Hi Richard
If you intend on building a three-way system, I guess there might be cheaper solutions for a midrange than an MSW. I do not doubt that it would work but it looks like a waste of resources, since the MSW goes definitely higher than 20 kHz.
To me it's the same waste of resources as if somebody would use the excellent Compact Driver from Stage Accompany as a super tweeter only (it actually plays from about 1 kHz to ultrasonics !!).
Regards
Charles
BTW: Tonight I'm going to test (i.e. audition) my breadboarded active subtractive crossovers on music. On the scope they worked perrrrrrrfectly (i.e they sum flat in terms of frequency response AND the summed output waveshape is exactly what is getting into them, whether sinusoidal, triangular or rectangular). What comes out of the divers in the end is a different story of course.
If you intend on building a three-way system, I guess there might be cheaper solutions for a midrange than an MSW. I do not doubt that it would work but it looks like a waste of resources, since the MSW goes definitely higher than 20 kHz.
To me it's the same waste of resources as if somebody would use the excellent Compact Driver from Stage Accompany as a super tweeter only (it actually plays from about 1 kHz to ultrasonics !!).
Regards
Charles
BTW: Tonight I'm going to test (i.e. audition) my breadboarded active subtractive crossovers on music. On the scope they worked perrrrrrrfectly (i.e they sum flat in terms of frequency response AND the summed output waveshape is exactly what is getting into them, whether sinusoidal, triangular or rectangular). What comes out of the divers in the end is a different story of course.
Hi Al
I can even post it before it is working ideally since I assume there are others out there who already have some MSW's and who might probably want to experiment a little as well (it could be even interesting for those who like to experiment with other FR/woofer combinations).
I was only able to listen for five minutes yesterday. So I will have to invest some more time for getting levels and positioning right. But the first impressions were positive. I hope I will find time next week to make some measurements as well.
For the first experiments I am using small IC amplifiers with approx 10 Watts only. So I ran straight into limitations on the woofer side (I am currently using a Dynaudio 20W75 in a 18 liter enclosure, combined with a Linkwitz transform circuit) when I was playing a CD with a lot of low-frequency content. So I might probably make it switchable, to get more flexibility with positioning and/or choosing the FR<->SPL tradeoff that matches personal taste best.
Regards
Charles
I can even post it before it is working ideally since I assume there are others out there who already have some MSW's and who might probably want to experiment a little as well (it could be even interesting for those who like to experiment with other FR/woofer combinations).
I was only able to listen for five minutes yesterday. So I will have to invest some more time for getting levels and positioning right. But the first impressions were positive. I hope I will find time next week to make some measurements as well.
For the first experiments I am using small IC amplifiers with approx 10 Watts only. So I ran straight into limitations on the woofer side (I am currently using a Dynaudio 20W75 in a 18 liter enclosure, combined with a Linkwitz transform circuit) when I was playing a CD with a lot of low-frequency content. So I might probably make it switchable, to get more flexibility with positioning and/or choosing the FR<->SPL tradeoff that matches personal taste best.
Regards
Charles
phase_accurate said:
A sealed electret is a pressure sensor by definition. I guess if you didn't place it sideways, the intertia of the membrane would generate a signal component in proportion to the acceleration. Is this the idea?
The Funkschau article (or just the month and year, as I will be vising the Deutsche Bibliothek on Friday) would be really helpful.
Greetings,
Eric
Hi Eric
I will try to search for it tonight, I hope I still have it.
You are right about the reason for mounting the microphone sideways. Otherwise you would end up with something that reacts to pressure AND acceleration !
Regards
Charles
P.S.:
Maybe I will have time this week to measure the response of my active-crossover/MSW/Dynaudio combination. The Dynaudio Woofer was only provisional (I do in fact have tweeters I want to use them with in the end). The final speaker will use a Auditechnology driver that is capapble of moving three times as much air as the Dynaudio driver is capable of. I received these new drivers last Friday and I must say they are simply astonishing.
I will try to search for it tonight, I hope I still have it.
You are right about the reason for mounting the microphone sideways. Otherwise you would end up with something that reacts to pressure AND acceleration !
Regards
Charles
P.S.:
Maybe I will have time this week to measure the response of my active-crossover/MSW/Dynaudio combination. The Dynaudio Woofer was only provisional (I do in fact have tweeters I want to use them with in the end). The final speaker will use a Auditechnology driver that is capapble of moving three times as much air as the Dynaudio driver is capable of. I received these new drivers last Friday and I must say they are simply astonishing.
Good luck with Audio Technology drivers. I understand they are custom made and cost a fortune.
I have dug out an old Elektor article from 87 about an active subtractive crossover that goes back to an article by Lipshitz and Vanderkoy. Apparently, you can get 4th order LR HP and LP slopes by using a 4 th order LR for the lowpass and subtracting this signal from the original signal after delaying it with a 2nd order LR allpass. The result seems to be phase accurate
and you get a higler slope for the tweeter.
Have to do some more reading and thinking to see if there is a catch (wonder why it is not used more widely).
Eric
I have dug out an old Elektor article from 87 about an active subtractive crossover that goes back to an article by Lipshitz and Vanderkoy. Apparently, you can get 4th order LR HP and LP slopes by using a 4 th order LR for the lowpass and subtracting this signal from the original signal after delaying it with a 2nd order LR allpass. The result seems to be phase accurate
and you get a higler slope for the tweeter.Have to do some more reading and thinking to see if there is a catch (wonder why it is not used more widely).
Eric
Hi Eric
The mentioned 4th order subtractive filter is just another way to implement a LR filter.
Unfortunately these are not phase accurate in the way I understand it, i.e. the drivers are not in phase throughout the crossover area, they are out of phase by 360 degrees !
Neither are the output signals of a subtractive filter in phase in the crossover region but the SUM of the output signals is exactly representing the input signals.
While an LR4 crossover has it's advantages (IMD, polar response and frequency response are all excellent) I don't want to use one within this project in order to get a transient perfect (whatever that means in the real world) response.
The woofers are indeed custom made and did also cost a small fortune.
I don't want to be regarded as snob and I must say that the active concept I am after should as well work with much cheaper drivers than Audiotechnology woofers and MSW fullranges, but I wanted to build a speaker that I will keep for many years.
I have been to High-End exhibitions a lot and I do of course like to hear (and see !) all that cost-no-object stuff (after all I am an EE who likes nicely built equipment).
But as Mr. Ishiwata from Marantz once pointed out, it isn't very difficult for a good EE to design and develop well performing cost-no-object equipment but it is rather difficult to design moderately priced equipment of reasonable performance and build-quality.
Having worked as an audio and video repair technician years ago, I really appreciated equipment that was nicely built and couldn't understand how any decent engineer could design really crappy stuff without beeing embarassed. By crappy stuff I don't even mean equipment with bad perfomance figures but but such that isn't meant to work properly for more than just a few years.
I have to admit this was alittle off-topic.
Regards
Charles
The mentioned 4th order subtractive filter is just another way to implement a LR filter.
Unfortunately these are not phase accurate in the way I understand it, i.e. the drivers are not in phase throughout the crossover area, they are out of phase by 360 degrees !
Neither are the output signals of a subtractive filter in phase in the crossover region but the SUM of the output signals is exactly representing the input signals.
While an LR4 crossover has it's advantages (IMD, polar response and frequency response are all excellent) I don't want to use one within this project in order to get a transient perfect (whatever that means in the real world) response.
The woofers are indeed custom made and did also cost a small fortune.
I don't want to be regarded as snob and I must say that the active concept I am after should as well work with much cheaper drivers than Audiotechnology woofers and MSW fullranges, but I wanted to build a speaker that I will keep for many years.
I have been to High-End exhibitions a lot and I do of course like to hear (and see !) all that cost-no-object stuff (after all I am an EE who likes nicely built equipment).
But as Mr. Ishiwata from Marantz once pointed out, it isn't very difficult for a good EE to design and develop well performing cost-no-object equipment but it is rather difficult to design moderately priced equipment of reasonable performance and build-quality.
Having worked as an audio and video repair technician years ago, I really appreciated equipment that was nicely built and couldn't understand how any decent engineer could design really crappy stuff without beeing embarassed. By crappy stuff I don't even mean equipment with bad perfomance figures but but such that isn't meant to work properly for more than just a few years.
I have to admit this was alittle off-topic.
Regards
Charles
Sorry to carry on off topic but just to put my tuppence worth in.....
I know of a certain pro-audio manufacturer who used a subtractive 4th order LR filter on one of their active monitors. I have to say that it sounded awful, although there were several other problems with the electronics as well. When i put my current prototype 4th order LR board and amps in instead, it was a definite improvement. I think the reason they used a subtractive filter was to try and cut down on component cost/board size. As I said, I suspect the main problem was more down to bad electronics design than any unsound theory of the subtractive principle.
phase_accurate - I'd be interested to know how your filters turn out, and what you've done about the phase issue with the subtractive method.
Incidentally, subtractive methods are definitely the way forward for dynamics processing, although i only know of 2 manufacturers who use it in their products.
I know of a certain pro-audio manufacturer who used a subtractive 4th order LR filter on one of their active monitors. I have to say that it sounded awful, although there were several other problems with the electronics as well. When i put my current prototype 4th order LR board and amps in instead, it was a definite improvement. I think the reason they used a subtractive filter was to try and cut down on component cost/board size. As I said, I suspect the main problem was more down to bad electronics design than any unsound theory of the subtractive principle.
phase_accurate - I'd be interested to know how your filters turn out, and what you've done about the phase issue with the subtractive method.
Incidentally, subtractive methods are definitely the way forward for dynamics processing, although i only know of 2 manufacturers who use it in their products.
MFB & subtractive crosovers
For those interested in the MFB sources: I posted these in Circlotrons "MFB from voice-coil former" thread.
For the ones intersted in other uses of the subtractive crossover than mine: There was an article in EW+WW written by a German guy, using reasonably priced materials to construct an active three-way system:
February 2000, page 105, Title "Adaptable active speaker system".
Regards
Charles
For those interested in the MFB sources: I posted these in Circlotrons "MFB from voice-coil former" thread.
For the ones intersted in other uses of the subtractive crossover than mine: There was an article in EW+WW written by a German guy, using reasonably priced materials to construct an active three-way system:
February 2000, page 105, Title "Adaptable active speaker system".
Regards
Charles
Hi all
For the impatient ones I post the schematic of the bare crossover network below.
Before someone starts throwing flames at me for using the 741 OPs:
I did of course not use these in the circuit but they were just handy to draw this diagram (i.e. I was too lazy to import another model into P-Spice).
The values are the current ones and are still experimental.
If you regard the frequency, where both outputs are equal in amplitude, as the crossover frequency, then it is 220 Hz approx with the values shown.
Because the MSW's lower cutoff frequency is only about an octave below the crossover frequency, it's inherent phase-shift disturbs the whole arrangement a bit.
Therefore a "gentle" notch filter is used around 100 Hz after the highpass output. I will publish values and circuit later on, as soon as I have been able to perform measurements.
This crossover topology might also be useful for users of other FR & woofer combinations if the crossover frequency is altered accordingly.
Depending on what power amplifiers are used, the summing OP-AMP and R6 may be omitted and the higpass output taken from the node between R4 and R5. For this purpose the amp should have high input impedance and 6dB more gain than with summing opamp. Additionally the high output would be in-phase and the low output inverted, whereas both outputs are inverted when the summing OP-AMP is used.
For those who would never dare to use any OP-AMP (though an OP-AMP in this configuration will have good THD and IMD performance): The circuit can as well be built with transistors or tubes but there would be some additional coupling caps needed.
Regards
Charles
For the impatient ones I post the schematic of the bare crossover network below.
Before someone starts throwing flames at me for using the 741 OPs:
I did of course not use these in the circuit but they were just handy to draw this diagram (i.e. I was too lazy to import another model into P-Spice).
The values are the current ones and are still experimental.
If you regard the frequency, where both outputs are equal in amplitude, as the crossover frequency, then it is 220 Hz approx with the values shown.
Because the MSW's lower cutoff frequency is only about an octave below the crossover frequency, it's inherent phase-shift disturbs the whole arrangement a bit.
Therefore a "gentle" notch filter is used around 100 Hz after the highpass output. I will publish values and circuit later on, as soon as I have been able to perform measurements.
This crossover topology might also be useful for users of other FR & woofer combinations if the crossover frequency is altered accordingly.
Depending on what power amplifiers are used, the summing OP-AMP and R6 may be omitted and the higpass output taken from the node between R4 and R5. For this purpose the amp should have high input impedance and 6dB more gain than with summing opamp. Additionally the high output would be in-phase and the low output inverted, whereas both outputs are inverted when the summing OP-AMP is used.
For those who would never dare to use any OP-AMP (though an OP-AMP in this configuration will have good THD and IMD performance): The circuit can as well be built with transistors or tubes but there would be some additional coupling caps needed.
Regards
Charles
Attachments
Manger disto
Klang und Ton has an active project with the MSW in their brand new 01/03 issue. From the SPL plots, I assume it is the Nd-magnet device even if they don't give the model number or the baffle conditions (sloppy article!).
Second harmonic at 80 dB/1m is 2.2% at 230 Hz and 0.65% at 1.7 kHz whereas third remains essentially below 0.3%-
At 90 dB/1 m, second is app. 10% at 230 Hz, with 3% peaks at 500, 800 and 1.7 kHz. Third is 3% at 290 Hz and 2.5% at 450 Hz and 1.2% at 700 Hz. They cross over at 160 Hz but don't give the function for the active XO (I suspect is is third order).
Step response does not look as nice as in other projects. I suspect part of the problem is the damping of the chamber. They have a description in words which is not entirely clear to me. There is no drawing, but I suspect they have damping material near the rear side of the MSW (which Daniela Manger advises against).
Still, they say they are absolutely in love with the sound...
Klang und Ton has an active project with the MSW in their brand new 01/03 issue. From the SPL plots, I assume it is the Nd-magnet device even if they don't give the model number or the baffle conditions (sloppy article!).
Second harmonic at 80 dB/1m is 2.2% at 230 Hz and 0.65% at 1.7 kHz whereas third remains essentially below 0.3%-
At 90 dB/1 m, second is app. 10% at 230 Hz, with 3% peaks at 500, 800 and 1.7 kHz. Third is 3% at 290 Hz and 2.5% at 450 Hz and 1.2% at 700 Hz. They cross over at 160 Hz but don't give the function for the active XO (I suspect is is third order).
Step response does not look as nice as in other projects. I suspect part of the problem is the damping of the chamber. They have a description in words which is not entirely clear to me. There is no drawing, but I suspect they have damping material near the rear side of the MSW (which Daniela Manger advises against).
Still, they say they are absolutely in love with the sound...
Re: Cool
Hi Christoph,
this passage escaped me when this thread was more active, but I found it when I reread the whole story.
Can you explain in more detail what this indirect tweeter is and why it would help?
Last night, I had my first listening experience with my own project. Boxes are particle board 24x23x35, lined with tiles (what a pain!) and dampening board (Trittschalldämmplatte), no rock wool or foam yet. Baffle edges are rounded with a 25 mm router bit.
I ran them without any XO, estimated f3 was 110 Hz. I had to crank up the volume quite a bit. Wonder if the stated sensitivity is correct. They lacked something (bass of course), and voices seemed different, more remote, but at the same time more natural and present. On pieces with a very clear female voice or with an American Indian flute, I couldn't detect any unnatural distortion. Actually, they seemed clearer than I had ever heard them.
I then connected the bass branch of just one of my T+A T160E speakers which are rolled off around 100 Hz. Bass seemed a lot too loud, but I got an idea of what they could sound like.
Regards,
Eric
cocolino said:
Hi Christoph,
this passage escaped me when this thread was more active, but I found it when I reread the whole story.
Can you explain in more detail what this indirect tweeter is and why it would help?
Last night, I had my first listening experience with my own project. Boxes are particle board 24x23x35, lined with tiles (what a pain!) and dampening board (Trittschalldämmplatte), no rock wool or foam yet. Baffle edges are rounded with a 25 mm router bit.
I ran them without any XO, estimated f3 was 110 Hz. I had to crank up the volume quite a bit. Wonder if the stated sensitivity is correct. They lacked something (bass of course), and voices seemed different, more remote, but at the same time more natural and present. On pieces with a very clear female voice or with an American Indian flute, I couldn't detect any unnatural distortion. Actually, they seemed clearer than I had ever heard them.
I then connected the bass branch of just one of my T+A T160E speakers which are rolled off around 100 Hz. Bass seemed a lot too loud, but I got an idea of what they could sound like.
Regards,
Eric
Hi Eric
So you got yourself a pair of MSWs as well ? What type are they - the rare-earth ones ?
BTW: The (almost frightening) clarity of female voices was an experience I also had when testing the drivers for the first time. Needless to say that I ran them fullrange as well.
I have also read the article in Klang & Ton and I was a little disappointed to see that the LF part uses "a box with holes" !!!!
The main culprits for the bad step-response for this project are the use of a vented box and the crossover type used IMHO.
Regards
Charles
So you got yourself a pair of MSWs as well ? What type are they - the rare-earth ones ?
BTW: The (almost frightening) clarity of female voices was an experience I also had when testing the drivers for the first time. Needless to say that I ran them fullrange as well.
I have also read the article in Klang & Ton and I was a little disappointed to see that the LF part uses "a box with holes" !!!!
The main culprits for the bad step-response for this project are the use of a vented box and the crossover type used IMHO.
Regards
Charles
Hi Charles,
yes, I got defective rare earth types and had them repaired at Manger. Actually, they had been back since beginning of November. I was going to work on my Neos first, but both home improvement stores here refused to cut parts below 20 cm because of the Berufsgenossenschaft (insurance for labor accidents). Must be a local thing, going to have my timber cut in Frankfurt instead.
The Neos, by the way, were even more eerie on female voices, but they will be difficult to work with because of directionality.
In the K&T article, they use a second order subtractive filter with delay à la Lipshitz and Vanderkooy (actually, they referred to the Duetta aktiv article where they more or less claimed the filter was transient perfect. This means the woofer has a second and the Manger a forth order slope => really strange! Actually, the Manger might benefit from a steeper XO, but at least you'd want to see matched slopes.
The whole construction description was poor. I wonder whether they want to use stuffing material right behind the Manger. Maybe that causes the bad disto behavior.
How's your filter coming along?
Regards,
Eric
yes, I got defective rare earth types and had them repaired at Manger. Actually, they had been back since beginning of November. I was going to work on my Neos first, but both home improvement stores here refused to cut parts below 20 cm because of the Berufsgenossenschaft (insurance for labor accidents). Must be a local thing, going to have my timber cut in Frankfurt instead.
The Neos, by the way, were even more eerie on female voices, but they will be difficult to work with because of directionality.
In the K&T article, they use a second order subtractive filter with delay à la Lipshitz and Vanderkooy (actually, they referred to the Duetta aktiv article where they more or less claimed the filter was transient perfect. This means the woofer has a second and the Manger a forth order slope => really strange! Actually, the Manger might benefit from a steeper XO, but at least you'd want to see matched slopes.
The whole construction description was poor. I wonder whether they want to use stuffing material right behind the Manger. Maybe that causes the bad disto behavior.
How's your filter coming along?
Regards,
Eric
Hi Guys,
New guy here. Interesting thread. I believe the higher order subtractive filters can be made transient perfect. To generate the highpass, you delay the full-range signal by a time equal to the group delay of the lowpass, before you subtract the lowpass from it. The catch is the delays are so large that it almost requires a digital circuit.
A Bessel lowpass, even a very steep one (4th - 10th order), with a flat group delay, gives a second order highpass and perfect summed response. An LR lowpass, with a small hump in the group delay, gives a 3rd order highpass and the signals sum "pretty well." Adding just a little extra delay, using the LR lowpass, you can get a 4th order initial rollof on the highpass and the summing isn't "too bad."
John Kreskovsky, of the FRD group, has a spreadsheet that lets you play with this stuff. It's based on the work of Lipshitz and Vanderkooy (spelling?).
http://www.geocities.com/kreskovs/new-xooz1.html
Read the notes in the spreadsheet. The example on the web page is the "not too bad" 4th order lowpass and 4th order highpass. The step response still looks pretty decent considering you will be applying it to real-world drivers.
All that said, it looks interesting but I haven't tried it. My plan is to get one of the DSP boxes (QSC, Rane, etc.) with a CAD-style interface that lets you draw your signal routing and give it a try.
New guy here. Interesting thread. I believe the higher order subtractive filters can be made transient perfect. To generate the highpass, you delay the full-range signal by a time equal to the group delay of the lowpass, before you subtract the lowpass from it. The catch is the delays are so large that it almost requires a digital circuit.
A Bessel lowpass, even a very steep one (4th - 10th order), with a flat group delay, gives a second order highpass and perfect summed response. An LR lowpass, with a small hump in the group delay, gives a 3rd order highpass and the signals sum "pretty well." Adding just a little extra delay, using the LR lowpass, you can get a 4th order initial rollof on the highpass and the summing isn't "too bad."
John Kreskovsky, of the FRD group, has a spreadsheet that lets you play with this stuff. It's based on the work of Lipshitz and Vanderkooy (spelling?).
http://www.geocities.com/kreskovs/new-xooz1.html
Read the notes in the spreadsheet. The example on the web page is the "not too bad" 4th order lowpass and 4th order highpass. The step response still looks pretty decent considering you will be applying it to real-world drivers.
All that said, it looks interesting but I haven't tried it. My plan is to get one of the DSP boxes (QSC, Rane, etc.) with a CAD-style interface that lets you draw your signal routing and give it a try.
I downloaded the QSC software and sketched out a possible digital implementation of JohnK's delayed, subtractive XO. Outside the XO, you need to EQ each driver as flat as possible, especially around the XO freq. As well, you need to time-align the acoustic centers outside the XO. The idea is you want to be sending a perfect waveform to "perfect" drivers. Baffle step, etc. would be handled by the "Room EQ" block on the input.
An externally hosted image should be here but it was not working when we last tested it.
This principle would indeed work and I was thinking of it years ago although for a mini P.A.
Personally I'd prefer an analog circuit simply because of issues like dynamics and frequency response.
If you want the full bandwidth of the newer media (SACD & DVD-A) then you have to go for 196 ksamples/second. Combined with the long filter length due to the low crossover frequency this would ask for a DSP with some hundred MIPS !
Regards
Charles
Personally I'd prefer an analog circuit simply because of issues like dynamics and frequency response.
If you want the full bandwidth of the newer media (SACD & DVD-A) then you have to go for 196 ksamples/second. Combined with the long filter length due to the low crossover frequency this would ask for a DSP with some hundred MIPS !
Regards
Charles