Correct Way to measure a horn loudspeaker
A question came up in the tapped horn thread about the frequency response of a horn built by one of the members Jean-Michel (jm_kzo). A good job to be sure was done of the enclosure. His frequency response was not what he expected. A couple of us recomended for him to do ground plane measurements.
That's all fine and dandy if you are working in the pressure field as per the directions. But immediately off a horn mouth can these methods produce accurate results?
I mean there is of course no reasonable corner loading of the mouth of the horn. So the response should reflect that.
I would think that modeling the horn in a half space environment should produce a closer result to a proper ground plane measurement. Then comparing the modeled response to the measured would be a better match.
As an observation when you are in a room with a good bass horn and standing right beside it there is not that much apparent output. But go a little distance from it and you get the full effect. The only exception I can think of is inside a car where you are always in the pressure field of the mouth of the box.
Or is this method of measuring at a distance that can be scaled to act as a 1 watt 1 meter response As in this comment "You didn't happen to do 28.3V at 10m did you?" by justaguy. That is a way to get the full effect of the horn into what? A ground plane measurement at that voltage and distance that can be considered as a 1 watt 1 meter measurement.
I have read up on this years ago and I thought that I had kept the papers. But for the love of good information I cannot find them!
Any thoughts would be appreciated.
Gawd, do I ever have had a lifetime of failure measuring loudspeakers. Some stuff works nice, like sharp pulses in an anechoice chamber but frequency response at home... forget it (at least with the kind of gear I've ever had). And all that goes double for anything under 200 Hz.
On good theoretical and practical grounds, I like pink noise or frequency sweeps using 1/3 octave bands - I have some signals off an old CBS-labs test record. The ear and the SPL meter show some agreement and the measurement experience makes seems coherent.
Although I remain a big fan of the famous Klipschorn bass (the two upper speakers are not at all to my taste), there are some nasty frequency deviations. I control them with a parametric equalizer tuned for my woofer pass-band around 25-140 Hz. Works like a charm. A great way to go.
Just as critical to lop-off the subsonics from a Klipschorn as any other speaker.
This is a problem area, one has to make some assumptions and so on.
In a perfect world, such as the inverse square law assumption that is made when you measure at one meter, there is no problem, the inverse square law holds constant and the SPL falls off as it is supposed to.
In the real world such as measuring a large subwoofer sitting on the ground, one finds that space the cabinet occupies distorts that 1 meter radius space between the microphone and the actual origin is not a theoretically tiny point source. To compensate for this, one can measure further away from the source so that the near field source size distortion is minimized.
A home size woofer can usually be measured at say 2 meters or more, a large woofer like we make at work is usually measured at 10 meters (-20dB RE: 1 meter) and for simplicity, driven at 28.3 V (+20dB over 2.83V) so the measured result is a conservative 1W1M.
As one is using the ground for a boundary for both the source and microphone, there are no reflections unless one is too close to a building etc. This is a perfect environment for measuring subwoofers unless it’s raining, snowing, windy, clouds of mosquitoes, loud cicada’s, it’s too hot or you just ran over the mic cable with the mower and so on.
Once one is above the subwoofer range, I prefer to use a measurement tower.
I built one, which raises the speaker and microphone up 20 odd feet with a mic position at 2 meters usually. This puts the closest reflection some 40 feet delayed behind and well below the test signal and rejected by the TEF machines windowing.
This (large reduction of any signals not coming directly from the drivers being tested) has proven to be a good way to get the raw data for working on crossover transfer functions, with the same outdoor considerations as above. Ben, for EQ'ing a speaker, the same applies, measure outdoors and generally speaking (especially if measured indoors), don't try to fill in dips, only cut the bumps.
Hope this helps,
The thing that I would love to find is if there are any papers that have shown a way to do a comparable measurement closeup like the thread about measureing in room response of a subwoofer.
diyAudio Forums > Top >Loudspeakers >Loudspeakers >Low frequency in room measurement
I love the line about running over the mic cable. lol You to!
The things we do to learn. And our appreciation is still increased and rewarded.
Thanks for your thoughts Tom.
Very nice of Tom to remind us of basic physical theory.
But my experience with sticking microphones in my listening room, like no-way to get much that's orderly and or much resembles my hearing experience. I am not talking about golden-eared hearing, just basic judgment of loudness. I suppose if I used lots of runs and averaged them or, as previously posted, 1/3 octave band noise, maybe some rough sense would arise but still the mic results would not become what I'd call primary source information.
In practice, I have a 45 year old bass sweep tone record (now on CD) that repeats over and over and with cricket chirps each 50 Hz. I've listened to that sweep (300-20) hundreds of times and have a good sense of what to listen for. After, I can set my parametric equalizer bands for each room, furniture change, or woofer mod. The high standards of courtesy of this board prevent me from offering a sarcastic "no thank you" that is so well deserved to Tom's outrageous and patronizing suggestion of hanging the Klipschorn out of doors, 20 feet in the air, in order to figure where the frequency curve bumps are.
Now Tom may be right about getting meaningful curves by hoisting the Klipschorn "20 odd feet" into the air (presumably with 10 feet of walls and floor too and with the mic held at 13.5 feet) and measuring out of doors. Haven't tried that yet... aside from the astounding hoisting challenges, there's that mega-buck TEF machine. BTW Tom, ever seen a Klipschorn?
Funny thing, maybe because I was a professor, I am especially allergic to ivory tower lecturing, even 20 foot high ivory towers.
Just to be clear I meant no personal digs on your comment. And I do appreciate your statements. It is good to have comments from someone with a good academic background. And looking at you lines on the bottom of your posts you have some experience.
But I to have the same problem that we all have when we try to measure a horn. What is available to me in papers that I have does not help me get accurate results next to a horn mouth. The same problem has troubled others measuring conventional cone drivers and they have found ways that approximate free field testing as described by Tom.
Granted we all have been making horns with drastically reduced mouth dimensions and this leads to a peaked response. These peaks tend to broaden and go away the farther we get from the undersized mouth.
But I wonder how far away from the mouth of a horn can a microphone be place and the response taken that represents the true output of the box? And if there are any mathematical manipulations that can help minimize the distance.
I was piqued at Tom's comments, not yours. You have asked a very good question.
My reply was that I've never been able to measure speaker frequency response in a room to any good degree of satisfaction. With fancy tools and fancy mics, might be possible. Or working with noise although that tends to feel vaguely unsatisfactory. And any obstacles you might find in a real room working with a boxed woofer, you'll find double with a horn.
If you are trying to make good sound, there are all kinds of other strategies. Listening carefully to favourite music and trying different equalization settings is one method that works pretty good. But attaching a mic to a VU meter has not worked for me except maybe with 1/3 octave noise. So why is that your goal?
If you want to document your system - and all us love data plots - I have no good advice except maybe to make a 20 odd foot tower strong enough to hold a horn and the room walls and floor, and convert your retirement savings into a TEF machine.
Hi Ben, Mark
I read your reply’s, perhaps more of an explanation is in order.
With eq or when designing a crossover, one can only deal with the things that are minimum phase. EQ can perfectly fix minimum phase problems because the phase and amplitude are what is needed to fix the inverse problem.
You cannot EQ things that are non-minimum phase without ending with a screwed up phase response.
The most common example of this problem is caused by a reflected signal being added back to the source, such as bouncing back from a far wall. These signals if arriving in anti phase, cause notches in the response. As they are caused by a delayed signal, one from a different time of origin, these cannot be fixed with EQ. These appear as notches in the response, hence the advice to only cut the peaks or add a tilt, don’t try to fill the notches.
Specifically, I am familiar with the K horns although I haven’t measured one.
If you are talking about trying to improve the lf portion, then you can do that without a tower. The object is to figure out what the speaker alone is actually doing and then based on measurements, which don’t include your room, fix that as best you can.
Fwiw, TEF machines (like the tef 20) appear on fleapay occasionally for hundreds, not thousands of dollars and other systems can also give the desired results.
Close measurements don’t tell you much either unless you have a correction file that relates the difference in response one sees at say 0 or 1/4 meter and several meters away.
The fact that the frequency response changes (on many speakers) up to some distance can make this issue confusing and is why “up close” measurements are mostly only good for comparing modifications where the speaker can be measured in exactly the same location each time. The problem is that the assumptions “assume” a point source that is infinitesimally small, not a source which occupies a good chunk of the origin’s volume and who’s outlet is approaching a meter square. This distorts what you measure “up close” relative to it’s actual response, measured farther away.
Once you have the speaker’s response ironed out, outside, then move it indoors.
You will see your room greatly alters your response and it is position dependent but keep in mind what can and can’t be fixed with further EQ.
Lastly, if your on the quest for deeper bass, approximate the horn is McBean’s “Horn Response”, compare the measured result (reality) to the prediction, then try other drivers in the program to see if there is one off the shelf that would give you a better response, lower cutoff or greater excursion limited output.
“Funny thing, maybe because I was a professor, I am especially allergic to ivory tower lecturing, even 20 foot high ivory towers.”
Funny thing, I’m an Inventor and not from academia at all and while I do this stuff for a living and have made a LOT of horns, I had to learn the hard way, on my own by doing, in reality, I wasn’t lecturing but attempting to save you time, hopefully the “why” part is more clear.
I've not had bad results just measuring groundplane in my garden. Here's a measurement of one of my labs taken with the mic on the ground at 2m.
I've put the 'official' measurements from the labforum underneath. It's the dark blue (single lab) one that compares with mine.
I think you'll agree that it's a fairly good likeness.
I've got measurements of nearly everything I've ever built and they all come up very close to the models. I'm perfectly happy with GP measuring. I think if I was to have 6 labs then I would have to back away from the subs, but that would get my mic closer to buildings etc which would interfere with the results.
So you get pretty good results. That means it is possible to do correctly. On this half of the planet we are going into spring. So I can finally do some decent measurements outside. The inside stuf is what drives me nuts. To bad I it is for most of four months! I'm wondering if you can still do the garden path measurements as you guys are going into the autumn season I think.
What I am looking at specifically is the horns where we have an undersized mouth. When I measure them the results are mostly hit and miss. With a measurement close enough to the mouth I get a the traditional ragged response that is the true response. (prior to the marketing filter) Many members are measuring their horns and getting close results to the simulations. What I am hoping for is that a little shop talk will get us the means and methods that work repeatedly. My experience is limited to the stuff I'm doing. Be it correct or not. But the comments from the other posts are well taken and appreciated. And I am learning different methods of doing some things and seeing some methods that are quite similar to the work that I have done.
|All times are GMT. The time now is 06:02 AM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2016 DragonByte Technologies Ltd.
Copyright ©1999-2016 diyAudio