Hello, I'm working around a self-designed crossover for a difficult cross between a Peerless 8'' P830869 and a Scan Speak 1'' D2905-950000. I was able to fight the 3khz peakness of woofer but I wasn't able to obtain a good sound from this beautiful tweeter. My question is:
how different kinds of crossover between Butterworth, Chebychev and Linkwitz-Riley affects sound, firmly remaining other conditions such order of crossover and flatness of the overall response?
In my experience I've had better results using a more resonant low-cut filter rather then a smoother one like Linkwitz-Riley, strange since I've had been sure of contrary..
how different kinds of crossover between Butterworth, Chebychev and Linkwitz-Riley affects sound, firmly remaining other conditions such order of crossover and flatness of the overall response?
In my experience I've had better results using a more resonant low-cut filter rather then a smoother one like Linkwitz-Riley, strange since I've had been sure of contrary..
Generally, experts consider Chebyshev to sound inferior. However, this shall not be misapplied, because as DougL put it, the electrical part is only a piece of the puzzle. An electrical Chebyshev on top of your driver's acoustic response may well sum up just right.
There has been a discussion of the relative merits of (odd-order) Butterworth and (even-order) Linkwitz-Riley. It appears that more people prefer the sound of Butterworth. The difference is that Butterworth has both a flat frequency response and a flat power response, but it has a skewed lobing pattern. Linkwitz-Riley also has a flat frequency response, and the main lobe is on-axis, but its power response has a dip. (We assume here, again, that "Butterworth" and "Linkwitz-Riley" is combined electrical/acoustic response.)
Now, here is the most confusing part: in reality the acoustic response of a driver can only be very irregular and unpredictable, that it is virtually impossible (nor is it necessary) to combine with an electrical filter to fit neatly into one of the aforementioned standard alignment. When people here talk about LR2 or LR4 crossover, they are not literally talking about LR2 or LR4. I would rather prefix them with "quasi". So a quasi-LR is an (even-order) crossover is one where at the crossover point the tweeter and the midrange are both -6dB and they are in phase, at least around the crossover region. In the same vein an (odd-order) quasi-Butterworth filter is one where at the crossover point both drivers are -3dB and they are 90 degrees apart in phase, at least around the crossover region.
There has been a discussion of the relative merits of (odd-order) Butterworth and (even-order) Linkwitz-Riley. It appears that more people prefer the sound of Butterworth. The difference is that Butterworth has both a flat frequency response and a flat power response, but it has a skewed lobing pattern. Linkwitz-Riley also has a flat frequency response, and the main lobe is on-axis, but its power response has a dip. (We assume here, again, that "Butterworth" and "Linkwitz-Riley" is combined electrical/acoustic response.)
Now, here is the most confusing part: in reality the acoustic response of a driver can only be very irregular and unpredictable, that it is virtually impossible (nor is it necessary) to combine with an electrical filter to fit neatly into one of the aforementioned standard alignment. When people here talk about LR2 or LR4 crossover, they are not literally talking about LR2 or LR4. I would rather prefix them with "quasi". So a quasi-LR is an (even-order) crossover is one where at the crossover point the tweeter and the midrange are both -6dB and they are in phase, at least around the crossover region. In the same vein an (odd-order) quasi-Butterworth filter is one where at the crossover point both drivers are -3dB and they are 90 degrees apart in phase, at least around the crossover region.
Before you rush out and dig into Butterworth, let me clarify further: in an anechoic chamber, on-axis, they sound the same (other than the sound difference between higher-order crossovers and lower-order crossovers). This is because you wouldn't be able to hear power response in an anechoic chamber and by definition both give flat on-axis responses.
The choice is as usual the standard answer: the best crossover type is the one which fits your speaker design the best.
For example, if you know ahead that your room acoustics has a major reflection with the main lobe of your crossover, then I guess it is a no-brainer: anyone would take a dip and choose LR to avoid that.
The choice is as usual the standard answer: the best crossover type is the one which fits your speaker design the best.
For example, if you know ahead that your room acoustics has a major reflection with the main lobe of your crossover, then I guess it is a no-brainer: anyone would take a dip and choose LR to avoid that.
Last edited:
It is a complex area, isn't it. I like to have an inductor (coil) as the last element before the tweeter and wired across the terminals so as to have a low impedence item in parallel and so helping to nullify possible impedance peaks from the tweeter's fundamental resonance. (I recall Vifa advocating this 25 yrs ago.) This will tend to make it easier to meet design goals for other components in the network if you're working on a design produced purely from theory.
(This doesn't commit you to even order Butterworth slopes either. Exploring parallel realizations of odd order series networks will give you the same result.)
Good luck......cross-overs are something of a "Black Art".....but good fun.
Cheers Jonathan
(This doesn't commit you to even order Butterworth slopes either. Exploring parallel realizations of odd order series networks will give you the same result.)
Good luck......cross-overs are something of a "Black Art".....but good fun.
Cheers Jonathan
Last edited:
As AllenB correctly describes, the SPL data for the 8" proves that a crossover under 1kHz to a dome tweeter is necessary for a smooth controlled directivity, a consistent polar response. With a crossover at higher frequencies, the listener will experience a confused SPL and soundstage as the 8" beams its energy over a narrow polar pattern to the listener, while the tweeter spreads its energy over a wide polar pattern around the room. confused soundstage
Attachments
Just one look at the ScanSpeak D2905 datasheet confirms that it required a >3000Hz crossover and is not suited for crossover to an 8" midbass. The 1.1" dome on the SB_Acoustics SB29RDC-4 has been used with 1400Hz LR4/LR4 crossovers with 8" midbass... not perfect but an acceptable compromise. Tymphany(Peerless) also sells a horn loaded 32mm dome tweeter DA32TX00-8 which has been used with ~1600Hz crossover frequencies... where the polar pattern of the DA32TX00 horn just matches the early beaming polar pattern of the 8" midbass.
So, find a robust tweeter which supports a low crossover ~1,400Hz like the SB29RDC-4 to reduce the 8" beaming effects. OR... accept the 8" beaming effects and find a horn loaded tweeter with a polar pattern which matches that 8" beaming polar pattern at ~1400-1600Hz crossover frequency. Same goal, a smooth controlled directivity function through the crossover.
So, find a robust tweeter which supports a low crossover ~1,400Hz like the SB29RDC-4 to reduce the 8" beaming effects. OR... accept the 8" beaming effects and find a horn loaded tweeter with a polar pattern which matches that 8" beaming polar pattern at ~1400-1600Hz crossover frequency. Same goal, a smooth controlled directivity function through the crossover.
Attachments
Yes you are right. Allow me to link to my own post again: Tweeter playing low is desireable ... WHY ?
Ok so near the beginning I directly answered the OP's question of whether Butterworth or LR sounds better. Then, when the topic of driver mismatch comes up I provided a link to my previous post on how to choose a crossover point (which indirectly covers driver selection).
I will add that crossover design is intricately linked to the enclosure design. Typically an expert will say I cannot begin a crossover design until I can measure the drivers in situ. That is true but why is this so? The two primary reasons are 1, the C2C (center-to-center) between midrange and tweeter decides the lobing pattern (but not lobing angle, go back to Butterworth vs LR on that), and 2, the offset between the tweeter and midrange introduces a phase shift (because if the time it takes for the tweeter to reach your ears is different from the time it takes for the midrange to reach your ears it is a phase shift). There are other reasons but I am unable to list them all.
So, each driver is subject to these phase shifts: 1, the acoustic phase shift of the driver itself (in enclosure), 2, relative phase shift arising from driver offset, and 3, the phase shift added by the crossover. The 3 simply sum together and becomes the total phase shift of each driver. (You may say 1 and 2 are one and the same thing, because they are almost never separately measured. I agree.)
Now, in the beginning I made a case that phase coherence (the tweeter playing in phase with the midrange, at least in the crossover region) is not a requirement, given that Butterworth sounds so good. But for whatever reason, the DIY community is heavily skewed towards phase coherence for the classic 2-way, and virtually every classic 2-way is a (quasi)-LR4. If you follow that trend, you will have to find some way for the 3 phase shifts for the tweeter and the 3 phase shifts for the midrange to be (almost) exactly the same - or, if they are (almost) exactly out of phase, then you simply invert the tweeter's polarity, and the design will be called (quasi)-LR2.
I will add that crossover design is intricately linked to the enclosure design. Typically an expert will say I cannot begin a crossover design until I can measure the drivers in situ. That is true but why is this so? The two primary reasons are 1, the C2C (center-to-center) between midrange and tweeter decides the lobing pattern (but not lobing angle, go back to Butterworth vs LR on that), and 2, the offset between the tweeter and midrange introduces a phase shift (because if the time it takes for the tweeter to reach your ears is different from the time it takes for the midrange to reach your ears it is a phase shift). There are other reasons but I am unable to list them all.
So, each driver is subject to these phase shifts: 1, the acoustic phase shift of the driver itself (in enclosure), 2, relative phase shift arising from driver offset, and 3, the phase shift added by the crossover. The 3 simply sum together and becomes the total phase shift of each driver. (You may say 1 and 2 are one and the same thing, because they are almost never separately measured. I agree.)
Now, in the beginning I made a case that phase coherence (the tweeter playing in phase with the midrange, at least in the crossover region) is not a requirement, given that Butterworth sounds so good. But for whatever reason, the DIY community is heavily skewed towards phase coherence for the classic 2-way, and virtually every classic 2-way is a (quasi)-LR4. If you follow that trend, you will have to find some way for the 3 phase shifts for the tweeter and the 3 phase shifts for the midrange to be (almost) exactly the same - or, if they are (almost) exactly out of phase, then you simply invert the tweeter's polarity, and the design will be called (quasi)-LR2.
Here is the link to the discussion thread: Pt. 2: Audibility of Crossovers
If you think about it, the claim is almost self-evident: people prefer both the frequency response and the power response to be flat, over only the frequency response to be flat of course!
However, even though reasonable-sounding it is, you must not rank it too highly in your priority list. Engineering, folks, it is all about priorities. There really are a lot of buts, like my example where the Butterworth off-axis main lobe hits a major reflection (depending on room) and you exactly want a power response dip to fix that (LR).
It is hypocritical of me to accuse other DIY'ers of adhering too rigidly to the phase coherence formula (LR) and point to your attention that the opposite camp (Butterworth) could sound just as good or even better. (Why else do people tout the virtue of first order if phase coherence is important?) But it really depends on the design. And the current one I am working on calls for phase coherence over flat power response.
Back to OP's proposal of a 8+1@3k, it violates the max C2C rule soundly, so I bet any phase coherence would become irrelevant. Butterworth it goes.
If you think about it, the claim is almost self-evident: people prefer both the frequency response and the power response to be flat, over only the frequency response to be flat of course!
However, even though reasonable-sounding it is, you must not rank it too highly in your priority list. Engineering, folks, it is all about priorities. There really are a lot of buts, like my example where the Butterworth off-axis main lobe hits a major reflection (depending on room) and you exactly want a power response dip to fix that (LR).
It is hypocritical of me to accuse other DIY'ers of adhering too rigidly to the phase coherence formula (LR) and point to your attention that the opposite camp (Butterworth) could sound just as good or even better. (Why else do people tout the virtue of first order if phase coherence is important?) But it really depends on the design. And the current one I am working on calls for phase coherence over flat power response.
Back to OP's proposal of a 8+1@3k, it violates the max C2C rule soundly, so I bet any phase coherence would become irrelevant. Butterworth it goes.
Last edited:
One major argument for LR is that humans find peaks more offensive than dips and LR had only dips off-axis. Butterworth has both. But as the other theory goes, all the energy radiated from a speaker will sooner or later reach your ear, so flat power response is important even if you listen perfectly on-axis.
Also, some people are crossover haters, thinking of them as all sources of (necessary) evil. This scores a point for LR because the power response dip means less likely to hear the imperfections of the crossover point. Butterworth boldly goes on as if nothing has happened. Of course, with a perfect implementation, this should not be a concern.
Also, some people are crossover haters, thinking of them as all sources of (necessary) evil. This scores a point for LR because the power response dip means less likely to hear the imperfections of the crossover point. Butterworth boldly goes on as if nothing has happened. Of course, with a perfect implementation, this should not be a concern.
Last edited:
I don't suspect Cyberstudio is meaning to refer to Butterworth and LR, but rather the power/response relationship.Doppler9000 said:
If the woofer and tweeter are involved in that reflection outside the crossover region, then through the crossover region as well would seem appropriate, even though a dip in that reflection through the crossover may not be the worst thing.
The lobing in question can be measured as a system with both drivers powered. This can be done with a preliminary crossover in place, making this a multi step process.
This is actually a time delay and is dealt with differently.
A minor point: the Scan 9500 is in fact perfectly capable of < 2KHz crossover frequencies in many systems; I've used it happily 1.8KHz LR4 in a 2-way standmount for example. Distortion performance is OK at such frequencies, not quite as low as some units, but still more than acceptable.
With that said, you should be able to get reasonable results using it and the Peerless. Yes, there will be 'better' solutions, but a quality design should still provide decent results. However, without having the measured on-baffle data (or as a minimum for approximate modelling, the baffle layout), exactly what will work well is an unknown. As has been noted, do not fixate upon specific electrical transfer functions; acoustical and electrical slopes are rarely one and the same, and even if they are, they do not account for the physical separation of the drive units in the XY and Z axis and the resulting time delays.
With that said, you should be able to get reasonable results using it and the Peerless. Yes, there will be 'better' solutions, but a quality design should still provide decent results. However, without having the measured on-baffle data (or as a minimum for approximate modelling, the baffle layout), exactly what will work well is an unknown. As has been noted, do not fixate upon specific electrical transfer functions; acoustical and electrical slopes are rarely one and the same, and even if they are, they do not account for the physical separation of the drive units in the XY and Z axis and the resulting time delays.
They are one and the same thing and certain variables are not independently adjustable. For example, if you "fix" the power response of LR its frequency response will no longer be flat. Now you "fix" the frequency response as well and viola, your crossover becomes a Butterworth.
Especially when everyone talks about LR here it is not the exact LR, but "quasi"-LR, which means nothing more than a shorthand for a design methodology where you aim for both flat frequency response and phase coherence. You have to sacrifice flat power response as a result.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.