@hifijim I see you have a passive and DSP notch on the M74A, which I think is a good idea.
Your choice of MF - HF crossover is also well done IMO. I'd run the M74A that high even with the T34B. I'm curious though, why you chose to cross the woofer that high with the mid being able to easily deal with a much lower crossover?
Your choice of MF - HF crossover is also well done IMO. I'd run the M74A that high even with the T34B. I'm curious though, why you chose to cross the woofer that high with the mid being able to easily deal with a much lower crossover?
Last edited:
Edit - I just noticed you had a HP at 500 hz in there which makes sense. It just didn't look like it was crossed that low on the individual driver response curves.
@profiguy thank you for the comments.
Yes, the midrange high pass is 500 Hz 2nd order with Q=0.5 (i.e. LR2). When combined with the natural rolloff of the driver, and the other EQ I have, this results in an acoustic LR4 at 500 Hz.
The woofer (with filter) has an acoustic response which is close to a 500 Hz 2nd order Q=0.7 (i.e. BW2). I found this combination to minimize the directivity bump at 500 Hz.
I have some wiggle room to try different woofer-mid crossovers. For example, I could make the woofer have an acoustic LR4 at 500 Hz. This causes a noticeable hump in the directivity index at 500 Hz, but I would need to listen to it to make a decision. How important are small changes in directivity in the 400 - 700 Hz range?
Here is a filter which has a slightly higher DI bump at 500 Hz, but it retains the asymmetric 4th order mid + 2nd order woofer crossover. I made the midrange high pass to be 500 Hz Q=0.707, and then adjusted to woofer low pass for flat response.
Would this sound better? or worse? The only way to know is to try it. The Hypex amps are being shipped to me now.
j.
Yes, the midrange high pass is 500 Hz 2nd order with Q=0.5 (i.e. LR2). When combined with the natural rolloff of the driver, and the other EQ I have, this results in an acoustic LR4 at 500 Hz.
The woofer (with filter) has an acoustic response which is close to a 500 Hz 2nd order Q=0.7 (i.e. BW2). I found this combination to minimize the directivity bump at 500 Hz.
I have some wiggle room to try different woofer-mid crossovers. For example, I could make the woofer have an acoustic LR4 at 500 Hz. This causes a noticeable hump in the directivity index at 500 Hz, but I would need to listen to it to make a decision. How important are small changes in directivity in the 400 - 700 Hz range?
Here is a filter which has a slightly higher DI bump at 500 Hz, but it retains the asymmetric 4th order mid + 2nd order woofer crossover. I made the midrange high pass to be 500 Hz Q=0.707, and then adjusted to woofer low pass for flat response.
Would this sound better? or worse? The only way to know is to try it. The Hypex amps are being shipped to me now.
j.
@hifijim Looking at the FR curves, it appears the woofer LP frequency is too high, but it may be a result of the baffle step being too aggressive. It can be difficult to dial in BSC and set a LP point when blending with a considerably smaller driver above this. I'd however leave the mid HP set to begin with.
I'd personally retain the 500 hz LR2 HP, as it improves transient response and phase relationship better between woofer and mid. The 4th order filter will worsen this and introduce a noticeable hard shift in phase, making the directivity swing further and linearity suffer. It will also affect FR balance in the nearfield, as higher order filters typically don't sum very smoothly in this area. This is a common trait with higher order filters used at lower frequencies, as timing will noticeably suffer.
The ear is quite sensitive in the range of 300 - 1k to relative phase and timing errors. Getting this area right is likely the most important part in blending the woofer with the mid. It will also have the most influence on how vocals sound, including many other acoustic instruments which are difficult to accurately reproduce.
I'd set the mid HP first and then dial the woofer LP in to blend. This usually ends up being easier to implement as a method of getting the LF - MF transition as best as it can be, both from a DI and phase POV. If after all this you can't achieve the desired DI uniformity, move the mid HP and then try again adjusting the woofer HP slope and frequency.
I'd personally retain the 500 hz LR2 HP, as it improves transient response and phase relationship better between woofer and mid. The 4th order filter will worsen this and introduce a noticeable hard shift in phase, making the directivity swing further and linearity suffer. It will also affect FR balance in the nearfield, as higher order filters typically don't sum very smoothly in this area. This is a common trait with higher order filters used at lower frequencies, as timing will noticeably suffer.
The ear is quite sensitive in the range of 300 - 1k to relative phase and timing errors. Getting this area right is likely the most important part in blending the woofer with the mid. It will also have the most influence on how vocals sound, including many other acoustic instruments which are difficult to accurately reproduce.
I'd set the mid HP first and then dial the woofer LP in to blend. This usually ends up being easier to implement as a method of getting the LF - MF transition as best as it can be, both from a DI and phase POV. If after all this you can't achieve the desired DI uniformity, move the mid HP and then try again adjusting the woofer HP slope and frequency.
Last edited:
@hifijim The more I look at your measurements, the more convinced I am about the BSC needing attention first. The LP point is already lower than I thought and may need to be initially dialed up closer to that of the mid to start with. This will make things easier to begin with as a starting point.
Drivers with as wide directivity as these are easily influenced by every lump and bump in their path. The midrange dip at about 1.5K looks to be diffraction related and is consistent with the baffle step shift shown in the polar. It could also be influenced by the transition thing (I don't have a good description as they would all sound unkind and not how I want to frame it). If you wanted to know which, you could test by placing some felt foam or other fairly aborbent material around the transition or on the final sharp edge.
Not sure if you saw it but @stoneeh printed and measured the waveguide and baffle shape I simulated for you earlier on.
https://www.diyaudio.com/community/threads/t25a-simple-waveguide-profile.423131/post-8008337
This is a truly amazing thread,
Inspiring, and really the crux of what makes this community so great.
I’m a novice, albeit an older one. I try to contribute only when I feel strongly about a particular topic. In the last couple years I’ve enjoyed a transition to an active system due to Pat McGinty of Meadowlark Audio. A lot of what is being discussed here is a path he has followed and even led for quite some time. If you are not familiar please see his website @ Meadowlarksings.com
I look forward to reading further and want to thank this community for a very interesting diversion.
Best
Inspiring, and really the crux of what makes this community so great.
I’m a novice, albeit an older one. I try to contribute only when I feel strongly about a particular topic. In the last couple years I’ve enjoyed a transition to an active system due to Pat McGinty of Meadowlark Audio. A lot of what is being discussed here is a path he has followed and even led for quite some time. If you are not familiar please see his website @ Meadowlarksings.com
I look forward to reading further and want to thank this community for a very interesting diversion.
Best
Yes that is so true. Both of the dome drivers have very wide dispersion. The T25A tweeter is noticeably wider than the similar sized SB26ADC aluminum dome for instance. The M74A mid is wider than the 4" SB12MNRX2, even though both have about the same Sd. As you say, every lump and bump in the path along the baffle has some effect.
The transition thing, I call it a transition block, protrudes 3 mm above the mid-tweeter baffle. I intended it to be flush, but tolerance build up and the practicalities of assembly worked against that goal. I just checked the prototype, and it is flush. Ah, the freedom of working with XPS foam, putty, and tape... it is very forgiving. Not so with rigid structural materials.
The transition block bump is 96 mm below the mid (center). The distance from mid center to the bump varies with the clocking angle, and at +/- 45 degree clocking angle, it is about 135 mm.
The half wavelength frequency of 96 mm is 1786 Hz, and 135 mm is 1270 Hz. My cancellation null lies neatly in that range. That is some compelling evidence for the culprit.
Early in the design phase I faced a choice on how to handle the baffle edge around the M74A. I could have used tapered bevels which would have allowed a one-piece baffle, or I could have used straight bevels with a transition block, which necessitates a 3 piece baffle. The diffraction sims showed that in order for the tapered bevels to be effective around the midrange, they would need to run the full length of the cabinet, below the woofer. This would be exceedingly difficult for me to fabricate, so I elected to go with this design.
j.
Some progress… I am working on the stands. These are white oak, and will contain the Hypex FA253 amps. As with my other designs, the speakers will bolt to the stands, and wiring will enter the speakers through the bottom. This modular concept allows me to build multiple speakers which all utilize the same Hypex amplifier.
I made the notch filters for the M74A midrange driver. These are small enough to fit into the midrange hole in the baffle. They will be wired in series the mid. I am sure that @wolf_teeth could have shrunk these down to 6mm x 6mm 🙂, but this is compact enough for my purposes...
I have been studying the M74A diffraction situation. As discussed in the post above, I am measuring a bit more diffraction between 1000 – 2000 Hz with the finished speaker than I measured with the prototype Post 257 . The prototype had a completely flush transition between the mid-tweeter baffle and the transition block. The finished cabinet has a 3 mm step. Another difference is the prototype cabinet has no woofer or tweeter, so the midrange has no opportunities to diffract off those drivers, whereas with the final cabinet the drivers will diffract off each other. It is possible that some of the diffraction differences between the two measurements are due to the other drivers, in addition to the 3 mm step.
Here is a comparison of the on-axis M74A response in the final cabinet (solid) and the prototype cabinet (dashed).
I have been running some VituixCad sims using the measured prototype M74A data in place of the final cabinet M74A data. What I am showing below is the best optimization for each situation. This is filter data I would use to start the subjective voicing process.
Because I built the transition block from solid cherry, it would be possible to modify in situ to remove the 3 mm step. I would use an 80 grit sanding block to transform the abrupt step into a gradual ramp, then I would carefully refinish the area. However, I am not convinced this effort would be justified. Both of the above sims look quite nice, and most of the differences are just that, differences. The M74A prototype sim does offer a very slight improvement in the look of the on-axis graph, in terms of sound power and PIR, I don’t prefer one over the other. The differences are within the precision of my measurements.
If I do make the modification, there is no going back. That is another aspect I need to consider.
j.
I made the notch filters for the M74A midrange driver. These are small enough to fit into the midrange hole in the baffle. They will be wired in series the mid. I am sure that @wolf_teeth could have shrunk these down to 6mm x 6mm 🙂, but this is compact enough for my purposes...
I have been studying the M74A diffraction situation. As discussed in the post above, I am measuring a bit more diffraction between 1000 – 2000 Hz with the finished speaker than I measured with the prototype Post 257 . The prototype had a completely flush transition between the mid-tweeter baffle and the transition block. The finished cabinet has a 3 mm step. Another difference is the prototype cabinet has no woofer or tweeter, so the midrange has no opportunities to diffract off those drivers, whereas with the final cabinet the drivers will diffract off each other. It is possible that some of the diffraction differences between the two measurements are due to the other drivers, in addition to the 3 mm step.
Here is a comparison of the on-axis M74A response in the final cabinet (solid) and the prototype cabinet (dashed).
I have been running some VituixCad sims using the measured prototype M74A data in place of the final cabinet M74A data. What I am showing below is the best optimization for each situation. This is filter data I would use to start the subjective voicing process.
Because I built the transition block from solid cherry, it would be possible to modify in situ to remove the 3 mm step. I would use an 80 grit sanding block to transform the abrupt step into a gradual ramp, then I would carefully refinish the area. However, I am not convinced this effort would be justified. Both of the above sims look quite nice, and most of the differences are just that, differences. The M74A prototype sim does offer a very slight improvement in the look of the on-axis graph, in terms of sound power and PIR, I don’t prefer one over the other. The differences are within the precision of my measurements.
If I do make the modification, there is no going back. That is another aspect I need to consider.
j.