I’m building a very simple omnidirectional speaker with a bass/mid firing upwards and tweeter firing forwards. Its now time to take some measurements in order to model a crossover, but I can’t decide how best to achieve this. I can measure in the garden, either well away from surfaces or up against a wall to introduce 2 reflective surfaces, or in a room to include all surfaces. But, because of the omnidirectional nature of the sound waves, I can’t decide how best to capture them. How do people normally approach measuring omnidirectional loudspeakers?
Member
Joined 2003
I would probably not measure it differently than any other speaker. Follow CTA-2034-A and measurement guides for VituixCAD, including both horizontal and vertical axis.
If your placement is fixed, measure from the speaker/listener axis.
This will allow you to correct for the direct/indirect combination in your crossover.
If your placement may vary, move it around and use the s/l axes. You are likely going to find
some placements with difficult problems, and some easier to ameliorate.
Commercial speakers like the Ohm Walsh series have to tell you how to place for best response, as everyone is using the same crossover
for a given speaker(note that they do offer versions for different placings, manipulating the response/radiation).
We have an advantage here, being able to adjust for our own rooms and placements.
This will allow you to correct for the direct/indirect combination in your crossover.
If your placement may vary, move it around and use the s/l axes. You are likely going to find
some placements with difficult problems, and some easier to ameliorate.
Commercial speakers like the Ohm Walsh series have to tell you how to place for best response, as everyone is using the same crossover
for a given speaker(note that they do offer versions for different placings, manipulating the response/radiation).
We have an advantage here, being able to adjust for our own rooms and placements.
If your XO is in the KHz area and you can place the speaker at least a couple of meters from any wall, gating will see to that what you see is the speaker drivers only.
//
//
@Ramista did you want a single measurement per driver or are you planning the complete power picture?
A speaker which is axissymetrical around the Y-axis will be the same for horizontally arranged angular measurements so vertical measurements are of interest.
Better to use a custom scheme, not sure that Vituixcad can be used for custom work.
A speaker which is axissymetrical around the Y-axis will be the same for horizontally arranged angular measurements so vertical measurements are of interest.
The horizontal weighting would not be the same. One could use a 'point' based system of measurement rather than an angular one.
Better to use a custom scheme, not sure that Vituixcad can be used for custom work.
Member
Joined 2003
Not sure I completely agree with that. CTA-2034 doesn't care what the speaker is, that's sort of the whole point. Wieghting is based on listener vs speaker for on-axis vs off-axis. You may change your own wieght of on-axis vs power response etc as needed for your preference. Data gathering process does not need to change just because one speaker points at the ceiling.
Ultimately, I need to be able to produce FRD files of the individual drive units for use in XSim. I’ve just looked at CTA-2034 and it compares a weighted average anachoic response to simply pointing a microphone at the ceiling and results were fairly similar. So I’m now wondering if measuring the drive units separately in the room with a fixed microphone position will give me what I’m looking for?
Potentially it can with some speakers, especially with some understanding of the DI relationship and a little luck.
If you are planning on using a single measurement axis, make sure you feel that the listening axis is reasonably representative of the sound as a whole, you don’t want to be sitting in a shadow. Measure again after you cross to see it worked.
Then get yourself an equaliser, or a software equivalent, or adjust your crossover accordingly and set the tonal balance by listening tests.
If you are planning on using a single measurement axis, make sure you feel that the listening axis is reasonably representative of the sound as a whole, you don’t want to be sitting in a shadow. Measure again after you cross to see it worked.
Then get yourself an equaliser, or a software equivalent, or adjust your crossover accordingly and set the tonal balance by listening tests.
Thank you. Maybe I could measure at a number of different positions in the room and choose the measurement from a position which is closest to the average?
Member
Joined 2003
That is one of the aims when doing multiple measurements. In fact, the more appropriately designed the speaker (consistent in the room), and the better it is utilised, the less critical it is to take multiple measurements for acceptable results... in a manner of speaking to make a point.
Since this speaker is designed to reflect a significant amount, this will still be a variable. Preferrably it would do so in a balanced way.
A set of horizontals is not a complete set of measurements unless the speaker is axissymetrical, where you weight like this.
Adding a set of verticals helps with a rectangular type speaker but doesn't simply allow interpolation of the missing points. It is reasonable, but incomplete.
Besides, a 360 degree radial ought to be weighted evenly around vertical axis (horizontal measurements), and multiple such horizontal measurements are pointless unless they are a part of a complete power set.
Determine where the speaker is to be used in the real world - in the garden or in a room. Measure accordingly.
Mr Kinky is absolutely correct. With the setup that is considered, on axis frequency response will be determined to a large degree by reflected sound in the lower regions.
OK! Can you explain why please? And what software/approach would you replace it with?
Can you see something poor or missing with the approach that has been used to measure omnidirectional speakers described in the link in post 2? No problem if you want to go about things differently but I am curious about the reasoning behind which quantities are the more relevant in a successful omnidirectional speaker.
My question was directed towards DcibeL who suggested ditching XSim.
However, thanks for reminding me about the Carlsson Cult link. There’s a paragraph which sums up what I’m thinking really well:
”If the loudspeaker has the ambition to be what Stig Carlsson called orthoacoustic, it must on the one hand radiate direct sound with an even tone curve to the listener, and on the other hand radiate sound towards reflective surfaces in a normal, furnished room so that the sum of direct sound and reflections becomes an even tone curve in a large part of the room. He devoted most of his life to this particular problem.”
The Google translation doesn’t make it particularly easy to understand, but I think the author is suggesting using something like a sample and hold FFT sound level meter, constantly moving around the room until no further changes are observed. One possible problem I see here is getting an over exaggerated low frequency response due to the effect of room boundaries. I may be able to overcome this by staying at least a metre away from all walls. Also, I’m most interested in the 500 to 8000Hz range, which will hopefully not be affected by room boundary issues.
At the one visit at Mr Carlssons living quarter aka. development studio, he had measurements running - it was B&K stuff and this particular long duration (hours? measurement was made with a rotating boom with a mic in the end (like this). The thing had an radius of maybe 0,75 meter so a big vertical turn. I remember I was amazed as I had never sen such a contraption. But I suppose it really measured a larger aspect of sound directivity and room interaction that is typical. I like his speakers. He was an interesting person.
//
//
