The next step was to pull on the thought that Fluid mentioned. I tried a quasi 3rd order LR filter, and again, I let VituixCad optimize the on-axis responses to match the idealized target curves. For a 3rd order filter, I cascade a 1st order and 2nd order filter together.
The result is the first graphic. Other than a -2 dB dip in the on-axis response at 1.8k, everything looks great. The DI and ERDI curves are impressive, and the horizontal off axis curves are extremely smooth and uniform all they way out past 120 degrees.
If we attempt to EQ the on-axis curve to be flat (2nd graphic), it looks like we might over do it. The listening window curve and the horizontal off-axis curves all have an undesirable peak at 1.8k. If instead, we try to EQ the listening window flat (3rd graphic), it seems much better, but the horizontal off-axis still looks a bit untidy.
As Fluid mentioned, the strength of this crossover lies in how it naturally handles the z-offset between a tweeter and mid driver. The next 3 graphics show the effects of a 32mm, 44mm, and 56mm z-offset. As a point of reference, when I built a system using an SB17CAC35 6” driver and a SB26CDC 1” dome mounted on flat baffle, the measured z-offset was 43mm.
It appears that both 32 mm and 44 mm look very good, there is nothing to complain about at all. When the z-offset is increased to 56 mm, the response starts to degrade just a bit, with the ERDI and DI curves developing more variance.
j.
The result is the first graphic. Other than a -2 dB dip in the on-axis response at 1.8k, everything looks great. The DI and ERDI curves are impressive, and the horizontal off axis curves are extremely smooth and uniform all they way out past 120 degrees.
If we attempt to EQ the on-axis curve to be flat (2nd graphic), it looks like we might over do it. The listening window curve and the horizontal off-axis curves all have an undesirable peak at 1.8k. If instead, we try to EQ the listening window flat (3rd graphic), it seems much better, but the horizontal off-axis still looks a bit untidy.
As Fluid mentioned, the strength of this crossover lies in how it naturally handles the z-offset between a tweeter and mid driver. The next 3 graphics show the effects of a 32mm, 44mm, and 56mm z-offset. As a point of reference, when I built a system using an SB17CAC35 6” driver and a SB26CDC 1” dome mounted on flat baffle, the measured z-offset was 43mm.
It appears that both 32 mm and 44 mm look very good, there is nothing to complain about at all. When the z-offset is increased to 56 mm, the response starts to degrade just a bit, with the ERDI and DI curves developing more variance.
j.
Attachments
3rd order Butterworth has a 90 degree phase difference, but the Quasi 3rd order should not, the phases should track at the intended crossover point this is probably due to not adding the right amount of z depth. The phase in the idealized graphs Jon Marsh made has the traces lining up.
How you know if you are right or wrong with this one?
I mean, if you have 3rd order butterworth with 90degree phase difference or same thing with quasi 3rd order and in both cases you have flat on axis. That should be OK not?
Oh, I am pretty sure that Fluid is right. I was curious so I verified it with a quick sim. This sim has no diffraction effects, so it is simply an ideal summation of two filter outputs. I blocked out the DI and off axis plots, since they have not valid here.
The BW3 has a 90 degree phase difference at the 1k crossover, and the phase difference stays almost constant across a very broad range.
The Quasi LR3 with no woofer delay also has a 90 degree phase difference across a broad range. Since both the low pass and high pass filters are down -6 dB at 1k, the summation is not a flat response, it has a dip.
The quasi LR3 with the right amount of woofer delay results in the phase meeting up at 1k, but the tweeter and woofer phase do not track each other over a broad range, they diverge above and below 1k. The summation is not perfectly flat, but it is more than adequate when used with real world drivers. This particular filter seemed to work well with a woofer delay anywhere from 40 mm to 100 mm, so this shows how the quasi LR3 filter is flexible and very useful.
Does this answer your questions?
The BW3 has a 90 degree phase difference at the 1k crossover, and the phase difference stays almost constant across a very broad range.
The Quasi LR3 with no woofer delay also has a 90 degree phase difference across a broad range. Since both the low pass and high pass filters are down -6 dB at 1k, the summation is not a flat response, it has a dip.
The quasi LR3 with the right amount of woofer delay results in the phase meeting up at 1k, but the tweeter and woofer phase do not track each other over a broad range, they diverge above and below 1k. The summation is not perfectly flat, but it is more than adequate when used with real world drivers. This particular filter seemed to work well with a woofer delay anywhere from 40 mm to 100 mm, so this shows how the quasi LR3 filter is flexible and very useful.
Does this answer your questions?
Attachments
More then enough, thanks. Some graphs tells a lot.
I didnt question Fluid, more myself.
BW sum -3db flat and have 90 degree phase difference between drivers.
Quasi LR3 sum -6db and have no phase difference between drivers at XO when apply Z-offset.
But in the first case, BW, phase track each other over a wider range right.
Is there any acoustic benefit, if you can archive that, wide range phase tracking?
I tried the BW on my design and DI is flatter as well. Or is that just coincides?
Hi,
I have a doubt:
1) In the Quasi 3rd order LR target given by Kimmosto here: http://www.htguide.com/forum/showthread.php?44128-VituixCAD-v2&p=639230&viewfull=1#post639230, the 3dB cut off frequencies of the 1st order and 2nd order filters are different For example, they are 890Hz and 1346Hz for the 2nd order and first order filters, respectively. But in pic 3 of your post, I see that both the filters have the same 3dB cut off of 1000Hz. Wouldn't this be causing the summation dips around the crossover frequency and potentially lesser phase tracking around the crossover point. Because I see very small SPL variation around crossover point in Kimmosto's pic. And by using FIR linear phase filters, he has eliminated the phase matching issue. When we want to use IIR filters like here, we want to make the phases match through the delays.
I was able to match the target response for quasi LR3 using a 1st order at Fc and a 2nd order at Fc with Q=0.707. See the first graphic; the magnitude and phase perfectly match the target. But you have me curious about what Kimmo was up to. So I did it both ways, and whether I use his filter selection or mine, it makes little difference (although there is a very slight difference).
In his example at the htguide forum, he is showing FIR filters. That is why his example sums totally flat. I am trying to keep my examples applicable to both active and passive designs, so I am avoiding any transfer functions which would be difficult to implement with passive components.
To quote Kimmo from that other forum: "It's probably common that designer plays with passive network until phase match and magnitude sum are okay. I have done that too especially with passive speakers. Target slopes are not necessarily available/visible so difficult to know what acoustical orders actually are and are slopes symmetrical or not. Result matters so who cares about academic details."
Words of wisdom !
j.
In his example at the htguide forum, he is showing FIR filters. That is why his example sums totally flat. I am trying to keep my examples applicable to both active and passive designs, so I am avoiding any transfer functions which would be difficult to implement with passive components.
To quote Kimmo from that other forum: "It's probably common that designer plays with passive network until phase match and magnitude sum are okay. I have done that too especially with passive speakers. Target slopes are not necessarily available/visible so difficult to know what acoustical orders actually are and are slopes symmetrical or not. Result matters so who cares about academic details."
Words of wisdom !
j.
Attachments
Last edited:
It's not about right or wrong in this case, it's about yes or no. Let's see what fluid has to say.
What are you looking for?
I am trying to answer vineethkumar01's question regarding the implementation of a quasi LR3 filter.
In my opinion, with my ears, there is no sonic benefit to having phase matched through the crossover region. It might be different at very low frequencies.
And to your other question, It is totally expected that BW3 will provide the best DI results with certain baffle layouts. Other baffles will need other solutions.
Am I right or wrong or is it yes or no, I'm not even 100 percent sure what the question isHow you know if you are right or wrong with this one?
I mean, if you have 3rd order butterworth with 90degree phase difference or same thing with quasi 3rd order and in both cases you have flat on axis. That should be OK not?
Yes the quasi 3rd order needs some z depth on the woofer to function as intended, Jim's later sims show the mechanics, like the Jon Marsh example.
There is a significant difference off axis between a BW3 with the phases in quadrature and a Quasi LR3 with the phase aligned at the crossover point.
Thinking flat on axis is like tunnel vision. It is the 3D radiation that does the best job of showing why something sounds the way it does.
Hi,
Thanks a lot for posting these comparison plots. It helps in clearing my doubt. I agree with your point that more focus should be on the end results rather than worrying about minor technical details.
I was thinking that Kimmosto might have used the specific configuration of filter cut offs because it was the result of some optimization done by the optimizer and hence might have had some advantage over the other configuration where all cut off frequencies are set to 1000Hz. At least in the case of ideal drivers as is the case here, it doesn't seem to be doing anything extra.
Also, Kimmosto might have used FIR filters just for illustrating the concept of Quasi LR filter targets having flat sum and each driver response down by 6 dB at crossover frequency, while assuming that the phase match is achieved by some means, like driver offset or something else.
Thanks
Vineeth
I am still curious about how a 1st order at 1364 plus a 2nd order at 890 Hz Q=0.7 can give the same response as a 1st order at 1000 Hz plus a 2nd order at 1000 Hz Q=0.7... My mathematical intuition says this should not be true... but clearly it is. On the other hand, this is math which I long ago forgot.
When you pointed to Kimmo's post I thought "well, once again I got it wrong", but it looks like both solutions give the same answer. I am suspicious that there must be a good reason for Kimmo to have selected the values he did, rather than how I did it.
You may have misunderstood my point, and I apologize for not being more clear. My point was that when we go beyond ideal drivers in a baffle simulation to using measured data to simulate a real filter network, there will be all kinds of deviations from the ideal. I was definitely not trying to discourage a discussion of the technical or academic aspects.
j.
When you pointed to Kimmo's post I thought "well, once again I got it wrong", but it looks like both solutions give the same answer. I am suspicious that there must be a good reason for Kimmo to have selected the values he did, rather than how I did it.
You may have misunderstood my point, and I apologize for not being more clear. My point was that when we go beyond ideal drivers in a baffle simulation to using measured data to simulate a real filter network, there will be all kinds of deviations from the ideal. I was definitely not trying to discourage a discussion of the technical or academic aspects.
j.
Technically, I see that Kimmosto's first order filter cut off is 1346 Hz instead of 1364 Hz as we have in above plot
(not that it will cause significant differences, but still pointing out). If we assume that the 1346 vs 1364 cut off doesn't matter, comparing the above the two plots I can see that the 1000Hz cut off filter based implementation has slightly more roll off at the lower frequencies where the lowpass starts and where the high pass turns flat. Kimmosto's implementation holds the filter response flatter both around the low pass roll off region and around the high pass filter roll off region thereby maintaining spectral flatness for more frequencies. So both configurations are not exactly the same.. The kind of filter cascade here (1st order + 2nd order) results in multiplication of the frequency response of the both filters to arrive at an asymptotical 3rd order roll off, which depends on the chosen cut off frequencies unlike a fixed 3rd order filter which has less flexibility in terms of tweaking the overall response.
Sorry that I also wasn't clearer in my response. I was specifically speaking with respect to the ideal kind of scenario we are playing around with here. As you said, with real world drivers the filter phase response and the driver phase response will interact together to form the overall phase response. Then, at least technically, there could be difference between the two topologies that we are comparing. But it could be implementation specific and not much important with ideal drivers.
Regarding phase matching, From my limited understanding, I see that Kimmosto has specifically sacrificed little phase matching for better vertical directivity here: https://www.diyaudio.com/community/threads/vituixcad.307910/post-6533097
So, while it may be good to have phase matching as much as required, it could also be sacrificed a little if there are other goals worth sacrificing it for, I think.
Great... Perfect answer and you didn't even fully understood the question.👍Am I right or wrong or is it yes or no, I'm not even 100 percent sure what the question is
Yes the quasi 3rd order needs some z depth on the woofer to function as intended, Jim's later sims show the mechanics, like the Jon Marsh example.
There is a significant difference off axis between a BW3 with the phases in quadrature and a Quasi LR3 with the phase aligned at the crossover point.
Thinking flat on axis is like tunnel vision. It is the 3D radiation that does the best job of showing why something sounds the way it does.
What is the difference between those 2 topologies soundwise?
How far from the xo point need the 2 phases track each other? I saw that with bw3 they stay parallel over a wider span. With quasi LR3 it is less depending on the z offset distance.
I am not able to give any answers regarding difference in sound of the two topologies as I personally feel it would implementation-specific. I would also wait for reply from fluid regarding that.
However, for learning more about the technical aspects about phase matching, directivity, power response, etc,
I would refer this post by Kimmosto: https://www.diyaudio.com/community/threads/vituixcad.307910/post-6513513 and some of the subsequent posts as well as some before. More importantly the entire thread of which the above post is a part of is a gold mine regarding speaker design information using VituixCAD. I just keep on going back to it again and again and then again so that those concepts that Kimmosto has explained over a long series of posts sinks deep enough into my mind..
I just wish all the important parts in that thread was captured in one document so that it is easy to refer again and again. I myself have copied and pasted random parts of it into a document but have not been able to refine and format it properly.
Last edited:
^Use the brand new bookmark feature of the forum platform!
edit. Kimmosto has vast knowledge and posts are rather detailed but you have to read his message carefully. Most of the posts relate to some specific context, like the link above to a two way speaker with direct radiating tweeter and probably with largish woofer. I mean, most of the stuff is quite easy (or at least possible ) to come up with just tinkering with VituixCAD, with real measurements (and some without, why not). All one has to "know" is what a good set of responses should look like ( CTA-2034-A standard is the best target currently in my opinion, for those of us who don't know better ) and forget anything else, just tinker around and try achieve that what ever it takes.
Even more important, which Kimmosto doesn't often say, is that real improvement is in the physical realm! I mean, the link above gives tips how to optimize the crossover for sub optimal situation, two way speaker with direct radiating tweeter! The whole trick becomes unnecessary with better system, wave guide that matches well with woofer below, better physical construct. But this is just my opinion, observation and perspective to things.
edit: for example Kimmosto message few posts before resonate to me a lot more https://www.diyaudio.com/community/threads/vituixcad.307910/post-6512977
edit. Kimmosto has vast knowledge and posts are rather detailed but you have to read his message carefully. Most of the posts relate to some specific context, like the link above to a two way speaker with direct radiating tweeter and probably with largish woofer. I mean, most of the stuff is quite easy (or at least possible
) to come up with just tinkering with VituixCAD, with real measurements (and some without, why not). All one has to "know" is what a good set of responses should look like ( CTA-2034-A standard is the best target currently in my opinion, for those of us who don't know better ) and forget anything else, just tinker around and try achieve that what ever it takes.Even more important, which Kimmosto doesn't often say, is that real improvement is in the physical realm! I mean, the link above gives tips how to optimize the crossover for sub optimal situation, two way speaker with direct radiating tweeter!
The whole trick becomes unnecessary with better system, wave guide that matches well with woofer below, better physical construct. But this is just my opinion, observation and perspective to things.edit: for example Kimmosto message few posts before resonate to me a lot more
https://www.diyaudio.com/community/threads/vituixcad.307910/post-6512977
Last edited:
Thanks a lot tmuikku.. I just discovered it.. Awesome feature..^Use the brandnew bookmark feature of the forum platform!
I think any answer to this question would be worthless. The whole design has to be considered and at that point it might be possible to make some educated guesses.
They don't have to track at all, it all depends on what you aim to achieve with the crossover and what the drivers and enclosure require. Vineeth gave an example from kimmo above where the phase was moved apart to tilt the forward lobe.
Take a look at this "sub optimal" design. The drivers and layout limit the options but the result speaks for itself.I mean, the link above gives tips how to optimize the crossover for sub optimal situation, two way speaker with direct radiating tweeter!
http://www.htguide.com/forum/showthread.php?44128-VituixCAD-v2&p=640418&viewfull=1#post640418