Not so much more I can do on the speakers before the second mid arrives, but I did 'glue' the magnets to the bracing today. I put masking tape on the magnets and a liberal amount of elastic (acrylic?) glue on the bracing, then mounted the drivers loosely and left it to dry. The intention is to tighten them down a bit on the gaskets after it has dried. The tape should come off the magnets if I want to remove the woofers. I guess I should check the impedance after for see what happens to the 250Hz basket resonance.
I have also done some side-project to prepare for running the active setup. I built the ESP signal detector https://sound-au.com/project38.htm to power up the amps when there is line level signal from the Flex Eight.
I did some sims on the L&R 'summing' on the input, and depending on the output impedance of the source, it can have a significant effect on channel separation (from memory 33dB separation with 500ohm source), so I would not use it as is on a stereo amp, but in my case I'm running a separate output from the Flex Eight, summing L&R to one channel in the DSP. Could be a future sub channel or similar.
Pic below, clearly marked so the better half knows it's not for hair dryers or whatever
Has some DC-blocking and other small stuff inside too. Switch on 0=auto, and 1=manual override
I have also done some side-project to prepare for running the active setup. I built the ESP signal detector https://sound-au.com/project38.htm to power up the amps when there is line level signal from the Flex Eight.
I did some sims on the L&R 'summing' on the input, and depending on the output impedance of the source, it can have a significant effect on channel separation (from memory 33dB separation with 500ohm source), so I would not use it as is on a stereo amp, but in my case I'm running a separate output from the Flex Eight, summing L&R to one channel in the DSP. Could be a future sub channel or similar.
Pic below, clearly marked so the better half knows it's not for hair dryers or whatever
Has some DC-blocking and other small stuff inside too. Switch on 0=auto, and 1=manual override
I measured the impedance with the supported magnets on both speakers today, and pulled up an old measurement (first graph) to compare. The scaling is wrong on the old measurement due to some problems with the 'measurement box' but I fixed that today. The 'blip' at 250Hz is basically gone, and I don't se any obvious ill effects higher up either. I realize now I have the marker on 54Hz in both graphs, there is something (minor) going on there too, but not sure what. I'm thinking the box is not long enough to have a standing wave at 54Hz..? Any ideas?
I think I have to disagree on that. If you look earlier in the thread, I had the woofer mounted with the magnet outside, and could feel the resonance in the magnet (but not in the box), and I compared impedance when I was pushing it with my hand, and it had a significant effect.
Did you make a write-up somewhere on your measurements? Curious which one would be best in a 3-way design, functioning from 80-1200 hz.
Presenting all the measurements would be a huge mess and take to much time, sorry.
I'm using a very new chassis from Kartesian acoustic - https://www.kartesian-acoustic.com/ cause I start from 80-90Hz for home cinema.
But in a real 3-way design I would cross over higher and probably would choose a PHL driver.
I'm using a very new chassis from Kartesian acoustic - https://www.kartesian-acoustic.com/ cause I start from 80-90Hz for home cinema.
But in a real 3-way design I would cross over higher and probably would choose a PHL driver.
Okay then, unfortunate for us
Regarding calling it a 3-way, it's actually more of a 2-way, but crossed over to a subwoofer at 80hz. Would you still recommend the PHL for that?
And may I ask which driver from Kartesian and why you choose it above the PHL?
There are so many measurements ... it would need a homepage to organise that. And I honestly prefer to use that time to get new projects done.
There is only one 200mm driver vom them, Wom200 something. It simply is better suited for getting down to 80Hz in a closed volume. And Qms is higher and it has a more "clear" midrange - but also a little more "sterile" midrange. My goal was to get as much resolution in the midrange as from an "high power" 8" driver as possible (and still be able to cross at 80Hz - so no pur PA midranges). Downside - it has an area of high H2 and I'm not sure about consistency of the brand. They change drivers and make them better, not all drivers are always available ... a risk I took for my private speakers but not sure if I would stick with them for a commercial project.
The PHL drivers are old, made the same way since years ... and still with the absolute best money can buy! Midrange in the (not perfect but as good as possible) listening comparison was more "warm" and pleasant, but with less details. Also lower Qms.
Here some of the many measurements. When there are topics related to these drivers I probably pick some and post them, but I'm not regularly reading here.
Kartesian Wom200:
Beyma old vs new:
Different drivers .... and so on.
There is only one 200mm driver vom them, Wom200 something. It simply is better suited for getting down to 80Hz in a closed volume. And Qms is higher and it has a more "clear" midrange - but also a little more "sterile" midrange. My goal was to get as much resolution in the midrange as from an "high power" 8" driver as possible (and still be able to cross at 80Hz - so no pur PA midranges). Downside - it has an area of high H2 and I'm not sure about consistency of the brand. They change drivers and make them better, not all drivers are always available ... a risk I took for my private speakers but not sure if I would stick with them for a commercial project.
The PHL drivers are old, made the same way since years ... and still with the absolute best money can buy! Midrange in the (not perfect but as good as possible) listening comparison was more "warm" and pleasant, but with less details. Also lower Qms.
Here some of the many measurements. When there are topics related to these drivers I probably pick some and post them, but I'm not regularly reading here.
Kartesian Wom200:
Beyma old vs new:
Different drivers .... and so on.
Well, the midrange finally arrived, so now I made a basic setup in the minidsp flex eight just to be able to get them playing in the garage. Linearized the drivers with some basic parametric filters in vcad and copied biquads to minidsp, and then added std LR2 filters on top of that (300 & 2kHz atm). Amps are from 'stock' Tweeters run by a DIY JLH class A, Mids from a (modded) HK670, bass from a DIY class D in a AV receiver case. I have been tuning delays between drivers by ear (found this to work well in the past), same with amplitude balance between drivers etc, so very basic still, but the 'new speakers honeymoon' can begin
First thing I could noticed was some 'thickness' of male voices from the bass. Could be some resonance, or just the preliminary filters, sounded a bit resonant though. The 'listening room' is what it is, but I want to find/correct problems before I drag the speakers home. I rent the garage, and it's not next doors to my apartment, so there is some effort in moving them between locations.
Since this is the first time I work with vcad and minidsp, hints are welcome. I did some experiments with rephase and eqapo some years ago, but this is different..
As I did it now, I copied each biquad separately from vcad, and pasted it to the minidsp, which seems a bit clumsy. I might be missing something here..
Seems REW is more 'adapted' to minidsp, where everything can be exported/imported as a whole set?
I found that channels could be linked in the minidsp, which I think means parameters should be 'copied' between channels that are linked. I linked L & R channels (woofer to woofer etc), but it is still a bit unclear to me which parameters are linked (copied to linked channel), and which are not.
I need to sort out my vcad project too, since I have it on a USB stick, and when I move it between computers, the drive letter is different and nothing works, all the measurements have to be imported again, and things can go wrong..
Since this is the first time I work with vcad and minidsp, hints are welcome. I did some experiments with rephase and eqapo some years ago, but this is different..
As I did it now, I copied each biquad separately from vcad, and pasted it to the minidsp, which seems a bit clumsy. I might be missing something here..
Seems REW is more 'adapted' to minidsp, where everything can be exported/imported as a whole set?
I found that channels could be linked in the minidsp, which I think means parameters should be 'copied' between channels that are linked. I linked L & R channels (woofer to woofer etc), but it is still a bit unclear to me which parameters are linked (copied to linked channel), and which are not.
I need to sort out my vcad project too, since I have it on a USB stick, and when I move it between computers, the drive letter is different and nothing works, all the measurements have to be imported again, and things can go wrong..
Hi, the program is accurate so if there is error its either in the accuracy of the measurements, possible error in measurements, mistake merging / windowing the measurements, or implementing the filters into DSP. Could also be how the speakers work with the room. If at all possible verify response is close to simulated by measuring the whole speaker, at least axial response or something.
First tip I had wrong at beginning, in VituixCAD options set DSP system to MiniDSP, it might affect some filters.
ps. Copy pasting seems luxyry, I have to manually adjust every filter
First tip I had wrong at beginning, in VituixCAD options set DSP system to MiniDSP, it might affect some filters.
ps. Copy pasting seems luxyry, I have to manually adjust every filter
I'm lazy, so copy/paste for each filter is too much It seems the filters are implemented correctly when vcad is set to 'minidsp 96kHz'.
There is a saying here in Sweden: S*** in,-s*** out. I guess that applies here too
I have only done nearfield measurements on the woofer, because I can't lift it, and I read about ground plane measurements, but it seems to require a lot of space.. I'm guessing the woofer is pretty omni up to 300Hz anyways, so off axis measurements should not be necessary? Mid and tweet are 1m measurements to 90deg hor. No verticals.
It is a bit of a mess with the woofer/mid integration in vcad because of this. Phase and amplitude does not match at all between measurements, so I have to manually adjust them to something that seems plausible. It will not be easy to verify, measurements will include the room more or less. So I guess the result will be a combination of in-room measurements and what sounds best to the ear. Delays between drivers are not that hard to guesstimate, and can partially be tuned by ear too I think. Inverted polarity measurements will help to verify that also.
It seems the filters are implemented correctly when vcad is set to 'minidsp 96kHz'. There is a saying here in Sweden: S*** in,-s*** out. I guess that applies here too
I have only done nearfield measurements on the woofer, because I can't lift it, and I read about ground plane measurements, but it seems to require a lot of space.. I'm guessing the woofer is pretty omni up to 300Hz anyways, so off axis measurements should not be necessary? Mid and tweet are 1m measurements to 90deg hor. No verticals.
It is a bit of a mess with the woofer/mid integration in vcad because of this. Phase and amplitude does not match at all between measurements, so I have to manually adjust them to something that seems plausible. It will not be easy to verify, measurements will include the room more or less. So I guess the result will be a combination of in-room measurements and what sounds best to the ear. Delays between drivers are not that hard to guesstimate, and can partially be tuned by ear too I think. Inverted polarity measurements will help to verify that also.
Yeah that is the hard part, getting the low end response into simulator without error, especially with physically big speaker as the measurement is hard and should be done farther than 1m away diminishing the already short window and so on. All in all some work to get accurate data into simulator.
The manual says low end can be done with near field measurement and diffraction tool generated off-axis responses. I've also struggled with all this, need to make adjustments by ear, then measure in room response and things seem off and so on I think my problem is all of the three ways have been developed and measured more or less separately in different dates and environments and there is probably error / difference in windowing, gain structure of the measurement setup and so on. Need to measure everything same day same rig to get stuff spot on, revisit simulations. Its process with many steps so easy to get errors in and then reality doesn't match simulation so, although very annoying its just something to live with, just do it all again until process is accurate enough. Have fun!
The manual says low end can be done with near field measurement and diffraction tool generated off-axis responses. I've also struggled with all this, need to make adjustments by ear, then measure in room response and things seem off and so on
I think my problem is all of the three ways have been developed and measured more or less separately in different dates and environments and there is probably error / difference in windowing, gain structure of the measurement setup and so on. Need to measure everything same day same rig to get stuff spot on, revisit simulations. Its process with many steps so easy to get errors in and then reality doesn't match simulation so, although very annoying its just something to live with, just do it all again until process is accurate enough. Have fun!
Well, the preference rating is something, for example you could have the automatic optimizer optimize the rating I think, smoothen out some of the curves like PIR. But, its not everything, its just a single number and you need to know how to utilize it. If you haven't read out the CTA-2034-A standard paper available here (for free): https://shop.cta.tech/products/standard-method-of-measurement-for-in-home-loudspeakers and of course VituixCAD help https://kimmosaunisto.net/Software/VituixCAD/VituixCAD_help_20.html#Preference_rating
Its hard to give specific advice what to do and what not A reasonable start could be to tune the crossover slopes and possible delays for phase matching crossover like LR2 or LR4. Then figure out if there is some things in the responses you can EQ smoother (that show up most measured axis, especially on the listening window) while keeping eye on Power Response and DI, basically optimally these are smooth without high Q bumps or dips at least so experiment with xo points and so on. When everything starts to look something reasonable then just listen and enjoy, make it as basis for further developments which you can compare against.
Then there is things like step response and ERDI you could look and try various crossover filters and see what you like and gauge how the system performs to ear. 0-axis response can be sacrificed some in order to get smoother power response for example, or you could optimize for 10-20deg off-axis or listening window rather than the 0-axis. Of course tweak looking at the preference ratings as well, but I think its not best place to start, perhaps bit later on as you have some kind of good basis established. Its iterative process for me at least
I like to say crossovers are trivial, and they are in sense the tools are so good its only matter of work to get as good xo as possible with the given measurements, with the given construct. When you tweak the system long enough and listen in your room you might notice you'd need better measurements to get better something so, then its another build. But, I suspect a big speaker like that with fine crossover is gonna sound good so all you have to do is have fun with it.
Its hard to give specific advice what to do and what not
A reasonable start could be to tune the crossover slopes and possible delays for phase matching crossover like LR2 or LR4. Then figure out if there is some things in the responses you can EQ smoother (that show up most measured axis, especially on the listening window) while keeping eye on Power Response and DI, basically optimally these are smooth without high Q bumps or dips at least so experiment with xo points and so on. When everything starts to look something reasonable then just listen and enjoy, make it as basis for further developments which you can compare against.Then there is things like step response and ERDI you could look and try various crossover filters and see what you like and gauge how the system performs to ear. 0-axis response can be sacrificed some in order to get smoother power response for example, or you could optimize for 10-20deg off-axis or listening window rather than the 0-axis. Of course tweak looking at the preference ratings as well, but I think its not best place to start, perhaps bit later on as you have some kind of good basis established. Its iterative process for me at least
I like to say crossovers are trivial, and they are in sense the tools are so good its only matter of work to get as good xo as possible with the given measurements, with the given construct. When you tweak the system long enough and listen in your room you might notice you'd need better measurements to get better something so, then its another build. But, I suspect a big speaker like that with fine crossover is gonna sound good so all you have to do is have fun with it.
Last edited:
You can do groundplane measurements in 1m or even 0,5m - but you have to tilt your speaker! (and correct the level of course). You can also but it sideways instead of standing to reduce distance to the floor. And tilt your mic so that the capsule touches the floor when you measure up to 20khz.
Realistic measurements including baffle wit 30-40cm start at 30cm and at 50cm it looks already pretty good. There still are changes in the area of 1-2dB when I go further away but that could be room influence too. How wide is your baffle?
It's better to get a good measurement and a little mistake from the baffle as get a shi??y measurement cause of to much distance.
Realistic measurements including baffle wit 30-40cm start at 30cm and at 50cm it looks already pretty good. There still are changes in the area of 1-2dB when I go further away but that could be room influence too. How wide is your baffle?
It's better to get a good measurement and a little mistake from the baffle as get a shi??y measurement cause of to much distance.
Thank you for the suggestions!
I will try to make a few variants, and load them into the minidsp flex eight. The beauty of it is that I can have 4 presets, and switch between them with the remote. Looking forward to some playtime
The baffle is 50cm wide. I think I should do some reading about ground plane measurements..
I will try to make a few variants, and load them into the minidsp flex eight. The beauty of it is that I can have 4 presets, and switch between them with the remote. Looking forward to some playtime
The baffle is 50cm wide. I think I should do some reading about ground plane measurements..
I made a few variants to try. These are a little bit based on what I would and would not do in a passive XO. Simple and complex LR2 is what I would probably try as a passive, probably not complex LR4. Anyway, here I can try whatever
Simple is using as little correction as possible, and just crossing over where it seems natural, and not worry so much about wiggles etc. Complex I'm trying to do more EQ to the response, that would require more components in a passive XO.
LF EQ is still a wild guess, something I ended up with yesterday that sounded reasonable in the garage.
'By tradition' I'm still focusing a lot on the on-axis, but I'm trying to do some compromises here.
Feel free to comment!
Simple is using as little correction as possible, and just crossing over where it seems natural, and not worry so much about wiggles etc. Complex I'm trying to do more EQ to the response, that would require more components in a passive XO.
LF EQ is still a wild guess, something I ended up with yesterday that sounded reasonable in the garage.
'By tradition' I'm still focusing a lot on the on-axis, but I'm trying to do some compromises here.
Feel free to comment!
Attachments
Yeah cool While there is doubt in accuracy of the measurements there is no reason to go and tweak every detail as it might be wrong thing to do. If the measurements are good then there is no reason not to EQ! EQ your hearts content, especially if the response is nice for more than one direction so that you are not making one exact position sounding good while sacrificing every else position. When DI is smooth the EQ fixes both axial and power response. This is where the "good measurements, good construct" thing I mentioned comes from, if the acoustic behavior of the system is nice you can pretty much EQ the system to what ever target curve and its nice anywhere. You took some lengths to mitigate the edge diffraction for example so enjoy its benefits.
Few simple observations, the last octave beams hard so you might want to make the on-axis as main listening axis, its up to you though how you want it. If you check out normalized directivity chart (horizontal) you might be able to spot if there is any edge diffraction going on, if not then you could just EQ the on axis what ever you like, try the optimizer and preference rating at some point. Some edge diffraction might be hiding when the measurements are windowed and you might be able to tune the sound better by sliding the lower crossover point for example. If you have power available you could boost the mid driver if you want to, to get lower crossover point there, for example. Basically you can sacrifice some of the system sensitivity for smoother response as you can pour gobs of power in and still get plenty loud enough.
The complex LR4 looks smoothest of the three as there is no sudden kink in power / DI, but there is difference how everything sounds for each of these and you might just as well prefer some of the other settings. It is how ever possible to do complex LR2 xo which might sound the best and even then make it better, what ever that is. It is good excersise to do tweak the current measurement set for very smooth responses, although if it doesn't represent reality the sound might not be there but you'll get experience using the program so its not time wasted.
ps. I had past 10 IIR filters for HF section, almost 10 for mid and only two for bass and then some for room EQ with my pro driver 3-way but had to reduce to 8 for limitations in current DSP. Response of the drivers is not exactly smooth so they can benefit some massaging. My mid driver for example has similarly bumpy ride for the whole pass band as yours, its weird thing and might benefit investigating if there is some problems with the driver / enclosure combination or something. If not then just EQ it smooth. If you have FIR capability in the DSP you could make single filter that makes everything to a target (replacing multiple IIR filter blocks).
While there is doubt in accuracy of the measurements there is no reason to go and tweak every detail as it might be wrong thing to do. If the measurements are good then there is no reason not to EQ! EQ your hearts content, especially if the response is nice for more than one direction so that you are not making one exact position sounding good while sacrificing every else position. When DI is smooth the EQ fixes both axial and power response. This is where the "good measurements, good construct" thing I mentioned comes from, if the acoustic behavior of the system is nice you can pretty much EQ the system to what ever target curve and its nice anywhere. You took some lengths to mitigate the edge diffraction for example so enjoy its benefits.Few simple observations, the last octave beams hard so you might want to make the on-axis as main listening axis, its up to you though how you want it. If you check out normalized directivity chart (horizontal) you might be able to spot if there is any edge diffraction going on, if not then you could just EQ the on axis what ever you like, try the optimizer and preference rating at some point. Some edge diffraction might be hiding when the measurements are windowed and you might be able to tune the sound better by sliding the lower crossover point for example. If you have power available you could boost the mid driver if you want to, to get lower crossover point there, for example. Basically you can sacrifice some of the system sensitivity for smoother response as you can pour gobs of power in and still get plenty loud enough.
The complex LR4 looks smoothest of the three as there is no sudden kink in power / DI, but there is difference how everything sounds for each of these and you might just as well prefer some of the other settings. It is how ever possible to do complex LR2 xo which might sound the best
and even then make it better, what ever that is. It is good excersise to do tweak the current measurement set for very smooth responses, although if it doesn't represent reality the sound might not be there but you'll get experience using the program so its not time wasted.ps. I had past 10 IIR filters for HF section, almost 10 for mid and only two for bass and then some for room EQ with my pro driver 3-way but had to reduce to 8 for limitations in current DSP. Response of the drivers is not exactly smooth so they can benefit some massaging. My mid driver for example has similarly bumpy ride for the whole pass band as yours, its weird thing and might benefit investigating if there is some problems with the driver / enclosure combination or something. If not then just EQ it smooth. If you have FIR capability in the DSP you could make single filter that makes everything to a target (replacing multiple IIR filter blocks).
Last edited:
Another tips: You might want to evaluate the system in mono at first, listen the delays, distortion (crossover points), resonances, frequency balance and so on, make sure the crossover works. While listening for the LR4 and LR2 filters for example, feel out if you can sense the sound with other than ears. Some filters can make the responses nice but somehow ruin the transient response or something, that everything sound flat and not as dynamic as they could. Kimmosto tips to have Excess group delay below 1ms down to 100Hz, basically its mellow xo between bass and mid at least.
Then at some point you might want to measure both speakers separately and tweak the response to match each other. Listen in stereo if phantom center is center and sound is not localizable to either of the speakers. If sound localizes to speakers there might be still some resonances there that make the sounds differ. I used white noise fed to both speakers for this test and found out the very HF do not correlate and do localize to the speakers somewhat. It might be the resonances my drivers have or just feature of hearing I have no idea. Phantom center localization to my speakers is reduced simply by tweaking EQ on the highs, mellow it out some. When the phantom center is good the imaging should be good. By good I mean you can literally make the speakers silent with mono white noise and all the sound localizes to the phantom center. Something the measurements help to tweak out but not show directly.
It can be a long process if you want to tweak after tweaking Most important advice is just have fun withit, its a DSP so back and forth and exploration is possible but not necessary
Then at some point you might want to measure both speakers separately and tweak the response to match each other. Listen in stereo if phantom center is center and sound is not localizable to either of the speakers. If sound localizes to speakers there might be still some resonances there that make the sounds differ. I used white noise fed to both speakers for this test and found out the very HF do not correlate and do localize to the speakers somewhat. It might be the resonances my drivers have or just feature of hearing I have no idea. Phantom center localization to my speakers is reduced simply by tweaking EQ on the highs, mellow it out some. When the phantom center is good the imaging should be good. By good I mean you can literally make the speakers silent with mono white noise and all the sound localizes to the phantom center. Something the measurements help to tweak out but not show directly.
It can be a long process if you want to tweak after tweaking
Most important advice is just have fun withit, its a DSP so back and forth and exploration is possible but not necessary
Last edited: