You need to have the far field response of each driver individually. Then for each driver, you will simulate the diffraction response for that driver's baffle, since each baffle is a different size.
When you do the merge for each driver, you will have a common near field response, but each driver will have a unique diffraction simulation and a unique far field response.
When you do the merge for each driver, you will have a common near field response, but each driver will have a unique diffraction simulation and a unique far field response.
Not quite, you only simulate the diffraction response with the nearfield measurements.
How far down that goes, depends on the time window you were able to take plus I always take the simulated f3 x 3 or 4 roughly as an additional point (or where the impedance goes to a minimum again).
It also depends where the simulated bafflestep will be.
To give some practical numbers, let's say that you lowest frequency with a time window is about 180Hz.
I would make the transition stitch between nearfield + baffle step simulation around 360Hz.
Although in the end it's mostly a bit of a feel thing where that transition is nice, even and smooth.
So no golden rules here.
With a 3-way system this can be a bit tricky, depending how low you want to cross things between the midrange and woofer.
You don't have to do the nearfield + bafflestep for the tweeter, since the far field measurement doesn't have the lower frequency issues with it.
The whole procedure is very well described in an appnote from Audio Precision;
(page 11 and further)
The issue here is how to use the near field responses of the 3 drivers, since they are placed in different sides of the box and the diffraction is different for each one of them.
One solution is to do as hifijim suggested above, but I would like to avoid measuring the polar response of each one driver, merging with each one near field response with its diffraction and then trying to find a good method to add the 3 responses correctly.
By measuring the far field response of all the 3 drivers together, I am are sure that I have the correct response at least 2 or 3 times the gating frequency.
The question is about the low frequencies response.
Any other thoughts?
One solution is to do as hifijim suggested above, but I would like to avoid measuring the polar response of each one driver, merging with each one near field response with its diffraction and then trying to find a good method to add the 3 responses correctly.
By measuring the far field response of all the 3 drivers together, I am are sure that I have the correct response at least 2 or 3 times the gating frequency.
The question is about the low frequencies response.
Any other thoughts?
Just measure this design on the your choosen reference/listening axis.
If they are all gonna do exactly the same in terms of frequency, I personally would just simulate the low-end first and see how that all adds up.
Below the bafflestep those three drivers will basically just act as one unit.
If you want to do it well, the downside of such systems is actually that you have to unfortunately measure the entire polar response.
In this case horizontally as well as vertically.
But again, I would first simulate it to get a better sense of what is going on and what to expect.
I don't really know what it's being used for?
Member
Joined 2003
3 drivers, same or different, measurement, diffraction, merge process is the same. If driver share same cabinet volume, nearfield must be completed with all 3 wires as intended for design, and adjust overall amplitude accordingly. If the drivers are are same, nearfield measurement in this case provides same result regardless of driver, so just one nearfield measurement is needed, 3 diffraction simulations for different baffle dimensions.
If you want to design with each individual driver's FR and diffraction, you will have to measure each separately FF.
Three responses. Merge three times with the the three BS-corrected NF (could be one NF measurement only if all are the same and share the volume).
Create a crossover with three drivers and position them accordingly (shows your drawing of the enclosure and drivers positions, assume all drivers on center panel horizontal axes):
Three responses. Merge three times with the the three BS-corrected NF (could be one NF measurement only if all are the same and share the volume).
Create a crossover with three drivers and position them accordingly (shows your drawing of the enclosure and drivers positions, assume all drivers on center panel horizontal axes):
- front: X,Y,Z,R,T = 0
- top : X=0, Y=+40, Z=+60, R=0, T=+90
- side : X=+100 (assumed, as per your drawing), Y=0, Z=+60, R=+90, T=0
Question: The AP appnote mentions M as main dimension (3d diagonal) of enclosure. For a bookshelf sizes i can imaginaire this, but for a tall floorstander (mine is 115x34x46cm) would imply more than 3.5 meter mic distance. Is this a necessity or would 1 meter suffice?
Yes, I agree.
But the question is:
Is there a way to manipulate the Near Field responses of the 3 drivers when performing the MERGE operation, if the measurement of the Far Field response includes all three drivers together?
Member
Joined 2003
I've already done that recommendation as measurement instructions for ARTA, CLIO and REW so there's not much to add or repeat. Requires just reading. Full design including multiple separate mid-woofers can be measured at once in large anechoic at 2 m. In that case near field measurements and Merger are not needed.
That kind of soundbar design is omni so unofficial answer is "no baffle loss" scaled by -6 dB.
That kind of soundbar design is omni so unofficial answer is "no baffle loss" scaled by -6 dB.
I don't think most people "just" have a large anechoic laying around?
With 3 drivers doing the same low end, wouldn't it be -9.5dB?
Last edited:
That's the reason why gdan should read and follow measurement instructions or messages of hifijim, Draki and Dcibel. Of course simultaneous measurement of all drivers is possible and has some advantages over acoustical summing in dual plane simulator, but different mechanical locations could/probably cause problems with limited time window especially when speaker is rotated 0...180...360 deg while measuring both planes.
Diffraction file reader in Merger tool assumes scaling of 0...+6 dB i.e. "baffle step" so omni radiator requires -6 dB scaling with 'No baffle loss' checked to be compatible with full space radiation. Number of connected drivers, electrical connection, input sensitivity, measurement voltage and radiating area need more scaling actions of course.
Hi kimmosto,
Thanks for your response.
I measured all 3 drivers together because, as you say, the acoustical summing in the simulator, is much more difficult and it depends a lot of the parameters (x, y, z, angles etc) which some times are not easy to measure and insert them with accuracy.
Thanks for your response.
I measured all 3 drivers together because, as you say, the acoustical summing in the simulator, is much more difficult and it depends a lot of the parameters (x, y, z, angles etc) which some times are not easy to measure and insert them with accuracy.
Hi,
A question, -when I have modelled the driver distances from traced responses, I have done this in relation to each other (two way speakers). For example, when the midbass och tweeter are mounted in the same baffle, the tweeter is tagged with a negative Z in the driver tab due to its closer proximity to the listener. The value I used was the relative difference between the midbass and the tweeter.
Q:
Is it better or any difference to tag both drivers in relation to the baffle?
For example, both drivers get a positive Z but the Z for the midbass is larger (ex: 2mm for the tweeter and 25mm för the midbass)
Last edited:
Member
Joined 2003
Z dimension by physical driver location would be correct process for measured data. However for minimum phase data (traced), you need to make some assumption of actual acoustic centre of the driver, it can be 20mm+ for a 6" midbass for example. Best to avoid traced data for any real design work due to too many inaccuracies and assumptions that have to be made.
Regarding the dimension relations, 0,0,0 would be straight in front of you, at listening distance set in the options. Generally this would be either tweeter axis, so tweeter is 0,0,0 and other driver coordinates are entered relative to the tweeter location. Alternatively, some prefer to use a location centred in between tweeter and midrange as 0,0,0, and driver coordinated are entered relative to this point.
Positive or negative Z coordinate is in direct relation to listening distance. For example, listening distance is 2000mm, and all drivers entered as Z = -500mm coordinate, will provide the same result as listening distance of 1500mm, and all drivers at Z = 0mm. I would use the driver closest to listener at 0, and other drivers are relative to this location.
Regarding the dimension relations, 0,0,0 would be straight in front of you, at listening distance set in the options. Generally this would be either tweeter axis, so tweeter is 0,0,0 and other driver coordinates are entered relative to the tweeter location. Alternatively, some prefer to use a location centred in between tweeter and midrange as 0,0,0, and driver coordinated are entered relative to this point.
Positive or negative Z coordinate is in direct relation to listening distance. For example, listening distance is 2000mm, and all drivers entered as Z = -500mm coordinate, will provide the same result as listening distance of 1500mm, and all drivers at Z = 0mm. I would use the driver closest to listener at 0, and other drivers are relative to this location.
I really don't understand the practical point of modelling from traced responses????
You have different measurement conditions - different size of the baffle, different location of the drivers on it, different volume of the box. Accordingly, you will have different responses.
You can just as effectively and accurately measure the distance from the moon to the earth.
You don't?
What about preliminary research to do some feasibility investigation?
To get some ideas what to expect and what problems and pitfalls there might be.
How certain responses (frequency as well as impedance) act on certain filters when passive filters are being used etc.
Absolute essential part of any proper and good design.
Is it accurate and usable for the real thing?
Obviously not, that's why it's called preliminary research.
When you gain more experience over the years, the less you have to do these things.
But I have done hundreds of designs in the last 10-15 years as a professional as well as hobby and there are sometimes still some small occasions where I just want to know a few things beforehand.
For beginners it's also a very useful tool to understand and learn the effects of certain choices, parameters and differences.
Way back in the day I spend hours and hours with Boxsim, these days people can use a lot more other programs like VituixCAD.
So calling it out so bluntly is something I don't understand.
Even less so when calling the entire thing BS.
What I do agree on, is that you can't go into exact details of the crossovers and the response.
For the reasons (and many others) you just mentioned.
So a proper answer to @revisorn , would be to make him aware that traced responses don't have the same reference point.
Therefor we have no idea how the phase is in relationship with each other (if any at all).
So depending on what he wants to do with it, we can give advise accordingly.
If the goal is making an actual practical filter out of it, that unfortunately is not going to work.
I say that with a tiny little asterisk, because if you have enough experience and the design is super standard, you can make it work as long as you have a microphone to verify or you have worked on a very similar design (with similar drivers).
Mostly because filters in those situations are almost always the same or very similar.
Obviously with an important note that this approach is FAR from te preferred method and NOT recommended by default!!
But sometimes you have to make things work and it's all you have at that moment.
Feasibility of?
Acoustic design is not dependent on the uniqueness of a driver response measurement, hence some don't even choose drivers until after the speaker is designed.