Question about phase measurement for the purpose of crossover design

[percy]

Can anyone please clear this up for me -

If the acoustic measurement of both drivers (woofer and tweeter) of a 2-way are taken with the mic as well as the drivers always in a fixed position, drivers mounted on the actual baffle, then for the purpose of crossover design do I need to remove the excess phase or convert it to minimal phase ? or does it not matter ? I am a bit unclear as to what type of phase should I export to the frd.

Measurements will be taken using ARTA in dual channel mode and exported to a frd file which will be used either in Speaker Workshop or PCD spreadsheet.
Please advise.
Thanks!

[a.wayne]

Convert to min phase ..............

[speaker dave]

We subtract phase to make the curve more legible. 1 or 2 meters of phase would be a lot of rotations that would dominate the curves and obscure the relative difference between drivers. There are two ways to do it and only one will work for the particular simulation you want to do.

If you do minimum phase on both drivers then you lose the view of how one unit (usually the woofer) has more phase shift than the other. This will be wrong... unless your program expects you to enter the effective depth of each unit. That is how far behind the baffle the acoustic center or voice coil center is.

If , on the other hand, your drivers are defined as being on the surface of the cabinet in your simulation, then the only proper way is to subtract exactly the same airpath delay for both units. This will retain the relative phase shift between the two. I usually take this approach and minimum phase the woofer (deeper unless you are using a horn upper unit). and then subtract exactly the same time or phase for the second unit.

You really have to decide which approach your software requires and follow these directions explicitly.

David S.

[AllenB]

Removing the excess delay is a good move, but it shouldn't matter. If you don't, however, it will be difficult to judge by eye when tweaking the crossover.

[dlr]

Originally Posted by percy

Can anyone please clear this up for me -

If the acoustic measurement of both drivers (woofer and tweeter) of a 2-way are taken with the mic as well as the drivers always in a fixed position, drivers mounted on the actual baffle, then for the purpose of crossover design do I need to remove the excess phase or convert it to minimal phase ? or does it not matter ? I am a bit unclear as to what type of phase should I export to the frd.

Measurements will be taken using ARTA in dual channel mode and exported to a frd file which will be used either in Speaker Workshop or PCD spreadsheet.
Please advise.
Thanks!

I have not used SpeakerWorkshop much, but I do know the PCD rather well. IMO you can simply use the measurements directly, but this is limited. The phase will show extra wraps, but I've not found that to be a problem. My comments will all relate to the PCD.

First, you have to decide whether or not you'll want to create the model so that you can investigate the response at points other than the measurement point. You can do that only if you create a minimum-phase model and then model the x,y,z axes correctly, the z-axis (acoustic offset) being the only difficult one (depth).

If you work only on the measurement point with direct measurements, then you would leave the x,y,z at (0,0,0) for both drivers and use the measured phase. Moving off-axis will introduce additional excess-phase between the drivers with increasing error as angle increases. Small changes probably are not a real issue.

You could do as David (speaker dave) says as well, removing some of the excess-phase (distance). It's a step that I have not found to be required, but you may want to do that.

You cannot use minimum-phase without other data. If you use that, then you'll need to then use the x,y,z for each driver in relation to the measurement point. If you measure on the tweeter axis, then you'll likely enter (0,0,0) for the tweeter and use appropriate offset values for the woofer. This makes the tweeter the reference of origin, so-to-speak. Other drivers are made relative to this. You could just as easily do it with the woofer. In either case, you'll need to determine the z-axis offset, either by estimating (can work) or measuring it.

On the latter point, it's really a fairly simple matter to determine the appropriate relative acoustic offset with a third measurement taken at the same time. There's an explanation of this method at my site here. Just be aware that I did that with a direct connection to the tweeter, so use low volume settings. That will work fine.

Once you have this done accurately, you can then consider any axis for design, with the caveat that when off-axis, there will be some driver rolloff and diffraction changes taken into account, but I wouldn't be concerned with those at this stage.

Dave

[percy]

Originally Posted by dlr

First, you have to decide whether or not you'll want to create the model so that you can investigate the response at points other than the measurement point.

I didn't follow this part. What do you mean "decide" ? One would want a speaker to perform equally well at any listening position, right ? What am I not understanding ?

Originally Posted by dlr

You cannot use minimum-phase without other data. If you use that, then you'll need to then use the x,y,z for each driver in relation to the measurement point.

you mean just the acoustic center(voice coil) offset with reference to the other driver, right ?

Also, what about total "system delay"(computing latency, flight time). Do I factor that in ? or should not ?

Originally Posted by AllenB

Removing the excess delay is a good move, but it shouldn't matter. If you don't, however, it will be difficult to judge by eye when tweaking the crossover.

what exactly would I be judging ? the frequency/amplitude response or the phase ? If amplitude response, and if that difference is in such miniscule range that its not noticeable to the eye on the chart, then would it be noticeable to the ear ?

[dlr]

Originally Posted by percy

I didn't follow this part. What do you mean "decide" ? One would want a speaker to perform equally well at any listening position, right ? What am I not understanding ?

It's about how you set up the model in the software and how flexible it is. One way will accurately predict only one point in space (measured phase). The other will reasonably predict any point in space with regard to the crossover and acoustic center.

you mean just the acoustic center(voice coil) offset with reference to the other driver, right ?

Yes, the absolute centers can't be accurately determined, but the relative offset (as it relates to the driver models used) can be. Once you create the model, such as the lowpass of the raw driver, don't change it after you determine the offset.

Also, what about total "system delay"(computing latency, flight time). Do I factor that in ? or should not ?

You won't need to do anything else once you've set up the model accurately. Keep in mind that none of these methods takes driver directionality nor baffle diffraction fully into account. It does for the measured point, though.

Dave

[AllenB]

Originally Posted by percy

what exactly would I be judging ? the frequency/amplitude response or the phase ? If amplitude response, and if that difference is in such miniscule range that its not noticeable to the eye on the chart, then would it be noticeable to the ear ?

The idea is only to have the drivers close to being in phase around the crossover. They don't have to be exact.

But lets say they are just over 90 degrees out of phase over some portion of the crossover region. If you adjust the on-axis crossover to achieve a level response, there will be more cancellation happening than not. It would work, but not in the best way.

So, when you go back to tweak that crossover, you'll probably be trying to get each driver's individual crossover to shift a little in phase in that region while at the same time dropping a little in amplitude. You'll be doing this using small adjustments and you'll want to see the phase plot clearly enough.

The problem with an unwrapped phase plot starts when the wraps begin to happen as often as the resolution of the plot. They won't look like wraps anymore, just random up and down lines. Even when unwrapped there can be some confusion when they reach past 180. At least if the plot is unwrapped you can estimate the curve.

[dlr]

Originally Posted by AllenB

The idea is only to have the drivers close to being in phase around the crossover. They don't have to be exact.

It should be pointed out that this works only for even order crossovers, such as LR2 and LR4. BW1, BW3 and higher, though less often used, are in phase quadrature throughout the crossover. The drivers are never in phase on the design axis at any point for Butterworth.

Dave

[a.wayne]

Originally Posted by dlr

. If you measure on the tweeter axis,

Why would you not want to measure on the tweeter axis ...?