Sometimes a passive crossover is just stupid...

[Waxx]

I love good passive crossovers, and know how to make them. I don't believe in DSP solved anything, altough i also see the advantages (and disadvantages) of DSP. But sometimes people are so hardheaded that no argument can stop them.

I got a guy who asked to design a passive crossover for his dual concentric compression driver. He will use it with an analog active crossover to a woofer at 400Hz. I told him to use dsp, but he insists on all analog. So i started to design with the measurments that he supplied. And it just got ridiculous complex and expensive very soon...

This is only for a top (18kHz to 400Hz) in a custom biradial horn loading to 300Hz, but the circuit needs so many components, that the bil per side became 500€ (also because of the components he chose) and the chance on failure so high that DSP is the only logic solution i think But he heared on internet...

He is still going to build this, and pays me for the design (so no details). So maybe i just should shut up but i won't... For high power please use DSP, the world will thank you.

 

[SIY]

For high power please use DSP

Not just for high power. Digital crossovers let you build in delays and use very high slopes, both major positives for controlling polar pattern.

 

[LSPhil]

formation is not show

show your client that the Linkwitz transformation is not passively possible. Actively looks great what you also hear in reality. I personally am not a fan of crossovers with a division at-6dB but at-3dB. I'm sorry for my English and I promise I'll learn it.

 

[Chris A]

...So maybe i just should shut up but i won't... For high power please use DSP, the world will thank you.

from Siegfried Linkwitz:

“The only excuse for passive crossovers is their low cost. Their behavior changes with the signal level dependent dynamics of the drivers. They block the power amplifier from taking maximum control over the voice coil motion. They are a waste of time if accuracy of reproduction is the goal.”

The only place where passive crossovers have a chance of competing with active digital (DSP) crossovers is at very low power--such as very high efficiency systems, i.e., fully horn loaded...even the bass driver(s). However, the configuration that most fully horn-loaded loudspeakers are designed using multiple apertures with one aperture for each "way"--there are always time alignment issues between ways. So even then, you really need active DSP/multi-amping to correct those time alignment issues.

However, if you use a multiple entry horn design, like the Danley Unity (expired) and Synergy series, then time alignment is assured but only if the crossover uses first order filters, i.e., 90 degrees of phase lag of the lower frequency drivers relative to the higher frequency drivers. Lobing of ways relative to each other isn't an issue with full-range MEHs since the horn provides a common aperture for all drivers.

Not just for high power. Digital crossovers let you build in delays and use very high slopes, both major positives for controlling polar pattern.

The problem with high slopes using IIR filtering is all-pass phase growth through the crossover interference bands: 90 degrees of phase lag for every order of the crossover filters employed, which is audible in loudspeakers/rooms having high direct/reflected energy and low amounts of early reflections. Recommend instead an MEH to avoid those issues and to assure uniform polar output, i.e., no lobing, as well as high efficiency to avoid audible modulation distortion issues, and point source performance.

Chris

 

[StigErik]

The problem with high slopes is all-pass phase growth through the crossover interference bands:

The typical way to implement such filters is to use linear-phase FIR filters, i.e. no phase shift.

 

[Chris A]

I caught this as you were typing. Recommend waiting 2-3 more minutes replying to my posts to catch the "little things"...

By the way, "typical" isn't even 50% of the cases, in my experience. None of the dozens of DSP dial-ins in which I've participated use FIR filtering, and the phase swings are all less than 90 degrees above 100 Hz.

Chris

 

[SIY]

The only excuse for passive crossovers is their low cost.

And these days, the cost isn't necessarily high when compared to the sort of elaborate passive solutions found in high end speakers.

 

[Chris A]

I've been waiting for a few years to see how low the cost of "low cost DSP" would go, and now the prices are beginning to compete with "higher fidelity" passive crossover parts, particularly when higher order filter designs are used.

The problem that I still see is that the type of loudspeakers that I deal with are typically above 100 dB/1w/m efficient, so the noise floor and the sampling rate of the DSP units I've used still tends to be the limiting factor in lowering those prices.

(I've found that 48 kHz internal sampling rate tends to correlate with "rounding off" the sound quality very slightly, while 96 kHz sampling rate seems to add back that last perceptible snap of sound quality. This may be due to other factors in the DSP crossover designs I've used, I might add, so this advice is extremely subjective.)

So passive crossovers are still perceived as less expensive--and "analog" (whatever that really means to the subconscious of the "analog-only" adherents). However, when all things are considered, using better DSP units always seem to win out over time when the resulting sound quality and listener satisfaction are factored in, in my experience.

Chris

 

[Keith_W]

“The only excuse for passive crossovers is their low cost. Their behavior changes with the signal level dependent dynamics of the drivers. They block the power amplifier from taking maximum control over the voice coil motion. They are a waste of time if accuracy of reproduction is the goal.”

I was not aware that passive XO's change with the "signal level dependent dynamics of the drivers". Of course, if Linkwitz said it, it must be true so I am not disputing that. But what is the mechanism of this?

As for "blocking the power amplifier from taking maximum control over voice coil motion", I thought that this was debunked. Some time ago I was doing some reading on back-EMF and how passive XO's prevent the amplifier from "seeing" the driver's back-EMF directly. I asked an amplifier designer if this was true, and he said it isn't. Again, not arguing - trying to learn and see some discussion over these points.

 

[SIY]

I've found that 48 kHz internal sampling rate tends to correlate with "rounding off" the sound quality very slightly

I'd want to see actual evidence for that extraordinary claim.

 

[terryforsythe]

from Siegfried Linkwitz:

“The only excuse for passive crossovers is their low cost. Their behavior changes with the signal level dependent dynamics of the drivers. They block the power amplifier from taking maximum control over the voice coil motion. They are a waste of time if accuracy of reproduction is the goal.”

I removed the passive crossovers from my speakers and went all active. This provided significant improvements - tighter bass, better imaging, better dynamics, better tonality, etc.

Also, making changes to the design is much simpler with DSP - change some numbers on the computer screen rather than change out inductors, capacitors and/or resistors.

A fact too often overlooked is that the most significant weak point of a passive filter is the use of inductors. Commonly used laminated steel core inductors rob the speaker of dynamics, especially in the upper midrange, due to non-linear core losses. Air core inductors don't have that issue, but oftentimes have high series resistance due to the length of wire needed, especially if used for a woofer of a 3-way speaker. That reduces the amplifier's control over the woofer, which is most apparent in the bass.

My speakers had two laminated steel inductors in series with the woofers; one inductor had 1.2 ohm resistance and the other had 0.4 ohm resistance, for a total of 1.6 ohms. The woofers themselves are 4.15 ohms. Pulling out the inductors significantly improved the damping factor, which made a quite noticable improvement to the bass response - it is tighter, more impactful.

 

[terryforsythe]

I was not aware that passive XO's change with the "signal level dependent dynamics of the drivers".

See my post above. (I started writing it before I saw your post.)

For further clarification, when inductor cores are used, inductor core losses due to eddy currents and hysterises increase both with increased frequency and with increased magnitizing force (current and number of coil turns). The core losses are non-linear. Also, the permeability of steel cores is non-linear with respect to magnetizing force and frequency.

 

[Chris A]

I was not aware that passive XO's change with the "signal level dependent dynamics of the drivers". Of course, if Linkwitz said it, it must be true so I am not disputing that. But what is the mechanism of this?

I can see two things, the first being the largest factor by a large margin:

1) thermal compression. (Think of how long it takes a 60w incandescent light bulb to heat up to white hot--it's nearly instantaneous. Most don't think about this when talking about 500w/channel amplifiers.
2) hysteresis of electro-magnetic devices that make up passive designs. (This is very low--probably difficult to measure.)

As for "blocking the power amplifier from taking maximum control over voice coil motion"

I think this is related to output impedance-related instances of the exact amplifiers used. I know that typical SET amplifiers (very high output impedance used without global feedback) have built-in feedback in the triodes themselves, so anything that puts more "distance" between the driver and the output terminals of the amplifier will affect the resulting amplitude and phase of the acoustic output of that driver.

Chris

 

[gberchin]

(I've found that 48 kHz internal sampling rate tends to correlate with "rounding off" the sound quality very slightly, while 96 kHz sampling rate seems to add back that last perceptible snap of sound quality. This may be due to other factors in the DSP crossover designs I've used, I might add, so this advice is extremely subjective.)

If the filters are IIR, and if they are converted from analog prototypes to digital implementations via Bilinear Transform, then you might be noticing the high-frequency warping that is introduced by the process. Those are two big "ifs".

 

[Chris A]

I'd want to see actual evidence for that extraordinary claim.

You can come listen to my old collection of 48 kHz DSP crossovers (including an EV Dx38) vs. a 96 kHz sampling rate Xilica and Yamaha SP2060, as well as the little miniDSP 2x4 HD that runs on my surround loudspeakers (AMT-1s on top of Belle bass bins). It's not subtle.

Chris

 

[gberchin]

1) thermal compression.

Related to that -- as the voicecoil temperature rises, the impedance of the driver changes.

 

[Chris A]

If the filters are IIR, and if they are converted from analog prototypes to digital implementations via Bilinear Transform, then you might be noticing the high-frequency warping that is introduced by the process. Those are two big "ifs".

I yield to the most eminent voice of experience and infinite knowledge on this point.

Chris

 

[SIY]

thermal compression. (Think of how long it takes a 60w incandescent light bulb to heat up to white hot--it's nearly instantaneous. Most don't think about this when talking about 500w/channel amplifiers.

That seems principally like a function of the drivers, not the crossovers. Non-ideality in passive crossovers is principally from series resistances and, as Terry said, the level-dependent inductance and hysteresis of iron core inductors.

You can come listen to my old collection of 48 kHz DSP crossovers

It's your claim, not mine. And absolutely needs listening with basic controls to be taken even vaguely seriously.

 

[gberchin]

I yield to the most eminent ...

Chris, you almost made me choke on my Mountain Dew ... but thanks.

 

[terryforsythe]

2) hysteresis of electro-magnetic devices that make up passive designs. (This is very low--probably difficult to measure.)

The first part is correct. Around 35 years ago I did an investigation into core losses of laminated steel inductor cores for use in passive filters. I used a power loss meter made for that purpose. Still, I had to do a more in-depth analysis to distinguish the eddy current losses from the hysterises losses. It was an eye opening experience. The power losses can be significant, depending on the magnetizing force and frequency.