Phase correction???

[jdlech]

I have a project in which the mid and tweeter are out of phase at the crossover frequency of 3K by about 45 degrees. I don't want to spend endless hours obsessing on my crossover if there's a simple physical solution.

The wavelength of a 3K signal is about 4.523 inches. 45 degrees is 1/8 wavelength. So if I place my tweeter 4.523/8=.565 inches back from the midrange, they should end up very nearly in phase.

Just over 9/16 inch? Is this right? Or am I hopelessly simplifying the problem?

 

[jupiterjune]

How do you know they are 45 degrees out of phase?

JJ

 

[dlr]

I have a project in which the mid and tweeter are out of phase at the crossover frequency of 3K by about 45 degrees. I don't want to spend endless hours obsessing on my crossover if there's a simple physical solution.

The wavelength of a 3K signal is about 4.523 inches. 45 degrees is 1/8 wavelength. So if I place my tweeter 4.523/8=.565 inches back from the midrange, they should end up very nearly in phase.

Just over 9/16 inch? Is this right? Or am I hopelessly simplifying the problem?

Yes, it's right, that's the most direct way. The distance is correct if you have measured responses that show this phase delta with the crossover in place and if you are making a Linkwitz-Riley crossover. Another benefit of an offset is that this physically aligns the drivers so that the off-axis response also maintains a better phase relationship between drivers.

Dave

 

[jdlech]

Thanks for the confirmation....

Xover Pro 6 and Basta both show phase predictions.

Alas, it's a predicted response, not actual measurements. But if reality is anywhere near prediction, then a small offset could do no harm - or so goes my thinking.

My next xo point is 743Hz, which is an 18.2" wavelength. Luckily, my prediction places that crossover point at only 4 degrees out of phase - not enough to be of any concern to me.

My third xo point (between mid bass and bass) is at 281Hz, which is a wavelength of 48.1". My software varies quite a bit with this one. Currently. the mid bass leads the sub by about 90 degrees. OK, this is a problem - but the model only shows a 2db sag (something I find perplexing - shouldn't it be more?). However, to fix that problem, the mid bass would have to be a full foot back from the bass drivers. Now we're talking some substantial distance - at least some of which I would want to correct via the xover. To be honest, I see this as being a good argument for separate subs and an equally strong argument against them. Just a half foot and bass gets all messed up. One can screw it up either way.

Secondly, let me know if I'm getting off track here too. I'm thinking on creating something on the order of a 2" x 9/16" slope in the face instead of a discreet step. Logically, to minimize edge interferance should I want the nearest edge to be longer than a full wavelength away from the center of the driver? So that the upper edge of the slope ends at least 4 1/2 inches from the center of the tweeter and the bottom ends least 4 1/2" from the center of the midrange driver? (actually, I'm thinking that it should be more for the mid)

I'm intuiting my way through this one, (throwing chicken bones again) could you please tell me if I'm on the right track or not?

This is somewhat inspiring as I imagine form following function but that's a matter for another thread. Meanwhile, what would be the cheapest method of taking actual phase meaurements? (I have yet to hit the lottery)

The midrange drivers I was looking at using are on sale , so I ordered 8 of them. That's about the only thing this project has set in stone so far.

 

[bzfcocon]

Thanks for the confirmation....

Xover Pro 6 and Basta both show phase predictions.

If that's the predicted response, can't you set a physical offset in the simulations ? I am not familiar with your software, but most of them allow you to specify the physical alignment of the drivers until the phase matches.

 

[jnb]

It sounds like you're on the right track but you don't have good data to work with. You could get a Behringer mic and build a phantom power supply and cabling, or you could buy a cheap RS SPL meter and download a compensation file. Whatever you do, keep up the good work but don't take your results as being accurate, yet...

Physically misaligning the drivers gives a frequency dependent phase change. IMO, you are (generally speaking) better keeping your voice coils aligned and massaging the phase in the crossover. This may get tricky but I have always found a way. Same goes for your 281Hz example.

BTW, you are building to a target response/phase. The components (values and number of) are irrelevant if you take the meaning.

 

[jdlech]

It's true that bassbox pro and xover pro assumes all drivers are on the same plane and I seriously doubt they take baffle step into consideration (there's no mention of it anywhere in the documentation). Nor does it use, or even allow for, actual phase measurments. Nor does xover pro model cascaded xovers; just all in parallel. Nor does it model anything over a 3 way. I'm having to do two models; one with the three lower drivers and one with the three upper drivers - each is missing a driver.

The software does accept downloaded measurments for frequency response, and I have done so for all but the tweeter I'm looking at using . (why, oh why do so many tweeter mfgs not think we need actual measurments?) But I did type in a rough approximation based upon the response graph they published. All else are using detailed F.R. curves in addition to T/S.

Even so, I agree that I should try to take some actual measurments with them all on the same plane before deciding to put any slopes in. Although it could get interesting to build... But I need to keep somewhat within my limited woodworking skills.

So I should get something like
this
and one of these
All of which seem pretty economical to me. But then, how would I make the measurments? Is there some software for that? I already have a PC tone and white noise generator. Though I think I'd be better off getting one of these earlier than expected. (let's face it, most all pc sound cards blow, imo, but it'll do until I get some good speakers together. But not if I'm going to use a PC as a test tone generator.) Or maybe....

As a BTW, I'm still learning to use the great tools recommended here and at the FRD consortium, but they have a learning curve. It'll be a while before I make the changeover, but I'm working on it.

 

[jnb]

As with many great tools, the FRD tools have a learning curve but some, I know, were written by enthusiasts and will help you build a good crossover.

If your software assumes all drivers are on the same plane, you can fudge it by applying the appropriate delay to the phase response. If your software doesn't do phase, it may not be the right tool for building this kind of crossover.

Any published curve can be converted to an FRD file using the SPLtools.

The Behringer is a good mic. For software, you could try Arta, and there are many others. Arta does spectrum analysis, CSD, MLS etc, and generates all the right noises.

WRT sound card quality, for response testing most sound cards will perform well enough.

 

[Iain McNeill]

It's difficult to do phase correction in an analog active crossover. It's much easier (but still not simple) to do it with a digital active crossover.

You need the actual phase measurements of your drivers in your cabinets. I'm not sure manufacturers published "typical" responses are good enough.

Having said that, the jury is still out on whether phase distortion is even audible (i.e. is it worth the hassle?)

My opinion is that it's very hard to hear any error due to 100us of driver misalignment due to acoustic centers offset. However, I think the errors caused by driver phase interfering with the crossover target phase is audible. I am still trying to build an experiment to prove this to myself.

If you're not experienced with phase measurements I wouldn't worry too much for now. They can make your brain hurt. Design your XO right (with BSC) to produce a complementary (constant voltage/perfect reconstruction) filter set and keep the crossover points away from driver high pass functions (i.e. bias XO point high) You'll be 99% there.

 

[jupiterjune]

It's difficult to do phase correction in an analog active crossover. It's much easier (but still not simple) to do it with a digital active crossover.

Not if you know how! I use crossover 3 pro, and have a relatively easy way to do it. But of course, you have to measure, and compare the results to where crossover 3 pro is telling you where the phase should be to make tweaks to bring everything into phase.

You can simply reverse the tweeter and check the frequency and depth of the null to see how closely you have the phases aligned.

JJ

 

[jdlech]

Well, after a bit of tweaking, here's what I have so far...
[IMGDEAD]http://home.comcast.net/~jdlech/Images/ResponseCurve.JPG[/IMGDEAD]
Don't worry about that 14K spike, I'm going to use a notch filter on it, but xover pro cannot incorporate notch filters in the graph (though it can help design them).

Plus, xover pro cannot manage more than a 3 way xover, so the subs are missing. this is the top 3 drivers in a 4 way xover. (ah, the limitations of design software) Which is why I'm looking to quit using it.

Here's the response curve with the midrange inverted.
[IMGDEAD]http://home.comcast.net/~jdlech/Images/Inverted.GIF[/IMGDEAD]
The deep sags at the xover points shows the drivers are nearly 180 Deg. out of phase - cancelling each other out.

Having said that, the jury is still out on whether phase distortion is even audible (i.e. is it worth the hassle?)

And phase alignment is not just important for the response; as I understand it, phase distortion between two drivers 'blurs' the psychoacoustic information. Imagine listening to a single violin, but with phase distortion, your brain is receiving mixed signals as to the location of the violin. You won't be able to tell whether the violin is 4 feet or 10 feet away from you. And in worst case scenarios, you wont be able to tell whether it's left of center or directly in the center of the soundstage. Whether you want to call it imaging or soundstage, phase alignment is apparently important to keeping it clear and distinct.
At least, that my understanding.
What is unclear is if we use the lower frequencies to derive any significant imaging information.

It's been proven that even minor phase shifting is detectable by the human ear. How and why are not as important as the fact that it is.
Try it for yourself
I guess this page is about as good an explanation as I've found in 5 minutes of looking.

 

[jnb]

Plus, xover pro cannot manage more than a 3 way xover, so the subs are missing. this is the top 3 drivers in a 4 way xover. (ah, the limitations of design software) Which is why I'm looking to quit using it.

The FRD consortium offers a crossover simulator which uses arbitrary circuits for crossovers and uses phase information properly for both impedance and acoustic sides, making it a true simulator. It was produced by an enthusiast, so it may not seem as smooth as a commercial offering but don't think this makes it less useful, quite the opposite.

And phase alignment is not just important for the response; as I understand it, phase distortion between two drivers 'blurs' the psychoacoustic information.

As I understand it, phase distortion is any deviation from flat (or equal group delay, e.g. where you may be sitting one bass wavelength away from the speakers which happens to be several treble wavelengths which causes your phase plot to "wrap" around the chart, but when unwrapped by removing the delay, the chart is actually still flat).

Making sure your drivers correlate in phase to each other may share a connection to phase distortion but really can be approached as an independent issue.

Correlation errors will manifest in lobing issues. Any null you witness on axis with the polarities reversed will happen somewhere off axis with the polarities correct. Consistency in this lobing is an issue considered when comparing different textbook slopes.

That said, phase correlation on axis alone is a good start.