Taking Measurements for a 3-way

[dkalsi]

A few weeks ago, I took a stab at building my first speaker. I used the measurement technique suggest in Jeff Bagby's paper found at the link below:

http://audio.claub.net/software/FRD_Blender/White%20Paper%20-%20Accurate%20In-Room%20Frequency%20Response%20to%2010Hz.pdf

When I built the bookshelf, I worked with files containing minimum phase (extracted via Frequency Response Blender).

I wanted to know whether I can used the same technique for a such a low crossover point (e.g. 200hz). I could try outside ground plane measurements, but it won't be easy to haul the cabinets outside (and find a quite place).

In my mind, I was thinking about:

1) Taking Tweeter, Mid, Woofer measurements from a distance of 40" on tweeter axis
2) Taking nearfield measurements of the Mid and adjusting for baffle step loss
3) Merge Mid's nearfield and farfield response measurements using Frequency Response Blender and extract minimum phase
4) Taking nearfield measuremnts of the Woofer and adjusting for baffle step loss
5) Merge Woofer's nearfiled and fairfield response measurements using Frequency Response Blender and extract minimum phase

Will the steps above give me workable file?

When taking measurements inside the house, I can probably get close to 4ms gating for the Tweeter and Mid and probably 3ms for the Woofer. I understand I would need the woofers gating to be extended, but given that all measurements must be taken without moving the mic, and I'm trying to place the Tweeter and Mid as close to the middle between ceiling and floor, the woofer will end up being much closer to the floor.

Any thoughts?

[dcibel]

What size is the woofer and how wide is your baffle? How close is the woofer to the floor?

[civit]

My understanding is that you should move the mic so it is on axis and the same distance from the baffle for each driver,when it is your intention to extract minimum phase measurements. These measurements will then need to be positioned in your simulation software (PCD, VituixCAD, LSPCad, whatever). The time when you don't want to move the mic is when you're taking a series of measurements to glean the relative acoustic offset of your drivers - in this case you don't want to extract minimum phase (someone correct me if I'm wrong).

The measurement protocol described by Bagby for in-room anechoic response works well. Here are my reactions to your notes:

1) Taking Tweeter, Mid, Woofer measurements from a distance of 40" on tweeter axis
First, 40" is further than necessary. I like to do 2x baffle width + a little bit more. This might get you a longer gate. Also, these should be done on each driver's axis. If you do them all from the tweeter axis you are essentially designing with a woofer and mid that have treble rolled off.

2) Taking nearfield measurements of the Mid and adjusting for baffle step loss
Depending on how low your mid goes this might not be needed but is probably a good idea. The nearfield response gives much better detail in the bass region.

3) Merge Mid's nearfield and farfield response measurements using Frequency Response Blender and extract minimum phase
Yes.

4) Taking nearfield measuremnts of the Woofer and adjusting for baffle step loss
I like the VituixCAD 'merger' tool for this, but the FR blender spreadsheet works too. This step is critical, and I don't find that the scaling recommendations (woofer size -> db decrease) work very well, which is a problem, because setting the nearfield too high or low will totally screw up the balance of your system.

5) Merge Woofer's nearfiled and fairfield response measurements using Frequency Response Blender and extract minimum phase
Sounds good. You can put port measurements in there too, but I find those are a bit trickier to merge.


I like to do different sets of measurements.

For my last project, I did one set from a relatively long distance (40"), all three drivers measured using the same gate and same mic position. I used these to establish the Z offset for the mid and woofer.

I did another set of measurements 18" from the baffle on axis with each driver, 180 degrees in 5 degree increments. This helped visualize the off axis behavior of each driver, which allowed me to choose crossover points/slopes. These measurements were not minimum phase but I adjusted the delay in my CAD program so that time of flight was pretty much removed (I don't know of a good way to extract minimum phase from 60 files at once, and I'm not sure if the results would be accurate if I did).

Then as a sanity check I did the same measurements but only on-axis, extracting minimum phase from each and positioning them in XYZ.

I am finding that the same crossover looks the same for each set of measurements, this helps build some confidence that my design is correct.

[dkalsi]

What size is the woofer and how wide is your baffle? How close is the woofer to the floor?

dcibel,

The cabinet is about 12.6" wide. The woofer is the 10" (its the SB Acoustics SB29NRX).

[wogg]

There was an excellent short post from a member recently that had an outline for getting a 3 way measurment executed for design. The member was isaeagle (SP??) and I can't seem to find it at all, which is sad because I wanted to refer to that later.

If I recall correctly it was something like...

1. Take farfield measurements, all on the design axis with adequate far field distance to capture the baffle signature with fixed speaker placement, mic placement and input levels.
2. Measure and Save min-phase FRD for: woofer, midrange, tweeter, mid+woof, mid+tweet
3. Set X and Y offsets in PCD
4. Bring the woofer and mid into PCD and overlay your mid+woof measurement
5. adjust the Z offset until the sum matches the measured overlay, that gives you your measured Z
6. Repeat 3-5 for the mid+tweet

At that point you should have solid FRD files and offsets ready for your mid and tweet, but since this is a 3 way you may want to merge near-field for the woofer. I'd do that after you're solid on the above measurements since it's critical the mic, speaker and signal levels don't change. Take the near field (and port / PR if applicable) and use blender to merge below 300 Hz and use that output for the design.

[jhollander]

Do not move the mic when taking measurements. The mic position is the summation and design axis. The driver frequency responses will change when you move the mic. The design software doesn’t correct the responses back to the design axis.

dkalsi steps to add in
Set volume level with the tweeter
Take Tw plus Mid and Tw plus woofer for the offsets before moving the mic. 5 measurements before moving the mic. I've been adding in some off axis measurements before moving the mic, but it's optional and more work.
Adjust the volume level down the for the near field measurements to avoid clipping the mic.

[civit]

OK - I have a question. When taking measurements to establish acoustic offset, should any processing of the FR be done? Minimum phase? Gating? I use arta, which doesn't give you a FR, it gives you an impulse, which you then gate and generate a FR.

[dcibel]

dcibel,

The cabinet is about 12.6" wide. The woofer is the 10" (its the SB Acoustics SB29NRX).

Ok, so if you read through Jeff's paper, he says for an accurate far-field you should be at the very least 3 piston diameters, and at least 2 times the cabinet width, which puts you at about 30" minimum in your case, so the 40" distance for your measurement should be okay for far field measurement. However since you are taking all your measurements at the tweeter axis, you may have some gating issues to completely remove reflections from a woofer that is near the floor. The upper limit of the near field measurement for a 10" driver will be about 430Hz, so keep that in mind when merging the response.

Additionally, consider that a woofer next to the floor will have some boundary reinforcement from the floor, so applying a full 6dB baffle step to the near field will not be correct in this case.

Taking measurements at the tweeter axis is apparently the correct way to go for PCD, since it does not correctly compensate for the off-axis movement when you input the driver offsets, only phase is adjusted, the frequency response and amplitude are not so all the far field measurements should be from the same point in space. I have found SoundEasy to be more accurate in this regard, allowing use of measurements that are on-axis with each driver, and correctly adjusting the response when the drivers are placed on the baffle offset from the mic axis. This is beneficial, as it allows you to bring up the woofer from the floor for measurement purposes, as well as proving more accurate off-axis simulations.

If you do have issues with a floor reflection when taking your far-field, you may consider trying to merge 2 measurements to create the far-field. Simply lift the driver off the floor (4ft is ideal if you have 8ft ceiling), and take the far field measurement that doesn't have the reflection, then merge that with the measurement on the tweeter axis, so you get a far field that doesn't include the floor reflection, but does include the correct upper frequency data from the measurement at the tweeter axis. Then merge this with the near field for the low frequency data. I haven't attempted this myself, so you may have to experiment a bit to see if this method can work well or not.

[dkalsi]

My understanding is that you should move the mic so it is on axis and the same distance from the baffle for each driver,when it is your intention to extract minimum phase measurements. These measurements will then need to be positioned in your simulation software (PCD, VituixCAD, LSPCad, whatever). The time when you don't want to move the mic is when you're taking a series of measurements to glean the relative acoustic offset of your drivers - in this case you don't want to extract minimum phase (someone correct me if I'm wrong).

The measurement protocol described by Bagby for in-room anechoic response works well. Here are my reactions to your notes:

1) Taking Tweeter, Mid, Woofer measurements from a distance of 40" on tweeter axis
First, 40" is further than necessary. I like to do 2x baffle width + a little bit more. This might get you a longer gate. Also, these should be done on each driver's axis. If you do them all from the tweeter axis you are essentially designing with a woofer and mid that have treble rolled off.

Civit,

Sincere thanks for responding and sharing your experience. From what you describe, it makes sense why PCD, winPCD, and VituixCAD require driver placement information and radiating diameter information.

I can't remember the exact tread, but I had saved the following post by Jeff B:

"Well, I wouldn't have placed the mic between the two drivers on the two-way. Personally, I feel any axis other than the tweeter axis is less than ideal. So, for a three-way, and I have done many, I would recommend that the mic be on the tweeter axis here too. The bottom line is that the design axis should be the axis you plan to listen on. Most people listen on the tweeter axis due to the tweeter’s roll-off as you move off-axis is much more significant than the woofer’s is. I wouldn’t recommend that someone design with the mic between the two drivers and then listen on the tweeter axis. Their response on this axis will be a little different than what they saw in their design work."

Again, there is a high likelyhood that I might be misinterperting Jeff's post. I assume Jeff was taking all measurements the design axis (i.e., tweeter level).

With typical close proximity between the tweeter and the mid, I would assume the mid treble would not terribly rolled off. I couldn't imagine the mid being more than 10 degrees off axis??!? In regards to the Woofer, even if we are 60 degrees off axis, it really shouldn't matter around 200hz - 300hz correct?

I guess I also need to interpret how the modeling software is analyzing the files. By taking all measurements on axis, does it calculate that the what the response should be on the design axis (based on placement info and radiating diameter info).

[civit]

Again, there is a high likelyhood that I might be misinterperting Jeff's post. I assume Jeff was taking all measurements the design axis (i.e., tweeter level).

With typical close proximity between the tweeter and the mid, I would assume the mid treble would not terribly rolled off. I couldn't imagine the mid being more than 10 degrees off axis??!? In regards to the Woofer, even if we are 60 degrees off axis, it really shouldn't matter around 200hz - 300hz correct?

I guess I also need to interpret how the modeling software is analyzing the files. By taking all measurements on axis, does it calculate that the what the response should be on the design axis (based on placement info and radiating diameter info).

I think you have a good handle on the tradeoffs. Measuring your mid and woofer on the design axis will probably not effect their FR in the range you intend to use them, but I would still follow the protocol I described (measure each on axis at same distance, merging with nearfield, minimum phase) simply because it allows more accurate simulation of off axis behavior and perhaps more importantly, a more accurate depiction of the power response. Again, however, if you measured from one point, and you were a good distance away, the sine error there would not be huge and you could probably get away with it (..but the distance might effect your gating...etc). If you're design is well thought out, the axial response might be enough, but if you're doing an WWMTM or a dipole or something, you'd want to measure on axis (and do polars.)

[civit]

Taking measurements at the tweeter axis is apparently the correct way to go for PCD, since it does not correctly compensate for the off-axis movement when you input the driver offsets, only phase is adjusted, the frequency response and amplitude are not so all the far field measurements should be from the same point in space.

I didn't know this. It seems like measuring from as far as is practical would be beneficial in this case. TIL.

[dcibel]

Civit,

Sincere thanks for responding and sharing your experience. From what you describe, it makes sense why PCD, winPCD, and VituixCAD require driver placement information and radiating diameter information.

I can't remember the exact tread, but I had saved the following post by Jeff B:

"Well, I wouldn't have placed the mic between the two drivers on the two-way. Personally, I feel any axis other than the tweeter axis is less than ideal. So, for a three-way, and I have done many, I would recommend that the mic be on the tweeter axis here too. The bottom line is that the design axis should be the axis you plan to listen on. Most people listen on the tweeter axis due to the tweeter’s roll-off as you move off-axis is much more significant than the woofer’s is. I wouldn’t recommend that someone design with the mic between the two drivers and then listen on the tweeter axis. Their response on this axis will be a little different than what they saw in their design work."

Again, there is a high likelyhood that I might be misinterperting Jeff's post. I assume Jeff was taking all measurements the design axis (i.e., tweeter level).

With typical close proximity between the tweeter and the mid, I would assume the mid treble would not terribly rolled off. I couldn't imagine the mid being more than 10 degrees off axis??!? In regards to the Woofer, even if we are 60 degrees off axis, it really shouldn't matter around 200hz - 300hz correct?

I guess I also need to interpret how the modeling software is analyzing the files. By taking all measurements on axis, does it calculate that the what the response should be on the design axis (based on placement info and radiating diameter info).

Well tweeters will have narrowing directivity in the top octaves, so keeping the tweeter as the design axis ensures that the high frequency response will only drop as you move off axis. If your design axis is not the tweeter axis, the high frequency response can be a bit less predictable. Your question about off-axis response I think are best answered by using your mic and taking some measurements so you can observe the real-world data for yourself.

Like I said, the offset values in PCD only affect the phase so that the simulation adds in the correct "time of flight" to the minimum phase that you've provided, the off-axis simulation when you move the mic around is an approximation at best.

[dcibel]

I didn't know this. It seems like measuring from as far as is practical would be beneficial in this case. TIL.

Yes, that's correct with the exception that the further you move away, the more reflections are introduced to the measurement, so you want to be as far away as possible while providing as large a gate as possible and still provide a reflection free measurement. It's a catch 22, so indoors you end up being as close to the speaker as possible while still qualifying for "far field" status.

[dkalsi]

Just to be clear, any extraction of minimum phase should occur ONLY after merging from far-field and near-field correct??? From what I understanding, minimum phase calculations are impacted by the "tails" of the FR.

I had saved the following from DLR's post:

"Remember that offset is relative, that is, it's a value required so that the two minimum-phase files sum to the measured sum. If you change either file, the relative offset needs to be updated to account for the change.

The absolute value of the offset isn't really important. All it amounts to is the difference between the two measurements that "aligns" the time delays between the measurements to correspond to the difference required for the generated phase of each driver file. There is no "correct" offset, only the offset needed for the measurement phase as generated for the SPL of the files."

[dcibel]

Just to be clear, any extraction of minimum phase should occur ONLY after merging from far-field and near-field correct?

Minimum phase is a direct result of the frequency response, so it is the last step of the process, always.