The reason that a nearfield measurement is "room response free" is that the relative SPL of the driver is much higher than the SPL of the room echoes when the mic is positioned very close to the driver cone. The room responses are still there, but are essentially insignificant, so you do not see their effect(s) on the frequency response. The baffle step diffraction "echoes" are not seen when the mic is very close (<1cm) to the dustcap for the very same reason.
First of all, Dave. Thank you for the endorsement. It is much appreciated.
The blender was Charlie's idea, and I added the diffraction model and box modeling - which could be used instead of near-field measurements if someone wanted a quick and dirty measurement. I tried it this way, and when compared with my NF measurement they were very close.
I wanted to write up the process in a way that anyone could follow and understand why the old way was flawed and how we can do it a bit better in easy steps. I hope the paper is useful for a wide audience this way.
I'd like to answer a couple of questions that were raised -
First, when measuring a driver with a phase plug I would still attempt to get the mic as close to the cone as possible even though now we will be offset to one side due to the presence of the plug. All that happens by no longer being concentric with the cone edge is that we lose a bit of useable bandwidth, but since we will be splicing / blending at a lower frequency anyway, so it really shouldn't impact us much. Being close to the cone would be more important than being centered in this case.
Second, although some programs have implemented a simple step to simulate the baffle diffraction step, this has issues too. Real diffraction is much more complex than a step, it also includes a peak just above the step frequency that can be 2-3 dB in amplitude. Being able to accurately simulate this peak along with the step is part of what allows for much better alignment between the near and far field data sets and allows for what I believe is a more accurate summed response. This, and the blending over a range, are what make our spreadsheet so useful.
Jeff
Next step is to produce response set including also off axis responses to calculate 'Power response approximation'. Plain on axis simulation or near field measurement with diffraction effect is not very usable for that when splicing frequency is above 200 Hz.
For example, the following off axis & power response approximation contains driver+box+diffraction simulation until 340 Hz, and windowed far field measurements above 340 Hz. Off axis responses at the low end are calculated from on axis response by subtracting difference between windowed off axis and on axis responses. This method is not extremely accurate, but extends at least some directivity information one...two octaves below splicing frequency.
Off axis responses are generated with diy application, designed to create ##extended data## -formatted response files for LspCAD 6. Program removes possible phase and magnitude steps at transition frequency.
For example, the following off axis & power response approximation contains driver+box+diffraction simulation until 340 Hz, and windowed far field measurements above 340 Hz. Off axis responses at the low end are calculated from on axis response by subtracting difference between windowed off axis and on axis responses. This method is not extremely accurate, but extends at least some directivity information one...two octaves below splicing frequency.
Off axis responses are generated with diy application, designed to create ##extended data## -formatted response files for LspCAD 6. Program removes possible phase and magnitude steps at transition frequency.
An externally hosted image should be here but it was not working when we last tested it.
Nice to hear you're getting settled in fine.
BTW I'm cringing at the thought of a Bose product that actually might be really good
I've an 27 year old RS 40-1197 fe103 look alike on the shelf, modified with a removable phase plug. Will have todo some tests. After doing a whole heap of on / off axis tests today started messing with plastaline clay (love the stuff) and made a few measurements with a 18mm ball attached to the center and low center of an Airborne RT4001 AMT. Simply wanted to find out first hand what the deal is with "phase plugs" (they aren't), very interesting results none the less. < put some perspective on the Mundorf AMT's with the varying slot widths. Want todo some tests using something absorptive than hard reflecting. Perhaps needled felt and or viscoelastic foam covered in microfiber. Multilayered approach seems best for minimum diffraction at a given frequency.
It's in my blood, can't help it
Mike
PS, would post up before / after ball plugging it, but sat back and thought, now just who in their right mind would actually go and do such a thang?
Very true but it is still good to have an accurate free field curve. The biggest advantage seems to be to properly know the bass level vs. the midrange and up level. Simple splicing methods tend to encourage you to elevate the bass level above its true level.
I know you can calculate the proper level based on the free field distance and the driver area, but people seldom do that.
Achieving an accurate portrayal of the 100 to 500 Hz region was always the most difficult problem and this seems to be a real improvement there.
David
Member
Joined 2009
Thank you for the explanation, Charlie!
I was falsely assuming that the intensity of a modal region is only the result of the physical distance to all sources (driver and reflective surfaces), while in fact the power of each converging wave is also a contributing factor. Moving towards a source affects not only the relative distance but also the relative power of converging waves.
The 4pi to 2pi correction will always tilt upwards to about 6dB higher in the upper frequencies. Since you splice at the top of the nearfield measurement you would splice at the levels of the higher section and the bass corner region would be inherently lower.
His examples illustrate the difference between the 2 methods.
David
^
The "old" method would produce this:
The "new" method would produce that:
The "new" method suggests that the lows need more boost hence there would be more low frequency energy radiated into the room compared to the "old" method.
The after allows a proper crossover to be designed to match the levels, eg no bass boost when completed from an accurate simulation. Example given looks like a good candidate for a 2.5way IMO.
I see what you are saying. (With the new method) you would predict the bass level is lower and compensate by EQing it louder. I was saying that, in the splicing process, and assuming that the upper range was fixed, that you would splice the bass curve at a lower relative level.
Potayto, potahto
David
Potayto, potahto
David
I have to admit that although I've done all the conventional close mic/gated mic measurement techniques in the past (learning from EE's at Tektronix and Dolby Labs), I now just put the mic where I'm going to sit, and look at pink noise, one driver at a time, and then one speaker cabinet at a time.
I run each driver separately to check for crossover accuracy and level matching, and I do move the mic around in the general listening area since high frequencies may have significant cancellation effects, but I generally don't bother with gating and fancy software any more. The room acoustics are usually much worse than any half way decent speaker. I end up calibrating the total system both electronically and acoustically. Then I use tone controls of my own design to make it even more enjoyable (custom loudness compensation). Knowing enough about acoustics to understand how to interpret the RTA display is important too.
Reasonably accurate diffraction modeling pretty much requires that you model one driver at a time. Each driver's center will be different distances from edge sources that it uses to calculate the diffraction from. If you want to do multiple drivers then you have to model each one individually and then do a vector summation of their responses.
ARTA just shows the direct SPL change when leaving halfspace, not the edge diffraction. Anyways, the method described here is about generating accurate response data over the widest range possible, for A driver. You can then use PCD or any number of programs to simulate the drivers running together. Some programs have their own diffraction simulation and whatnot, eg. boxsim, so you'd not be doing all the steps necessarily. Cat-skinning and all that. Just got to keep track of what you're doing and why.
P.S. Thanks for the write up guys.
P.S. Thanks for the write up guys.
Last edited:
Hi Jeff,
Thanks for the prompt response. Unfortunately, I think I may not have phrased my question correctly.
On page 12 of your white paper, it shows a graph entitled “Modeled baffle diffraction response” with a small graphic underneath and to the left entitled “coordinates”, which I assume means the coordinates of the driver mounted on the front baffle. With this in mind, if I have an MTM design that has, for example, 2 x 8” bass/mid-range drivers that are wired in parallel with both mounted symmetrically on the front baffle, how are these two drivers’ coordinates entered into the software? I ask this because it appears to only have provisions to enter coordinate data for one driver. Is there an alternative procedure for this?
Regards
Peter
Thanks for the prompt response. Unfortunately, I think I may not have phrased my question correctly.
On page 12 of your white paper, it shows a graph entitled “Modeled baffle diffraction response” with a small graphic underneath and to the left entitled “coordinates”, which I assume means the coordinates of the driver mounted on the front baffle. With this in mind, if I have an MTM design that has, for example, 2 x 8” bass/mid-range drivers that are wired in parallel with both mounted symmetrically on the front baffle, how are these two drivers’ coordinates entered into the software? I ask this because it appears to only have provisions to enter coordinate data for one driver. Is there an alternative procedure for this?
Regards
Peter
In a multidriver system each driver might have its own relationship to the cabinet edges and its own diffraction curve profile. In the particular case of an MTM with a symmetrical baffle (top to bottom and left to right) then both woofers have the same correction curve.
Combined response of the two woofers ends up being the response of one driver plus 6dB.
David
Combined response of the two woofers ends up being the response of one driver plus 6dB.
David
I did understand that this was what you were asking, and my answer remains the same. You calculate diffraction first by treating the driver as a point source on the baffle and then determine the length of multiple "rays" from that point to the various edges. You then calculate the diffracted response for each ray and then sum all of them together to approximate the driver's diffraction response. For multiple drivers you have to do this multiple times, because each may have its own unique diffraction.
However, if you had the forethought to make your baffle perfectly symmetrical, so that each woofer is exactly the same distance from the edges, then each woofer will have the identical diffraction response. You can calculate it for one, add that to the near-field, and then merge that with the far-field for both woofers together. You can then use this response data in your crossover modeling.
Jeff
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.