DcibeL said:
PCD assumes that a 0 offset axis is the simulated measurement axis. If you set the tweeter's vertical offset to a nonzero value and the woofer's to 0, then the measurement axis becomes the woofer axis. PCD uses this information when computing the resulting acoustic offset for drivers.
When using SWshop, we have to calculate this offset manually. Take a look at Roman's explanation in the link below. You'll understand what's going on in PCD.
http://www.rjbaudio.com/Audiofiles/driver offset calculator.html
Member
Joined 2003
Ok, I understand that. So what would be the drawback of designing the crossover using the woofer as the center axis. I had originally designed this way since I estimated the woofer would be about head height, and I didn't think at my listening distance (2 to 2.5m) it would really matter that much. I guess it does matter, since the response is noticeably different if I model it with the tweeter as the center axis.
I traced the manufacturer's impedance plot rather than Zaph's because Zaph's only went to 3kHz. I think if there's any reason for my impedance to be higher, it would be due to the trace inaccuracy of tracing a somewhat low resolution plot. I visually inspected them just now and they look pretty close to the same to me.
And yes I did get your email. Looking at the crossover right now. Looks like the RLC on the tweeter helped a lot. I'm going to leave out the RLC on the woofer though, as it adds $35-$40 to the cost of my crossover with not a lot of gain. I'll post the result here later tonight once I've finalized it with real obtainable values.
Augerpro, I had considered just putting the tweeter on the bottom, and it does work well as a solution, but I don't want to turn my speakers into stumps by designing them shorter than average. I don't have any tower speakers around to compare, so I compared to the dimensions of commercial Totem speakers which are 36" high for a tower with a 6.5" woofer. I'd rather not add width to my speakers, so the only way to make them short would be to have them be very deep, so I'm not sure this is the solution I want, but as far as the crossover is concerned, yes it does work.
Member
Joined 2003
Ok, the final crossover PCD files are attached. I think we have a winner. The cost if anyone is interested will be CAD$84 plus taxes and shipping.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
DcibeL said:
Using tweeter axis as the listening axis is simply conventional because tweeter is prone to off-axis rolloff. But using woofer axis as listening axis 2 to 2.5 m away doesn't really make the tweeter's high freq SPL roll off very much. In your case, this will give you better phase alignment.
Impedance at 1kHz in your zma file is about 8.75 ohms. Inspection of Zaph's and manufacturer's plot, it is just a hair above 8 ohms. This makes some difference in xo sim. I attach my zma file I modeled to Zaph's zoomed plot. If you change your zma to this file in PCD, you'll see some difference in the result of your xo sim---not much, though.
Reid, the reason why I help you is that I somehow happened to! When I decide to help someone, I want to do it well.
Yes, I just saw your xo above. It looks pretty good. I'm pretty sure that by some fine-tuning, you'll certainly achieve the sound you're looking for. The 6.2 ohm resistance in the RLC notch in the tweeter net is 6 ohm resistor plus .2 ohm DCR of the inductor. So, use simply a 6 ohm resistor. What I meant by 1.22 uF in the notch is 1 uF plus .22 uF in parallel. You won't be able to find a .3 uF cap.You'll get much better phase tracking if you set the tweeter's vertical offset to .165 and set the woofer's to zero. Leave the woofer's Z offset as is (.035). Then this setting will simulate the woofer axis as the listening axis. Try it.
P.S. Forgot to say this. When you order xo parts, be sure to order a few extra parts. You'll need 3, 3.3, 3.6 ohm resistors to try different levels of tweeter padding. According to my experience of the HDS tweeter, you'll certainly need these
This is not a waste of money at all. Also, do not order a single 12 uF cap for the primary capacitance in the tweeter net. Order 10 uF and four 1 uF caps for each speaker. This way, you can try 10 to 14 uF for different levels of tweeter's lower rolloff. Don't believe your current xo is final. I know you WAnt to do so. But trust me. I've been on the same road as you.
Member
Joined 2003
Ahh, my real name on the interwebJay_WJ said:Reid, the reason why I help you is that I somehow happened to! When I decide to help someone, I want to do it well.
. I had changed the 6.2 Ohm resistor to 6.32 Ohm because the inductor I chose had a 0.12 Ohm DCR. Doesn't affect the crossover much though. The 1.22uF cap in the notch I changed to a 1.3uF Solen cap. I don't think I'll need to use multiple in parallel for this value.Of course there will always be some tweaking. The resistors are only 50 cents each so it's just pocket change really. That is a good idea for the 12uF cap as changing this value does affect the crossover in a fairly dramatic way.Jay_WJ said:P.S. Forgot to say this. When you order xo parts, be sure to order a few extra parts. You'll need 3, 3.3, 3.6 ohm resistors to try different levels of tweeter padding. According to my experience of the HDS tweeter, you'll certainly need these
To conclude this thread, we have discussed three different ways of adjusting phase alignment in crossover networks:
- Asymmetrical Rolloff: An asymmetrical rolloff can be used to adjust the phase relation of each driver, and the resulting summed response can still be flat.
- All-pass Delay Networks: An odd circuit arrangement that is quite uncommon since phase can usually be accounted for with an asymmetrical rolloff.
- Location Of Drivers On Baffle: Not really a crossover adjustment, but a valid point was made that if the tweeter is moved from on top of a driver to underneath the phase response changes dramatically due to movement of the listening axis from on axis to off axis. Listening location has to be considered in this instance.
[/list=1]
Thanks again everyone for your help.
The above list was generated as a summation of information. There are still questions of how we know this and is this correct?
When we design a multi-driver (two-way or greater) loudspeaker the sound is reproduced by multiple discrete sources. In most instances, the apparent points of origin for the sources are not equal. Thus a measurement of the transient response shows the sound coming from more than one discrete source. This has been called a "difference in air path delay."
Under extreme limits, these differences in air path delay can be described in terms of electrical "phase." Electrical phase, however, is not an analogue of air path delay.
If you attempt to correct for only electrical phase, you will not correct for air path delay. You can demonstrate this by testing for transient performance. Using either asymmetrical crossover slopes or electrical "all pass filters" with transient testing the loudspeaker will still show the response to be from multiple discrete sources.
The only way to account for air path delay is to physically offset the drivers to make the air path delay equal, or to use a true electrical delay line to account for different air path delays.
If you choose to use electrical delay lines, then the optimal path is to do the delay in the digital domain and bi-amp (for a two-way). There are commercially available devices to do this and they often include digital filters. They allow for adjustment or tweaking of the final design. Thus they seem like the best option for amateur loudspeaker designers/builders.
Mark
Even if the drivers are physically aligned, there should be air path delay due to phase shifts caused by drivers' rolloffs, unless the crossover is analog 1st-order or digital transient perfect. There is a debate over whether a speaker has audible benifit from a transient perfect crossover or not. My take on this issue is, the effect shouldn't audible. It's only measurable in unrealistic square wave reproduction.
-jAy
-jAy
Here's my short write-up about this issue: dealing with driver offset on flat baffle. Someone asked this question a while ago and this was my reply to him.
-------------------------------------------
... As far as I know, there's no book or article that explains in detail how we can deal with driver offsets on a flat baffle in a crossover design. For LR4 designs, the method is really simple. In short, on a flat baffle, we don want to use theoretical LR4 rolloffs due to the driver offset (i.e., drivers are not time aligned; the midwoofer is farther from the listenter than the tweeter). What we normally do is "relax" the drivers' rolloff rates.
To see how this works, consider this fact. When a midwoofer is low-passed, there occurs an increasing "phase lag" over a frequency range. This delay begins far before the crossover frequency and gets larger as the frequeny increases. A fact is, as the rolloff rate gets higher, the phase lag rate also gets higher. For the theoretical LR4 low-pass filter rolloff, the delay amount reaches 180 degrees at the crossover frequency. The opposite happens with a high-pass filter. When a tweeter is high-passed, there occurs a "phase lead." This phase lead also begins far before the crossover frequency and gets larger as the frequency decreases. As the rolloff rate gets higher, the phase lead also gets higher. For the theoretical high-pass rolloff, the lead amount reaches 180 degrees at the crossover frequency. Then what's the phase difference between low and high-pass rolloffs at the crossover point? Yes, 360 degrees. That is, the sound wave from the woofer is exactly one wavelength slower than that from the tweeter. That's why we have a full, 6 dB gain (double the SPL of each driver) at the xover frequency.
Now, what will happen if the midwoofer is a little farther from the listener than the tweeter? Then, there will be an additional amount of phase lag for the midwoofer, or equivalently, an additional amount of lead for the tweeter. As a result, their phase is misaligned. The solution is simple. If we relax the rolloff rate of either the midwoofer or the tweeter a little bit, the corresponding lag or lead will decrease. That's why we call this technique "asymmetric LR4." But this is a misnomer because, sometimes, we can relax both drivers' rolloff rates.
What will happen if we don't use this technique but use theoretical, exact LR4 rolloffs on a flat baffle? Then, the "in-phase" listening axis that ensures phase alignment around the xover point will be located below the speaker, often times below the woofer axis. Why? Because that is a point where the tweeter and midwoofer are time-aligned.
Lastly, this technique usually don't work for LR 2nd order crossovers. The reason is, 2nd order rolloffs are already very shallow, not leaving a room for further rolloff rate reduction due to the drivers' naturall, limited bandwidth rolloffs.
Note that we can use a flat baffle without worrying much about this phase alignment between woofer and midrange in a 3-way. The reason is that at a lower frequency wavelength is long enough so that the driver offset won't cause a large phase misalignment.
Hope this explanation helps you to better understand this issue.
-------------------------------------------
... As far as I know, there's no book or article that explains in detail how we can deal with driver offsets on a flat baffle in a crossover design. For LR4 designs, the method is really simple. In short, on a flat baffle, we don want to use theoretical LR4 rolloffs due to the driver offset (i.e., drivers are not time aligned; the midwoofer is farther from the listenter than the tweeter). What we normally do is "relax" the drivers' rolloff rates.
To see how this works, consider this fact. When a midwoofer is low-passed, there occurs an increasing "phase lag" over a frequency range. This delay begins far before the crossover frequency and gets larger as the frequeny increases. A fact is, as the rolloff rate gets higher, the phase lag rate also gets higher. For the theoretical LR4 low-pass filter rolloff, the delay amount reaches 180 degrees at the crossover frequency. The opposite happens with a high-pass filter. When a tweeter is high-passed, there occurs a "phase lead." This phase lead also begins far before the crossover frequency and gets larger as the frequency decreases. As the rolloff rate gets higher, the phase lead also gets higher. For the theoretical high-pass rolloff, the lead amount reaches 180 degrees at the crossover frequency. Then what's the phase difference between low and high-pass rolloffs at the crossover point? Yes, 360 degrees. That is, the sound wave from the woofer is exactly one wavelength slower than that from the tweeter. That's why we have a full, 6 dB gain (double the SPL of each driver) at the xover frequency.
Now, what will happen if the midwoofer is a little farther from the listener than the tweeter? Then, there will be an additional amount of phase lag for the midwoofer, or equivalently, an additional amount of lead for the tweeter. As a result, their phase is misaligned. The solution is simple. If we relax the rolloff rate of either the midwoofer or the tweeter a little bit, the corresponding lag or lead will decrease. That's why we call this technique "asymmetric LR4." But this is a misnomer because, sometimes, we can relax both drivers' rolloff rates.
What will happen if we don't use this technique but use theoretical, exact LR4 rolloffs on a flat baffle? Then, the "in-phase" listening axis that ensures phase alignment around the xover point will be located below the speaker, often times below the woofer axis. Why? Because that is a point where the tweeter and midwoofer are time-aligned.
Lastly, this technique usually don't work for LR 2nd order crossovers. The reason is, 2nd order rolloffs are already very shallow, not leaving a room for further rolloff rate reduction due to the drivers' naturall, limited bandwidth rolloffs.
Note that we can use a flat baffle without worrying much about this phase alignment between woofer and midrange in a 3-way. The reason is that at a lower frequency wavelength is long enough so that the driver offset won't cause a large phase misalignment.
Hope this explanation helps you to better understand this issue.
Jay_WJ said:
Jay,
To clarify a bit of what you said, even with 1st order acoustic slopes (or any of the other transient perfect filter schemes) in a theoretical world you still have to deal with the inherent bandpass nature of all real drivers. At some point that woofer or tweeter acoustic slope is going to start to roll off faster than the target slope. This steeper rolloff will introduce additional phase shift which, in theory, should be compensated for with additional physical offsets of the drivers to bring the phase at fc back into alignment with the target phase.
Please note that I say that you should do this 'in theory' because while this additional phase shift will mess up a transient perfect design, its actual impact on non-TP systems is probably negligible.
regards,
Dennis
Baffle spacing, phase angles and time alignment, revisited
When drivers are stacked vertically, nulls form above and below the loudspeaker at predictable locations. Using constant directivity horns, one can design a speaker having a radiation pattern within the nulls. This provides uniform response within a specific coverage angle in the horizontal and vertical planes.
When drivers are stacked vertically, nulls form above and below the loudspeaker at predictable locations. Using constant directivity horns, one can design a speaker having a radiation pattern within the nulls. This provides uniform response within a specific coverage angle in the horizontal and vertical planes.
djarchow said:
Good point, Dennis. That's why a system with a full range driver can have a larger group delay than a multi-driver, acoustic 1st order system, due to its limited bandwidth at both ends. Right, there is no true T-P design in practice. But we can have one close to it. Perhaps, the most sophisticated T-P design, based purely on analog/physical methods, is the Thiel CS3.7.
Jay_WJ said:
Hi Jay,
I pretty much agree with everything you've said. I'm just not sure how important it is to correct for the phase. You're not really having your cake and eating it too. If you stay true to the LR4 target, the phase is off a little and there will be some minor vertical lobing issues. The flip side is that if you have asymmentric filters, well, you no longer have "true" LR4 filters. Something else gives. You may have to accept nonflat FR? Which is audibly better though, a mild on axis rise, a slight degradation of lobing around the xover?
Sure, in the RS22528A the phase tracking isn't great, but it's over a much shorter frequency interval d/t the higher order filter and the response is commendably flat. Perhaps this is a better tradeoff, or perhaps either tradeoff is, well, a tradeoff
Of course, Mark, symmetric lobing or perfect phase alignment at the listening position shouldn't the foremost goal of crossover designinig. Depending on a system design with various factors, we can try many things, some of which may give a better result than others. In any case, it's rare that we have perfect, theoretical LR4, LR2, or BW3 rolloffs.
P.S. Even with asymmetric LR4 (or something between LR4 and BW3), we can have a response close to flat around the xover point by adjusting a frequency point where each rolloff begins.
P.S. Even with asymmetric LR4 (or something between LR4 and BW3), we can have a response close to flat around the xover point by adjusting a frequency point where each rolloff begins.
Jay_WJ said:
Jay,
The new Thiel looks great. I haven't had a chance to hear this model yet. If nothing else, it sure looks cool
I think the best implementation of TP speakers I have heard were from Dunlavy. Great sound from relatively cheap drivers and cabinets.
John Kreskovsky did a nice implementation of a TP 3 way using the filler driver approach a few years back and his results show just how good the results can be with a TP design.
http://www.geocities.com/kreskovs/TP3-project1a.html
I also designed a TP 3 way design using higher end drivers (SS Revelator and Millennium) which measured and sounded great. However, the MFTFM driver layout resulted in a speaker that was over 5' tall and just too big for the space I had them in.
The TP design philosophy has always intrigued me, if nothing else for the technical challenge. I still have a goal to design and publish a TP design using relatively inexpensive drivers (TB, Vifa, etc) along the lines of the old Dunlavy Cantata and may try to get to it later this year.
Regards,
Dennis
So is it true, that if the drivers are perfectly aligned, the phase will be perfect as well when textbook acoustic target curves are used?
What if different order electrical networks are used on each driver in order to achieve the same order acoustic slope? As I understand it, the acoustic slope is what matters, not the electrical network.
Or do other things also effect phase? Such as impedance and impedance phase?
What if different order electrical networks are used on each driver in order to achieve the same order acoustic slope? As I understand it, the acoustic slope is what matters, not the electrical network.
Or do other things also effect phase? Such as impedance and impedance phase?
Maybe we have to clarify something here. Making a couple of sources that radiate from slightly different Z planes at some fc to SUM on an arbitrary axis, is another thing than getting them into coherent rise time. That takes physical alignment.
Also, achieving good summing on an arbitrary axis, lends little value if correct summing is not continued for a smooth projection cone around that axis.
Also, achieving good summing on an arbitrary axis, lends little value if correct summing is not continued for a smooth projection cone around that axis.
Salas said:
Maybe we have to clarify something here. Making a couple of sources that radiate from slightly different Z planes at some fc to SUM on an arbitrary axis, is another thing than getting them into coherent rise time. That takes physical alignment.
Also, achieving good summing on an arbitrary axis, lends little value if correct summing is not continued for a smooth projection cone around that axis.
Hi guys
Boy i know the Quest your on very well.
While involved, it can be done. Here is a solution i came up with some years ago, here is a current example of a speaker which can reproduce the square wave between about 220Hz to 2600Hz (spanning two crossovers) and radiates into space without interference lobes, all the drivers acting as one single source.
Look at the part on the SH-50, somewhere here in the archives are the oscilloscope photo's.
Download / read the White paper on the Synergy horn, download the .CLF file for the SH-50 and look at the spherical radiation pattern (3d polar). Try the various tool bars to examine various measured parameters. These measurements are taken by an independent company set up to do this kind of data gathering (a huge number of points).
http://www.danleysoundlabs.com/technical downloads.html
Best,
Tom
Maybe we have to clarify something here. Making a couple of sources that radiate from slightly different Z planes at some fc to SUM on an arbitrary axis, is another thing than getting them into coherent rise time. That takes physical alignment.
Also, achieving good summing on an arbitrary axis, lends little value if correct summing is not continued for a smooth projection cone around that axis.
Hi guys
Boy i know the Quest your on very well.
While involved, it can be done. Here is a solution i came up with some years ago, here is a current example of a speaker which can reproduce the square wave between about 220Hz to 2600Hz (spanning two crossovers) and radiates into space without interference lobes, all the drivers acting as one single source.
Look at the part on the SH-50, somewhere here in the archives are the oscilloscope photo's.
Download / read the White paper on the Synergy horn, download the .CLF file for the SH-50 and look at the spherical radiation pattern (3d polar). Try the various tool bars to examine various measured parameters. These measurements are taken by an independent company set up to do this kind of data gathering (a huge number of points).
http://www.danleysoundlabs.com/technical downloads.html
Best,
Tom
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.