[ctviggen]

http://www.diyaudio.com/forums/showthread.php?s=&threadid=15943&perpage=10&pagenumber=6

See, eg, post 57 (actually, the whole thread is interesting, but this post deals specifically with digital crossovers). I agree with this assessment.   Digital crossovers aren't taking into account actual driver response (eg, Qts, xmax, impedance, etc.) or box (eg, Qtc) response.  Therefore, while beneficial, digital crossovers have substantial faults, too.  I'm looking for the day when you plug your driver and box response into the equations and the digital crossover is adjusted therewith.

[JoshK]

I had read that thread before.  I think Thunau is referring to a box like Behringer and not so much DEQX/TACT although some of what he is saying is still applicable.  But this is nothing new, and people like Rick Craig have been saying it all along, that simply buying a digital xo, plopping speakers in a box and setting text book electrical crossovers is not going to do it.  They are both right.  

You still need to know what you are doing in terms of how to measure, what measurements to take and how to compensate for this in your crossover.  Then you go to your digital crossover and program accordingly.  So the only thing the digital crossover in this case buys you is the ability to not do it via passive components.  However, if you don't have the ability to learn passive crossover topology and can solder it up, chances are you probably aren't going to get the digital implementation right either.  Linkwitz more or less has said the same thing.

The Behringer can still be programmed, from what I gather, to do the appropriate xo.  I have seen some threads where the Behringer is made to do some pretty complicated slopes and tailoring, so it isn't that the Behringer is limited but that its software isn't setup to be a plop in the box solution.  The UI of most of the applications I know of fall short of being dummy proof, and require some knowledge on the users part to assist in the design of the filters.  I think though that the software is getting easier and easier to use.  Will it ever be a simply plop in the box application?  Don't know, maybe, maybe not.

[Davey]

That's not a "fault" with digital crossovers.  It's inherent because of the "open-loop" nature of a typical system.  It's up to the user to factor all that other information into the setup and adjust accordingly.

Only when a microphone is used and you have a "closed-loop" system does the final result resemble a "plug-and-play" result like you're interested in.  And even that road is littered with many potholes.  

The days when a system can be optimized fully and hold your hand during every step of the process are aways off me thinks.

Cheers,

Davey.

Well, you posted when I was typing Josh.  Thanks for making me look like an idiot.  

[art]

For $300, it can turn any hobbyist into a speaker designer. At least in his own mind. As long as it gives them the satisfaction of twiddling knobs to their hearts content and keeps them occupied, then I am all for it.

No, they won't want to hear about driver phase response, polar radiation patterns, and the like. But with a seemingly endless array of possible bad XO combinations, they won't ever have to, will they?

Pat

[timbley]

I've been playing with a pair of DCX2496 and a DEQ2496 to run a set of 5 way active speakers I built last year. I've heard some great sound out of this system, but it's increasingly un-reliable. This Behringer gear gets all out of whack from time to time. So, I "re-boot" everything and it's better, but not necessarily 100%. Sometimes it's worse with clearly audible distortion and digital grungy noise. One unit had an intermittent noise problem that's grown into a continuous noise problem. Sometimes I get a deep notch in frequency response around 10K that simply won't go away until the gear is re-started or the digital signal is switched off and back on. Every time a digital signal is lost and re-connected there's a good chance something will be out of whack.

When the gear is all working right, I don't think it's that hard to design a good sealed box system with decent drivers and get quite excellent sound with a digital active crossover and some half way decent amps. I'm highly skeptical of the need for much of the more sophisticated measurements and corrections. But I'd like to get some better stuff just in hopes that it'd actually work correctly.
My system can sound fantastic, but it's become a headache and I'm getting tired of it. Lately something has gone wrong again. It's developed an edginess that's irritating.  It might be that 10K notch thing permanently stuck, or maybe something new. I'm demoralized and don't feel like messing with the mic and RTA any more.

If I were a professional musician I would not trust this Behringer gear as it could decide to have a bad day and totally wreck the sound of a live performance.

[skrivis]

Digital x-overs don't eliminate the need for knowledge and measurement, but they do eliminate some of the problems of passive high level x-overs.

For instance, you may not have to worry so much about some of the driver parameters since you're driving it straight from the power amp and can have a much lower impedance looking at the driver.

You will have an amplifier channel per driver, so there will be less interaction. (At the amplifier side and the driver side of things.)

If you set a given slope, you will get very close to that slope electrically. That isn't as true with passive high level x-overs.

It may also be easier to work with enclosure effects, since they won't be throwing off your x-over nearly as much.

Where it really gets interesting is that there's the possibilty of using drivers that don't work well with 1st-order x-overs. (I feel that standard high-order x-overs have major problems, especially with phase response.) We may be able to use a very steep x-over and correct the phase and time response so that it sums to a piece of wire, like 1st-order x-overs do.

For right now, I lean towards well-damped drivers that are "stiff enough." These won't have the peaky resonances that most stiffer drivers do. (Yes, they do have a broader, low amplitude resonance.) I feel this is less objectionable than the approach of using drivers with peaky resonances. It works out quite well with 1st-order x-overs too.

The stiffer diaphragms that you get with metal cones or domes can be more linear than the "stiff enough" drivers - within a limited frequency range.

Something like DEQX may well make it possible to employ more linear drivers, but still accurately reproduce a transient waveform.

You'll have to know the characteristics of the drivers and enclosure. That will allow you to make intelligent choices. You've got sharp x-overs to keep the drivers from getting into trouble. Then you have to adjust _the entire system_ for the best performance. It's still a "sum of the parts" situation, so you have to know what the end result is. As someone pointed out, you can't just slap in a "perfect" x-over and expect the end result to be perfect too just because of that.

A further issue is the increasing availability of computer software for designing things. There's a marked tendency to build the expert knowledge into the software. It makes things easier in some ways for good engineers. But it can also to some extent cover up the mistakes of a duffer. It can even reduce flexibility for the sake of uniformity and ease of use.

I think word processors are a good example. They allow almost anyone to produce uniformly mediocre documents. They correct some of your mistakes, but you never gain the knowledge needed to produce really excellent documents, and they may not even allow you to produce anything that's superb.

The moral of this story is that there are no panaceas.

[dwk]

I agree with the observation that none of this is new. The point about filter slopes etc has nothing to do with digital vs analog or even active vs passive, but is all about the features available in standard off-the-shelf active xovers (both analog and digital).  I don't think that anyone with a real understanding of xover design would have any illusions about this - I suspect it's only the beginner or layman that may be confused and taken in by some level of 'DSP is better' hype.
 This shouldn't undermine the value proposition of the DCX, though - by using a measurement package and carefully combining the built-in xover filters with parametric stages, it's still possible to get very good systems out of it. The point is only that you can't dial in a textbook 4th order eletrical xover and expect world-class results.

BTW - Jan (Thuneau) is at least backing up his position. He has produced a PC-based xover package that I'm beta testing that appears so far to be excellent.  Far more flexibility than the DCX et al in defining your xover biquads - multiple stages each with their own Q and freq, plus you still get parametric for handling in-band and/or breakup problems.  You still need a measurement package, but the nature of the software makes integrating this into a more closed-loop workflow reasonably easy - I still have much more experimenting to do in that area, though.

[timbley]

I've been playing with a pair of DCX2496 and a DEQ2496 to run a set of 5 way active speakers I built last year. I've heard some great sound out of this system, but it's increasingly un-reliable. This Behringer gear gets all out of whack from time to time...

All this is fixed with waxed paper under the XLR boards to keep the circuits from touching the steel chassis. For now I'm back to loving the Behringer DCX crossovers. Still, shame on Behringer for not preventing this problem in the first place.