How to measure drivers for xo design at low Fc?

[tcpip]

Dear all,

I'm new to speaker design. I've just successfully finished designing, building, and fine-tuning a 2-way MTM (with a lot of help from many of you). I've used a Panasonic mic capsule, SW software, an external USB sound module, and my laptop to do the measurement and xo design. I think I "know how to do it now", for Fc higher than, say, 1KHz.

But I want to design a 3-way now, and the woofer-mid Fc will be fairly low: maybe 300Hz. At that point, I'm in near-field measurement territory. Therefore, I guess I'll have to splice the nearfield with the farfield SPL curves to get the full curve for both the midrange and the woofer, before I can begin designing the xo.

I have a few questions, and I can't seem to get clear answers to them even after having read Claudio Negro's tutorial and getting some clear instructions from some of you:

• When I splice the nearfield and gated-farfield SPL curves for a driver, will the phase information be accurate in the resultant graph?
• When I did gated farfield SPL measurements, I used the time delay in the pulse measurement between 0msec and the first pulse to remove excess delay and get the phase plot cleaned out. In a nearfield measurement, since there's no pulse, how do I do this?
• Till now, I've never specified any X, Y or Z anything about driver placement in SW when doing my SPL measurements. I've just put the mic 1.2m in front of the speaker and taken both midbass and tweeter readings with the mic in one place. However, for nearfield, I'll obviously have to move the mic between midrange and woofer readings. How does the software then know how the two waveforms will integrate (in the phase attribute) at a summing point, say at 1 metre? Unless it knows how the waves will integrate, it can't simulate the xo's total SPL, right?
• I intend to build a 3-way with two midrange drivers (MTM) crossing over to a single woofer. In that case, how do I tell SW to combine the radiation from three drivers when simulating the woofer-mid xo?

Hope these questions are not stupid or FAQ. If there's anything you want me to read, please point my nose at it, and I'll go.

Incidentally, I don't have any large field or parking lot where I can take gated farfield measurements down to 50Hz or anything like that. I'll have to do it all in my modest-sized living room, hence nearfield measurements are a must, I guess.

Please help.

PS: In case you are curious, I already have some inexpensive drivers lying around, and I feel they can't be used well in a two-way, and it'll be great to use them to learn how to do a 3-way design, hence this project. So the drivers are a given; please don't suggest "Why do you need a 3-way at all?" The mids will be TB W4-657 drivers, xo'ing at perhaps 3KHz to a local fabric dome tweeter made by Peerless India. The woofer will be some unknown 8" aluminium-cone driver with a huge magnet, cast frame, fat rubber surround and (presumably) huge Xmax, which I picked up from the local market. It's apparently made by a company called "Alda". I'm hoping that I'll be able to xo these Alda 8-inchers at something like 300-500Hz to the W4-657s without too much cone breakup problems with that Al cone.

[tcpip]

Can anybody please help?

Is it that my questions don't make sense?

[nunayafb]

First off patience, we all know what its like to hit a stopping point and have to wait for some replies to our questions, its a part of the hobby, for noobs anyway. In addition, some posts take a couple of days for an educated response, not everyone checks here daily.

All right to the Q's:

When I splice the nearfield and gated-farfield SPL curves for a driver, will the phase information be accurate in the resultant graph?

To my knowledge the only phase info that is of any use is the phase that is generated when mic'd at one distance with all drivers in the baffle. Any disruption to the mic or speaker position, will throw off the phase (relative and absolute). So my answer would be no, the spliced phase will be meaningless.

When I did gated farfield SPL measurements, I used the time delay in the pulse measurement between 0msec and the first pulse to remove excess delay and get the phase plot cleaned out. In a nearfield measurement, since there's no pulse, how do I do this?

I do not have any latency problems so I dont use the delay, maybe I should? What I do is set the first marker right before the second pulse and the second marker before the third pulse. The first pulse is the MLS warmup cycle and if I understood everything correctly should not be included. Nearfield does not have room interaction issues therefore does not need gating, so the pulse measurement is meaningless. There is an upper frequency limit to nearfield though due to the physics of this type of measurement. To do nearfield just put the mic as close to the dust cap as possible and click "nearfield"

I'm hoping that I'll be able to xo these Alda 8-inchers at something like 300-500Hz to the W4-657s without too much cone breakup problems with that Al cone

Just do a gated measurement and look for cone breakup, I bet it will be at around a couple of khz and wont be an issue for woofer use.

Gotta run, hope someone else answers the rest of your questions soon.

[lndm]

Originally posted by tcpip [*]When I did gated farfield SPL measurements, I used the time delay in the pulse measurement between 0msec and the first pulse to remove excess delay and get the phase plot cleaned out. In a nearfield measurement, since there's no pulse, how do I do this?

You could try adding delay in steps until you notice the phase plot at the top end bending upward. Then reduce the delay until the plot appears to make the most sense, or is at least readable. Apply delay to each driver consistently as delay is a simulated substitute to moving the microphone.

[*]Till now, I've never specified any X, Y or Z anything about driver placement in SW when doing my SPL measurements. I've just put the mic 1.2m in front of the speaker and taken both midbass and tweeter readings with the mic in one place. However, for nearfield, I'll obviously have to move the mic between midrange and woofer readings. How does the software then know how the two waveforms will integrate (in the phase attribute) at a summing point, say at 1 metre? Unless it knows how the waves will integrate, it can't simulate the xo's total SPL, right?

Have you considered taking your measurements from the listening position? Room interactions will therefore be a desired part of the measurement, and correction could be a part of the crossover. I'm not sure who'll agree but it works for me at the bottom end.

If you want more info, you could go from gated, to pink noise, to swept sine testing to include reflection and standing wave information respectively, then trawl throught the plots and smooth them by hand using best judgement.

While tweaking, I find it useful to connect my mike (at the listening position) to a spectrum analyser (computer based) and play pink noise, changing crossover values whilst watching the screen.

[tcpip]

Originally posted by nunayafb
To my knowledge the only phase info that is of any use is the phase that is generated when mic'd at one distance with all drivers in the baffle. Any disruption to the mic or speaker position, will throw off the phase (relative and absolute). So my answer would be no, the spliced phase will be meaningless.

Okay, then, how do I take measurements of the woofer and the midrange in such a way that the phase information in them is relevant?

Alternately, can I design the woofer-to-mid xo by just working on the (spliced) SPL curves, and ignore the phase issue altogether. Would that result in phase-accurate xo? (I'd guess not, because the Fc could be as high as 200-400Hz, where phase would, I'm sure, be very important.)

I do not have any latency problems so I dont use the delay, maybe I should? What I do is set the first marker right before the second pulse and the second marker before the third pulse. The first pulse is the MLS warmup cycle and if I understood everything correctly should not be included. Nearfield does not have room interaction issues therefore does not need gating, so the pulse measurement is meaningless. There is an upper frequency limit to nearfield though due to the physics of this type of measurement. To do nearfield just put the mic as close to the dust cap as possible and click "nearfield"

Whatever you've said is right, but my question was somewhat different. My question was: how do I get the phase curve to "straighten out" for the mid and woofer, while ensuring that the "straightened out" phase curves remain phase-coherent?

Just do a gated measurement and look for cone breakup, I bet it will be at around a couple of khz and wont be an issue for woofer use.

Yes, I too am hopeful. It's just that severe cone breakups even at 1KHz will require a steep xo at 300Hz. But that's a bridge we'll cross when we come to it.

Gotta run, hope someone else answers the rest of your questions soon.

Thanks for the help.

Originally posted by lndm
You could try adding delay in steps until you notice the phase plot at the top end bending upward. Then reduce the delay until the plot appears to make the most sense, or is at least readable. Apply delay to each driver consistently as delay is a simulated substitute to moving the microphone.

Yes, this is the only way out, as far as I can see now (and currently I'm suffering from severe myopia on this issue. )

So, based on the answers and some of my own newbie thinking, I'm wondering whether the following will work:

• I take a nearfield measurement of the mid, keeping the mic aligned with the baffle surface. I don't follow the "normal" practice of keeping the mic a few mm from the dustcap. I splice this with the gated farfield midrange graph.
• I take a nearfield measurement of the woofer, again keeping the mic aligned with the baffle surface. I splice this with the gated farfield woofer graph.
  
  I now have two SPL+phase measurements which should be "time aligned", in the sense that the phase recorded in the two should be coherent. This way I'll also be bypassing the question of "where is the acoustic centre of the driver?"
  
  Is my belief correct?
• I add "Delay" from the SW menu to one of these graphs to see where the "unwrapping" is optimum. I note down how much delay I've added.
• I add exactly the same amount of delay to the other graph.

Will this work? Can I design the xo with these two graphs?

Have you considered taking your measurements from the listening position? Room interactions will therefore be a desired part of the measurement, and correction could be a part of the crossover. I'm not sure who'll agree but it works for me at the bottom end.

Have you designed xo's around ungated farfield measurements of woofer and midrange? Such graphs look so terribly peak-and-trough-filled that I wouldn't know how to even begin designing a flat-SPL xo for it.

Thanks a lot,
Tarun

[tcpip]

I have a refinement to what I suggested earlier.

• I take gated farfield measurements of woofer and
  midrange, and "unwrap" their phase graphs by adding delay to both in equal amount, to get as close to minimum phase (isn't that the term?) as possible. This then gives me a close approximation to what the SPL and phase would look like, with a mic close to the acoustic centre.
• I take a nearfield measurement of the mid, as described above with the mic aligned with baffle surface. Then I unwrap its phase curve.
• I take a nearfield measurement of the woofer and unwrap its phase curve too, by exactly the same amount as the mid.
• Then I splice the mid's farfield with its nearfield, and the woofer's farfield with its nearfield.

Will this work?

[lndm]

Originally posted by tcpip
I take a nearfield measurement of the mid, keeping the mic aligned with the baffle surface. I don't follow the "normal" practice of keeping the mic a few mm from the dustcap. I splice this with the gated farfield midrange graph.

I take a nearfield measurement of the woofer, again keeping the mic aligned with the baffle surface. I splice this with the gated farfield woofer graph.

I now have two SPL+phase measurements which should be "time aligned", in the sense that the phase recorded in the two should be coherent. This way I'll also be bypassing the question of "where is the acoustic centre of the driver?"

Is my belief correct?

Yes. In doing so you'll want to choose only one of the measuring distances for phase info. I would use the far field phase plots and ignore the nearfield phase plots.

One reason is that the near field plots become invalid at higher frequencies, and another reason is that diffraction effects may contribute phase information that is needed.

[*]I add "Delay" from the SW menu to one of these graphs to see where the "unwrapping" is optimum. I note down how much delay I've added.[*]I add exactly the same amount of delay to the other graph.

Will this work? Can I design the xo with these two graphs?

Yes. It will actually be the same whether you unwrap or not, just more difficult to see. Unwrapping also helps to expose driver anomalies, and more easily shows group delay at steep parts of the plot (for example at the crossover region or woofers fundamental resonance).

Have you designed xo's around ungated farfield measurements of woofer and midrange? Such graphs look so terribly peak-and-trough-filled that I wouldn't know how to even begin designing a flat-SPL xo for it.

I do like to start the way that you are starting. I just find that things don't always come out exactly as planned. I think we all like to tweak, and this is my preferred method.

If you want to do this, you'll need to set the analyser to wide band. Third octave spectrum analysis is an excellent tweaking tool, but I use between twelfth octave and full octave analysis as I find each step seems more useful in dealing with specific problems.

For example (broadly speaking), full octave is more useful for tonal balance issues. Third/sixth is useful for crossover issues as it gives more data but ignores the high Q peaks/troughs. Twelfth octave is useful for finding resonances.

EDIT: Just noticed your second post.

[claudio]

Hi all,
Tcpip I would suggest you to calculate minimum phase from the driver's frequency response using FR Combiner, and then use the resulted curves (spl+phase) for crossover simulation. I describe this procedure in my home page (folder SW Phase).
In short:
1) place mic on mid axis and measure woofer and mid (far field).
2) measure NF of each driver and splice it with the far field.
3) Try to use the higher sample rate your card supports.
4) Use FR Combiner to calculate the driver's minimum phase and export the resulting curve.
5) Import the curve in the simulation software.

You can also try the Ralph's technique, if you like, always using the calculated minimum phase curves: I descibe it in my Home Page (folder SW Acoust. Measure/Far Field/Gated).


Regards,
Claudio

[tcpip]

Originally posted by claudio
Tcpip I would suggest you to calculate minimum phase from the driver's frequency response using FR Combiner, and then use the resulted curves (spl+phase) for crossover simulation. I describe this procedure in my home page (folder SW Phase).

Got it. Basically, this way, I can forget worrying about whether my nearfield and farfield will have matching phase. I basically derive the phase data from the SPL data. Fine. I'll start reading up FRC manual now.

One more question: do I need to worry about baffle diffraction and apply BDS before I use FRC?

Thanks a lot.

[claudio]

Originally posted by tcpip

One more question: do I need to worry about baffle diffraction and apply BDS before I use FRC?

IMHO, don't worry about baffle step.

Regards,
Claudio