional slope which is first order above the nominal crossover point and third order below...
Anyway, enough boring you with details...
Yes, also to get a speaker X-Over optimised... For that I typically make near-field gated measurements in precise positions where I compare against the simulation and make sure I get the expected responses. I find this is more useful than the traditional 1m at tweeter high measurements, when combining these "extreme closeups" with far-field measurements.
Ciao T
Last edited:
You do not bore me at all but discussions with you need a lot of time. That can get very intense and then i do not find the time to alaborate on my ZDL any futher 🙂
Anyway, again i agree. The attractiveness of series xovers is in the simplicity if you get it working. If a crossover can be made with less parts i think that helps transparency but a series crossover is just another option for me and not the holy grail.
I do measurements at various distances too out of a lot of reasons but you find me recently more listeing to music. When i find something wrong then, i go back to measurements so my listening inspires measurements and not so much the other way around. If you had asked me 20 years ago i whould have told you that i will not even listen to a speaker that does not measure well but that has changed over time.
Anyway, again i agree. The attractiveness of series xovers is in the simplicity if you get it working. If a crossover can be made with less parts i think that helps transparency but a series crossover is just another option for me and not the holy grail.
I do measurements at various distances too out of a lot of reasons but you find me recently more listeing to music. When i find something wrong then, i go back to measurements so my listening inspires measurements and not so much the other way around. If you had asked me 20 years ago i whould have told you that i will not even listen to a speaker that does not measure well but that has changed over time.
- Status
- Not open for further replies.