You've got some excellent ideas here Bill, really like the idea of having
vendor's drivers accessed right within the program.
There are baffle response calculators out there, perhaps one of the authors
can provide a script for standard FRD in, on specified baffle response out.
User provides baffle dimensions, square, rounded, beveled edge, etc.
Or perhaps as a plug-in to your program.
Pete, thanks. Jeff Bagby has given me code and tips on baffle simulation, but I am dreadfully behind on implementing something like that. I'm stuck mostly on the user interface, it has to be VERY simple to operate, without users having to read any notes, or else it won't get much use.
I'm also behind on getting vendors on board for access to component and driver information. I hope to get back with Parts Express before too long -- it's not nearly as easy as it was meeting with them when I lived about 15 minutes away from them!
Bill
I'm also behind on getting vendors on board for access to component and driver information. I hope to get back with Parts Express before too long -- it's not nearly as easy as it was meeting with them when I lived about 15 minutes away from them!
Bill
Hi Bill,
I've been simulating a design with a notch at 9 KHz and it is difficult to
see what is going on in the stop band because of the 20KHz upper limit.
I found where to change the upper and lower plotting freqs but it seems
to be limited to 20K or did I miss something? Can we get 50 or say 100K?
Thanks!
I've been simulating a design with a notch at 9 KHz and it is difficult to
see what is going on in the stop band because of the 20KHz upper limit.
I found where to change the upper and lower plotting freqs but it seems
to be limited to 20K or did I miss something? Can we get 50 or say 100K?
Thanks!
Hi Bill,
I want to make sure that I'm understanding how your programs works, is it
correct that we should think of your current model as treating the drivers as
point sources, with the mod delay parameter used to account for difference
in acoustic centers (with an FRD that has the time of flight delay removed)?
So we have the Z dimension handled by the "mod delay".
Then, if we want to model the delay due to drivers having non-coincident X,Y
positions on the baffle, assuming our observation point is 1m on axis for the
tweeter (could be any other driver or position relative to the baffle) then we
have to compute the additional delay for the path length being longer to the
observation point and add it into the mod delay parameter that we enter?
It would be nice to be able to specify X,Y positions for the drivers, perhaps
relative to the tweeter and have the program add delay in the same way so
that it does not have to be done outside of the program. The lower cost
version of CALSOD works this way IIRC but it also adds some flat attenuation
for increased path length difference IIRC.
This would be a first good step for improving the program. Next would be to
model the drivers as flat pistons of a specified diameter, and use the off axis
angle to modify the driver's response - this mode being an option of course.
This is how the Pro version of CALSOD works IIRC. Obviously, there is always
the option of using actual measured, on the actual baffle at the observation
point FRDs but these features are good to have during the exploration phase
of design.
I want to make sure that I'm understanding how your programs works, is it
correct that we should think of your current model as treating the drivers as
point sources, with the mod delay parameter used to account for difference
in acoustic centers (with an FRD that has the time of flight delay removed)?
So we have the Z dimension handled by the "mod delay".
Then, if we want to model the delay due to drivers having non-coincident X,Y
positions on the baffle, assuming our observation point is 1m on axis for the
tweeter (could be any other driver or position relative to the baffle) then we
have to compute the additional delay for the path length being longer to the
observation point and add it into the mod delay parameter that we enter?
It would be nice to be able to specify X,Y positions for the drivers, perhaps
relative to the tweeter and have the program add delay in the same way so
that it does not have to be done outside of the program. The lower cost
version of CALSOD works this way IIRC but it also adds some flat attenuation
for increased path length difference IIRC.
This would be a first good step for improving the program. Next would be to
model the drivers as flat pistons of a specified diameter, and use the off axis
angle to modify the driver's response - this mode being an option of course.
This is how the Pro version of CALSOD works IIRC. Obviously, there is always
the option of using actual measured, on the actual baffle at the observation
point FRDs but these features are good to have during the exploration phase
of design.
Hi Pete,
As it currently stands, Xsim operates more or less as you say, it doesn't handle any baffle positioning other than differences in depth (distance from microphone). The intention, when I get back to programming on it, is for it to use x, y, z positions as well as measured off-axis response infinite baffle curve sets from each driver (interpolating between provided curves). When off-axis curves aren't available, there should be an option to generate simulated curves from piston diameters. At least that's the plan.
That still wouldn't be terribly accurate because of the baffle diffraction effects, floor bounce, ceiling bounce, etc. The plan is to also include baffle diffraction modeling (this part is particularly hampered by the need to come up with an intuitive way to set it all up in the user interface).
Sorry this isn't happening faster. Since retirement and moving, I've gotten distracted by a load of other things (like getting a listening room set up well), but I should be catching up on those before too long.
To date, I've mostly used it with measurements of drivers on the actual baffles, so accuracy is very good that way. I make some curve sets from different off-axis points to get an idea of directivity (just run a couple sets of Xsim in parallel). That doesn't help you decide where to put drivers on baffles or baffle shapes yet, though.
As it currently stands, Xsim operates more or less as you say, it doesn't handle any baffle positioning other than differences in depth (distance from microphone). The intention, when I get back to programming on it, is for it to use x, y, z positions as well as measured off-axis response infinite baffle curve sets from each driver (interpolating between provided curves). When off-axis curves aren't available, there should be an option to generate simulated curves from piston diameters. At least that's the plan.
That still wouldn't be terribly accurate because of the baffle diffraction effects, floor bounce, ceiling bounce, etc. The plan is to also include baffle diffraction modeling (this part is particularly hampered by the need to come up with an intuitive way to set it all up in the user interface).
Sorry this isn't happening faster. Since retirement and moving, I've gotten distracted by a load of other things (like getting a listening room set up well), but I should be catching up on those before too long.
To date, I've mostly used it with measurements of drivers on the actual baffles, so accuracy is very good that way. I make some curve sets from different off-axis points to get an idea of directivity (just run a couple sets of Xsim in parallel). That doesn't help you decide where to put drivers on baffles or baffle shapes yet, though.
I would be very pleased with just having it handle the X,Y positions on the baffle, and then allow a constant radius arc from whatever the observation point is to do polars in the vertical and horizontal. The reason for having horizontal is for say situations where for example you might have a TMW with the M being two Ms side by side or any other unusual driver arrangement. The X,Y could go right in the tune option for the drivers.
I would consider it good enough to leave baffle edge diffraction up to either touching up in the voicing phase, or taking actual measurements on the intended baffle and producing new FRDs. It would be nice to have a simple module that provides a simple shelving transfer function to either put in baffle loss, say for when the FRDs are IB and you want to design the XO for a finite sized baffle but it might as well be general to also go the other way.
This was available through the BDS module in CALSOD.
It would be really nice to have general transfer function modules also as in CALSOD.
The CALSOD manual is free and available here, you might want to take a look whenever
you get back to thinking about Xsim:
CALSOD user's manual
I would really like many of the modules that are available, BDS, BES, BUT, HLR, and the
ones for specifying impedance without a ZMA file, IVB, ICB, IMP.
One that I wish had another option is the XYZ module that specifies the observation point, there should have been an option to arc in the horizontal and vertical by a specified number of degrees from the XYZ observation point.
I don't like to rely on measured FRDs other than the published ones since by the time
I measure them on the actual baffle, I've bought the drivers and I can then tweek the
actual hardware rather than going back to simulation. By the time I'm at this point I just
want to finish the design. Certainly for difficult problems it is always nice to have the
option of going back to simulation, I just don't think I'd use it that often.
Anyone know of a driver to observation point path length calculator spreadsheet, I'd guess
that someone has done it. Thought I'd ask before doing one.
I would consider it good enough to leave baffle edge diffraction up to either touching up in the voicing phase, or taking actual measurements on the intended baffle and producing new FRDs. It would be nice to have a simple module that provides a simple shelving transfer function to either put in baffle loss, say for when the FRDs are IB and you want to design the XO for a finite sized baffle but it might as well be general to also go the other way.
This was available through the BDS module in CALSOD.
It would be really nice to have general transfer function modules also as in CALSOD.
The CALSOD manual is free and available here, you might want to take a look whenever
you get back to thinking about Xsim:
CALSOD user's manual
I would really like many of the modules that are available, BDS, BES, BUT, HLR, and the
ones for specifying impedance without a ZMA file, IVB, ICB, IMP.
One that I wish had another option is the XYZ module that specifies the observation point, there should have been an option to arc in the horizontal and vertical by a specified number of degrees from the XYZ observation point.
I don't like to rely on measured FRDs other than the published ones since by the time
I measure them on the actual baffle, I've bought the drivers and I can then tweek the
actual hardware rather than going back to simulation. By the time I'm at this point I just
want to finish the design. Certainly for difficult problems it is always nice to have the
option of going back to simulation, I just don't think I'd use it that often.
Anyone know of a driver to observation point path length calculator spreadsheet, I'd guess
that someone has done it. Thought I'd ask before doing one.
Last edited:
I have a gerenal question about crossover design. I don't have that big of box of crossover parts and I am trying to redo the crossover for a pair of DIY speakers I picked up a while ago. So, how close will the software get me to the best crossover I need for these speakers? Can I just go with what is modeled and buy those components? How exacting are these software programs?
It's a completely accurate modeler, within the restraints of computer math (no restraints, in effect) and the principle of garbage-in-->garbage-out.
IOW, if you have good data on the drivers, relative to each other and how they are mounted in the box, accuracy should be right on. If not, it can't be any more accurate than the data you give it.
Regards,
B
IOW, if you have good data on the drivers, relative to each other and how they are mounted in the box, accuracy should be right on. If not, it can't be any more accurate than the data you give it.
Regards,
B
To reinforce what Bill said, if you take FRD and ZMA measurements with the drivers mounted in the final enclosure, then add the crossover in a program like Xsim. Note that by measuring them in the box you _have_ the response of the drivers as influenced by baffle step, and baffle diffraction so you do not have to add it in the simulation program and it is completely accurate since it is actual. Then the program simply adds (cascades them with the driver FRD) the transfer function's of the crossover as loaded by the driver's impedance and vector sums them. Some have posted simulated in this way vs. built and the agreement was nearly dead on, as would be expected since the program just crunches the math.
If you get FRDs off the web, or trace them from data sheets then you are at the mercy of the quality of those measurements.
What is your speaker, was it the one posted on the AK forum?
Last edited:
Zaph shows excellent agreement between simulated and actual here:
Zaph|Audio - ZD5 - Scan Speak 15W8530K00 and Vifa XT25
Zaph|Audio - ZD5 - Scan Speak 15W8530K00 and Vifa XT25