Accoustic Offset/Minimum Phase in Speaker Workshop

[mpotoka]

In Manual 2 for Speaker Workshop, it talks of a spreadsheet for converting a speaker workshop file into one usable by FRC combiner. Is that still the method to use for extracing minimum phase from measurements? If so, can someone send me a link to that specific spreadsheet?

It seems that with taking both measurements (2 way design) from the same location, the offset is already "in" the data. How do you actually go about using the data to figure out the proper offset?

[dlr]

In Manual 2 for Speaker Workshop, it talks of a spreadsheet for converting a speaker workshop file into one usable by FRC combiner. Is that still the method to use for extracing minimum phase from measurements? If so, can someone send me a link to that specific spreadsheet?

It seems that with taking both measurements (2 way design) from the same location, the offset is already "in" the data. How do you actually go about using the data to figure out the proper offset?

I don't know about the spreadsheet. You are correct about the offset if, and only if, the measurement system also measures phase. If you have to generate phase then that data is lost if you don't take measures to re-create the phase to match measured phase.

If you have measured phase there are a couple of ways to create a model that recovers the phase. Visit John K's site for a discussion of one way to do it that may be easier and only requires the measurement for each driver alone.

John K's site

Drivers and Phase

You can, however, find the AC without ever having measured phase. My site has an article that shows how you can do this. You need three measurements for this, the two drivers individually and one with both drivers connected.

Keep in mind that the model you create (assumed highpass and lowpass slopes and Fc) alter the phase generated, so be sure to establish that up front. My article does not touch on this aspect.

[mpotoka]

Thanks for the links.. It has helped me understand phase and its issues, and will continue to do more reading.

I was hoping someone could help me take a picture like this:



And get the minimum phase from it. Any SW users out there? Perhaps it is just time to bite the bullet and get Soundeasy or lspcad

[ThomasW]

Sorry I've never used SW I'm just dropping in links and we'll make this a reference thread..

http://chadgray.info/soundeasy/index.cfm

http://www.rjbaudio.com/Audiofiles/FRDtools.html

[Jed]

Thanks for the links..


And get the minimum phase from it. Any SW users out there? Perhaps it is just time to bite the bullet and get Soundeasy or lspcad

Using the FR combiner from the FRD consortium, import the FRD file, then click on combine response, then click on calculate minimum phase, then process the zma file by clicking the import impedance extrapolate button. You'll have data for FRD and ZMA with minimum phase when you are done. Then in speaker workshop you'll have to calculate the acoustic center distances, and you can find a neat calculator for that over at www.rjbaudio.com under the audiophiles section.

Hope this helps,

Jed

[Dave Bullet]

I do exactly what Jed describes, that is export the dataset from SW as an FRD, then import into the FRC spreadsheet and "extract minimum phase".

The question I have is why do you want to extract minimum phase if SW has already measured the phase for you?

If you want to just use SW to measure drivers then model a crossover in another program, then ok.

You probably already know this.... the reason your picture has lots of phase wrapping, is your start marker for your measurements is too early. Try and place your start marker closer to the first impulse spike (take a pulse measurement). You may reduce resolution of the FR measurement, but your phase will wrap less (less flight time between signal being sent and measurement being received and processed by SW) making it easier to pick a crossover point where both drivers relative phase is tracking "best". (in addition to all the other criteria when choosing a crossover point).

Cheers,
David.

[Dave Bullet]

In Manual 2 for Speaker Workshop, it talks of a spreadsheet for converting a speaker workshop file into one usable by FRC combiner. Is that still the method to use for extracing minimum phase from measurements? If so, can someone send me a link to that specific spreadsheet?

I think you are referring to the program (not a spreadsheet) emedded in page 148 (5th volume) of the SW manual. It is a self extracting EXE. Written in italian. After installing, run the program, and choose "File -> Lingua" from the menu, and open up the English.LG file to get an English user interface.

David.

[rc white]

In speaker workshop If you put a microphone at the point you want the acoustic centers to coincide, and then put a sine wave at the nominal crossover frequency into each driver alternately, the two traces you obtain will be offset in time equivalent to the acoustic offset, the offset is then..

(343*longer time) - (343*shorter time)

[dlr]

I do exactly what Jed describes, that is export the dataset from SW as an FRD, then import into the FRC spreadsheet and "extract minimum phase".

The question I have is why do you want to extract minimum phase if SW has already measured the phase for you?[/B]

I was thinking that SW measured phase. The measurements can be used directly, if so, of course. In that case, ignoring the wrapping is the thing to do. Display of phase can be turned off in SW, correct?

There is one other reason to create a model with minimum-phase and the correct AC. This would be if you want to change the design point to somewhere other than the measurement mic position. In that case, the change in excess-phase would have to be calculated manually and entered, an easy area for mistakes.

[complet]

In Speaker Workshop you can get minimum phase, by choosing "Transform" -> "Delay" and mark "Remove excess delay" and then enter the time of flight in mS.
But why?
I recommend measuring both drivers from one mic location. Usually 1m at the tweeter axis (listening height).
First measure pulse respons, set the first time marker at 0, and the second marker just before the first reflection. (Its true that the phase wrappes a lot when measuring from 0mS, but it also give you a lot better resolution of the frequency plot.)
Then measure "on axis".
Turn off the phase data in the plot, so it dont confuses you. SW uses it in the calculation anyway.
This way you have the measured real phase to work with, not a calculated one that can be more or less acurate.
It can be tricky to place the mic in the exact same distance from both drivers, and if you miss, minimum phase will be wrong.

[mpotoka]

Ok--well thanks for all the information/replies. A couple things

1. I did know that a lot of my phase wrapping is supposedly a cause of having my first marker too far before the initial impulse. In the past, no matter how close to the impulse I have moved the marker, I still have not gotten rid of much of the wrapping.

2. I have attempted to use FRC with an exported .frd file, whenever I try to extract minimum phase, I always get the red-light error at 11.9% completed. I tried it with two different computers, same issue on both.

3. From my reading, it is suggested that phase is an important thing to look at when choosing a crossover point. When I see major wrapping like there is above, it is quite difficult to learn much from it. Or am I supposed to use the phase date on an impedance measurement?

4. What exactly is the function of the Hilbert transform then, or don't I need it for anything when designing in SW?

5. Is the phase alignment of measurements important only for purposes of modeling crossovers?

[rc white]

Acoustic offset also varies with frequency, I measured one nominal 5.5 inch driver as having 14mm. with respect to a metal dome tweeter at 1.5kHz. 22mm. at 3kHz.
As you might expect this is due to the outer cone sections decoupling and the acoustic center moving closer to the cone apex, it is also a non minimum phase effect since it is in effect an all pass filter added to the acoustic transfer function.
If you just take the Hilbert transform of the frequency response then this is not being taken into account and 8mm. is 25degrees of phase shift at 3kHz., this is quite a sizable error.

[Dave Bullet]

Ok--well thanks for all the information/replies. A couple things
2. I have attempted to use FRC with an exported .frd file, whenever I try to extract minimum phase, I always get the red-light error at 11.9% completed. I tried it with two different computers, same issue on both.

I think this is a bug with FRC (well - software shouldn't "crash" like that IMHO). To fix this, click Normaliser, and under the response curve where you have loaded your FRD file, click the top button that says "extend top slope" until you get "resample only - preserve bandwidth", then click "execute" under the same response curve. Once resampling is complete, go back and extract minimum phase - and it won't crash - i guarantee it :B

4. What exactly is the function of the Hilbert transform then, or don't I need it for anything when designing in SW?

You need to know the phase of each driver so you can design a crossover where each driver where each driver is in phase through the crossover region. Drivers are typically minimum phase devices, which means the phase can be calculated from the FR curve (via a HBT). Some software uses the HBT approach and as DLR correctly stated this can be useful if you want to change (in simulations) the mic axis to simulate different off-axis responses.

However - in SW case - it measures the actual phase for each driver. You therefore do not need the FRD tools or to do any post processing. It's all there ready for you to model crossovers.. with 3 important cavaets:
1. Your pulse responses must be consistent between measurements. A wandering impulse spike in a pulse measurement means measured phase will vary therefore cannot be used to model a crossover

2. You must not move the start marker between measuring the woofer and tweeter, otherwise the phase changes between measurements making them useless for crossover design.

3. You must not move the mic or relative driver position. For example if you measure a tweeter with microphone on axis to the tweeter, then disconnect the tweeter to measure the woofer - leave the microphone where it was (ie. still on tweeter axis) and measure the woofer).

5. Is the phase alignment of measurements important only for purposes of modeling crossovers?

Yes. In regards to frequency response as I described above. When it comes to impedance phase, high swings indicate a difficult load for the amplifier, especially if the impedance is low and it is in the bass region. but happy to stand corrected here.

David.

[dlr]

In Speaker Workshop you can get minimum phase, by choosing "Transform" -> "Delay" and mark "Remove excess delay" and then enter the time of flight in mS.
But why?

It's really only necessary to use a model with minimum-phase if the design is be to made or checked at some point other than the mic position. This can be useful.

[mpotoka]

The FRC worked like you said--Thanks.

It looks like I can stick with SW, if i can resovle my latency issues with my firewire card. I really need a better room for measuring in anyway.

One other question concerning speaker workshop--when you disconnect the tweeter, for instance, to then measure the woofer. Do you re-match so the VU meter values are the same, or do you make say the tweeter match, then when the woofer is measured will it give you the relative sensitivity?

[dlr]

Acoustic offset also varies with frequency, I measured one nominal 5.5 inch driver as having 14mm. with respect to a metal dome tweeter at 1.5kHz. 22mm. at 3kHz.
As you might expect this is due to the outer cone sections decoupling and the acoustic center moving closer to the cone apex, it is also a non minimum phase effect since it is in effect an all pass filter added to the acoustic transfer function.
If you just take the Hilbert transform of the frequency response then this is not being taken into account and 8mm. is 25degrees of phase shift at 3kHz., this is quite a sizable error.

Actually, the AC as defined and used in CAD software does not vary with frequency. If it did, then the Hilbert-Bode transform would fail, since the FR would not be minimum-phase and this is a requirement for the HBT. Any variations are likely due to improper model setup prior to executing the HB transform. In addition, no one has yet been able to precisely determine the AC of a driver. The AES working group assigned to that some years ago closed with no success. The problem is the dependence of the HBT on the model where assumptions must be made.

I have yet to have any piston driver exhibit non-minimum-phase behavior.

[dlr]

You need to know the phase of each driver so you can design a crossover where each driver where each driver is in phase through the crossover region. Drivers are typically minimum phase devices, which means the phase can be calculated from the FR curve (via a HBT). Some software uses the HBT approach and as DLR correctly stated this can be useful if you want to change (in simulations) the mic axis to simulate different off-axis responses.

Be careful of the phase requirements in a crossover design. The crossover region may not have in-phase drivers. You're probably thinking of even-order XOs such as BW2, BW4, LR2 and LR4. However, if the design is other than these, the two XO legs will not be in phase. Odd-order Butterworth is in phase quadrature (90 degrees out-of-phase) and other XO types (usually not used for speakers, but possible) will have other phase requirements.

[Dave Bullet]

One other question concerning speaker workshop--when you disconnect the tweeter, for instance, to then measure the woofer. Do you re-match so the VU meter values are the same, or do you make say the tweeter match, then when the woofer is measured will it give you the relative sensitivity?

No - don't change your pre-amp or sound card levels between measurements. This is becasue you want to measure the tweeter as it is, then the woofer as it is. Most tweeters are more efficient than most woofers. You want to see a higher tweeter measurement, which you then pad down in the crossover. for example below are 2 raw measurements I took. The RED is a Wharfedale 8" woofer, the BLUE is a Seas 27TBFC/G tweeter. I kept the same mic placement, speaker enclosure placement, sound card and pre-amp levels throughout.

David.