High Order XOs

[sibi1865]

I have been surprised recently by some favourably reviews I have seen of loudspeakers that I later find use high order crossovers. Some I have heard myself and I have been impressed. Still, it has made me wonder why manufactures are using them, especially with 3-way designs, in light of problems such as tolerance requirements and voice coil temperature. However, these loudspeakers both measure and review well. Clearly, some manufacturers have gone some way to overcome these technical difficulties. Would anyone know how they have done this, and why they would choose to use a high order crossover anyway when they could use one of lower order?

 

[sibi1865]

Apologies if I wasn’t clear! From my understanding, increasing the order of a crossover increases the complexity and the tolerances of components required. If I also understand correctly, there are also increased variations caused by the voice coil temperature on increasing order.

 

[Kvalsvoll]

Would anyone know how they have done this, and why they would choose to use a high order crossover anyway when they could use one of lower order?

It is actually easier to make a high-order filter work properly. Circuit topology and component value calculators can be found for free on-line today - anyone capable of posting on this forum can do it, if you have a degree in electric circuit design you can do it yourself from scratch.

There is only one rational reason for choosing a low order filter, and that is cost. Saving a couple of components in the crossover adds up when you scale up production.

 

[DonH56]

A higher-order crossover provides more flexibility to match driver characteristics for the desired sound without adding distortion caused by sending those same drivers energy outside their frequency range.

Example: Take a woofer with a 300 Hz crossover, drive it with 100 W, and see how much power there is at 600 Hz (an octave above);

First order = 6 dB/octave = 25 W at 600 Hz
Second order = 12 dB/octave = 6.3 W at 600 Hz
Third order = 18 dB/octave = 1.6 W at 600 Hz
Fourth order = 24 dB/octave = 0.4 W at 600 Hz

That energy at 600 Hz may be above what the woofer can handle, increasing its distortion. The energy it is putting out also combines with that from the midrange driver, so you have to ensure they produce a coherent waveform (can be difficult when physical placement and signal phase is varying through the overlap region). The woofer's energy also become more directional as frequency increases, making it harder to control dispersion for smooth on- and off-axis sound. Etc. Same principles apply for the other drivers. The greater the overlap among drivers, the harder it is to get clean sound.

To me it makes sense to use higher-order crossovers to minimize distortion and interaction among drivers. That is one of the big benefits of using active (DSP) speaker systems. It costs more to make, but provides much greater flexibility to the speaker designer, and potentially much better sound.

FWIWFM - Don

 

[Kvalsvoll]

Interesting.
I had no idea that increased voice coil temperature caused greater variations in performance with higher order crossovers compared to the same voice coil temperatures WRT lower order crossovers. Of course, you are talking about all other things being equal, correct?
Could you please provide a link?
Thank you.

Filter cut-off frequency will not change, filter q and slope will not change significantly, even with extreme variations in voice coil resistance. What will change, is the overall level - you get thermal compression.

Since I happen to have a simulation of a system running here right now, I could verify that quite easily. Testing for increased coil resistance shows it behaves exactly like that - at 10% increase there is a significant reduction in overall level, no change in cut-off frequency, no change in slope. For a 20% increase - same, even for a 50% increase - same. By now this tweeter has burned to destruction, so what happens to slope or cut-off is by now irrelevant.

A different filter circuit topology may behave different, but fact is, if properly designed, increased voice coil resistance due to thermal heating does not affect crossover cut-off.

 

[thewas]

Also higher order crossovers reduce the cancellation lobing width at the vertical radiation angles at non coax drivers.
On the other hand lower order crossovers can better "smear" the directivity difference at the crossover region between different drivers at the horizontal angles.
Pretty much everything in this world is a compromise.

 

[sergeauckland]

A higher-order crossover provides more flexibility to match driver characteristics for the desired sound without adding distortion caused by sending those same drivers energy outside their frequency range.


To me it makes sense to use higher-order crossovers to minimize distortion and interaction among drivers. That is one of the big benefits of using active (DSP) speaker systems. It costs more to make, but provides much greater flexibility to the speaker designer, and potentially much better sound.

FWIWFM - Don

Exactly this. I can't think of any reason for using low-order crossovers. (or frankly, passive crossovers at all) With DSP-based crossovers, 48dB/octave is normal, so there's no reason to use anything less. There's this 'less is more' attitude in audiophila, less or no feedback, less or no oversampling or reconstruction filters, less or no short-circuit protection or output zobel networks, no tone controls, all things claimed to improve sound, but, guess what, most of them are actually cheaper to implement than doing it properly. Funny that!

S.

 

[Plcamp]

I believe higher order crossovers to be more susceptible to the precise timing alignment of the drivers...a thing that’s easy to overcome if you are using a dsp for the crossover, and more difficult if you are using post amp crossovers.

The other thing is the loss of amp control of drivers from the inevitably higher series resistance that arrives with higher order crossovers.

I personally think higher order cross is best done with pre-power-amp electronic crossovers, using dsp to “get the parameters right”, and then implement it in a dedicated opamp crossover with multiple power amps.

That way your low output impedance amps have maximal control of the load, and are perfectly time aligned as well.

Hey I even bought the opamps...but have yet to get to the project of designing and implementing the crossover,

 

[sergeauckland]

I believe higher order crossovers to be more susceptible to the precise timing alignment of the drivers...a thing that’s easy to overcome if you are using a dsp for the crossover, and more difficult if you are using post amp crossovers.

The other thing is the loss of amp control of drivers from the inevitably higher series resistance that arrives with higher order crossovers.

I personally think higher order cross is best done with pre-power-amp electronic crossovers, using dsp to “get the parameters right”, and then implement it in a dedicated opamp crossover with multiple power amps.

That way your low output impedance amps have maximal control of the load, and are perfectly time aligned as well.

Hey I even bought the opamps...but have yet to get to the project of designing and implementing the crossover,

Keeping the crossover in DSP has the great advantage of accuracy and repeatability. A dedicated opamp crossover will have tolerances in the component values and it's going to be a lot harder to get 48dB/octave using opamps than with DSP. Noise is also more likely to be an issue using several opamps than DSP. I accept that if your crossover slopes are only going to be 12 or perhaps 18dB/octave, opamps will probably do a 'good enough' job, but why limit it to such low slopes?

Regarding time alignment, this is trivially easy with DSP, much harder to do with opamps.

S.

 

[Plcamp]

I own PAP Trio 15 TB’s, and it’s those where I experimented on a vast array of possible crossover topologies, use of IIR and FIR (and both together). I reached some conclusions about what sounds best...and IMO the slight favorite was a Harsch topology, which isn’t very high order. The one thing I can point at with that topology that is unlike the higher order ones is that is has a naturally small phase shift across the crossover region...maybe that contributed to what I liked.

In my experience, once the drivers were time aligned physically (you can’t time aligned drivers electronically in an open baffle) they all .... from second order through 8th order... sounded pretty much identical. The higher order ones were more susceptible to driver physical alignment, and I perceive but did not test that their off axis response was more erratic.

Most important of all, and this surprised me, was that the fullrange needed to be crossed (as measured acoustically) much higher than I wanted. That fullrange, and I suspect all fullrangers, just can’t deliver down to 200 hz and maintain high end fidelity at the same time. Mine cross at 750 hz acoustically now as a result. Much higher than I want.

 

[HooStat]

So, what about the extreme case with DSP? Why not use a brick wall at appropriate relevant crossover frequency and no overlap at all between two drivers?

 

[Plcamp]

The dsp I have (minidsp2x4 HD) cannot achieve brick wall...except maybe at very high frequencies where you wouldn’t cross anyway ... plus IMO ... don’t see any possible and significant advantage above about 4th order.

 

[sergeauckland]

So, what about the extreme case with DSP? Why not use a brick wall at appropriate relevant crossover frequency and no overlap at all between two drivers?

There's always an overlap, however small, as a true 'brick-wall' filter isn't possible, although one can get close with enough taps. What's important is for the overlap to be short such that the out-of-band energy going into each driver is minimised, which reduces distortion and helps with dispersion. How much difference there is in practice between 24dB/octave and 48dB/octave depends on the drivers' behaviour, but as I can't think of any downside to higher rates, if they're available on the DSP crossover, one might as well use them.

S.

 

[FeddyLost]

Would anyone know how they have done this, and why they would choose to use a high order crossover anyway when they could use one of lower order?

They wouldn't do this if they would use lower order crossover just due to low count of expensive details...
For example, a lot of good pistonical drivers require high-order crossover to get good distortion numbers in breakup region.
There's no sense of making high-end speakers with i.e. ceramic drivers and destroy sound with cones' ringing.
Also, high order crossovers increase safe SPL.

 

[BenB]

I see a lot of benefits to low order crossovers:
Smooth transitions between drivers (whatever character they may have that can be different, like directivity).
Lower distortion in the crossover region, because more drivers are contributing to the sound, reducing the load on any one driver (or set of drivers).
Overlapping drivers can be used to increase system sensitivity (as with the newer version of the BMR compared to the old one.)
Minimum phase implementation with no pre-ringing (compared to DSP brick walls mentioned here).
Low amounts of phase distortion (not usually an issue at 4th order or less, but if it does get worse with higher order).

I do think first order is problematic, since it typically asks too much of the high frequency drivers.

I use cascaded 2nd order filters in my multi-way line arrays to achieve constant directivity (when paired with particular driver layouts). I really like this type of filter. For example, in my skylarks, the tweeter starts off with a 2nd order filter at 3,500 Hz, but transitions to a 4th order filter at 1,000 Hz. This provides a smooth transition, and extra protection.

You do need to choose well behaved drivers to get away with lower order filters, but I think that's usually for the best.

 

[BillH]

I see a lot of benefits to low order crossovers:
Smooth transitions between drivers (whatever character they may have that can be different, like directivity).
Lower distortion in the crossover region, because more drivers are contributing to the sound, reducing the load on any one driver (or set of drivers).
Overlapping drivers can be used to increase system sensitivity (as with the newer version of the BMR compared to the old one.)
Minimum phase implementation with no pre-ringing (compared to DSP brick walls mentioned here).
Low amounts of phase distortion (not usually an issue at 4th order or less, but if it does get worse with higher order).

I do think first order is problematic, since it typically asks too much of the high frequency drivers.

I use cascaded 2nd order filters in my multi-way line arrays to achieve constant directivity (when paired with particular driver layouts). I really like this type of filter. For example, in my skylarks, the tweeter starts off with a 2nd order filter at 3,500 Hz, but transitions to a 4th order filter at 1,000 Hz. This provides a smooth transition, and extra protection.

You do need to choose well behaved drivers to get away with lower order filters, but I think that's usually for the best.

Based upon the discussion I read going on about the Dynaudio Emit M10 crossover, it would seem that low order cross-overs have little to be desired in a 2-way.

The reality of driver summation is quite complex, or perhaps you might say messier. In practice the slopes of the two drivers might not match, and so the 'crossover point' is just an arbitrary frequency where they happen to be playing at the same level. If you're approximating a second order LR slope you will have to invert one of the drivers. If summation at the crossover point is a bit high, you can bring the frequency of the low pass filter up a bit, or the high pass filter down a bit, or you can change the level of the tweeter, or the slope of either driver. In other words, you use whatever electrical components shape the response to get the response you want in space, while hopefully keeping impedance high (since the drivers are playing in parallel, their impedance at frequencies where they play together will be lower - just like a 4 ohm resistor in parallel with another 4 ohm resistor will create a net resistance of two ohms, two speakers with an impedance of 4 ohms at the crossover point will have 2 ohms at that point.)

Since the crossover point tends to be where the greatest directivity discontinuity is, having a peak here might not be a bad idea, since off axis the response may be smoother, and on axis it might sound 'detailed'. If you want actual smooth accurate response through the crossover region you're going to either want a two way with a small woofer and big tweeter, a three or four way speaker, or use a waveguide.

The first order approach does blend the drivers really well, which can reduce directivity issues, but you then run into another problem where the drivers drift out of phase due to the enormous bandwidth they cover. The lobing of first order networks is no bueno, and there will be axis where the response is really bad.

 

[617]

Based upon the discussion I read going on about the Dynaudio Emit M10 crossover, it would seem that low order cross-overs have little to be desired in a 2-way.

It was a good solution in the 80s. Greetings from Somerville, by the way, I grew up in Concord near Bedford.

 

[headshake]

I was just adjusting the timing of my inverted tweeter the other day. Here is a pretty shot of the LR6 I am using. The top purple line is the tweeter inverted back to normal with the 110us timing.

Pretty much everything in this world is a compromise.

I agree. It is easy to get lost in the pros and never know the cons.

I have not spent one second thinking about heat issues until this thread.

Isn't the big downside of low orders is that the vertical directivity is really ugly?

 

[dasdoing]

intresting paper http://www.acourate.com/freedownload/XOWhitePaper.pdf

 

[Plcamp]

intresting paper http://www.acourate.com/freedownload/XOWhitePaper.pdf

I like the subtractive filter approach.