Or not. He'll also learn about gating plus splicing of near-field, as well as how to sum different near-field sources (e.g., woofer and port). I had an anechoic chamber at my disposal for a little while and was disappointed/delighted that my measurements there looked extremely close to the ones I got in my home.
For interpretation, there's a lot of other sources (Dickason's Loudspeaker Design Cookbook is a nice practical one, and do not neglect the treasure trove at Liberty Audio's site- in a just universe, Bill Waslo would be the most famous guy in audio), but first, you have to get good data.
We had one in the technical center that I manage. It had been used for measurements of machine noise, but has since been taken down and sold off, since we're not doing that any more. My pleas that it made my speaker measurements easier did not impress upper management.
Yes, the supplements come with the book in a CD. The book itself, while technical, gives everything one needs to understand measurements. It is a superb book.
Hello Charlie
It's not up to me to tell the developers what term to use. ARTA uses Start and End markers for their IR and calls them "Gates". The resulting FR within these markers (Gates) is therefore a "Gated Frequency Response".
I disagree with them because of methodology. A Gated measurement is derived from a specific process of obtaining measurements. To say it's a Gated Frequency Response when that process is not adhered to causes confusion. The OmniMic that I have does the same thing as ARTA. They avoided using this term in their GUI and user manual.
Let me put it another way.
If I were to list in my restaurant menu a "magnificently marbled Ribeye Steak GRILLED to perfection" and what I served to the customer is a Ribeye that's PANFRIED, he has the right to reject it. Both are cooked but the method of cooking is very different.
Attachments
I don't really start a design with softwares. This is how I approach a 2-way.
Firstly, I run them active (bi-amp) at a crossover frequency I think they'll sound best at. This is a very Raw response. My intention is to simply have an idea of what the woofer and tweeter sound like. If the music is not exciting, it stops there. It doesn't matter how well it measures. If they sound good, then simple SPL sweeps are done to see the behavior of the crossover and the system response.
Having decided on the crossover frequency, the next stage is tuning or EQ. Say I want a flat response from 200Hz to 20kHz (+/- 3dB). I use passive components to tune the system response. SPL sweep will tell me how flat the response is and what is happening in the crossover region. If the speaker sounds good enough, it can stop here.
If I want to go passive all the way, then I use PCD to ballpark the network components to fit my targetted response. Usually, I have to manually fine tune the values until the new SPL sweeps of the individual drivers match that of the ones done with the active crossover with passive tuning.
Basically, it's music, measure, music, measure. It can measure nice and yet the music doesn't sound right. This is not uncommon. It simply means that I'm not getting it right. That's why the speaker is not "Singing".
Well, of course you can do what all of us do (near field measurements, gating) but this doesn't remove all the uncertainty. Multi driver speakers have a near and a far field, there are diffraction effects.
nannoo I suggest that you post your impluse response along with your spl measurement. preferably showing where your gating is set.
I have successfully done indoors measurements down to about 300Hz, below that is difficult in a typical house.
Try to get as far away from walls/objects as you can and preferably angle a bit across the room (so you are not exactly parallel or perpendicular to the walls).
Attached is a measurement of a morel DMS37 tweeter done at 1M Room size was about 4.5M X 7M with about 2.5M ceiling. second picture was taken of the measurement setup at the time of this measurement.
I'd also second SY's recommendation to get Joe D'appolito's book. I have it, but unfortuately I don't think I got (or I lost) the CD.
For desiging a two way speaker I'd say that in room measurements provided done properly (and you can get consistent results) should be more than adequate.
Note that the gating on that measurement should be set slightly lower than what it is in the attached picture, you can see the ripple in the high frequencies as a result of including the bump just before 130CM.
Tony.
I have successfully done indoors measurements down to about 300Hz, below that is difficult in a typical house.
Try to get as far away from walls/objects as you can and preferably angle a bit across the room (so you are not exactly parallel or perpendicular to the walls).
Attached is a measurement of a morel DMS37 tweeter done at 1M Room size was about 4.5M X 7M with about 2.5M ceiling. second picture was taken of the measurement setup at the time of this measurement.
I'd also second SY's recommendation to get Joe D'appolito's book. I have it, but unfortuately I don't think I got (or I lost) the CD.
For desiging a two way speaker I'd say that in room measurements provided done properly (and you can get consistent results) should be more than adequate.
Note that the gating on that measurement should be set slightly lower than what it is in the attached picture, you can see the ripple in the high frequencies as a result of including the bump just before 130CM.
Tony.
Attachments
Last edited:
1. Repeatability. When you can measure from one day to the next and get consistent results then you know you are measuring reliably.
2. Measure the individual drivers. Measure (without moving anything) the drivers with their crossover. Now simulate the crossover with the individual driver measurements. Compare the simulated result to the actual result. If they are very close then you can probably rely on those measurements.
You may find that certain places in the room have a null or a peak from a standing wave. If you see some unexpected dips or peaks try moving the speaker or the mic a little and try again.
Tony.
First, nothing can be more difficult and frustrating than trying to measure the response of a speaker system.
The book by D' Appolito mentioned by others is an excellent source to get up to speed. Read it and then come back and ask questions.
After that you will realize that the challenge to accurately define the performance of your creation is going to be at least an order of magnitude more than you thought.
Measuring inside is not going to be possible due to room reflections, which you can not tame. All you measure is the room with an unknown speaker. Your plan will simply not work.
Software 'gating' helps reject first reflections, but it seriously degrades the resolution of the measurement, too. A gate of 200Hz will mean that everything below 200 Hz is not measured, but it also means the plot's resolution is also 200Hz, which is useless for measuring what is happening below 1 kHz.
I use inside measurements to determine what is happening at the crossover frequencies in the mid and upper frequency ranges where room response is not so much an issue via gating, but not anything more. I certainly can't measure my bass response and do not even try.
Outside measurements are much better if you have the open space and it is quiet (no wind, ambient noise, etc.). Aside from noise and wind, the limit here is where the nearest objects are that can cause a refection in relation to the device under test.
Outside there are a number of methods that work depending on what you are trying to measure (via ground plane, infinite baffle, or suspending the speaker very high in the air). Read the recommended books to understand them, as explaining them here is just too involved.
The book by D' Appolito mentioned by others is an excellent source to get up to speed. Read it and then come back and ask questions.
After that you will realize that the challenge to accurately define the performance of your creation is going to be at least an order of magnitude more than you thought.
Measuring inside is not going to be possible due to room reflections, which you can not tame. All you measure is the room with an unknown speaker. Your plan will simply not work.
Software 'gating' helps reject first reflections, but it seriously degrades the resolution of the measurement, too. A gate of 200Hz will mean that everything below 200 Hz is not measured, but it also means the plot's resolution is also 200Hz, which is useless for measuring what is happening below 1 kHz.
I use inside measurements to determine what is happening at the crossover frequencies in the mid and upper frequency ranges where room response is not so much an issue via gating, but not anything more. I certainly can't measure my bass response and do not even try.
Outside measurements are much better if you have the open space and it is quiet (no wind, ambient noise, etc.). Aside from noise and wind, the limit here is where the nearest objects are that can cause a refection in relation to the device under test.
Outside there are a number of methods that work depending on what you are trying to measure (via ground plane, infinite baffle, or suspending the speaker very high in the air). Read the recommended books to understand them, as explaining them here is just too involved.
I don't find it difficult or frustrating. A bit time consuming because I rather listen to what I'm doing to the speakers than measuring them.
I think the "gating" you're referring to is what is used in ARTA. In an actual Gated SPL Sweep, there's no loss in resolution because no post processing is done. In a Precision Gated Sweep mode, the user can choose as fine as 300 points per decade. The software allows higher resolution but it will only slow down the sweep.
It is possible to measure a quasi-anechoic bass response in the confines of a room. Do a SPL Sweep at the mouth of the port (for BR), do a Pressure SPL Sweep of the woofer, "Sum" the two plots, then "Referenced" it to 1 meter. Now that the bass plot is made, "Splice" it to the other plot for 20Hz to 20kHz bandwidth.
After all these are done, you may well find that your final plots have no correlation with what you're hearing.
The simple answer is there is no simple way to measure in your room.
Putting padding on the wall only absorbs the highest frequencies of the reflections, the ones that can be gated out, but does nothing for the ones that matter. Additionally, in the impulse response, the high frequencies are the ones that most clearly show when the reflections arrive, so absorbing them makes it more difficult to set where to gate out the reflections.
In a room that size you will simply not get a farfield measurement of any accuracy or resolution. The only way to get some idea of the complete performance is from a complex set of nearfield, midfield farfield, inbox, calculations from impedance fitting, diffraction response calculations etc. Way beyond the capabilities of most hobbyists.
I have spent more than 30 years working on refining speaker measurement in hostile environments, including compensating for the limitations of even some of the best anechoic chambers around the world.
The availability of modern measurement tools at low prices just allows one to more quickly get the wrong answer! Nearly all of them are limited by the same considerations of measurement space, regardless of whether they use real impulses, pseudo-random noise, slow sweeps, fast chirps, or anechoic sweeps.
On top of that each choice of excitation signal has further consequences of how it interacts with the test device to alter the apparent frequency response.
The best you will get in your room is a very rough approximation. Use that as a guide, and then tune by ear to give you what you want to hear!
Andrew
Putting padding on the wall only absorbs the highest frequencies of the reflections, the ones that can be gated out, but does nothing for the ones that matter. Additionally, in the impulse response, the high frequencies are the ones that most clearly show when the reflections arrive, so absorbing them makes it more difficult to set where to gate out the reflections.
In a room that size you will simply not get a farfield measurement of any accuracy or resolution. The only way to get some idea of the complete performance is from a complex set of nearfield, midfield farfield, inbox, calculations from impedance fitting, diffraction response calculations etc. Way beyond the capabilities of most hobbyists.
I have spent more than 30 years working on refining speaker measurement in hostile environments, including compensating for the limitations of even some of the best anechoic chambers around the world.
The availability of modern measurement tools at low prices just allows one to more quickly get the wrong answer! Nearly all of them are limited by the same considerations of measurement space, regardless of whether they use real impulses, pseudo-random noise, slow sweeps, fast chirps, or anechoic sweeps.
On top of that each choice of excitation signal has further consequences of how it interacts with the test device to alter the apparent frequency response.
The best you will get in your room is a very rough approximation. Use that as a guide, and then tune by ear to give you what you want to hear!
Andrew
That room is just too small to get the quasi-anechoic FFT deal to work much below treble range, unless you're measuring a very small speaker (and can thus get pretty close with the mic). Even then, you're not going to get very low in frequency. Assuming the 8' is vertical, the middle of the room on the floor pointing up is probably your best bet.
Bingo.....
Drag them outside for some good old ground plane measurements , measuring them indoors in a house is a waste of time below 800hz ...
Steady state noise is ignored very well, but impulsive noise will affect the response in a particular frequency range - the frequencies that are played when the impulse occurs.
The noise is uncorrelated to the input signal so its effect is spread evenly across the 64000 data samples after the sweep measurement is cross-correlated with the input signal. An impulse will have all frequencies at and at some time they will be briefly correlated to the measurement.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.