[andyr]

OK, Xover gurus ... I currently run my Magneplanars 3-way active, using the stock XO slopes (18dB bass LP, 12dB mid BP and 12dB ribbon HP).  Various people are using the (digital) DEQX unit with their 3-way Maggies, using 48dB and 96dB slopes ... however, it should be understood the DEQX provides some degree of group delay correction, in addition to straight active filters.

As I understand it, group delay increases dramatically with higher order filters than 24dB ... in fact some people refuse to go even this high because of its negative effect!    So I was wondering whether 48dB analogue filters would be worth doing ... or will the resulting group delay make the system sound bad?

IE. is this something you really need to do in the digital domain because of the associated DSP facilities which you have available?  (An analogue implementation of 48dB L-R Xovers would be simply that ... there would be no associated group delay or phase compensation!  )

Regards,

Andy

[JoshK]

Go to htguide.com/forum and search in the Mission Possible DIY sub-forum on cauer elliptic filters.  There are a few examples that use them.  In my gallery, you can see my Modula MTM monitors which use this CE filter.

Basically the CE filter is very similar to a 8th order LR filter but has less parts count and less group delay.  Jon Marsh is a bit of a brainiac and he brought this topology back to life for xo design.   I find my Modula's sound pretty nice.   They also can tolerate pretty loud levels since the tweeter is shielded from virtually all lower freq's outside of passband.  In the Modulas the tweeter is crossed over very low as well to avoid the lobing the MTM setup creates.

[JoshK]

In response to your question about whether or not to do it in the digital domain, I'll add what I know.  Your assumption is correct that this is best done in the digital domain.  Analogue filters are IIR type filters and thus they can do no phase correction. 

The FIR filters OTOH can and often do do phase correction.  There is a price to pay, however, and that is with the pre-ring (shown to be audible in the pro-scene, but lots of happy DEQX users).  That is where my knowledge on the matter ends. 

Not all digital xo's use FIR filters, many use the standard IIR and therefore they can do nothing for the group delay.  I don't know of many digital xo/eq boxes that use FIR outside of Tact and DEQX.  But there are a few software programs that use FIR (like BruteFIR) for systems that involve a computer and pro audio sound cards with multiple outputs.  That, by the way, is the cheaper solution but is much less simplistic.   I chose to make an investment in the expensive DEQX since it has many uses and virtues.

I know a number of folks who held out for the prices of DEQX competing boxes to drop in price and become more available.  Well I've had my DEQX for over 2 years and they've only gotten more expensive and the competition is still nill. 

[andyr]

Thanks for your help, Josh.

I'll spend some time researching the htguide.com/forum site for those Cauer filters. 

Regards,

Andy

[Davey]

Andy,

I think an answer to your question might be in a practical experiment.  Try to borrow a DCX2496 Behringer crossover and program it for a generic 48db/octave LR crossover.  Then sum the two outputs together to create a fourth-order all-pass function.  You can insert this into your playback chain (probably with headphones) and see what type of audible difference you can discern relative to the reference.  The "reference" should be a direct feedthrough setup of the DCX so you can eliminate any contribution the ADC/DAC of the Behringer might contribute to the sound.

If you think the result is acceptable then you have another project to build. 

Dave.

 

[andyr]

Andy,

I think an answer to your question might be in a practical experiment.  Try to borrow a DCX2496 Behringer crossover and program it for a generic 48db/octave LR crossover.  Then sum the two outputs together to create a fourth-order all-pass function.  You can insert this into your playback chain (probably with headphones) and see what type of audible difference you can discern relative to the reference.  The "reference" should be a direct feedthrough setup of the DCX so you can eliminate any contribution the ADC/DAC of the Behringer might contribute to the sound.

If you think the result is acceptable then you have another project to build. 

Dave.

Hi Dave,

Thanks for the suggestion - I agree using a Behringer for an experiment would be a good thing to do.  However, I currently don't have any headphones! 

Can you just confirm the following is what you meant ...
* program the DCX2496 Behringer crossover for a generic 48db/octave LR crossover - let's say we try 3 examples, at 100hz, 1Khz, 10Khz.
* add the LP & HP together - which gives, essentially, no roll-off at all?
* the Behringer has a bypass switch, to send the input signal ("the reference") straight through to the output, with no processing?

Thanks,

Andy

[Davey]

Andy,

If you sum the two outputs together the result is a flat amplitude response, but phase response (distorted) the same as an 8th-order crossover.

The DCX doesn't have a straight bypass switch, but you can send the input to the output with no processing.  (Via the ADC and DAC stages and all the rest of the electronics.)

This way when you A/B the difference is only the phase distortion of the particular crossover you have setup.

Of course, if you ultimately use speakers in a real room then all sorts of other variables come into consideration. 

Cheers,

Dave.