I think things are a little greyer than that statement sounds. There is no need to dispute Toole but there are certainly techniques that can be used to make the processing of in room measurements as a basis for correction in a way that I don't think Toole ever experienced. There are some speakers for whom anechoic measurements really don't help in getting the best sound from, that fall outside the standard box with increasing size cones and domes paradigm. I don't say this to suggest that in most situations anechoic measurements aren't the better choice, just not to completely discount everything else that can be done in their stead.
So same, same then.. Everything is a model of the physical world how you get there is not the point, it’s what the model describes and the validity of that data.
Ok, I hear you and will take your statement as fact, but could you educate me further on this?
Not sure I wholly understand your point but the near field scanner is taking a consistent measurement of the the far field response. That is, although it inevitably contains measurement errors they will reduce and converge on the far field response with increased resolution. This is important because it indicates the basis/foundation for the measurement is reliable. This is not necessarily the case for one or two other suggestions in the thread.
As far as I can see the measurement technique is both valid and of the far field. The only problems I can see is that it is far too expensive for DIY use at home and should a DIYer wish to pay substantial sums to get an accurate/reliable/impartial/legal/... set of measurements doing so in the far field is likely to be preferable.
Just to make sure I get your idea:
- Measure the impulse response as usual (without the absorber in place).
- Measure the impulse response with the absorber between the speaker and the microphone (the direct sound from the speaker will be attenuated once it reaches the microphone; whereas the room echoes will not be affected much by the absorber, except the ones that come from behind the speaker).
- Determine the room echoes from the "difference" of measurements (1) and (2), whereby this will require a bit more than just subtracting (2) from (1). Something correlation method will be required.
- Subtract the room eoches from measurement (1) to get an anechoic impulse respone (or, at least reduce the echoes in the data).
Sorry if I was vague. I am not disagreeing with what you are saying in principle.
Last edited:
If I actually do it sure, I have plenty of fibreglass but I'll need to find the right materials that have increasing gas flow resistivity for the rest 🙂
Not really although I suppose that could be tested. The point was to make as anechoic a floor as possible (in between the speaker and mic) at least down to about 150 or 200Hz so the gate time could be increased in whatever size space there is for measuring.
That could be used with the beamforming technique as it works better with reduced reflections
This doesn't sound like my understanding of what the NFS is doing, the Part 3 document in the thread linked below has a good explanation of Sound Field Separation where the the reflections can be determined by working out what comes from outside the measurement plane and what comes from inside. The near field measurements can them be computed to far field data through the Helmholtz equation and Spherical harmonics, Klippel might vary due to using dual cylinders rather than dual sphere's. Part 2 explains this. These two documents are part of a series made by NTK at ASR.
https://www.audiosciencereview.com/...hematics-and-everything-else.9970/post-269403
Good to see this sort of thing on ASR. There was none when I gave up on the place a few years back. Possibly worth another look?
There are a range of ways of presenting how to get from the near field to the far based on putting the details of the methodology to the fore but underneath they are all required by what is physically and mathematically valid to involve a surface enclosing the sources and knowing the pressure and particle velocity on that surface. I was saying nothing about the details because I haven't looked them up for either the commercial scanner or a DIY attempt. The link you provided involves ways of estimating the room contribution which I was not intending to include in my comment about evaluating the near field in order to get the far field which has been a fairly standard technique not only in acoustics for 150 years or so. So I think we may have got our lines crossed a bit.
From what I understood from the initial documents that FM NTK has authored, the technique used in the nearfield scanner (atleast his version of it) is to express the acoustic pressure field as a weighted linear combination (infinite sum) of spherical harmonics. The infinite summation may be truncated to a finite summation in the appropriate regime of frequencies where the error involved is within acceptable limits. In common mathematical terms, we are trying to do a basis expansion. Given that we have fixed the basis functions, we now need the information to compute the weighing coefficients. This information is obtained via aplying appropriate transfromation on the the high SNR nearfield SPL measurements.
Point to note here is that the nearfield measurements are used to compute the weighing coefficients (with relatively lesser error) for the basis expansion which is very different from our regular use of it to plot the nearfield SPL vs frequency plots. Here multiple nearfield measurements are used to understand how the waves at different frequencies try to expand into surrounding 3D space (and hence sort of the entire 3D balloon of SPL variation vs frequency) instead of getting that single dimensional SPL vs frequency plot that we normally generate with measurement at a single point close to the driver).
Point to note here is that the nearfield measurements are used to compute the weighing coefficients (with relatively lesser error) for the basis expansion which is very different from our regular use of it to plot the nearfield SPL vs frequency plots. Here multiple nearfield measurements are used to understand how the waves at different frequencies try to expand into surrounding 3D space (and hence sort of the entire 3D balloon of SPL variation vs frequency) instead of getting that single dimensional SPL vs frequency plot that we normally generate with measurement at a single point close to the driver).
You seem to be stating that all the things the Klippel NFS analyzes well are desirable, but also arguing that the hardware is not necessary. It feels kind of like saying no one needs airplanes because people can drive or take a boat wherever they want.
The Klippel system generally gives more accurate data down to lower frequencies for full radiation than anything except very involved measurement methods or anechoic chambers. It does this at far lower cost and with greater ease than truly competing methods. That's what it's for. The price doesn't have to make sense for the vast majority of individuals.
Erin's tests and commentary about professionally designed speakers that don't measure well also seem to indicate that measuring accurately without a Klippel is not something that's easy or done as often as most would like. If Klippel makes it easier for manufacturers to achieve flat on-axis frequency response and smoothly trending directivity, then it does help the end user. Or if it causes some manufacturers to try harder to achieve this, that's also good.
The idea that it's easy to EQ a room properly seems to be refuted by the Toole paper "The Measurement and Calibration of Sound Reproducing Systems." Many also seem to agree that there are limits to how far you can push in-room equalization (given how it's typically done now) before you start causing side effects. Starting with more optimal speakers helps.
Exactly what I've been doing for over *20 years since I purchased the Speakerlab-bundle (Madisound forum).
*well, those rare times when I want loudspeaker IR data.
Center (tweeter) loudspeaker in a room free from objects (..place loudspeaker on a modest platform to get the tweeter vertically mid-point in room).
Pile many of the pillows in my home between the mic and all around the loudspeaker.
Start test.
I do have access to a moderate size room for this though (13/20/8 ft) where I can pull-out everything and place it in my hallway, and not everyone is going to have this luxury - those that don't will have to "window" lower in the IR (higher in freq.) with the attendant loss in resolution at lower freq.s for this "gated" result.
I don't think TNT is arguing with any of the above.
He's basically saying you don't need that sort of accuracy to get a decent or even good result: "Drive a car because you can afford it and it actually works to get you somewhere - you don't need a plane" ..to carry the analogy.
You do however need enough polar data to achieve a very good result, and that can be done with standard techniques applied well.
Did he say EQ the room?
I believe he means the polar response of the loudspeaker needs to be uniform to get a reasonably uniform result in-room. That the loudspeakers freq. response is itself not linear, (perhaps +/3db), is for most (with DSP receivers) largely irrelevant (barring major deviation).
This of course all goes back to getting a good polar measurement at higher freq.s (assuming the design is largely omni at lower freq.s - like most loudspeakers).
Last edited:
How are you going to get accurate polar data below 300 Hz (or even higher) with standard measurement techniques in typical environments? Even frequency response in this range can be difficult to do well.
I don't get the "hate" for Klippel.
You have a guy that's measured more speakers than most of us ever will saying the improvement provided by his NFS purchase was large, but a bunch of other people who've never touched one saying it doesn't matter. I find it odd.
https://www.diyaudio.com/community/...u-pay-for-klippel-service.390177/post-7122903
I don't get the "hate" for Klippel.
You have a guy that's measured more speakers than most of us ever will saying the improvement provided by his NFS purchase was large, but a bunch of other people who've never touched one saying it doesn't matter. I find it odd.
https://www.diyaudio.com/community/...u-pay-for-klippel-service.390177/post-7122903
Last edited:
Is far as I read it, there is no hate for the Klippel. Most here seem to agree for 20-200 Hz NF is just fine, and so is the range above 1000 Hz with windowing. What is left is the problematic 200-1000 Hz, yes. But accuracy in that range, plus the convenience, speed and off axis accuracy in general come at a very high price: 100K.
As I read the thread, there have been some questions about the relevance of high accuracy in that range, given the large impact of rooms in that range plus the fact the usual suspects are pretty omni at 1000Hz. I am not sure the jury is out yet: a daft designer will still screw up with the help of the NFS measurements.
As I read the thread, there have been some questions about the relevance of high accuracy in that range, given the large impact of rooms in that range plus the fact the usual suspects are pretty omni at 1000Hz. I am not sure the jury is out yet: a daft designer will still screw up with the help of the NFS measurements.
By knowing how sound radiates at low frequencies?
Hate? It is a cracking bit of kit that I'm fairly sure everyone would like to have around to use but, given a typical hobby DIY budget, one that is several orders of magnitude too expensive to be of relevance given the available alternatives and applying a bit of knowledge/common sense.
We may be starting to talk past each other. I think we all agree that Klippel NFS data has value, but we disagree substantially on how much value. It its current form, Klippel NFS is unobtainable for most of us. But the idea that some of us would pay $500 for the data, while others are going to the substantial effort of engineering a DIY-equivalent, has left others questioning our sanity, or at least questioning our perspective. Those on the other side of the fence see the Kippel NFS data as possibly useful, but hardly crucial... more of a nice to have, and certainly not worth paying hundreds of dollars for.
I think the root of this disagreement is how we look at the scope of our problem...
If we are focused on the next speaker project, we probably won't see much value. In my first graphic I show the typical phases of designing and building a new speaker. From this perspective, NFS data becomes available after the cabinet is constructed. At this point everything about the speaker has already been fixed except the crossover design... so from this perspective, the question is "Will the NFS data allow me to make a better crossover?" If my baffle step assumptions/calculations are off by a few dB, the NFS data will help me. But honestly, a well performed outdoor ground plane measurement will verify the baffle step and the bass performance. If the NFS data identifies a port resonance or a cabinet resonance, it is probably too late to fix that. And there are other ways to identify those kinds of problems. If the NFS data suggests that I should have spaced the drivers differently, or used more baffle edge radius, or changed the depth of the cabinet... well, it is far too late for that knowledge to be useful.
If, on the other hand, we are focused on how our speaker projects evolve over time, we will probably see much more value. In my second graphic I show this process. With each new project, we learn something new, and we incorporate that knowledge into the next project. So from this perspective the question is "Will the NFS data allow me to learn more, learn faster, and achieve a higher performance on future projects?" Unequivocally yes. There is no doubt that NFS data, over time, over a series of projects, would help me make better/faster progress. So even though NFS is currently unavailable, it is worth pursuing options to engineer a substitute, or looking for ways to rent the service.
So our perspective matters. When we disagree, it is not because some of us are unhinged, it is likely because of where our focus lies.
j.
I think the root of this disagreement is how we look at the scope of our problem...
If we are focused on the next speaker project, we probably won't see much value. In my first graphic I show the typical phases of designing and building a new speaker. From this perspective, NFS data becomes available after the cabinet is constructed. At this point everything about the speaker has already been fixed except the crossover design... so from this perspective, the question is "Will the NFS data allow me to make a better crossover?" If my baffle step assumptions/calculations are off by a few dB, the NFS data will help me. But honestly, a well performed outdoor ground plane measurement will verify the baffle step and the bass performance. If the NFS data identifies a port resonance or a cabinet resonance, it is probably too late to fix that. And there are other ways to identify those kinds of problems. If the NFS data suggests that I should have spaced the drivers differently, or used more baffle edge radius, or changed the depth of the cabinet... well, it is far too late for that knowledge to be useful.
If, on the other hand, we are focused on how our speaker projects evolve over time, we will probably see much more value. In my second graphic I show this process. With each new project, we learn something new, and we incorporate that knowledge into the next project. So from this perspective the question is "Will the NFS data allow me to learn more, learn faster, and achieve a higher performance on future projects?" Unequivocally yes. There is no doubt that NFS data, over time, over a series of projects, would help me make better/faster progress. So even though NFS is currently unavailable, it is worth pursuing options to engineer a substitute, or looking for ways to rent the service.
So our perspective matters. When we disagree, it is not because some of us are unhinged, it is likely because of where our focus lies.
j.
Attachments
Nicely framed Jim.
The other thing is rather than talking past each other, it would be good if one is able to take a step back, think about WHY they have the position they have.
Probably only a handful of people like augerpro and I have NFS measurements of our speakers cabinets, to compare directly to our traditional "baffle-diffraction compensated near field measurements blended with far field measurements"
*I plan to write a post about the quasi anechoic measurements vs NFS measurements shortly
The other thing is rather than talking past each other, it would be good if one is able to take a step back, think about WHY they have the position they have.
- I can afford it- I don't have $90K 'burning a hole in my pocket', but I'd rather buy this, than an expensive imported luxury automobile, which other people seemingly have no problems buying.
- Lost opportunity cost - I could afford it, but there's other things that I would rather do with that money
- I can't afford it eg. I'm retired, and no longer have an income, or this is a hobby that I typically spend hundreds (not thousands) of dollars on, in my spare time. EU90K? Get outta town!
- It's not about the cost, there's just nothing really wrong with the measurements I've got now - eg. "what I've got is good enough" *
Probably only a handful of people like augerpro and I have NFS measurements of our speakers cabinets, to compare directly to our traditional "baffle-diffraction compensated near field measurements blended with far field measurements"
*I plan to write a post about the quasi anechoic measurements vs NFS measurements shortly
I’ve been following this with interest and here’s my contribution. I did this many years ago, even before REW had vector averaging, sliding windows, etc. This is what can be achieved by measuring outside and stacking good quality absorption to take out the floor reflection. It is like an anechoic chamber. No need to raise the speaker several feet off the floor. In fact, you can use less insulation than shown here but raise the insulation off the ground and it has the effect of acting like there is insulation all the way to the ground. A great way to get more resolution and clean measurements in the 100-1000 Hz range.
More details of these measurements here:
https://www.avsforum.com/threads/1-inch-cd-comparison-seos18.1477921/
As I look back at these images from 10 years ago, the only thing I have today is the mic and mic stand. The drivers, cabs, horns, and even the pillows and insulation are gone
More details of these measurements here:
https://www.avsforum.com/threads/1-inch-cd-comparison-seos18.1477921/
As I look back at these images from 10 years ago, the only thing I have today is the mic and mic stand. The drivers, cabs, horns, and even the pillows and insulation are gone