I have a pair of magi 3.7i and would like to convert the crossover from serial to parallel.
So I can bi-amp them and play around with better parts.
From the schematic it looks like, and I am guessing 😊
12 db on the tweeter
6 db band pass on the midrange
6 db on the woofer
Am I right?
Found the schematic here http://www.integracoustics.com/MUG/MUG/tweaks/neolith/
I don't know where I can find neolith so I hope it's okay to use his material
So I can bi-amp them and play around with better parts.
From the schematic it looks like, and I am guessing 😊
12 db on the tweeter
6 db band pass on the midrange
6 db on the woofer
Am I right?
Found the schematic here http://www.integracoustics.com/MUG/MUG/tweaks/neolith/
I don't know where I can find neolith so I hope it's okay to use his material
I started with the assumption that you want to match the response of the original design. I left out the optional resistors in my model, then used standard parallel crossover elements to get a reasonable match for the first 10 dB of roll-off for the individual sections. All planar elements were modeled as flat resistance at the values listed on the schematic (which happens to be a seriously confusing way to draw the circuit, to me).
Redrawn/simplified schematic:
My take on the original crossover curves (values approximate and open to interpretation):
Tweeter - 12 dB per octave, 3.2 kHz, Q of 0.6. That matches pretty well down to 1.5 kHz, then deviates where the bend is in the original.
Midrange highpass - 12 dB/octave at approximately 390 Hz, Q of 0.85. Same issue as above - the tail deviates some farther down, and the bump up above the reference level at 600 Hz isn't there.
Midrange lowpass - 12 dB/octave at 5.5 kHz, Q of 0.62
Woofer lowpass - 12 dB/octave at 450 Hz and Q of 0.92, deviations again from original
Converting to a parallel crossover is going to cause some changes you may not have anticipated. Just doing basic matching of the individual driver responses with simple crossovers doesn't work, even with a phase inversion of the midrange (which is standard for 12 dB/octave parallel crossovers). You will also lose the time/phase coherent aspect of the current crossover. Whether all that matters to you, I can't say, but you should realize you will be making significant changes to the way the speaker operates to do this.
This is what happens when you try to use the same basic curves in the parallel configuration:
I believe the 3.6 version used a parallel crossover. If you really just want to bi-amp, that may be a more suitable version to start with.
If you want to proceed, you have to decide how you want to approach the project. Trying to use more standard Q's at the original crossover frequencies, with inverted midrange polarity seems like a logical approach, but it's not my speaker.
Redrawn/simplified schematic:
My take on the original crossover curves (values approximate and open to interpretation):
Tweeter - 12 dB per octave, 3.2 kHz, Q of 0.6. That matches pretty well down to 1.5 kHz, then deviates where the bend is in the original.
Midrange highpass - 12 dB/octave at approximately 390 Hz, Q of 0.85. Same issue as above - the tail deviates some farther down, and the bump up above the reference level at 600 Hz isn't there.
Midrange lowpass - 12 dB/octave at 5.5 kHz, Q of 0.62
Woofer lowpass - 12 dB/octave at 450 Hz and Q of 0.92, deviations again from original
Converting to a parallel crossover is going to cause some changes you may not have anticipated. Just doing basic matching of the individual driver responses with simple crossovers doesn't work, even with a phase inversion of the midrange (which is standard for 12 dB/octave parallel crossovers). You will also lose the time/phase coherent aspect of the current crossover. Whether all that matters to you, I can't say, but you should realize you will be making significant changes to the way the speaker operates to do this.
This is what happens when you try to use the same basic curves in the parallel configuration:
I believe the 3.6 version used a parallel crossover. If you really just want to bi-amp, that may be a more suitable version to start with.
If you want to proceed, you have to decide how you want to approach the project. Trying to use more standard Q's at the original crossover frequencies, with inverted midrange polarity seems like a logical approach, but it's not my speaker.
I think some of the components in the crossover are really okey, like the film capacitors. What type of inductors is used?
If bi-amping is the goal. Is there a way to keep a series crossover mid-to-tweeter? I find the mid-to-tweeter integration is better on the 3.7 than the 3.6. For bi-amping, the bass-to-mid should be active.
The crossover and the bass and mid drivers of the 3.6 are really different to the 3.7. The ribbon tweeters are the same.
If bi-amping is the goal. Is there a way to keep a series crossover mid-to-tweeter? I find the mid-to-tweeter integration is better on the 3.7 than the 3.6. For bi-amping, the bass-to-mid should be active.
The crossover and the bass and mid drivers of the 3.6 are really different to the 3.7. The ribbon tweeters are the same.
There are many different ways to do all of this, but none of them are going to give you what was there in the same way as the original, but now with 2 amps driving it.
It's going to be a more involved project than a straight conversion to bi-amping would be with a speaker starting off with a parallel crossover configuration. It's likely to need measurement, adjustment, etc. For those with the capabilities, it might be an interesting project. For someone looking for an easy conversion, I think that's less likely.
Here's what it looks like with a simple first order parallel configuration, since I didn't include that previously. I tried to keep the cross points at the same frequencies.
Here are the midrange and tweeter in isolation with the original series elements on them. It's reasonably close to the original model, so would likely be OK in this configuration.
Dealing with the woofer/midrange transition while maintaining the same time/phase relationship of the original is likely to need DSP.
A standard 12 dB/octave crossover with a polarity inversion on one of them would probably work too, but that loses the time/phase coherence of the original for the bass/lower mid frequencies. Again, each person has their own ideas about what aspects are important to retain, so I'm not offering an opinion on that.
Crossover part info. from the Integra link: "the 24.0 uF caps (bank of 3) had a change of manufacturer (Solen to RTI, respectively). The other caps remain the same (also supplied by RTI). Also, the 3.7 uF RTI caps are film & foil. The inductors are steel laminate which have DCRs of 0.1 ohm or less."
It's going to be a more involved project than a straight conversion to bi-amping would be with a speaker starting off with a parallel crossover configuration. It's likely to need measurement, adjustment, etc. For those with the capabilities, it might be an interesting project. For someone looking for an easy conversion, I think that's less likely.
Here's what it looks like with a simple first order parallel configuration, since I didn't include that previously. I tried to keep the cross points at the same frequencies.
Here are the midrange and tweeter in isolation with the original series elements on them. It's reasonably close to the original model, so would likely be OK in this configuration.
Dealing with the woofer/midrange transition while maintaining the same time/phase relationship of the original is likely to need DSP.
A standard 12 dB/octave crossover with a polarity inversion on one of them would probably work too, but that loses the time/phase coherence of the original for the bass/lower mid frequencies. Again, each person has their own ideas about what aspects are important to retain, so I'm not offering an opinion on that.
Crossover part info. from the Integra link: "the 24.0 uF caps (bank of 3) had a change of manufacturer (Solen to RTI, respectively). The other caps remain the same (also supplied by RTI). Also, the 3.7 uF RTI caps are film & foil. The inductors are steel laminate which have DCRs of 0.1 ohm or less."
Thanks everybody, for taking the time to simulate the filter and give such great feed-back.
I have one question. Shouldn't the woofer/midrange be 3 db down at XO? like the midrange/tweeter?
I have one question. Shouldn't the woofer/midrange be 3 db down at XO? like the midrange/tweeter?
For a textbook basic filter response, yes.
In this case, no.
Series crossovers can be adjusted to achieve sharper or more gradual roll offs. When this is done, the shape of both curves around the cross point will change also. They still sum flat though.
An impedance mismatch between the legs can also cause the curves to deviate from textbook response.
With parallel crossovers you have more ways to force things to match a target curve. With series crossovers, it's a little more "take it as it comes" because things are interacting more across the whole circuit. One of the benefits though, is that it does all sum flat electrically even when crossover parts aren't perfect.
In this case, no.
Series crossovers can be adjusted to achieve sharper or more gradual roll offs. When this is done, the shape of both curves around the cross point will change also. They still sum flat though.
An impedance mismatch between the legs can also cause the curves to deviate from textbook response.
With parallel crossovers you have more ways to force things to match a target curve. With series crossovers, it's a little more "take it as it comes" because things are interacting more across the whole circuit. One of the benefits though, is that it does all sum flat electrically even when crossover parts aren't perfect.
Last edited:
I don't see the point in converting this system to parallel.
If you're interested in "improving" the scheme, the only direction that makes sense is to remove the high-level crossover entirely and drive the system as tri-amp with line-level crossover.
That can be done with either DSP or ASP. And, your logical starting point should be the same electrical curves Magnepan is using for the stock system. You then have a relative "reference" to alter from as you see fit.
I performed this sort of process for the 3.6 and posted this on this on Audioasylum website over twenty years ago. 🙂
http://www.integracoustics.com/MUG/MUG/tweaks/davey/mg3.6xo.htmThat's not to say the stock transfer functions can't be improved on........but it does give you a starting point.
Dave.
If you're interested in "improving" the scheme, the only direction that makes sense is to remove the high-level crossover entirely and drive the system as tri-amp with line-level crossover.
That can be done with either DSP or ASP. And, your logical starting point should be the same electrical curves Magnepan is using for the stock system. You then have a relative "reference" to alter from as you see fit.
I performed this sort of process for the 3.6 and posted this on this on Audioasylum website over twenty years ago. 🙂
http://www.integracoustics.com/MUG/MUG/tweaks/davey/mg3.6xo.htmThat's not to say the stock transfer functions can't be improved on........but it does give you a starting point.
Dave.
I do agree with @Davey - going active is the only thing that makes sense if you expect a significant improvement. If you do it right the Maggies will become a far better loudspeaker. I have both 3.6 and 3.7 converted to active using Hypex FusionAmps. XO points are 700 and 4000 Hz, 4th order.
Yeah, that's quite different than the stock configuration of 3.7i's. You've made a significant change to the system using those electrical slopes.
Dave.
Hello Davey from AudioAsylum! I posted as EdG twenty years ago.
I still have the MG 1.6. After a lot of crossover re-engineering, I eventually concluded that Magnepan knew what they were doing. 🙂
Ed
Yes indeed, I remember. The external series resistor(s) modification to alter the midrange response was very clever.
Dave.
True. I'm also using DSP to improve the frequency response. No Magnepan speakers measure well. This is Stereophile's measurement of the 3.6/R. We can do a lot better than this ...
I hope you know why it measured like that or the way they measured it. That is not how they measure in a listening room.
As Roger mentioned, that's not a valid/realistic measurement. (Certainly not representative of how they sound in a room.) If you're making DSP corrections based on that, you're not going in the right direction.
The problems with the measurement technique are explained by Siegfried Linkwitz in the comments section:
https://www.stereophile.com/content/magnepan-magneplanar-mg36r-loudspeaker-more-comments
Dave.
Yeah, I know (at low frequencies). But the dips at 2, ~6 and 20 kHz will be there in some way or another.
I do DSP corrections based on my own measurements of course.
I still have one 3.6 in storage with the original passive XO in place. Maybe I'll drag it out and measure it in my room 🙂
I do DSP corrections based on my own measurements of course.
I still have one 3.6 in storage with the original passive XO in place. Maybe I'll drag it out and measure it in my room 🙂
Last edited:
No, not necessarily, in my opinion.
Any of those wiggly dips/peaks would most likely be altered just by moving the microphone position.
These large panel speakers have a complicated radiation pattern that is difficult to characterize unless the measurement is performed in a fully anechoic environment.
You can certainly perform "room correction" via DSP in your listening room, but the corrections are not representative of issues with the speakers, nor would they work for other users in other rooms.
Dave.
This is my 3.6/R with the passive XO.
1/6 oct smoothing and full time window.
Below is what Linkwitz recommended - 50 ms time window and 1/2 oct smoothing.
1/6 oct smoothing and full time window.
Below is what Linkwitz recommended - 50 ms time window and 1/2 oct smoothing.