The sonic benefits of an active crossover. A discussion.

[DonH50]

Hello Don

The complexity of the midrange drivers voltage drive curve. Any out of the box analog could not replicate that curve. Typically they are all limited to simple slopes, some adjustable in the crossover region over one or more crossover points. You can't do notch filters or any complex curve.

Now digital you can do it all.

Rob

Hey Rob,

I think I misunderstood. I thought Tim was asking for any proof that a passive speaker-level crossover could provide better performance than using a line-level crossover, and I did not see how your example provided that. It certainly highlights the benefits of DSP...

Tnx - Don

 

[DonH50]

Frantz -- Whilst agreeing with most of your points, I must note a DSP is a pretty complicated circuit, it just hides the complexity in a little chip...

 

[opus111]

From an implementation viewpoint, don't active filters provide better approximation of said filter (crossovers) functions?

In general yes, active filters provide closer to the theoretical frequency response as they can be built with defined loads (terminations) - passive XOs have non-linear terminations - drive units.

Aren't passive crossover very wasteful? (For the most part anyway)?

Its true that a passive XO can't provide boost so normally a tweeter padding resistor is included which is wasteful. Similarly with baffle step correction there's bound to be a waste of amplifier power. So I'd say definitely wasteful (of the order of 3dB or so potentially, which is a halving the effective amplifier power) but perhaps not very wasteful.

Can we compare the distortion induced by passive crossover to those of active? IOW which of the two distort more?

If the active XO is implemented with opamps then the THD will normally be extremely low. The THD of a passive XO might be rather high by virtue of adding source impedance and hence the non-linear load of the drive unit imposes the distortion. However THD isn't highly audible if low-order (which it normally will be). The opamp distortion will be more likely lower-level IMD which will be audible on transparent sources. It reduces the tonality, adding a greyness (or coldness) to the sound if implemented well, and adds sibilance and potentially other colouration in addition if implemented badly.

 

[FrantzM]

Opus111

From the above it looks that I don't need to change my stance concerning Active vs passive. Thanks. Reinforce my view that digital active filters with active driver correction may be the way to go.

 

[FrantzM]

Frantz -- Whilst agreeing with most of your points, I must note a DSP is a pretty complicated circuit, it just hides the complexity in a little chip...

I agree Don but in the digital world we seem to be at ease with utterly complicated circuit

 

[DonH50]

True, "but it's just one chip", but I am an old analog dinosaur...

 

[Phelonious Ponk]

FWIW, while I'm sure digital has the potential for cleaner performance, my humble monitors use analog active crossovers, and they definitely have "the sound." Makes me wonder if it comes more from the direct coupling of amps to drivers, the careful matching of amp/power to drivers, the lack of "competition" for power during critical passages...than it does from the crossover circuitry itself.

But it's all speculation on my part...

Oh, and by the way, it occurs to me that there is a place where you can hear and compare almost exactly the same cabinets and drivers executed in active and passive versions -- pro sound reinforcement. Not "high end," but I think the high end might be surprised by what they hear. Just pack up your $30k turntable, your $10K phono pre, and head for Guitar Center. See you there!

Tim

 

[microstrip]

All I read until now only states that an active speaker will probably measure better according to some unknown criteria.
No one has proved that low power amplifiers sound better than high power ones, that damping factor is a critical aspect, and good quality passive components degrade significantly the sound quality.

Not enough to change my opinion already stated in post #48.

 

[FrantzM]

All I read until now only states that an active speaker will probably measure better according to some unknown criteria.
No one has proved that low power amplifiers sound better than high power ones, that damping factor is a critical aspect, and good quality passive components degrade significantly the sound quality.

Not enough to change my opinion already stated in post #48.

ok ...

 

[treitz3]

One other point: inductors can exhibit hysteresis much worse than capacitors when driven hard, and those were usually the largest source of distortion in high-power crossovers.

Hello, Don. Would this be measurable or audible distortion, perhaps both?

Tom

 

[DonH50]

Measurable, yes, easily, but probably masked by the drivers themselves.

Audible? I am not sure... I suspect the answer is the same as above except when significantly overdriven. It's been too long since I did those tests; at the time we thought we could hear a difference, but I do not recall at the moment if we did much in the way of DBT. And, since we went from passive to active, there were too many changes to point to a single component. I did measure the crossovers alone, but listening tests were of the entire systems before and after conversion.

There are a lot of folk who claim to hear significant differences after crossover mods. In some cases, in the past, I was able to figure out why, or thought I did, and there were real, measurable reasons. On this, like wires and fuses, I am skeptical but hesitant to dismiss a large body of empirical evidence.

 

[treitz3]

Fair enough. Thanks, Don.

I think the biggest thing I have read so far that I can relate too regarding the possible deficiencies of an active is the sonic signature of the unit itself and added/associated wires. One other thing was that some folks had mentioned that you can't take a passive speaker and make it active, that you [EDIT:]*should* start building the active system from the ground up. Why is that?

It has also been said that you can do anything with digital. So this leaves me wondering why someone would have to start from scratch with an active to achieve the best sonic end result.

Tom

 

[microstrip]

Hello, Don. Would this be measurable or audible distortion, perhaps both?

Tom

This subject was covered by speaker manufacturers long ago - I remember seeing papers and measurements in Wireless World (now called Electronics World) when I was still a student - I think that the studies were carried by B&W and the BBC. Almost all good quality inductor would show negligible distortion, only some chokes using transformer type inadequate EI lamination measured poorly. Some manufacturers however choose to use cores that have characteristic distortions and introduce some euphonic coloration - I think that its is not we are addressing here.

At some time in the 80's it was fashionable for the manufacturers to advertise the superior saturation properties of the cores they used in the coils. At that time crossover boards became a lot bigger!

An important aspect of inductors in passive speakers is susceptibility to vibration. As far as I know, only Mundorf publishes data on this effect, but only relative measurements.

 

[DonH50]

• I have no idea why you couldn't rip out the crossover and replace it with active. Perhaps it is better to start from ground zero but I don't see a clear reason why that would be a requirement. The tests I did so long ago started with (IIRC) Polk, B&W, and Infinity speakers; we bypassed the internal crossovers and replaced them with (analog) active crossovers and connected the amps direct to the voice coils. I did similar with my Maggies for many years, but only replaced the external box so the mid/tweeter crossover was still passive.
• As for starting and targeting a digital crossover-based design, I am not a speaker designer so am not sure what the trades would be, nor how much active or passive crossovers would drive things like driver selection and cabinet design. Certainly the digital crossover gives you a lot more flexibility.
• Solid-core inductors exhibit the worst hysteresis by far and it happens mainly at high currents. Same thing happens with transformers (which are coupled inductors). I agree with microstrip that any decent design would have that masked by other things (driver nonlinearity being the major one in my mind).
• Inductors and capacitors are sensitive to vibration (and heat, etc.), certain types more so than others. I am not sure the sonic impact.

 

[microstrip]

(...) It has also been said that you can do anything with digital. (...)

Tom

Tom,

Yes, but still a lot to debate on this aspect. Unless you one those members who thing that all DACs sound the same, you will have no problems accepting that the algorithms to implement anything will have a strong sonic signature. And perhaps those who are masters in the coding of the algorithms lack the expertise to analyses their sonic characteristics.

 

[microstrip]

• I have no idea why you couldn't rip out the crossover and replace it with active. Perhaps it is better to start from ground zero but I don't see a clear reason why that would be a requirement.

• 
  
  Don,
  
  Many designers ( I would risk almost all of them) voice their speakers to have a non flat response, introducing the small irregularities in the crossover design. IMHO it would not be a fair challenge to a digital crossover to have to reproduce all these small effects. Also the designers use passive elements to damp the dynamic response of the units in a way it is not easy to measure and model - they really must know how to deal with back EMF of the units in passive designs.
  
  Perhaps speakers having simple crossover circuits, such as your Magneplanars, having panels that are mostly resistive and most older designs will be good candidates. But if you read the Toole papers on crossover design, you will find that, perhaps in part due to the availability of computer simulators and quality measuring systems, many crossovers of modern speakers are becoming very complex.

 

[GaryProtein]

• I have no idea why you couldn't rip out the crossover and replace it with active. Perhaps it is better to start from ground zero but I don't see a clear reason why that would be a requirement. The tests I did so long ago started with (IIRC) Polk, B&W, and Infinity speakers; we bypassed the internal crossovers and replaced them with (analog) active crossovers and connected the amps direct to the voice coils. I did similar with my Maggies for many years, but only replaced the external box so the mid/tweeter crossover was still passive. . . . .

That's exactly what I did. I didn't have to rip out the old, I just disconnected the passive crossover between the midranges a tweeters and added an analog active crossover and more power amps. (In the IRS-V there was already an active crossover between the woofer and midrange) I use a PassLabs XVR1 three-way active crossover. I initially set the crossover points and slopes the same as the original passive crossover as a starting point and then made slight adjustments in the crossover points between the woofer and midrange. That took a fantastic speaker and made it fantasticker.

As Robert Frost said, "Two roads diverged in a wood and I took the one less traveled, and that has made all the difference."

 

[FrantzM]

Don,

Many designers ( I would risk almost all of them) voice their speakers to have a non flat response, introducing the small irregularities in the crossover design. IMHO it would not be a fair challenge to a digital crossover to have to reproduce all these small effects. Also the designers use passive elements to damp the dynamic response of the units in a way it is not easy to measure and model - they really must know how to deal with back EMF of the units in passive designs.

Perhaps speakers having simple crossover circuits, such as your Magneplanars, having panels that are mostly resistive and most older designs will be good candidates. But if you read the Toole papers on crossover design, you will find that, perhaps in part due to the availability of computer simulators and quality measuring systems, many crossovers of modern speakers are becoming very complex.

microstrip

Any crossover has a transfer function. In the case of a speaker crossover it is not always trivial to find what the transfer function is but that can be achieved. From there replicating with digital is eminently possible I would venture to say ALWAYS possible and in much finer and repeatable ways than passive components. As for damping the back EMF a low impedance like an amp even tubes could be very welcomed by most drivers.

 

[microstrip]

microstrip

Any crossover has a transfer function. In the case of a speaker crossover it is not always trivial to find what the transfer function is but that can be achieved. From there replicating with digital is eminently possible I would venture to say ALWAYS possible and in much finer and repeatable ways than passive components. As for damping the back EMF a low impedance like an amp even tubes could be very welcomed by most drivers.

Frantz,

Hypothetically everything is possible. As I often say with infinitum resources and infinitum time we should all create the perfect product. It seems that, as I have some limited experience in audio measurement and always think about the practical aspects and you just argue with the theory without any constrains you will always win.

We can always carry a poll - which of our members is able to measure the response of the crossover (modulus and phase), fit a mathematical function to this data and then develop the digital algorithms to implement this transfer function? And remember that if you are reproducing a measured curve, being finer and repeatable will only mean you will reproduce 100 times exactly the same error.

 

[DonH50]

Don,

Many designers ( I would risk almost all of them) voice their speakers to have a non flat response, introducing the small irregularities in the crossover design. IMHO it would not be a fair challenge to a digital crossover to have to reproduce all these small effects. Also the designers use passive elements to damp the dynamic response of the units in a way it is not easy to measure and model - they really must know how to deal with back EMF of the units in passive designs.

Perhaps speakers having simple crossover circuits, such as your Magneplanars, having panels that are mostly resistive and most older designs will be good candidates. But if you read the Toole papers on crossover design, you will find that, perhaps in part due to the availability of computer simulators and quality measuring systems, many crossovers of modern speakers are becoming very complex.

You could certainly design a line-level crossover to provide whatever response is desired, flat or not.
Connecting the amp directly to the drivers should handle the damping and back-EMF issues nicely.
Are they becoming more complex to better comensate the response? If so, why not move to active line-level crossovers?

There are many program that simplify the task of converting a transfer function into a filter. For that matter, the analysis program I use (RplusD) will create a filter function from the measured room response or T-S parameters that can then be fed directly into several DSP units.