Another simulating adventure - the Acoustic Reality option
I noticed that PCD7 comes with an example template for the ARSXO series crossover design. I have used that layout before with success and it has a relatively low parts count so I thought I would give it a try on this MTM project. The template uses 4th order Bessel target slopes so that should be fine for this case where the mids should filter down fairly steeply.
It seemed impossible initially when I was trying to use values for the main cap and resistor that are close to the example specs, but when I started dialing in numbers you would usually think were far out of range it started to work. It turns out that if you aren't going to use a coil on the wfr with a value intended to provide BSC then things are very different from the typical design.
I found that a coil of about 0.5 mH and a cap of 35-40 uF with a resistor of 1-2 ohms makes a nice match to the target slope. The tweeter just needs a 0.07 mH coil. Things are improved with a zobel of sorts on the woofers, but the roll-off is still shallows out up towards 8 kHz. I have attached screenshots from PCD for this case called "Basic ARSXO".
Adding a cap in parallel to the zobel produces a more complete roll-off curve, however the wfr-twtr phases won't align with that circuit. Eventually I found that a resistor and a cap both in parallel to the zobel on the woofer will produce a nice deep roll-off and still allow a good phase match to occur. It is pretty sensitive to the values of all the components at that point but it can be done with a minimum number of coils. I would guess those extra caps in parallel can be NPEs, or maybe NPEs combined with small value film caps.
I have attached screenshots for this case called "Mod ARSXO".
It is interesting how both of these circuits produce impedance profiles that are quite flat end-to-end, but the impedance is lowish in the 1.8-4 ohm range. If that is a workable impedance profile for a solid state amp then this circuit just might be worth a try.
1.) A question in my mind is how close does the phase alignment have to be for good results? In the "Basic" case it is dead on, but it is a few degrees off but still nicely parallel in the "Mod" case. The trade-off being weighed there is phase alignment vs low pass filter depth. I'm guessing that for an MTM advantage goes to the deep roll-off, but maybe someone in the audience has experience that says otherwise.
2.) Another general question I have is - Why does the woofer phase curve suddenly turn up at a certain point, 3 kHz in this case? I don't understand the physical reason for that. Is it inherent to the drivers? Is it maybe related to baffle width?
I noticed that PCD7 comes with an example template for the ARSXO series crossover design. I have used that layout before with success and it has a relatively low parts count so I thought I would give it a try on this MTM project. The template uses 4th order Bessel target slopes so that should be fine for this case where the mids should filter down fairly steeply.
It seemed impossible initially when I was trying to use values for the main cap and resistor that are close to the example specs, but when I started dialing in numbers you would usually think were far out of range it started to work. It turns out that if you aren't going to use a coil on the wfr with a value intended to provide BSC then things are very different from the typical design.
I found that a coil of about 0.5 mH and a cap of 35-40 uF with a resistor of 1-2 ohms makes a nice match to the target slope. The tweeter just needs a 0.07 mH coil. Things are improved with a zobel of sorts on the woofers, but the roll-off is still shallows out up towards 8 kHz. I have attached screenshots from PCD for this case called "Basic ARSXO".
Adding a cap in parallel to the zobel produces a more complete roll-off curve, however the wfr-twtr phases won't align with that circuit. Eventually I found that a resistor and a cap both in parallel to the zobel on the woofer will produce a nice deep roll-off and still allow a good phase match to occur. It is pretty sensitive to the values of all the components at that point but it can be done with a minimum number of coils. I would guess those extra caps in parallel can be NPEs, or maybe NPEs combined with small value film caps.
I have attached screenshots for this case called "Mod ARSXO".
It is interesting how both of these circuits produce impedance profiles that are quite flat end-to-end, but the impedance is lowish in the 1.8-4 ohm range. If that is a workable impedance profile for a solid state amp then this circuit just might be worth a try.
1.) A question in my mind is how close does the phase alignment have to be for good results? In the "Basic" case it is dead on, but it is a few degrees off but still nicely parallel in the "Mod" case. The trade-off being weighed there is phase alignment vs low pass filter depth. I'm guessing that for an MTM advantage goes to the deep roll-off, but maybe someone in the audience has experience that says otherwise.
2.) Another general question I have is - Why does the woofer phase curve suddenly turn up at a certain point, 3 kHz in this case? I don't understand the physical reason for that. Is it inherent to the drivers? Is it maybe related to baffle width?
Attachments
The knee in the FE83 phase curve - further investigation
Since no one has offered any suggestions yet I decided to try another PCD simulation to address the question #2 above. This seems like a plausible approach since Fountek makes a twin Sister to the FE83, the FE85. It has almost the identical Fs and sensitivity and maybe the same cone or at least a very similar cone. I posted the FR curves of the two drivers earlier in the thread so you can scroll up and compare them. The main difference is that the FE85 has the smoother FR curve, whereas the FE83 has a rise at about 2kHz and then drops rather suddenly at about 3kHz. Which just happens to be where the phase curve suddenly breaks upwards. So I'm thinking that maybe this rise and drop is caused by some cone resonance problem and as such it might affect the phase fairly strongly. What if I modeled the smoother FE85 in this plan in place of the FE83?
Well, it turns out that if I do that the phase curve of the FE85s does not suddenly turn up at 3kHz and you can achieve good phase tracking between the woofs and tweeter all the way to 5kHz and then it more slowly begins to turn up. This works the same whether you put one or two drivers in the simulation, for either the FE83 or the FE85, so it is not an artifact of comb interference. The problem seems to be inherent to the driver.
On a related note - I made the first measurements of the frequency response of the FE83 using my low-tech home made mic over the weekend. This driver actually sounds kind of nasty during the chirp that FuzzMeasure runs through it. It seems to be mostly due to a strong cone break-up peak at about 12kHz. The peak is about 15db tall as I measured it! (OK, I was using a rather loud signal that might have driven Manuel Noriega out of the building). The rise from 2-3kHz that you see in the OEM curve was more pronounced in my measurement. This is clearly going to need some work, so while I had it strapped to the board I thought I'd see what I could do to make it talk more clearly.
I got out the dental instruments
eek and selected a pick with a smooth curve on one side. I used that to emboss some ridges into the cone at 4 points around the circumference from just inside the rubber surround about 1/2 way down towards the dust cap. This did change the shape of the 2-3kHz hump, rolling off the 3kHz side some so that the hump was more triangular than rectangular, and it changed the cone break-up. Before the high pitch could easily be measured 30 degs off-axis, and afterwards it was almost gone at that angle. Still a big on-axis peak. The sound also changed to something a bit less piercing. It looks like this simple means of stiffening the cone without adding any mass has some promise. Maybe that combined with some bits of carbon fiber glued on can be used to get this thing to stop screaming. I'm thinking if I succeed in suppressing or moving the big break up peak the 2-3 kHz hump might also be reduced and I might get rid of the phase anomaly. Then this bargain driver should sound a lot better and be easier to work into an affordable crossover.
Since no one has offered any suggestions yet I decided to try another PCD simulation to address the question #2 above. This seems like a plausible approach since Fountek makes a twin Sister to the FE83, the FE85. It has almost the identical Fs and sensitivity and maybe the same cone or at least a very similar cone. I posted the FR curves of the two drivers earlier in the thread so you can scroll up and compare them. The main difference is that the FE85 has the smoother FR curve, whereas the FE83 has a rise at about 2kHz and then drops rather suddenly at about 3kHz. Which just happens to be where the phase curve suddenly breaks upwards. So I'm thinking that maybe this rise and drop is caused by some cone resonance problem and as such it might affect the phase fairly strongly. What if I modeled the smoother FE85 in this plan in place of the FE83?
Well, it turns out that if I do that the phase curve of the FE85s does not suddenly turn up at 3kHz and you can achieve good phase tracking between the woofs and tweeter all the way to 5kHz and then it more slowly begins to turn up. This works the same whether you put one or two drivers in the simulation, for either the FE83 or the FE85, so it is not an artifact of comb interference. The problem seems to be inherent to the driver.
On a related note - I made the first measurements of the frequency response of the FE83 using my low-tech home made mic over the weekend. This driver actually sounds kind of nasty during the chirp that FuzzMeasure runs through it. It seems to be mostly due to a strong cone break-up peak at about 12kHz. The peak is about 15db tall as I measured it! (OK, I was using a rather loud signal that might have driven Manuel Noriega out of the building). The rise from 2-3kHz that you see in the OEM curve was more pronounced in my measurement. This is clearly going to need some work, so while I had it strapped to the board I thought I'd see what I could do to make it talk more clearly.
I got out the dental instruments
eek and selected a pick with a smooth curve on one side. I used that to emboss some ridges into the cone at 4 points around the circumference from just inside the rubber surround about 1/2 way down towards the dust cap. This did change the shape of the 2-3kHz hump, rolling off the 3kHz side some so that the hump was more triangular than rectangular, and it changed the cone break-up. Before the high pitch could easily be measured 30 degs off-axis, and afterwards it was almost gone at that angle. Still a big on-axis peak. The sound also changed to something a bit less piercing. It looks like this simple means of stiffening the cone without adding any mass has some promise. Maybe that combined with some bits of carbon fiber glued on can be used to get this thing to stop screaming. I'm thinking if I succeed in suppressing or moving the big break up peak the 2-3 kHz hump might also be reduced and I might get rid of the phase anomaly. Then this bargain driver should sound a lot better and be easier to work into an affordable crossover.I believe you've been using RM to extract minimum phase from the drivers' frd and zma files. That should tell you that phase is simply a mathematical function of the FR and impedance responses respectively.
So what you're seeing is a result of the FR and if you can change that by altering the cone, then when you extract minimum phase again, you'll see a corresponding change in the phase as well.
Dental instruments to those delicate cones -. You're braver than I am!
So what you're seeing is a result of the FR and if you can change that by altering the cone, then when you extract minimum phase again, you'll see a corresponding change in the phase as well.
Dental instruments to those delicate cones -
. You're braver than I am!A driver having a continuously rising phase response doesn't exist in the real world.
If you measure the phase of a driver at a distance of 1 meter then the actual phase response is completely unreadable, because the phase shift due to the delay of 2.9 ms is very much larger than the phase shift caused by the driver's frequency response. Hence it's necessary to subtract the phase shift of the delay from the measurement in order to get the desired result. If you now overdo this (ie if you subtract more delay than actually is) then you get a phase response which begins to rise at a certain frequency.
This makes it sound like explanation for the different phase curves that the Hilbert transform extracts from the frequency response graphs of the two Fountek drivers I have compared is that Fountek may have done the measurements on the two drivers, FE83 and FE85, differently.
Seems unlikely, but I suppose it is a formal possibility.
The point you raised implies that one should only try to extract phase information from near field measurements that you might make yourself. I will keep that in mind.
What is special about the 3rd order BW X/O in an MTM?
I have been trying to answer this question by simulating in PCD7. I am finding something interesting.
I tried the following crossover plans with the same set of inputs, which are the Vifa TC9FD and the Vifa OT19 tweeter, where two 8-ohm TC9FDs are in parallel in an MTM. All models use the same Y-axis driver spacing and 2mm of Z-axis offset, which is to say very close to perfect mechanical alignment of acoustic centers. 2mm just to allow for not getting the AC lined up exactly perfectly in practice.
1. 3rd order Butterworth target acoustic slopes for both drivers, normal polarity.
2. 4th order Bessel target acoustic slopes on both drivers, whichever polarity gives best phase alignment..
3. 4th order Linkwitz-Reilly acoustic slopes for both drivers, whichever polarity gives best phase alignment.
All of the above were achieved using 3rd order electrical filters. Some can also be achieved reasonably well with 2nd order electrical on the tweeter. The observation I make works out the same either way.
That observation is this:
If I align the crossovers to get a near-perfect phase match between Ms and T I can get nice smooth flat frequency response that looks great on "paper" (screen). However, the power response always droops around the crossover point, a bit more to the lower side than to the upper side. I see that same thing with pretty much every crossover design I try to fiddle with.
However, with the BW3 crossover, if I keep it aligned so that as much of the phase curve as possible maintains a 90 degree separation between the M's and T then I get a power response which exactly follows the FR curve for as far as that 90 degree separation holds. At the extreme ends where one driver or the other is filtered down to a low level it doesn't matter what the relationship of phase curves is any more. Of course.
Is this essentially the "magic" of the original D'Appolito array configuration?
A speaker that has flat and smooth frequency response and flat and smooth power response is supposed to be the overall design goal, as I understand it.
OTOH it certainly seems to make sense that when two drivers are both operating at the same frequency and db level they would sound best if in phase.
My questions -
1. Does the MTM with perfect power response, achieved using the BW3 XO, really sound good, or does the odd phase relationship cause some sort of coloration?
2. Is there some way to get an in-phase crossover to maintain a power response that is coincident with the frequency response? I can't seem to do it and I don't understand why that should be so difficult.
I have been trying to answer this question by simulating in PCD7. I am finding something interesting.
I tried the following crossover plans with the same set of inputs, which are the Vifa TC9FD and the Vifa OT19 tweeter, where two 8-ohm TC9FDs are in parallel in an MTM. All models use the same Y-axis driver spacing and 2mm of Z-axis offset, which is to say very close to perfect mechanical alignment of acoustic centers. 2mm just to allow for not getting the AC lined up exactly perfectly in practice.
1. 3rd order Butterworth target acoustic slopes for both drivers, normal polarity.
2. 4th order Bessel target acoustic slopes on both drivers, whichever polarity gives best phase alignment..
3. 4th order Linkwitz-Reilly acoustic slopes for both drivers, whichever polarity gives best phase alignment.
All of the above were achieved using 3rd order electrical filters. Some can also be achieved reasonably well with 2nd order electrical on the tweeter. The observation I make works out the same either way.
That observation is this:
If I align the crossovers to get a near-perfect phase match between Ms and T I can get nice smooth flat frequency response that looks great on "paper" (screen). However, the power response always droops around the crossover point, a bit more to the lower side than to the upper side. I see that same thing with pretty much every crossover design I try to fiddle with.
However, with the BW3 crossover, if I keep it aligned so that as much of the phase curve as possible maintains a 90 degree separation between the M's and T then I get a power response which exactly follows the FR curve for as far as that 90 degree separation holds. At the extreme ends where one driver or the other is filtered down to a low level it doesn't matter what the relationship of phase curves is any more. Of course.
Is this essentially the "magic" of the original D'Appolito array configuration?
A speaker that has flat and smooth frequency response and flat and smooth power response is supposed to be the overall design goal, as I understand it.
OTOH it certainly seems to make sense that when two drivers are both operating at the same frequency and db level they would sound best if in phase.
My questions -
1. Does the MTM with perfect power response, achieved using the BW3 XO, really sound good, or does the odd phase relationship cause some sort of coloration?
2. Is there some way to get an in-phase crossover to maintain a power response that is coincident with the frequency response? I can't seem to do it and I don't understand why that should be so difficult.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.