Hi everyone!
I am an audio novice and haven't heard SL's LX521 speaker, but am contemplating building the LXmini or LX521. However, looking over the details of the speaker and reading the long threads about it, I was wondering if someone could shed some light on what makes it so good or different from everyone else's designs. There's quite a number of dipole designs on here from various members, but what can be learned or adapted from SL's creation?
1. Driver selection. I understand that SL tends to use SEAS drivers because he has a working relationship with them and they are of course high quality drivers. Is there something particular about his driver selection? I doubt that there's that much difference between designs with drivers of approximately equal quality driven within their limits. Are the drivers a particular range of qts, frequency range or sensitivity? How is CD involved in the driver choice?
2. Baffle shape. SL apparently arrived empirically at his somewhat unusual baffle design, but it looks to generally be a narrow baffle approximately following the contour of his drivers, but with some sharp edges and an extension on top. Does the baffle really contribute much to the sound? Again, I can't imagine it really does very much given its minimal nature.
3. Bass module. I'm sure the bass module is great, but this part also doesn't seem revolutionary. I'm sure it functions amazingly, but I assume it could be replaced with any number of dipole bass designs available on Diyaudio.
4. Digital crossover and DSP. I assume that most of the 'magic' comes from his crossover choices and DSP/EQ settings, which would also make the driver choice much less critical. Is there anything special that SL has done here to make the speaker sound so good? In his choice of frequency response which is particularly pleasing to the ear rather than EQing the speaker razor flat?
I'm sure all this has been discussed previously, but is probably buried in the bottom of other threads and thought it might be nice to start a more concise discussion where newbies like myself could learn a few things from the musings of the more learned persons on this forum.
Thanks! 🙂
I am an audio novice and haven't heard SL's LX521 speaker, but am contemplating building the LXmini or LX521. However, looking over the details of the speaker and reading the long threads about it, I was wondering if someone could shed some light on what makes it so good or different from everyone else's designs. There's quite a number of dipole designs on here from various members, but what can be learned or adapted from SL's creation?
1. Driver selection. I understand that SL tends to use SEAS drivers because he has a working relationship with them and they are of course high quality drivers. Is there something particular about his driver selection? I doubt that there's that much difference between designs with drivers of approximately equal quality driven within their limits. Are the drivers a particular range of qts, frequency range or sensitivity? How is CD involved in the driver choice?
2. Baffle shape. SL apparently arrived empirically at his somewhat unusual baffle design, but it looks to generally be a narrow baffle approximately following the contour of his drivers, but with some sharp edges and an extension on top. Does the baffle really contribute much to the sound? Again, I can't imagine it really does very much given its minimal nature.
3. Bass module. I'm sure the bass module is great, but this part also doesn't seem revolutionary. I'm sure it functions amazingly, but I assume it could be replaced with any number of dipole bass designs available on Diyaudio.
4. Digital crossover and DSP. I assume that most of the 'magic' comes from his crossover choices and DSP/EQ settings, which would also make the driver choice much less critical. Is there anything special that SL has done here to make the speaker sound so good? In his choice of frequency response which is particularly pleasing to the ear rather than EQing the speaker razor flat?
I'm sure all this has been discussed previously, but is probably buried in the bottom of other threads and thought it might be nice to start a more concise discussion where newbies like myself could learn a few things from the musings of the more learned persons on this forum.
Thanks! 🙂
Back in the day before miniDSP the analog EQ module that was the heart of the LX521 made it what it is. The baffle shape uses a combination of extended trapezoidal shapes to avoid equidistant edges from driver centers to reduce diffraction and improve imaging and smoothness. There are many driver choices available today that are very cost effective and can approach or exceed the Seas drivers chosen. Dipole W push pull woofers could be substituted with slot loaded dipoles or even large H frames. It's easy to say today that you can probably make something similar hat would sound quite good with help of miniDSP. But SL did this years ago with analog op amps and hand sketches. Quite a remarkable feat that we now benefit from. But in short, you can't really improve it too much if you want a dipole speaker with great imaging. One area would be use of quasi transient perfect XO like a Harsch vs LInkwitz Riley. Transient detail and attack would sound better with Harsch or even FIR filters and next gen Sharc miniDSP HD.
I have made a multiway dipole speaker with minidsp (see signature link)
I believe that the magic of LX521 and alike NaO dipoles By John Kreskowsky lie in smart baffle geometry (extending dipole radiation unusually high), well thought and set crossover points and careful setting of delay etc. parameters in dsp fine details.
Baffle details like oblique sides and material are not that important, nor driver name.
I believe that the magic of LX521 and alike NaO dipoles By John Kreskowsky lie in smart baffle geometry (extending dipole radiation unusually high), well thought and set crossover points and careful setting of delay etc. parameters in dsp fine details.
Baffle details like oblique sides and material are not that important, nor driver name.
Yes, more info here:
http://www.diyaudio.com/forums/multi-way/277691-s-harsch-xo.html
It is the target electro-acoustic transfer function, not the electrical only function. So it may require for example a first order electrical to get second order Bessel for high pass. The 4th order BW lowpass is usually one and the same for a woofer.
First EQ all responses as flat as possible with no filters and minimize use of boost eq.
On a 4 way it would go like this:
Subwoofer: BW4 LPF @ Freq_XO1, (optional LPF for infrasonic protection but avoid as it adds group delay)
Midbass: BES2 HPF @ Freq_XO1, BW4 LPF @ Freq_XO2, delay = 1/(2*Freq_XO1),
Midrange: BES2 HPF @ Freq_XO2, BW4 LPF @ Freq_XO3, delay = {1/(2*Freq_XO2)+1/(2*Freq_XO1)},
Tweeter: BES2 HPF @ Freq_XO3, delay = {1/(2*Freq_XO3)+1/(2*Freq_XO2)+1/(2*Freq_XO1)},
Typical Frequencies would be Freq_XO1=80Hz, Freq_XO2=800Hz, Freq_XO3=2.5kHz to 4.5kHz (to taste and to avoid tweeter distortion)
All drivers +ve polarity.
Last edited:
Buff, is "Why Harsch sound so harsh?" the title of this file, or is Harsch ******* xover paranoia damned out off topic in this file? 🙄
Last edited:
While I agree that the Harsch type crossover will provide better transient response, in theory, there are several problem with it in general, and in particular with speakers like the LX521 and NaO Note II RS. First, flattest response is obtained when the drivers acoustic response sums in phase quadrature. That means the vertical polar response will not be symmetric but will have a null to one side and a 3dB peak to the other. Second is the difficulty in obtaining a true 2nd order acoustic Bessel high pass. This can be done with a box speaker by pole shifting. However, with either of the above speakers, due to the dipole nature, a second order acoustic high pass on the lower mid would result in excursion continuing to increase below the cut off frequency. A minimum of a 3rd order acoustic roll off is required simply to maintain constant excursion below the cut off and a 4th order is desirable to assure excursion decreases below the lower mid high pass cut off.
By flattest response are you referring to the "overall" combination of loudspeaker and room such that a system having a flat power response is preferred?
With DSP it's not that hard to achieve close to 2nd order Bessel on the high pass filter. True, there will be more excursion than a 4th order LR but the slightly higher distortion from the lower order filter is not so noticeable if you pick the right driver having low distortion to start with. For me, the realism of the percusiion sounds and sounds of piano, stand up bass, drums, guitar, are well worth the effort.
Just because the phase doesn't sum up to perfect quadrature isn't such a big deal. Sure, there is about 1dB of ripple through the Xo region, but look at the flat phase that just rises gently by 55deg and us otherwise flat - that is, no 360deg phase wrap.
Just because the phase doesn't sum up to perfect quadrature isn't such a big deal. Sure, there is about 1dB of ripple through the Xo region, but look at the flat phase that just rises gently by 55deg and us otherwise flat - that is, no 360deg phase wrap.
With DSP it's not that hard to achieve close to 2nd order Bessel on the high pass filter. True, there will be more excursion than a 4th order LR but the slightly higher distortion from the lower order filter is not so noticeable if you pick the right driver having low distortion to start with. For me, the realism of the percusiion sounds and sounds of piano, stand up bass, drums, guitar, are well worth the effort.
Just because the phase doesn't sum up to perfect quadrature isn't such a big deal. Sure, there is about 1dB of ripple through the Xo region, but look at the flat phase that just rises gently by 55deg and us otherwise flat - that is, no 360deg phase wrap.
Sure, this looks good when you sim the filters alone. Add in the drivers (including their phase response) and the picture gets a little cloudier.
Here is a Harsch XO done by Bushmeister:
http://www.diyaudio.com/forums/multi-way/285030-bookshelf-multi-way-point-source-horn.html
Phase:
Step response:
http://www.diyaudio.com/forums/multi-way/285030-bookshelf-multi-way-point-source-horn.html
Phase:
Step response:
No. How the two band pass filters sum to achieve flat summed response on axis. The combination off a B4 LP and a Bessel 2 HP will never sum perfectly as witnessed by the plots XRK971 posted. I.E. they sum like Butterworth filters with the same resultant asymmetric polar response.
As you know, I used to be a big believer in TP crossovers and speakers. I built a number for them and always thought they sounded better than regular crossovers. The problem was that it was always an apples to oranges comparison because changing crossovers meant changing polar response, overlap between drivers, as well as small changes in frequency response. Just too many variables changing. Once I got involved with Brodan and the Ultimate Equalizer software my opinion changed because for the first time I could design a speaker and change between regular crossover and linear phase system without any changes in polar or frequency response, overlap, etc. The only change was system phase response. On certain type of music I was still able to hear a difference, but in no way could I declare on better than the other. The differences were extremely subtle. At that point I lost interest in TP crossover in favor of polar and power response issues. And yes, I did test a version of the Note II RS with FIR filters yielding a linear acoustic phase system.
For what its worth, both the LX521 and the Note II RS have very flat GD through the lower to upper mid crossover. In both systems those two drivers are effectively coupled to act as one, although different approaches were employed.
For guys to which phase matters, it seems to me the most simple way. But of course they will always be guys claiming that pre response sucks, min phase IIR fetichists that claim to hear ( or simply have heard about on forums...) pre echoes...😀
Last edited:
At what point in time did you linearize the phase? In my humble opinion that's the important part to get it right. I see a lot of people linearizing phase with tools like rePhase, but there's a little more to it to get it right.
Phase is a moving target, the timing is important. It took me half a year of extensive tests to figure that one out.
I've listened to pure linear phase and phase mimicking the minimum phase band pass of my speaker, I prefer the latter, and even though the phase plots look very similar I can easily tell them apart.
In both cases it is important to apply a frequency dependent gating of about ~5 to 6 cycles and adjust the phase there at that point in time (at the listening position). It gives power to the midrange, much like you get with live music.
The lower FR being in time supports the higher notes when the timing is right.
You need something better than the frequency dependent window as used in REW, that FDW has a lot of smoothing.
Something like APL_TDA can be used to show the timing:
Recorded at the listening position, both left and right channels playing.
This is a normalized view, but the timing part is obvious.
I wouldn't call it a subtle change, I would call it worth the trouble, once you get the timing right. Plus I'd suggest not going for a straight line phase.
Above 1 kHz is a different story though. The phase gets messed up by our 2 ears summing the stereo signal and it get's way harder to tell any difference. That doesn't mean it's not important. As it will determine how the inter aural comb filtering, happening at the two ears, changes the overall FR balance.
Last edited:
Something like APL_TDA can be used to show the timing:
Recorded at the listening position, both left and right channels playing.
This is a normalized view, but the timing part is obvious.
I wouldn't call it a subtle change, I would call it worth the trouble, once you get the timing right. Plus I'd suggest not going for a straight line phase.
Above 1 kHz is a different story though. The phase gets messed up by our 2 ears summing the stereo signal and it get's way harder to tell any difference. That doesn't mean it's not important. As it will determine how the inter aural comb filtering, happening at the two ears, changes the overall FR balance.
What you try to explain is so hairy complicated, that i am absolutely convinced it's a whole waste of time and money and that all that jazz only works in your head!😀
Last edited:
You mean what year? I had been working with TP crossover for the last 20 years and with linear phase systems since around 2010. A system is either linear phase, where phase is either zero or a constant times frequency across the full spectrum; minimum phase, where crossover induced phase distortion is removed by implementing a TP crossover with resulting phase being the minimum phase of the speakers bandpass (note that such systems are no transient perfect though they behave that way at higher frequency); or nonlinear phase, which is a typical speaker. The problem with any of these systems is that the phase response varies with listening position and the design phase is only achieved at a single design point in space. How anyone feels about what sounds better is, of course, subjective. But unless the speaker is fixed and the phase response is manipulated by preprocessing of the input, so that only the phase response of the system changes, no conclusions as to what the cause of the difference in sound is can be made. For example, why would you expect a speaker with a 1st order crossover to sound anything like a speaker with a 2nd order LR crossover when aside from the phase response being difference so are the polar response, driver overlap in the stop bands, driver excursion, hence distortion, etc also different. You may prefer one speaker to the other, but to attribute it to phase response differences would be an inappropriate conclusion. It may play a roll, but it is far from the only factor.
It seems complicated - but certainly not a waste of time if you understand the true meaning if that spectral decay plot measured at listening position. That is basically about as close to ideal behavior as possible and makes for great sounding music. To dismiss it as such shows quite a bit of disrespect for Wesayso's pioneering achievement in making a DIY speaker this extraordinary. Anyone who follows the Two Towers thread knows the level of detail and study that Wesayso has put into this to make it possible. He gives away the steps and process to those willing to try freely. Accept it or not - it is anything but a waste of time.
That's why I used a speaker without any crossover to do these experiments.
The only change has been phase. But I wasn't getting at a year. The timing alludes to when do you want your phase to be right.
As you said, and me too, phase is a moving target. But not only does it change with position, it also changes in time. What I'm suggesting is to look at the phase plot from a frequency dependent windowed FR of about 5 to 6 cycles. Use that as the base plot to linearize the phase. Though care should be taken in a room, you can't just bully the phase into shape. Room anomalies that exist over a wider area shouldn't be fixed at all by EQ-ing phase. That won't sound anywhere near right.
The APL_TDA plot shows the timing at the microphone and is a gated frequency dependent view. A very good example of a second order speaker looks like this:
This one was taken in a room that's has been treated well above average, which is obvious, just by looking at the plot. Courtesy of jim1961.
There are a couple of requirements to get a chance of getting the phase somewhat right, one of them being a low level of reflections in (at least) the first ~20 ms. You can see the reflections in my APL_TDA plot. Though absolute level cannot be seen on that plot.
So what did you linearize, was the measurement gated, if so, how was it gated. I've experimented with the gating to find an optimal setting. As said, on speakers without any crossover whatsoever. Now I'll probably get heat on using no crossover 😀.
Here's an example of the measured IR from my speakers (though a standard view of an IR isn't telling the whole story):
Looking at an early waterfall plot (3ms with 0.1 ms rise time) gets you a clear midrange, bass extends to 17 Hz in room.
The speakers look like this:
Frequency and phase have been linearized with FIR filters, though both with a different length frequency dependent window.
The only change has been phase. But I wasn't getting at a year. The timing alludes to when do you want your phase to be right.
As you said, and me too, phase is a moving target. But not only does it change with position, it also changes in time. What I'm suggesting is to look at the phase plot from a frequency dependent windowed FR of about 5 to 6 cycles. Use that as the base plot to linearize the phase. Though care should be taken in a room, you can't just bully the phase into shape. Room anomalies that exist over a wider area shouldn't be fixed at all by EQ-ing phase. That won't sound anywhere near right.
The APL_TDA plot shows the timing at the microphone and is a gated frequency dependent view. A very good example of a second order speaker looks like this:
This one was taken in a room that's has been treated well above average, which is obvious, just by looking at the plot. Courtesy of jim1961.
There are a couple of requirements to get a chance of getting the phase somewhat right, one of them being a low level of reflections in (at least) the first ~20 ms. You can see the reflections in my APL_TDA plot. Though absolute level cannot be seen on that plot.
So what did you linearize, was the measurement gated, if so, how was it gated. I've experimented with the gating to find an optimal setting. As said, on speakers without any crossover whatsoever. Now I'll probably get heat on using no crossover 😀.
Here's an example of the measured IR from my speakers (though a standard view of an IR isn't telling the whole story):
Looking at an early waterfall plot (3ms with 0.1 ms rise time) gets you a clear midrange, bass extends to 17 Hz in room.
The speakers look like this:
Frequency and phase have been linearized with FIR filters, though both with a different length frequency dependent window.
John K, I see your points and from looking at your website and the very nice NaO Note speakers I see you are of the school of polar power response is everything. I have tried similar dipole designs to yours with 4th order LR filters and they certainly sound clean and are quite easy to implement in miniDSP. But for me, and I think many others, there is a certain "unnatural" sound from the phase wrapping that occurs with a LR4. When I tried the Harsch XO or even a 1st order flat phase transient perfect XO, the un-naturalness went away and the music was just real. It became a "you are there" experience vs a speakers feeling like "they are here". It's a subtle difference but once I have heard it, I know I want it. What many people don't know they don't miss. How many multiway speakers are transient perfect? Not many, as a result many people have no idea what they are missing. I do not have a problem with the vertical polar response not being smooth like you say - over a similar listening window as you describe its not a problem at all.
Btw, what is the origin of the NaO name? For me with a science background I cant help but say "sodium oxide" everytime I see it. 🙂
Last edited:
- Status
- Not open for further replies.