True, but I would like to keep this all in context.
The process of taking a near field response, modifying it with a simulated baffle diffraction effect, and then merging it with a far field gated response, is considered the current best practice. Every aspiring speaker designer needs to master this process. I use VituixCad and that software helps me to get it right.
At the very highest level of design there is a trend to push into a new state of the art, a more advanced process known as Klippel Near Field Scan. The NFS process is without a doubt more accurate. It is also unavailable at this time, except to a select few. Fast foward a few years and things may be different, but for now, Klippel NFS is not a realistic option for most of us.
i wish i could believe that, but so far i can't.
Measurements themselves, whether near-field or far-field are difficult, and always deserve at least a little bit of skepticism.
As best i can tell, simulations use ideal drivers, and ideal assumptions.
The more pieces to the puzzle, such as stitching together different measurements, applying simulated corrections,..... just means more room for error and more potential distance from reality, imo.
So I trust a simple far-field measurements, as reflection free as I can get them. They are the only ones that i feel give valid relative phase info, which i think is critical for crossover and timing work.
I dunno...i don't mean to be down on sims and measurement 'put togethers' ..... guess i'm just an old codger who's seen reality disagree with models too many times (sometimes In devastating ways). So I stick with simple direct measurements/observations.
I understand being skeptical of the process, and your concerns make sense. We all have to do what we are comfortable with.
This is just one data point, but it helped me get comfortable with the process of taking a near field response, modifying it with a simulated baffle diffraction effect, and then merging it with a far field gated response. The outdoor ground plane measurement took place on pavement, 19 ft from the nearest reflective surface.
This is just one data point, but it helped me get comfortable with the process of taking a near field response, modifying it with a simulated baffle diffraction effect, and then merging it with a far field gated response. The outdoor ground plane measurement took place on pavement, 19 ft from the nearest reflective surface.
Attachments
I tend to gravitate towards the most difficult solutions in many cases, if there's enough gain to be had by it. Though I don't think I'll be using this again anytime soon as a friend of mine has 3 lifts and the tallest one can reach 60ft, so I'll be welding up a jig and hauling my speakers to his place for the final measurements when it's not in use. I built the sound system for the spin room at the gym his wife owns earlier this year, so there are deals to be made.
Height is another line of thought to ponder as you consider the assets you may have available to you.
Height is another line of thought to ponder as you consider the assets you may have available to you.
Attachments
I agree regarding the simulations, which is actually why I haven't spent much time perfecting getting measurements below 400 Hz- I don't see a perfectly reliable method (that will yield much help in the actual design process) with taking the time to actually do all those steps and blend things (and take that statement with a grain of salt, obviously, based on this thread I haven't tried them all). To me it's easier to just flip capacitors and inductors when I'm working in the 400Hz and lower region until it sounds correct, and still look at FR and impedance to make sure nothing is crazy. Even with floor bounce and room interference, I've measured enough to be able to see if there are any glaring issues below 400 Hz at this point, but it's the more subtle issues I'm trying to isolate. All that said, I've never tried the "outside method", I should give this a whirl and see if that's beneficial to the design. The problem I see here is- what's quicker and better?... When I'm working in the 400Hz and lower region is just flipping parts and listening until it sounds correct good enough? Or Hauling everything outside, getting the speaker a couple/few feet off the ground and measuring? Or learning the blending and getting a simulation going to yield better results? Guess I won't know until I try them all.
I don't even spend a whole lot of time in the simulations using WinPCD anymore. I use it to rough in a crossover, and sometimes beyond that I'll use it just to see how a resistor, cap swap, maybe an inductor change... might effect the response of a driver- but I find WinPCD is not always an accurate reflection of reality, it's off enough that I have play around anyhow. So at this stage (having a plethora of spare parts) it's easier for me to just flip parts, measure and listen to reality than it is to horse around in a simulation.
Phase is a giant can of worms, another one that I don't particularly get too hung up on. IMO, I'll see or hear phase issue(s) (that actually matter) somewhere in the response when I take measurements. This is where off axis response comes in. And I'll be completely transparent... in 2 different designs in the past I've sacrificed better phase for better sound quality. The same applies to a frequency response- a flatter response doesn't always sound better to me. For example, a mild dip in the 2khz region might sound better with certain drivers than a flat response.
From the paper:
"A loudspeaker test technique is described which depends on nearfield pressure measurements made in a nonanechoic environment. The technique allows extremely simple measurements to be made of frequency response, power response, distortion, and electroacoustical efficiency."
Curious if the later tech ... including the Klippel NFS isn't fundamentally based off of work(s) described in this paper.
"A loudspeaker test technique is described which depends on nearfield pressure measurements made in a nonanechoic environment. The technique allows extremely simple measurements to be made of frequency response, power response, distortion, and electroacoustical efficiency."
Curious if the later tech ... including the Klippel NFS isn't fundamentally based off of work(s) described in this paper.
Try dialing all in as best you can inside, then go to the trouble of taking it outside. If you're going to look at low end outside, it has to be a very calm day. I tend to create unique shapes and want to see exactly how they perform.
You can also garner information from the comparison of these different measurements, as far as seeing what is real and what is an artifact of environment. In the end though, that last half db dial in of the low end is going to be subjective and unique to where the speaker is going to live.
You can also garner information from the comparison of these different measurements, as far as seeing what is real and what is an artifact of environment. In the end though, that last half db dial in of the low end is going to be subjective and unique to where the speaker is going to live.
Yes, I think at this stage I have the "best I can" dialed in as far as measurements in-room go. I think my next step will be to take outside measurements of a speaker I've already designed and built and look at them to see if there's any useful information that justifies this extra step. After that I'll figure out the blended simulation method and see which is more beneficial.
The bold for sure, I can admit I think I tend to enjoy what might be considered a slightly bass heavy speaker, and definitely more than a bass lean speaker. And this is subjective, there is no technical correct or incorrect here, at least for now, until we can find a perfect way to measure every frequency anyhow. And even at that, someone's room will change everything since I'm pretty sure no one here is listening to their stereo in an anechoic chamber.
FWIW, It was a while ago but I have played around with my OmniMic using different locations while measuring to see how much reflections effect the response. I've moved since I did this, but at my old house I used to measure all my speakers in my 23x27 foot garage with 8' ceilings. It was actually pretty dead aside from the cement floor because I had lumber racks, wood working equipment, shelves, etc on all the walls. Using the same impulse response (IIRC it was 3.5), and 1/24 smoothing, with the mic on axis at the tweeter, 1 meter away on all the different measurement locations...
-I measured in what I considered the ideal location in my garage (right in the middle of the room, using a blanket on the concrete floor between the mic and the speaker), with the speaker facing the open garage door.
-Then I measured with the door closed and blanket removed- no difference at all, at least nothing that would justify opening the door anymore.
-Then I moved the speaker about 4 feet away from a wall (or from a shelf against the wall) and measured and was shocked but basically no difference.
-Then I took all of it down to what was my listening room at the time (pretty terrible room, 7' ceilings, tile floor over cement with a small rug, pool table, and roughly about a 25' deep x 13' wide room with a large entry opening to the room on one side of the speakers... and again was pretty shocked at how much the measurement was basically the same.
A mild blip of a change here and there, but nothing that I would say would cause me to make any change in a crossover.
But again, I was also ignoring everything below 400 ish Hz, so I don't recall what changed in this region. It would be fun to do this experiment again but also pay attention down there this time.
-I measured in what I considered the ideal location in my garage (right in the middle of the room, using a blanket on the concrete floor between the mic and the speaker), with the speaker facing the open garage door.
-Then I measured with the door closed and blanket removed- no difference at all, at least nothing that would justify opening the door anymore.
-Then I moved the speaker about 4 feet away from a wall (or from a shelf against the wall) and measured and was shocked but basically no difference.
-Then I took all of it down to what was my listening room at the time (pretty terrible room, 7' ceilings, tile floor over cement with a small rug, pool table, and roughly about a 25' deep x 13' wide room with a large entry opening to the room on one side of the speakers... and again was pretty shocked at how much the measurement was basically the same.
A mild blip of a change here and there, but nothing that I would say would cause me to make any change in a crossover.
But again, I was also ignoring everything below 400 ish Hz, so I don't recall what changed in this region. It would be fun to do this experiment again but also pay attention down there this time.
Thanks for that Jim,
I can see how sims and ground plane measurements can match up well for a single driver in a sealed box.
And can picture how they can match for simple two-ways.
Don't get me wrong, I totally use modeling to help get a prototype started.
But once sawdust is made, I've found I can both measure and actually HEAR the speaker.... something I've yet to do with a sim Lol
Excellent. Love the pole. I've thought about a pair of telephone poles with a cross piece at top.
Yeah, I hear you. In the end, it's just about pleasing the ears. We all have our ways of getting there, and if we're happy...well, I guess they're all working.
For me , phase is a biggie because of the way I've learned to do xovers, and tune in general. If i were more mainstream, I doubt it would matter as much to me...
I definitely wasn't saying you were wrong, hopefully it didn't come across that way.
Oh no, not at all.
I took it as a simple sharing of what's been working for us....and what we haven't felt a need to address much (like me with sims past drawing up a prototype). thx
Hi,
For best time alignement between two drivers, should we measure the impulse response of each at the wanted (symetric) cut off, then move the Xo point of one driver (asymetric XO) until their impulse response is close enough after mic measurement again at each driver center and also off axis in relation to the filter order chosen ?
Is it ok in REW with its half dual channel protocol ?
For best time alignement between two drivers, should we measure the impulse response of each at the wanted (symetric) cut off, then move the Xo point of one driver (asymetric XO) until their impulse response is close enough after mic measurement again at each driver center and also off axis in relation to the filter order chosen ?
Is it ok in REW with its half dual channel protocol ?
I have a 9' 6" ceiling in my shop, so I'm clean to about a 6ms gate, whereas at my old house it was about 3.5ms. Still, only usable to around 500hz as there aren't many data points defining the curve below that. Once the measurement gate is extended to 500ms, you'll have gain all the way up into the low thousands of hertz before it tapers off into the gated measurement.
If you flatten an anechoic/outside/theoretically perfect measurement and then bring it inside it's likely to be warm. The best way to judge what level is correct at that point is ask the person who's going to be listening to it, they're always right. But seriously, if you like it a tad bass heavy there's nothing wrong with that we're all unique in what we enjoy, and measurements aren't the same as listening. That last half db is a judgement call IMO.
That's usually my wife or my younger daughter. Preferably my daughter because at this point, unfortunately for me because she has great ears, I think my wife's pretty sick of it. Now I almost always get a little grumbling when I ask her to listen to something I'm building and her response is almost always, "They sound great" lol. Thankfully my younger daughter still seems to enjoy it but at 17 years old she's not always around, she'll still tell me if something sounds off.
There's no need to grab the tin foil hat whenever the "S" word is mentioned
In this case it is a simple correction of the 2pi to 4pi response. At least from 200Hz on down an average woofer in a box is very predictable. ARTA applies a simple filter based on the baffle dimension to compensate for the fact that the nearfield measurement is 2pi. Vituix models the size of the baffle to come up with the same answer. Correcting the levels is a basic ratio of sizes.The only assumption in this is that the woofer and box are omnidirectional enough, so no dipoles or cardioid's.
Below 200Hz it could well be argued that this is actually the most accurate method of easily available measurement.
Anyone can make a bad direct measurement just as easily as a bad simulation. The skill, knowledge and care of the person doing it is what tends to make the difference. I trust you to make a good measurement but would be sceptical of anything you simulated
No, the NFS is based on Nearfield Acoustic Holography and Sound Field Separation which is prior art. If you search those two terms you'll find a lot of information. The NFS patent is about making the whole process easier to do when measuring a speaker.
Hi fluid, late reply...been traveling.There's no need to grab the tin foil hat whenever the "S" word is mentioned
The only assumption in this is that the woofer and box are omnidirectional enough, so no dipoles or cardioid's.
Below 200Hz it could well be argued that this is actually the most accurate method of easily available measurement.
Anyone can make a bad direct measurement just as easily as a bad simulation. The skill, knowledge and care of the person doing it is what tends to make the difference. I trust you to make a good measurement but would be sceptical of anything you simulated
Thou art indeed a wise man to not trust a sim from me !!!
Yep, i agree about below 200Hz....it's been the easiest range for me to sim frequency response (or measure in 2pi space) .
It makes sense modeling is pretty much about simple 2pi to 4pi baffle ramp adjustment.
True to the time vs frequency tradeoff however, the <200Hz range continues to prove to be the hardest to measure the time domain.
I've had good luck with Hornresp and subs; reasonable luck with HR on synergies...which would no doubt improve if i waded in deeper.
Vituix worked well for FIR filters doing impulse inversion....heck of a program huh?
If i felt it could model synergies well, I'd could see getting serious about learning it. Any thoughts there?
Vituix can't model a Synergy in the way Hornresp can, but it would be perfect for crossover modelling. That would allow you to see the effect on all of the measurement curves simultaneously. Whether you value this over sticking to your existing technique is up to you, but it will give you a much better idea of what you have without the need for remeasuring.
The reinforcement is 6 dB over the whole frequency range. It lifts the whole bode plot 6 db up. Otherwise, ground plane measurements wouldn't be a reliable means to measure speakers. But gp measurements obviously go wrong at higher frequencies, because the drivers will start to interfere with their mirror images.
Do you think it could model polars from multiple drivers in the conical horn? Along with reflections back into the throat and mouth terminations?
On-axis and near on-axis xover work is easy...I know exactly what i will get after tuning.....which will be close perfect no matter what xover freq I choose within reason. (it's become almost silly easy with driver flattening pre-xover technique, and then using steep lin phase xovers).
it's the xover effects on off-axis I'd love to be able to model.
Especially if there is a way to separate the xover effects on polars from throat reflections and mouth terminations.
I've been under the assumption Vituix is meant for flat baffles/mounting surfaces. And that it wouldn't be able to hold up for off synergy work.
It would be awesome to learn if i'm wrong about this