Took my first swing at it. LR2s across the board. The woofer breakup peaks are still only 25db and 35db down though. Maybe try 3rd order on the woofer to squash it? Or does that do bad things in 3-way designs? Notch instead?
I'd leave it 2nd order on the woofer and add a parallel cap across the inductor for a bottomless notch centered around 2500 Hz. That might kill the peak with a <$2 part.
Everybody's ears are different but you might find that rising response above 7K kind of hissy. A simple solution could just be a small (like 0.05 mH) inductor between the current filter and the tweeter.
Is there a way to model the parallel cap in WinPCD?
The peak was to compensate for the narrow dispersion of the ring radiator. 7k is right around where the responses break away from each other off-axis on the factory graph. It seemed folks like to leave a little heat on-axis in this type of case for power response? IDK I'm just monkey-see-monkey-do on some of this stuff right now lol. Now that I look at it again. It probably is too much. I'll look into reducing it a bit at least.
OK nm I think I figured it out. You gotta trick WinPCD into doing the tank cap by setting no crossover and using the alternate series and parallel circuits.
There is - you set it up as a Parallel RLC notch filter in the section right after the initial filter values. Just set the R Ohms value to 9999. I can't remember what you use for the original L value in the Low Pass section. I think 0 since the inductor isn't technically being used in that position anymore. It's been a while since I used PCD, so someone please correct me if I got that wrong.
And something I learned from Wolf years ago - make sure you have some resistance in series with either your shunt cap in the LP filter itself or add a little after the cap in the notch filter. 4 ohms is probably enough. If you don't have any resistance, you may be shunting high frequencies right to ground and some amps don't like that.
I just tried to fire up WinPCD and it's generating an error. Can't install a newer version either!
Not all amplifiers dislike the circuit without the R present, but it can be an ultrasonic short as described. If I am using an NPE for the shunt cap, I feel the higher ESR will swamp the issue enough to not really become a problem.
Alright Here's what I've got now. 1.2uf (plus 4ohm) punches between the breakup peaks and drags them both almost 40db down. Probably about as good as that is going to get. Unless there are any other tricks?
I did some more massaging while I was at it and drooped the top end a db or two as well.
Also, this is my first time trying to splice in the nearfield woofer measurement so I'm not really used to looking at it this way. The splice is around 350hz. The woofer alignment naturally creates ~1.5db hump which might make it seem like there is enough BSC when there may not be. My best guess is I've only got ~2db BSC right now, ignoring the hump. I'll see if I can drop the m/t down a hare without falling in a rabbit hole.
Are there any modeling software[s] that will allow you to plug-in the room dimensions/volume and speaker positions to simulate an in-room-response using this data?
Hopefully someone will correct me if I'm wrong... But IF you have measured your woofer far field in the cabinet AND you merged far field and near field taking into account your baffle step loss (which you would in Jeff Bagby's program or VituixCAD) then your model should basically be flat down to the 40's or wherever it starts to roll off (assuming you also merged the port near field).
Alright last one tonight. I promise. Not sure I can do much better anyway. Got the circuits playing a little nicer together and added a small notch on the tweeter's 4k bump since it kept sticking out.
Now If the real thing comes out like this I'll be amazed. My track record has not been great, but I'm learning that I need to pay more attention to details when taking the measurements. Hopefully it pays off.
@a4eaudio said:
Hopefully someone will correct me if I'm wrong... But IF you have measured your woofer far field in the cabinet AND you merged far field and near field taking into account your baffle step loss (which you would in Jeff Bagby's program or VituixCAD) then your model should basically be flat down to the 40's or wherever it starts to roll off (assuming you also merged the port near field).
Nope I'm probably doing it wrong. I just plonked the nearfield measurement onto the far gated response with FRDblender. In doing so I assumed I'd need to compensate for baffle step myself. I'm still in the stone age here.
@Steve_Lee said:
Are there any modeling software[s] that will allow you to plug-in the room dimensions/volume and speaker positions to simulate an in-room-response using this data?
Jeff Bagby's BDBS (Baffle Diffraction & Boundary Simulator) will do this. It has the Roy Allison power response equations built into one quadrant and the overall room pressure in another quadrant. You can sum these quadrants together with a baffle diffraction model of your speaker.
@a4eaudio said:
Hopefully someone will correct me if I'm wrong... But IF you have measured your woofer far field in the cabinet AND you merged far field and near field taking into account your baffle step loss (which you would in Jeff Bagby's program or VituixCAD) then your model should basically be flat down to the 40's or wherever it starts to roll off (assuming you also merged the port near field).
Nope I'm probably doing it wrong. I just plonked the nearfield measurement onto the far gated response with FRDblender. In doing so I assumed I'd need to compensate for baffle step myself. I'm still in the stone age here.
The port gets "summed" with the near plus far field in PCD or XSim.
The nearfield is spliced to show the baffle step drop from the far field, so your FR graph should be flat to the woofer.
@a4eaudio said:
Hopefully someone will correct me if I'm wrong... But IF you have measured your woofer far field in the cabinet AND you merged far field and near field taking into account your baffle step loss (which you would in Jeff Bagby's program or VituixCAD) then your model should basically be flat down to the 40's or wherever it starts to roll off (assuming you also merged the port near field).
You are correct David. On a side note I like to use the FRD Blender 2.0 that Charlie Laub and Jeff Bagby created.
First XO prototype as modeled in the last post above. With only one tweak so far of trying a resistor on the woofer shunt cap to soften the knee. I think the notch on the tweeter is a little more potent than modeled so I'll try a larger resistor value or maybe omitting the notch entirely if it doesn't prove necessary. No more tonight though.
So far this is the closest I've gotten a sim to replicate real life. So that is a win.
That spot in my room really murders the upper bass. It lines up perfectly with the ~130hz spike below so I'll write that off as an artifact.
I back burner'd this because I kept dickin with the prototype crossover and not being satisfied by hearing just the one. Gosh looking back at the measurement the more I'm thinking I should have just done a quick flying lead solder job and given some real listening time as-is to the pair. I was sweating the small stuff way too much. A db here and there is probably not something I can pick out and know what to do about for my skill/experience level. Not to mention I have yet to build a turntable to do accurate off-axis measurements.
Comments
Took my first swing at it. LR2s across the board. The woofer breakup peaks are still only 25db and 35db down though. Maybe try 3rd order on the woofer to squash it? Or does that do bad things in 3-way designs? Notch instead?
I'd leave it 2nd order on the woofer and add a parallel cap across the inductor for a bottomless notch centered around 2500 Hz. That might kill the peak with a <$2 part.
Everybody's ears are different but you might find that rising response above 7K kind of hissy. A simple solution could just be a small (like 0.05 mH) inductor between the current filter and the tweeter.
Is there a way to model the parallel cap in WinPCD?
The peak was to compensate for the narrow dispersion of the ring radiator. 7k is right around where the responses break away from each other off-axis on the factory graph. It seemed folks like to leave a little heat on-axis in this type of case for power response? IDK I'm just monkey-see-monkey-do on some of this stuff right now lol. Now that I look at it again. It probably is too much. I'll look into reducing it a bit at least.
OK nm I think I figured it out. You gotta trick WinPCD into doing the tank cap by setting no crossover and using the alternate series and parallel circuits.
There is - you set it up as a Parallel RLC notch filter in the section right after the initial filter values. Just set the R Ohms value to 9999. I can't remember what you use for the original L value in the Low Pass section. I think 0 since the inductor isn't technically being used in that position anymore. It's been a while since I used PCD, so someone please correct me if I got that wrong.
Yep, that is how I do it in Jeff's PCD, but I don't know if WinPCD has the same utility present.
InDIYana Event Website
And something I learned from Wolf years ago - make sure you have some resistance in series with either your shunt cap in the LP filter itself or add a little after the cap in the notch filter. 4 ohms is probably enough. If you don't have any resistance, you may be shunting high frequencies right to ground and some amps don't like that.
I just tried to fire up WinPCD and it's generating an error. Can't install a newer version either!
Not all amplifiers dislike the circuit without the R present, but it can be an ultrasonic short as described. If I am using an NPE for the shunt cap, I feel the higher ESR will swamp the issue enough to not really become a problem.
InDIYana Event Website
Alright Here's what I've got now. 1.2uf (plus 4ohm) punches between the breakup peaks and drags them both almost 40db down. Probably about as good as that is going to get. Unless there are any other tricks?
I did some more massaging while I was at it and drooped the top end a db or two as well.
Also, this is my first time trying to splice in the nearfield woofer measurement so I'm not really used to looking at it this way. The splice is around 350hz. The woofer alignment naturally creates ~1.5db hump which might make it seem like there is enough BSC when there may not be. My best guess is I've only got ~2db BSC right now, ignoring the hump. I'll see if I can drop the m/t down a hare without falling in a rabbit hole.
That looks perfect.
Are there any modeling software[s] that will allow you to plug-in the room dimensions/volume and speaker positions to simulate an in-room-response using this data?
Alright, ignoring the woofer bump, this looks more like ~4db of BSC if I'm interpreting this correctly.
Hopefully someone will correct me if I'm wrong... But IF you have measured your woofer far field in the cabinet AND you merged far field and near field taking into account your baffle step loss (which you would in Jeff Bagby's program or VituixCAD) then your model should basically be flat down to the 40's or wherever it starts to roll off (assuming you also merged the port near field).
Alright last one tonight. I promise. Not sure I can do much better anyway. Got the circuits playing a little nicer together and added a small notch on the tweeter's 4k bump since it kept sticking out.
Now If the real thing comes out like this I'll be amazed. My track record has not been great, but I'm learning that I need to pay more attention to details when taking the measurements. Hopefully it pays off.
Nope I'm probably doing it wrong. I just plonked the nearfield measurement onto the far gated response with FRDblender. In doing so I assumed I'd need to compensate for baffle step myself. I'm still in the stone age here.
Jeff Bagby's BDBS (Baffle Diffraction & Boundary Simulator) will do this. It has the Roy Allison power response equations built into one quadrant and the overall room pressure in another quadrant. You can sum these quadrants together with a baffle diffraction model of your speaker.
The port gets "summed" with the near plus far field in PCD or XSim.
The nearfield is spliced to show the baffle step drop from the far field, so your FR graph should be flat to the woofer.
You are correct David. On a side note I like to use the FRD Blender 2.0 that Charlie Laub and Jeff Bagby created.
First XO prototype as modeled in the last post above. With only one tweak so far of trying a resistor on the woofer shunt cap to soften the knee. I think the notch on the tweeter is a little more potent than modeled so I'll try a larger resistor value or maybe omitting the notch entirely if it doesn't prove necessary. No more tonight though.
So far this is the closest I've gotten a sim to replicate real life. So that is a win.
That spot in my room really murders the upper bass. It lines up perfectly with the ~130hz spike below so I'll write that off as an artifact.
I back burner'd this because I kept dickin with the prototype crossover and not being satisfied by hearing just the one. Gosh looking back at the measurement the more I'm thinking I should have just done a quick flying lead solder job and given some real listening time as-is to the pair. I was sweating the small stuff way too much. A db here and there is probably not something I can pick out and know what to do about for my skill/experience level. Not to mention I have yet to build a turntable to do accurate off-axis measurements.
Are you going to damp the outside of that cup? As is, it will pass through the sounds from the woofer.
Honestly, looking at your last measurement the only thing I would personally do is drop the top octave a bit and let 'er rip.
I guess I could try with what I have on hand
Tonight I messed around a bit with the last sim I was previously working on to get a sloped response.
(like before, ignore most of the BSC hump).