Hi everyone,
If you recall, I'm a big fan of XSim, but I'm embarking on converting a 2-way (AMT tweeter + 6.5" mid-woofer, ported) to a 3 way with the addition of a bass cabinet and 3-way plate amplifier. I took my original FRD measurements for the tweeter and mid, and have spliced in the Dayton reference data for the woofer. None of the levels or delays are probably accurate, they are just here to see how far I can get in this design while I consider whether or not to proceed.
Based on my good luck with LR4 in my center channel I'm trying to use that slope again here. Ignore the phase plots, this is VirtuixCAD under Wine/Linux so the graphics are not 100%.
Thoughts?
If you recall, I'm a big fan of XSim, but I'm embarking on converting a 2-way (AMT tweeter + 6.5" mid-woofer, ported) to a 3 way with the addition of a bass cabinet and 3-way plate amplifier. I took my original FRD measurements for the tweeter and mid, and have spliced in the Dayton reference data for the woofer. None of the levels or delays are probably accurate, they are just here to see how far I can get in this design while I consider whether or not to proceed.
Based on my good luck with LR4 in my center channel I'm trying to use that slope again here. Ignore the phase plots, this is VirtuixCAD under Wine/Linux so the graphics are not 100%.
Thoughts?
Don't know what you want to achieve, but that 6.5 probably could be crossed higher to 10. If I pick pretty good 6.5, but still random woofer data:
https://www.dibirama.it/home-page/mid-woofer/219-audax-hm170mn0-mid-woofer-6-5-8-ohm-120-wmax.html
https://www.dibirama.it/home-page/m...io-rs180-8-mid-woofer-6-5-8-ohm-120-wmax.html
https://www.dibirama.it/home-page/mid-woofer/816-scan-speak-18w-4434g00.html
Most of them are not "happy" below 100Hz, or even below 120-150Hz, Yours is crossed at ~90Hz.
Also, that bump on 200-400Hz kinda asks for -2dB
But this probably is best done by ear, right? 😉
https://www.dibirama.it/home-page/mid-woofer/219-audax-hm170mn0-mid-woofer-6-5-8-ohm-120-wmax.html
https://www.dibirama.it/home-page/m...io-rs180-8-mid-woofer-6-5-8-ohm-120-wmax.html
https://www.dibirama.it/home-page/mid-woofer/816-scan-speak-18w-4434g00.html
Most of them are not "happy" below 100Hz, or even below 120-150Hz, Yours is crossed at ~90Hz.
Also, that bump on 200-400Hz kinda asks for -2dB
But this probably is best done by ear, right? 😉
I'd suggest your practice session has reached it's objective. I think it's time to consider your acoustic position, and then to measure accordingly.
I'd like to have excellent off axis performance if possible. The tweeter I think is at the lower safe limit, and any higher and the woofer will beam I think. I'll measure off-axis before committing to the crossover point. On-axis you are correct, the ScanSpeak 18W revelators have excellent high frequency performance.
Good catch! 🙂
Thanks Allen!
I'm about 2 months away from having the bass cabinets built, so hope you and others are patient and don't feel I'm just asking random questions on a project that won't be built. 🙂
And apologies that I can never spell "VituixCAD" properly. I wonder if I became a supporter if I could have them change the name so I'm not wrong all the time?
Hmm... How much off axis is important? Up to +-20 degrees pretty much everything works if you can manage all the other reflections, for what you need seriously treated room (that is unergonomic, expensive and not necessary imo). From my limited experience, baffle shape is the king of influence on the uniformness of response at bigger degrees. The 3 best choices are in no particular order: 1) round very small baffle, like Ikea bowls, KEF blade etc. Compromise could be as narrow as possible classic cuboid with as big bevel/rounding of edges as practically possible 2) huge baffle, even better if with big round over on the sides, like Grimm Audio LS1, Poor Man's Strad, original Stradivari and so on. 3) in wall speaker
I have played with "1)" option pretty extensively, can confirm that it influences A LOT: even everything what is in the speaker shape at ~90 and more than 90+ degrees. You not just can measure, but hear the difference, wiggles are of +-10dB difference at -20dB levels or bigger. Crossovers need to be reworked even if you go from narrow bowl-like shape to the slightly bigger bowl-like shape. Also if going from full sphere/droplet shapes, to something truncated on the back. The difference not just audiophilia-esoteric level - it is dramatic!
And after you decide shape - then you should work on mid to tweeter crossover. That is not important for 10 to 6,5 frequency.
Another thing is tweeter waveguide: that is important, because most likely you need more shallow waveguide if you want to use midwoofer pretty high till its limit, and deeper, bigger if you want tweeter to go very low. I am not a pro at this, there are threads dedicated to waveguides.
Also most of the classic knowledge about stuffing, BR port location etc also apply BEFORE you do your crossovers - both of them influence up to 1kHz for sure. As you are playing with active stuff, you can play with your x-overs for years, every single day. You can make aggressive slopes which are almost impossible with passive and all the other DSP goodies. That is irrelevant. This forum is not needed for that, adjust for your taste and that's it.
Well I just went back to my original XSim crossover, I'm not remembering it well at all. 😀 Looks like my original design used a crossover much closer to 2.8kHz, not 1.8 kHz. Probably for good reason. 🙂
I'm going to keep adding to this thread rather than cluttering up the board with a new one. Solen has begun working up the design for the bass cabinets and its coming along quite nicely! I'll show the back view which in this case is the most interesting. 🙂
Erik,
Please advise of the exact driver model numbers. I think I can be support here, having done my own twin 12" woofers, 6" mid to tweeter 3-way, prior to studying directivity here
In the directivity study is TLDR, in short, in cabinet like yours (direct radiators on front baffle), bass is omnidirectional til about 150Hz (-3dB) so you could certainly move the crossover point higher, and relieve the 18W of bass which will reduce amplitude modulation distortion. @DcibeL has done some work on this with FSAF.
In addition, although I have never measured the RSS265HF-8, it's amplitude and impedance graphs suggest that it is capable of playing up to 500Hz without issue, so you have a wide margin to play with. Finally, if you cross the any mid-driver above the baffle step loss eg. 500Hz, you can preserve voltage sensitivity, which will certainly help if you ever decide to a convert to a passive 3-way...
Please advise of the exact driver model numbers. I think I can be support here, having done my own twin 12" woofers, 6" mid to tweeter 3-way, prior to studying directivity here
In the directivity study is TLDR, in short, in cabinet like yours (direct radiators on front baffle), bass is omnidirectional til about 150Hz (-3dB) so you could certainly move the crossover point higher, and relieve the 18W of bass which will reduce amplitude modulation distortion. @DcibeL has done some work on this with FSAF.
In addition, although I have never measured the RSS265HF-8, it's amplitude and impedance graphs suggest that it is capable of playing up to 500Hz without issue, so you have a wide margin to play with. Finally, if you cross the any mid-driver above the baffle step loss eg. 500Hz, you can preserve voltage sensitivity, which will certainly help if you ever decide to a convert to a passive 3-way...
I've unboxed the woofers and done basic impedance sweeps. These were on the living room table with basic stands used to ensure the magnet vent hole was not blocked. We can see here that there is good agreement between the Dayton data files and the measured values. My only goal at this point is to make sure all 4 drivers are probably electrically working as expected.
Hi @tktran!!
Thanks for that. I intend to play them as high as the off-axis measurements allow me to. 🙂 Thank you for your posts. I am thinking about 100 Hz right now. That takes plenty of load off the mid-woofers. We'll see when in cabinet how they behave. I'm never going to a passive 3-way. 😀
They don't have to be true omnis, just give me 45 degrees of perfect synergy with the mid-woofers. I'll stick to 4th order filters so hopefully this will give me excellent vertical response as well. I'm not sure if I posted it elsewhere but yes I'm using RSS265HF-8s.
My 2-way design had about 86 dB of sensitivity using a 4 Ohm midwoofer. These drivers should be around 90 in parallel. Overall I think the drivers are reasonably well matched given the amplifier sizes used and relative efficiencies. That all changes however when we consider music power per octave, in which case the tweeter's 100W amp is way too much. 🙂
In my modest living room I'll barely put 10W into each driver anyway.
Thanks for that. I intend to play them as high as the off-axis measurements allow me to. 🙂 Thank you for your posts. I am thinking about 100 Hz right now. That takes plenty of load off the mid-woofers. We'll see when in cabinet how they behave. I'm never going to a passive 3-way. 😀
They don't have to be true omnis, just give me 45 degrees of perfect synergy with the mid-woofers. I'll stick to 4th order filters so hopefully this will give me excellent vertical response as well. I'm not sure if I posted it elsewhere but yes I'm using RSS265HF-8s.
My 2-way design had about 86 dB of sensitivity using a 4 Ohm midwoofer. These drivers should be around 90 in parallel. Overall I think the drivers are reasonably well matched given the amplifier sizes used and relative efficiencies. That all changes however when we consider music power per octave, in which case the tweeter's 100W amp is way too much. 🙂
In my modest living room I'll barely put 10W into each driver anyway.
One more piece of advice. Excess group delay with LR4 might be audible when xo is in 100-400Hz range. This applies for 3-way W/M crossovers and with multiway-dsp I use it also for HT subs.
I prefer LR2, another benefit is a bit increased vertical directivity around xo which might reduce floor bounce effect. LR2 requires outdoor groundplane measurements to get right, because overlap is so wide.
I prefer LR2, another benefit is a bit increased vertical directivity around xo which might reduce floor bounce effect. LR2 requires outdoor groundplane measurements to get right, because overlap is so wide.
(E)GD is function of phase rotation delay in milliseconds. LR4 rotates 360deg, LR2 only 180, and this difference can be audible in bass, as smearing of transients eg. upright bass picking and kickdrum sound. As well a bass reflex makes delay 360deg worth, that is the problem with them, so I'm glad that you are planning to make a sealed subwoofer! VituixCAD author Saunisto also promotes low order crossovers in bass region.
https://www.audiosciencereview.com/...ility-of-group-delay-at-low-frequencies.8571/
https://acris.aalto.fi/ws/portalfil...udspeaker_Group_Delay_Characteristics_AAM.pdf
https://acris.aalto.fi/ws/portalfiles/portal/66449704/Audibility_of_Group_Delay_Equalization.pdf
https://www.audiosciencereview.com/...ility-of-group-delay-at-low-frequencies.8571/
https://acris.aalto.fi/ws/portalfil...udspeaker_Group_Delay_Characteristics_AAM.pdf
https://acris.aalto.fi/ws/portalfiles/portal/66449704/Audibility_of_Group_Delay_Equalization.pdf
Hi,
yeah 4ms excess group delay around 100Hz is quite much, and it's due to the low xo point. FIR filter to fix this needs quite a lot of taps at this low frequency, not all processors have enough.
Few more observations:
There seems to be a resonance around 500Hz on the mid response. Or, could be from your data splicing as well if you did near+ far field measurements and merged them. A resonance would need to be fixed by modifying the box, or if it's data processing error then would yield erroneous FIR filter so if you do FIR filters you must have very good data so it doesn't "fix" issues that actually do not exist, iow create actual problems out of bad data.
If I read properly you have ported mid, I suggest make the mid sealed and move up bass / mid xo as high as reasonable.
Another observation is that phase slopes do not align, so you have 4th order electrical filters but the acoustic slopes end up being something else. You should do 4th order acoustical slopes using what ever electrical filters to achieve that, instead! You can use optimizer in vituixcad to show target responses so it's quite easy task.
If you don't use FIR you'd likely get better sound using low order acoustic HP and LP as Juhazi suggests, to maintain nice tangible feel to the sound. Assuming the tweeter can handle some extra lows and the woofer doesn't have bad cone resonance.
If you use FIR you can use steep slopes, just make sure the data is good so the transfer function that makes the filter ends up good.
Tips:
Judge the system also by using relatively unintuitive metrics. Frequency response balance is perhaps the most obvious thing to judge a system with, but don't forget to judge by tangible feel as well. Too much group delay with the system or some issue with implementation and there is no tangible feel to the sound. It can sound right in many ways, but feels kinda lame. If you don't want the tangible sound, and listen relatively far away from speakers to utilize loud early reflections to make "wide sound stage", then almost anything goes as long as the "house curve" alias system balance is nice to your ear, this is really low bar to get "acceptable" sound and about any project with any implementation would reach this.
For very good hifi sound though, aim to get the tangible feel. Aim for high xo points for good SPL headroom, aim for good physical structure that has no resonances and yields nice directivity, here xo points have a role as well. Pay attention to group delay / phase stuff so that tangible sound happens. Weed out any resonances from the system, these draw attention to themselves even though not necessarily obviously audible they just seem to make feeling that "something is not quite right but I dont' know what" kinda feel robbing your attention from enjoying the music.
Pay attention to edge diffraction and match both speakers to each other very closely. Listen at close enough proximity to hear these minute details, move around in the room, play with positioning and toe-in, these really contribute to nice calming sound and enable you to optimize positioning and toe-in to get stereo image you want. Besides the frequency balance and tangible feel try to get sense if the sound feels calming (steady) or nervous (always changing) as you move around in the room? Is something drawing attention, or are you drawn into the music in awe? Do you hear the speaker or just the music?
In general experiment a lot, which the DSP enables you to do to improve listening skill to be able to detect any issues and do something about them. There is million opportunities to make humanly errors, it's relatively easy to get ideal performance in simulator but the data might not represent reality or implementing the simulator into DSP might have some humanly error ruining it. These aren't necessarily easily heard and could go unnoticed until you finally do. You could implement the exact same DSP settings again just in case you could notice AB testing between the previous that there was some error. If it still sounds the same (somehow wrong), you'd then redo measurements and data processing and simulation and then implement the DSP again. If it still sounds the same, perhaps your data doesn't have enough resolution, or perhaps your "taste" just needs it a it different, different balance. Perhaps it's in the data, just in very minute detail and overlooked.
In general, expect there could be a humanly error somewhere and inspect the system from many angles over time to identify and weed them out, at the same time your listening skill improves and you get more familiar with what you actually like, how your room sounds and so on. Also, be happy what you've got at any moment as it could be the best you can do. Basically all you need to do is stay open for it and adjust if necessary, only way to "improve" a system in my current opinion is just spend time with it, the system improves with listening skill.
TIP for VituixCAD, you can put the filter blocks touch each other so you don't have to use wire between them.
And most important advice, have fun!🙂 the second the task seems a burden reconsider what you are doing, and perhaps be happy with what you got by then.
yeah 4ms excess group delay around 100Hz is quite much, and it's due to the low xo point. FIR filter to fix this needs quite a lot of taps at this low frequency, not all processors have enough.
Few more observations:
There seems to be a resonance around 500Hz on the mid response. Or, could be from your data splicing as well if you did near+ far field measurements and merged them. A resonance would need to be fixed by modifying the box, or if it's data processing error then would yield erroneous FIR filter so if you do FIR filters you must have very good data so it doesn't "fix" issues that actually do not exist, iow create actual problems out of bad data.
If I read properly you have ported mid, I suggest make the mid sealed and move up bass / mid xo as high as reasonable.
Another observation is that phase slopes do not align, so you have 4th order electrical filters but the acoustic slopes end up being something else. You should do 4th order acoustical slopes using what ever electrical filters to achieve that, instead! You can use optimizer in vituixcad to show target responses so it's quite easy task.
If you don't use FIR you'd likely get better sound using low order acoustic HP and LP as Juhazi suggests, to maintain nice tangible feel to the sound. Assuming the tweeter can handle some extra lows and the woofer doesn't have bad cone resonance.
If you use FIR you can use steep slopes, just make sure the data is good so the transfer function that makes the filter ends up good.
Tips:
Judge the system also by using relatively unintuitive metrics. Frequency response balance is perhaps the most obvious thing to judge a system with, but don't forget to judge by tangible feel as well. Too much group delay with the system or some issue with implementation and there is no tangible feel to the sound. It can sound right in many ways, but feels kinda lame. If you don't want the tangible sound, and listen relatively far away from speakers to utilize loud early reflections to make "wide sound stage", then almost anything goes as long as the "house curve" alias system balance is nice to your ear, this is really low bar to get "acceptable" sound and about any project with any implementation would reach this.
For very good hifi sound though, aim to get the tangible feel. Aim for high xo points for good SPL headroom, aim for good physical structure that has no resonances and yields nice directivity, here xo points have a role as well. Pay attention to group delay / phase stuff so that tangible sound happens. Weed out any resonances from the system, these draw attention to themselves even though not necessarily obviously audible they just seem to make feeling that "something is not quite right but I dont' know what" kinda feel robbing your attention from enjoying the music.
Pay attention to edge diffraction and match both speakers to each other very closely. Listen at close enough proximity to hear these minute details, move around in the room, play with positioning and toe-in, these really contribute to nice calming sound and enable you to optimize positioning and toe-in to get stereo image you want. Besides the frequency balance and tangible feel try to get sense if the sound feels calming (steady) or nervous (always changing) as you move around in the room? Is something drawing attention, or are you drawn into the music in awe? Do you hear the speaker or just the music?
In general experiment a lot, which the DSP enables you to do to improve listening skill to be able to detect any issues and do something about them. There is million opportunities to make humanly errors, it's relatively easy to get ideal performance in simulator but the data might not represent reality or implementing the simulator into DSP might have some humanly error ruining it. These aren't necessarily easily heard and could go unnoticed until you finally do. You could implement the exact same DSP settings again just in case you could notice AB testing between the previous that there was some error. If it still sounds the same (somehow wrong), you'd then redo measurements and data processing and simulation and then implement the DSP again. If it still sounds the same, perhaps your data doesn't have enough resolution, or perhaps your "taste" just needs it a it different, different balance. Perhaps it's in the data, just in very minute detail and overlooked.
In general, expect there could be a humanly error somewhere and inspect the system from many angles over time to identify and weed them out, at the same time your listening skill improves and you get more familiar with what you actually like, how your room sounds and so on. Also, be happy what you've got at any moment as it could be the best you can do. Basically all you need to do is stay open for it and adjust if necessary, only way to "improve" a system in my current opinion is just spend time with it, the system improves with listening skill.
TIP for VituixCAD, you can put the filter blocks touch each other so you don't have to use wire between them.
And most important advice, have fun!🙂 the second the task seems a burden reconsider what you are doing, and perhaps be happy with what you got by then.
Last edited:
@DcibeL has data to LISTEN to the audible benefit of crossing over midwoofer higher here
The same principle applies to the 18W- cross where the distortion starts to rise, or if you don't measured distortion, where the frequency response naturally starts to drop.
eg.
Reference:
https://hificompass.com/ru/speakers/measurements/scan-speak/scanspeak-18w/8531g00
Based on this graph alone, I would crossover around 150Hz.
However, I also agreed with @tmuikku and @Juhazi.
High excess group delay due to low XO point. Reason 2 for move the crossover point higher.
And the RSS265HF-8 can manage it.
My woofers, OTOH, have dip in the nearfield at 500Hz, and my midranges cannot cross much lower than 400Hz as it runs of out of excursion. As a compromise, I chose acoustic LR4 @~350Hz.
I managed EGD of 1.5ms @ 100Hz (see below)
PS.
You should have a close look at sample the projects included in VituixCAD2- it will allow you to learn what is achievable with different crossovers, including use of passive, active IIR or FIR. And best of all, you don't even need to take your own measurements. The polar measurements are a big part of the challenge. Garbage in, garbage out, so please take good polar measurements in both planes.
The same principle applies to the 18W- cross where the distortion starts to rise, or if you don't measured distortion, where the frequency response naturally starts to drop.
eg.
Reference:
https://hificompass.com/ru/speakers/measurements/scan-speak/scanspeak-18w/8531g00
Based on this graph alone, I would crossover around 150Hz.
However, I also agreed with @tmuikku and @Juhazi.
High excess group delay due to low XO point. Reason 2 for move the crossover point higher.
And the RSS265HF-8 can manage it.
My woofers, OTOH, have dip in the nearfield at 500Hz, and my midranges cannot cross much lower than 400Hz as it runs of out of excursion. As a compromise, I chose acoustic LR4 @~350Hz.
I managed EGD of 1.5ms @ 100Hz (see below)
PS.
You should have a close look at sample the projects included in VituixCAD2- it will allow you to learn what is achievable with different crossovers, including use of passive, active IIR or FIR. And best of all, you don't even need to take your own measurements. The polar measurements are a big part of the challenge. Garbage in, garbage out, so please take good polar measurements in both planes.
Last edited: