"The last niggling doubt or unease I have is that the reverse-polarity summing is not giving me a deep notch... the notch I'm getting is barely 4-5dB deep."
This is where you can start tweaking your component values to improve the phase alignment at the crossover region.
The chart below is kind of hard to read due to all the graphs. But if you can set up chart showing the phase and amplitude of each driver, plus the total amplitude, you can see how the phase changes as you adjust component values. The dotted lines are phase, and once you get them to overlay each other, you will find your reverse notch will get much deeper.
You may also have to compromise between the best phase and the flattest amplitude...getting one may not give you the other.
Congratulations so far.
This is where you can start tweaking your component values to improve the phase alignment at the crossover region.
The chart below is kind of hard to read due to all the graphs. But if you can set up chart showing the phase and amplitude of each driver, plus the total amplitude, you can see how the phase changes as you adjust component values. The dotted lines are phase, and once you get them to overlay each other, you will find your reverse notch will get much deeper.
You may also have to compromise between the best phase and the flattest amplitude...getting one may not give you the other.
Congratulations so far.
An externally hosted image should be here but it was not working when we last tested it.
This is the kind of clear, operational-level instruction which I think I can safely try out even if my concepts are not clear.jbateman said:
Thanks. I'll try this tonight and see what it shows.BTW, I've put up the new SWD file with the phase-coherent acoustic measurements. You can see the xo and how the reverse notch is very shallow. Check here.
I looked at phase information and tried to make some sense of it.
I checked the delay between 0msec and the start of my MLS pulse in my SPL measurements, and that turned out to be 4.5msec. So I applied the "Delay..." function to both the tweeter and woofer, unchecked the "Remove Excess Delay" option, and applied 4.5msec of delay removal to both the acoustic graphs. The magnitude and phase of the woofer now looks like this:
The tweeter mag+phase looks like this:
I then plugged them into the 2K LR4 xo's that I'd designed, and computed the SPL and phase again for those networks. What I got for the final mag+phase is shown below. The low-pass xo is:
The high-pass xo is:
At the xo point of 2K, the low-pass xo has a phase of 115 degrees and the high-pass xo has either -180deg or +180deg, depending on how you look at it. Therefore, I can say that the phase difference between the low-pass and high-pass xo at the xo point is 65deg.
If I now switch the polarity of the tweeter, the phase difference increases, which shows up in the small notch in the SPL curve.
Have I handled the delay removal etc correctly?
How do I get the phase of the woofer and tweeter to align better than 65deg?
I checked the delay between 0msec and the start of my MLS pulse in my SPL measurements, and that turned out to be 4.5msec. So I applied the "Delay..." function to both the tweeter and woofer, unchecked the "Remove Excess Delay" option, and applied 4.5msec of delay removal to both the acoustic graphs. The magnitude and phase of the woofer now looks like this:
An externally hosted image should be here but it was not working when we last tested it.
The tweeter mag+phase looks like this:
An externally hosted image should be here but it was not working when we last tested it.
I then plugged them into the 2K LR4 xo's that I'd designed, and computed the SPL and phase again for those networks. What I got for the final mag+phase is shown below. The low-pass xo is:
An externally hosted image should be here but it was not working when we last tested it.
The high-pass xo is:
An externally hosted image should be here but it was not working when we last tested it.
At the xo point of 2K, the low-pass xo has a phase of 115 degrees and the high-pass xo has either -180deg or +180deg, depending on how you look at it. Therefore, I can say that the phase difference between the low-pass and high-pass xo at the xo point is 65deg.
If I now switch the polarity of the tweeter, the phase difference increases, which shows up in the small notch in the SPL curve.
Have I handled the delay removal etc correctly?
How do I get the phase of the woofer and tweeter to align better than 65deg?
It sounds, and seems to look good.
What you'll do next might need you to forget your target slopes, they are a good starting point, but you need the control to change the phase.
For starters, the crossed over woofer has a much steeper phase slope than the tweeter. Try removing one order of rolloff from its crossover. Fix any obvious problems with the woofer with a quick tweak, then try to find the components in the tweeter crossover that make both the phase and level simultaneously closer to the mark.
This is just a suggested method, I think it is really more of an art.
I would ignore any tiny notches.
BTW, you may already be on top of this (or this might not be the time), but don't forget your baffle step.
What you'll do next might need you to forget your target slopes, they are a good starting point, but you need the control to change the phase.
For starters, the crossed over woofer has a much steeper phase slope than the tweeter. Try removing one order of rolloff from its crossover. Fix any obvious problems with the woofer with a quick tweak, then try to find the components in the tweeter crossover that make both the phase and level simultaneously closer to the mark.
This is just a suggested method, I think it is really more of an art.
I would ignore any tiny notches.
BTW, you may already be on top of this (or this might not be the time), but don't forget your baffle step.
I downloaded your latest files, and took the liberty of simulating a crossover which looks more familar to me than the ones you provided. I just used typically available component values to demonstrate...no doubt this can be improved upon. In particular, this version would give a couple dB emphasis in the 2k to 5k region which may make it sound too forward or enven sibilant. And the impedance dips to about 3.5 ohms at 3kHz. It also deviates from your original HP and LP targets somewhat.
But it does give about 20dB null when you reverse the tweeter polarity, which is what this thread originated about. I think you may have some difficulty perfecting this crossover because the tweeter's acoustic center is quite a bit forward of the woofers' acoustic center.
Anyway, hopefully this will give you some ideas to ponder.
But it does give about 20dB null when you reverse the tweeter polarity, which is what this thread originated about. I think you may have some difficulty perfecting this crossover because the tweeter's acoustic center is quite a bit forward of the woofers' acoustic center.
Anyway, hopefully this will give you some ideas to ponder.
An externally hosted image should be here but it was not working when we last tested it.
lndm said:
I tried doing some more phase unwrapping experiments. I can see how this thing works. But I have one doubt.
I'm building an MTM. When I take measurements, my mic is at the same horizontal level as the tweeter. This means that flight path from mic to tweeter is shorter than for woofer to mic. I am taking measurements with my mic at 1.2m distance, and the distance between the (physical, not acoustic) centres of woofer and tweeter is about 15cm. This gives me about 29microsecond as the difference in the time of flight between tweeter-mic and woofer-mic.
Do I ignore this difference? The woofer curve with and without this 29usec difference look quite different, and will definitely have an impact on where I land up tuning my xo.
After your suggestion, I decided to try fine-tuning my delay in steps of 0.001ms (one microsecond) when I approached the optimal unwrapped curve. I found that even a single microsecond difference in delay changes the phase readings quite visibly. So I've begun to treat these delays with a great degree of care.soongsc said:
Lovely! I'm totally thrilled that you took so much trouble.jbateman said:
I can't figure out the topology of that woofer xo, though... did you put in BSC with that L and R in parallel?
How did you know about the acoustic centre of the tweeter and woofer? What, in the data, tells you this? And just so that I know which data you're looking at, can you tell me the name of the file that you downloaded, please? (I have quite a few .SWD files available on that Web server for you guys to look at, and only "asawari-pulse-data.swd" contains correctly time-aligned readings.)
Secondly, what do you think about applying different delays for the woofer and tweeter xo's, to compensate for the difference in the flight paths of the two to the mic?
You will want to calculate the path difference between the tweeter and woofers from your listening distance, subtract that from the path difference at the measurement point, and correct for the remainder.
Your drivers may not be time aligned, but in my opinion this does not prevent you from designing a proper crossover. My preference would be to work as jbateman has, rather than add dedicated delay componentry.
For my system, my drivers are a little lower than the listening height so I have put the tweeters below the woofers rather than tilting the baffle or building steps into it.
I see you are building MTM's. Aha. I am not the expert here but I believe I read Joseph D'Appolito once saying that it was desirable to be exactly 90 degrees out of phase across the crossover region in order to obtain a perfect(?) total combined lobing pattern. You are supposed to be able to sit and stand without noticing any sound difference. This is supposed to ensure the balanced frequency response of reflections from the room walls.
Your drivers may not be time aligned, but in my opinion this does not prevent you from designing a proper crossover. My preference would be to work as jbateman has, rather than add dedicated delay componentry.
For my system, my drivers are a little lower than the listening height so I have put the tweeters below the woofers rather than tilting the baffle or building steps into it.
I see you are building MTM's. Aha. I am not the expert here but I believe I read Joseph D'Appolito once saying that it was desirable to be exactly 90 degrees out of phase across the crossover region in order to obtain a perfect(?) total combined lobing pattern. You are supposed to be able to sit and stand without noticing any sound difference. This is supposed to ensure the balanced frequency response of reflections from the room walls.
tcpip said:
Normally, depending on the bandwidth of the woofer, the acoustic center would be behind the tweeter. Since the wave front will not be pefectly cylindrical, you can realistically only design an XO for one point. If you use a tweeter with less vertical dispersion, then the interaction effects can be reduced.
What I would recommend is to select the one design point, get a minimum phase for the tweeter, use the same market location to measure the woofer and see how the phase would look like. Depending on how much difference there is between the tweeter and woofer acoustic center, you may or may not choose to time align the drivers to some extend.
I understand what you mean here. The difference between the two path differences will probably be not worth keeping track of, but I'll calculate it anyway. The difference at the measurement point is 1.01cm. The difference at the listening point is 0.487cm. The difference between the two differences is 0.523cm. If I remember right, this amounts to about 15usec of time-of-flight difference.lndm said:
I'm confused here. How has jbateman worked? Has he added equal delay to both drivers or is there something else that you're referring to? I read through his posts, but I'm still confused.
IMHO, this entire thing of physically time-aligning drivers is not necessary. One can tune the xo to get the drivers to time-align. I feel it's better to listen on-axis, and get the phase alignment in the xo, than try to build tilted baffles and then live off-axis all our lives.
This is way too difficult for me for my first project. Therefore I'll just focus on the on-axis behaviour for now, and let the listener decide what to do when he decides to stand up.
What I meant by this is that he did not add any delay compensation in a direct sense. He simply applied a relatively normal crossover to each driver, each of which gave the right amount of added phase shift to create the balance.tcpip said:I'm confused here. How has jbateman worked? Has he added equal delay to both drivers or is there something else that you're referring to? I read through his posts, but I'm still confused.
To be specific, he used a more or less second order electrical filter to the woofer and third order to the tweeter. You could get similar results using a third order low pass and a forth order high pass, or even a first order low pass and a second order high pass.
I didn't just do this for time aligning purposes, I did it to make my life easier during the design phase. BTW, I'm only seven degrees off axis.
tcpip said:
The R just lessens the effect of the L at higher frequencies...it allows for a bit of adjustability, that's all.
The fact that the phase is very different on the tweeter and woofer measurements indicates the sound is not coming from the same vertical plane. Since the acoustic center of most drivers is approximately where the voice coil attaches to the cone or dome, it's pretty obvious that the tweeter's acoustic center is forward of the woofer's AC.tcpip said:
FWIW I uploaded the swd file after I was done playing with it here:
http://www.diyspeakercabinets.com/New_Folder/asawari-pulse-data-pecker.swd
The circuitry for your HP an LP filters will add delay in and of themselves. In the tweeter circuit I posted, L1 will shift the phase quite a bit when you vary it. And the RW1 will adjust phase a bit in the woofer circuit.
I don't think it would be worth your time right now to creat a delay circuit with the specific intent of getting the drivers phase to align prior to developing the rest of the crossover.
I wouldn't be able to help you with that circuit anyway. In the end, when you listen to your speakers you are primarily going to notice their frequency response. Small changes in amplitude will be be detectable, but you will struggle to hear any difference in small amounts of phase change.
No, no, I think I've not been able to explain what I had in mind.jbateman said:
What I meant was that if I am going to try to match the phases of woofer and tweeter, shouldn't I first try to ensure that the phase data of the raw FR measurements of the two drivers are correct in the time domain? And once that's done, I'd like to do phase unwrapping, to make it easy to read. So, if I have to add delay to the phase data to unwrap them, should I add the same amount of delay to both or should I take the differences in flight path into account?
I wasn't talking about designing a delay circuit at all. I wouldn't know how to do it even if I wanted to.
Yes, I understood this much. But in order to make the phase data readable, one has to add delay using the SW "Delay..." function, to "unwrap" the phase graph. And if he did that, I wanted to know whether he added equal amounts of delay to both woofer and tweeter, or he took the time-of-flight difference into account.lndm said:
tcpip said:
All I did was subtract 4.5 msec from both of these plots: L-tweeter.Pulse.Fft and L-woofers.Pulse.Fft
Those were the measurements that that were listed as the frequency response for your drivers (when you right click on the driver, then select DATA) I then "save file". Then when I created the new network, this newer data is what gets used whenever I right click the network, then select "Calculate Response"
All this did was remove the time of flight to allow you to see the phase data more clearly. It does not affect the relationship of the drivers to each other, since I removed 4.5msec from BOTH.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.