Hi all,
So I have ARTA set up with a calibrated microphone and gated input, looking to achieve accurate FR measurements for my drivers in the speaker cabinet.
I don't have much space (approx 10'x11'x8' room) but I have lots of furniture around the edges to break up reflections, and have arranged a duvet/blanket/cushions against the wall directly opposite the speaker in the hope of controlling the reflections.
I can't see any reflections in the impulse response, but have tried with both a 3ms gate and a long gate with similar results.
My problem is the measurements 'at driver' are very different to those at a distance. If I take repeated measurements from alternative distances, the frequency response seems to move all over the place. I have attached images of the frequency response measured up close then a few moves directly back from the driver (with volume adjust up a little) to show what I mean.
The reason for doing this is I was thinking of building the crossovers based on nearfield measurements, then measuring the overall response of the complete loudspeaker at 1m.
Obviously this is pointless if my measurements aren't accurate.
I've googled 'how to accurately measure FR' and found a few guides which I have read through, but besides talking about room nodes and reflections, I can't seem to find a guide on how to know when you're making accurate measurements, and how to fix the problems if you're not...
Does anybody have advice on how to know when your measurements are reliable?
Thanks.
So I have ARTA set up with a calibrated microphone and gated input, looking to achieve accurate FR measurements for my drivers in the speaker cabinet.
I don't have much space (approx 10'x11'x8' room) but I have lots of furniture around the edges to break up reflections, and have arranged a duvet/blanket/cushions against the wall directly opposite the speaker in the hope of controlling the reflections.
I can't see any reflections in the impulse response, but have tried with both a 3ms gate and a long gate with similar results.
My problem is the measurements 'at driver' are very different to those at a distance. If I take repeated measurements from alternative distances, the frequency response seems to move all over the place. I have attached images of the frequency response measured up close then a few moves directly back from the driver (with volume adjust up a little) to show what I mean.
The reason for doing this is I was thinking of building the crossovers based on nearfield measurements, then measuring the overall response of the complete loudspeaker at 1m.
Obviously this is pointless if my measurements aren't accurate.
I've googled 'how to accurately measure FR' and found a few guides which I have read through, but besides talking about room nodes and reflections, I can't seem to find a guide on how to know when you're making accurate measurements, and how to fix the problems if you're not...
Does anybody have advice on how to know when your measurements are reliable?
Thanks.
Attachments
You should take your measurements at your listening position. Near field response is only going to tell you what's happening right at the speaker. unless you listen in near field, that will not give you the results you are looking for. If you develop a crossover in near field, in a deadened room, once you are in listening position and have removed the extra damping, the response will be totally different. If you want it to sound a certain way in your listening position, take your measurements there and make adjustments to your crossover accordingly. Every part of the room will have a different response, but the most important one is where you are going to spend the largest amount of time listening. Set the room up like it will normally be when you are listening, then take your measurements. This will give you the most accurate picture of your response curve.
You should make (quasi) anechoic measurements of the loudspeaker. You have the ability to do this already, but it is a multi-step process. The measurement has to be done in parts: high frequencies (e.g. >200Hz) and low frequencies (e.e. <200Hz) and then the results merged together. Here's how to do it:
This is how to do the high frequency measurements:
Making gated-impulse frequency measurements using ARTA
Positioning the speaker and microphone for high frequency measurements:
The lower limit of the high frequency measurement is determined by the distance to the closest "boundary" (floor, walls or ceiling) - the farther away these are the lower in frequency you can measure accurately. So, you need to get the speaker into a space with high ceilings and elevate it off the floor. I use a heavy duty flying speaker stand but you could put the speaker on top of a 6-foot-tall ladder. The orientation of the speaker is not important, so place it "sideways". Now you need to position the microphone so that it is on the "listening axis" just as if you were sitting in front of the speaker. If you listen to the speakers off axis (not toed in) then put the microphone at that axis. You will need a mic stand for this and may need to elevate the mic stand as well. The distance between the microphone and loudspeaker should be around 4-6 feet. You also need the microphone to be as far away from any and all "boundaries" as possible.
Now make a gated impulse frequency response measurement following the tutorial at the link above. You should be able to get valid frequency response data down to about 200Hz, 100Hz if you have a big room with 15' ceilings to use. This is the quasi anechoic response of your speaker and if you don't care about the lower frequencies you can stop there.
To get the rest of the frequency spectrum you need to resort to another technique. The room becomes "too small" to measure lower frequencies directly in the far field because you will also be measuring the room response. At these frequencies only the woofer should be playing so one approach is to measure the frequency response of the woofer using a nearfield measurement. To convert this to the "farfield' measurement, you need to add back in the baffle step and scale the SPL level out the the farfield. This can be done in ARTA, as outlined in the following links or you can use freeware to do it.
The combined high and low frequency measurements are the loudspeaker's anechoic response. What you hear is the loudspeaker's (anechoic) response combined with the "room response", which varies with position in your room and includes resonances and frequency dependent adsorption. This might be what you are measuring now, and it is not a good way to really gauge the frequency response of a loudspeaker because every room has a different response, which is also position dependent.
-Charlie
This is how to do the high frequency measurements:
Making gated-impulse frequency measurements using ARTA
Positioning the speaker and microphone for high frequency measurements:
The lower limit of the high frequency measurement is determined by the distance to the closest "boundary" (floor, walls or ceiling) - the farther away these are the lower in frequency you can measure accurately. So, you need to get the speaker into a space with high ceilings and elevate it off the floor. I use a heavy duty flying speaker stand but you could put the speaker on top of a 6-foot-tall ladder. The orientation of the speaker is not important, so place it "sideways". Now you need to position the microphone so that it is on the "listening axis" just as if you were sitting in front of the speaker. If you listen to the speakers off axis (not toed in) then put the microphone at that axis. You will need a mic stand for this and may need to elevate the mic stand as well. The distance between the microphone and loudspeaker should be around 4-6 feet. You also need the microphone to be as far away from any and all "boundaries" as possible.
Now make a gated impulse frequency response measurement following the tutorial at the link above. You should be able to get valid frequency response data down to about 200Hz, 100Hz if you have a big room with 15' ceilings to use. This is the quasi anechoic response of your speaker and if you don't care about the lower frequencies you can stop there.
To get the rest of the frequency spectrum you need to resort to another technique. The room becomes "too small" to measure lower frequencies directly in the far field because you will also be measuring the room response. At these frequencies only the woofer should be playing so one approach is to measure the frequency response of the woofer using a nearfield measurement. To convert this to the "farfield' measurement, you need to add back in the baffle step and scale the SPL level out the the farfield. This can be done in ARTA, as outlined in the following links or you can use freeware to do it.
- Low-Frequency Loudspeaker Assessment by Nearfield Sound-Pressure Measurement D.B. Keele, J. Audio Eng. Soc., (April 1974): http://www.xlrtechs.com/dbkeele.com/PDF/Keele (1974-04 AES Published) - Nearfield Paper.pdf
- Nearfield Measurement with Multiple Drivers and Port http://www.klippel.de/fileadmin/kli...ement_with_Multiple_Drivers_ and_Ports(2).pdf
- Merging Near and Farfield Measurements http://www.klippel.de/fileadmin/kli...39_Merging_Near_and_Farfield_Measurements.pdf
- Merging Near and Farfield Measurements using ARTA: http://www.artalabs.hr/AppNotes/AP4_FreeField-Rev03eng.pdf
The combined high and low frequency measurements are the loudspeaker's anechoic response. What you hear is the loudspeaker's (anechoic) response combined with the "room response", which varies with position in your room and includes resonances and frequency dependent adsorption. This might be what you are measuring now, and it is not a good way to really gauge the frequency response of a loudspeaker because every room has a different response, which is also position dependent.
-Charlie
Last edited:
What are you actually measuring here???
Can you provide some more details on your set up (where is the mic, where is the loudspeaker, etc.) and the loudspeaker (how many drivers, is there a crossover, etc.). Pictures are worth many words, so if you can post a pic of your measurement set up, that would be helpful for giving you feedback.
If you PM me with you EMail I will send you a link to a PPT file that might clear up some of the confusion for you.
-Charlie
If you're testing your speakers where you're planning to use them then it's relevant. If you're designing them for a different environment, anything apart from anechoic is liable to be polluted by the acoustic impedance, standing waves, comb filtering and other esoteric nasties which almost all rooms generate, and your ear very cleverly works to overcome.
When tuning a control room in a recording studio, usually a bunch of measurements are taken, usually at listening position, then steps of a few feet either side, above and below, then the graphs are overlaid and averaged.
When tuning a control room in a recording studio, usually a bunch of measurements are taken, usually at listening position, then steps of a few feet either side, above and below, then the graphs are overlaid and averaged.
ARTA and similar FFT based systems, are not capable of Gated SPL measurements. The microphone is ON throughout the duration of the sweep.
In a Gated SPL sweep, the microphone is switched ON and OFF continuously. The ON time is adjusted to allow the oscillator sinewave burst to pass through only. The microphone then switches OFF before the first reflected sound from the nearest boundary arrives. By preventing the refected sound from entering the microphone, a Gated Sweep produces measurements similiar to one conducted in an anechoic chamber, hence the term Quasi-anechoic.
In a Gated SPL sweep, the microphone is switched ON and OFF continuously. The ON time is adjusted to allow the oscillator sinewave burst to pass through only. The microphone then switches OFF before the first reflected sound from the nearest boundary arrives. By preventing the refected sound from entering the microphone, a Gated Sweep produces measurements similiar to one conducted in an anechoic chamber, hence the term Quasi-anechoic.
Michael,
On and off state are easily achieved with post processing. A swept sine or MLS based IR measure may be gated by simply silencing the data containing reflections. Of course this obscures the low frequency information, which is room/speaker interaction, which may be thought of as a secondary baffle step.
Measurement of LF from very close improves direct to reflected signal strength, and allows somewhat larger measurement windows. Below 100-200Hz room effect is difficult to ignore, but omnidirectional characteristic of most speakers for use in domestic living rooms may be used to advantage; All modal peaks in room have common peak at speaker location. Measurement of LF at <1/4 wave from speaker with equalization bringing all peaks down works very well.
On and off state are easily achieved with post processing. A swept sine or MLS based IR measure may be gated by simply silencing the data containing reflections. Of course this obscures the low frequency information, which is room/speaker interaction, which may be thought of as a secondary baffle step.
Measurement of LF from very close improves direct to reflected signal strength, and allows somewhat larger measurement windows. Below 100-200Hz room effect is difficult to ignore, but omnidirectional characteristic of most speakers for use in domestic living rooms may be used to advantage; All modal peaks in room have common peak at speaker location. Measurement of LF at <1/4 wave from speaker with equalization bringing all peaks down works very well.
There will be differences in the measurements near and far, but there are two effects here and sorting them out can get complicated.
The real response will vary in the near field out to some distance at which point it stabilizes. It is important that you be somewhat close to this distance or the results are not valid. But as you move further back more and more of the room response will enter into your data and change the results.
It is highly unlikely that you can absorb the room reflections sufficiently to make them a non-factor - I have tried as much as 18 inches of absorption and it is not sufficient.
You must position the line between the speaker and the mic such that it is as far away from the enclosure surfaces as possible. This is usually diagonal. This is the best that you can do. Remember that both the speaker and the mic need to be as far away from surfaces as possible.
The real response will vary in the near field out to some distance at which point it stabilizes. It is important that you be somewhat close to this distance or the results are not valid. But as you move further back more and more of the room response will enter into your data and change the results.
It is highly unlikely that you can absorb the room reflections sufficiently to make them a non-factor - I have tried as much as 18 inches of absorption and it is not sufficient.
You must position the line between the speaker and the mic such that it is as far away from the enclosure surfaces as possible. This is usually diagonal. This is the best that you can do. Remember that both the speaker and the mic need to be as far away from surfaces as possible.
Your statement is false.
ARTA is certainly capable of doing a gated SPL measurement, by gating the measured impulse response so that only the portion containing the direct sound is processed into the frequency response. It's NOT a gated sweep per se as you have explained it, but it still produces a reflection free, quasi-anechoic SPL measurement because the non-direct sound is eliminated.
Maybe you are describing a continuous SPL measurement technique? The gated impulse response measurement is not continuous.
So many opinions! Our original poster is probably thoroughly confused by now. I have some, but not a lot of experience measuring speakers for the purpose of designing an X-over. Based on my limited experience, I think that CharlieLaub is giving the best advice. I recommend following it and add some suggestions.
1. The shortest distance in you room is 8' (presumably floor-to-ceiling), so put your speaker (lying in its side) with the center 4' off the floor to maximize the shortest distance to a reflecting surface.
2. I calculate that a 4' wave is 280Hz, so make your high-frequency measurements above 280Hz and low-frequency measurements below 280Hz. When making the high-frequency measurements, set your gate to just above 280Hz.
3. If your expected X-over point is far enough above 280Hz, you may be able to get away with only the high-frequency measurements.
4. Don't bother with the absorbing materials on the walls. It isn't gaining you anything.
Have fun!
1. The shortest distance in you room is 8' (presumably floor-to-ceiling), so put your speaker (lying in its side) with the center 4' off the floor to maximize the shortest distance to a reflecting surface.
2. I calculate that a 4' wave is 280Hz, so make your high-frequency measurements above 280Hz and low-frequency measurements below 280Hz. When making the high-frequency measurements, set your gate to just above 280Hz.
3. If your expected X-over point is far enough above 280Hz, you may be able to get away with only the high-frequency measurements.
4. Don't bother with the absorbing materials on the walls. It isn't gaining you anything.
Have fun!
OK... so what do you call the process of "selecting" a certain portion of the impulse response (the part without reflections, aka the direct sound)? What do you call the resulting frequency response? Is calling this a "gated impulse derived frequency response" incorrect? What are the correct semantics here?
-Charlie
@nannoo:
Run, do not walk, to Amazon and get this book:
Testing Loudspeakers: Joseph D'Appolito: 9781882580170: Amazon.com: Books
Loudspeaker measurement is a fascinating and involved subject that can't be summarized in a few paragraphs on a forum. Dr. d'A does a superb job of making the ins and outs understandable and showing exactly how choices in measurement techniques affect the results.
Run, do not walk, to Amazon and get this book:
Testing Loudspeakers: Joseph D'Appolito: 9781882580170: Amazon.com: Books
Loudspeaker measurement is a fascinating and involved subject that can't be summarized in a few paragraphs on a forum. Dr. d'A does a superb job of making the ins and outs understandable and showing exactly how choices in measurement techniques affect the results.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.