[woodsyi]

After reading a few threads on several forums and seeing a couple of manufacturers either whole heartedly or begrudgingly (my guess is the latter) offering direct drive units to accomodate the latest affordable DSPs,  I have a question about super steep x-overs.  Is it desirable to have a single driver handle specified frequency?  I ask this because I have played around with different x-over frequency between the ribbons and the woofers on VMPS RM40's.  What I found is that ribbons alone down to 200Hz give very clean sound, but it lacks timber without the woofers overlapping to some degree.  I have settled on 300Hz 4th order low pass for the woofers on modern recordings but for old ('50s and 60's) opera LPs, I bump it up to 500 Hz which boosts the sound from the orchestra pit a little bit to where you can hear them.  I have commented before that even 4th order seems too steep and I appreciate Brian's quasi 1st/2nd order slope which seems to blend quite well.  

To recap, if different drivers (especially if they are different type) have different sonic "texture" at the same frequency, wouldn't it be beneficial to blend the two sounds?  Just the same way great Bordeauxs are blends of Cab and Merlot?

[Rick Craig]

After reading a few threads on several forums and seeing a couple of manufacturers either whole heartedly or begrudgingly (my guess is the latter) offering direct drive units to accomodate the latest affordable DSPs,  I have a question about super steep x-overs.  Is it desirable to have a single driver handle specified frequency?  I ask this because I have played around with different x-over frequency between the ribbons and the woofers on VMPS RM40's.  What I found is that ribbons alone down to 200Hz give  ...

OK, I'll bite  

Some overlap is desireable for good driver integration. The acoustic rolloff of each driver must also be considered. Crossover points in the 200-300hz range can be tricky because that's typically where the microphone will pick up the floor bounce (reflection). This is also more complex because of the upper woofer being far away from the floor boundary. This creates a cancellation that will affect your DSP optimizer. Also, the driver arrangement used in the RM40 will make it more difficult. The planar drivers will have limited displacement crossed that down that low and the woofers may not tolerate a higher crossover point.

Having two drivers of different tonal characteristics overlap in the same area can work quite well and there are many good examples of this in other designs. Optimizing the driver material for specific frequency ranges is beneficial IMO.

[jeffreybehr]

woodsyi:  "  if different drivers (especially if they are different type) have different sonic "texture" at the same frequency, wouldn't it be beneficial to blend the two sounds?"

None of these or any analog filter ('crossover') is 'brick-wall' in operation; there's ALWAYS overlap.  A filter 'slope' is just that--a line heading down (or up) as frequency changes.  That line is NEVER vertical.   The 'slowest' (ie first-order) filters have the most overlap between adjacent drivers, the steep ones (ie 3rd- and 4th-order) filters the least.  Even a 4th-order filter is down only 27dB one octave beyond and 51dB two octaves beyond.  Fifty-one dB down is quite audible.

High-order filters in multidriver speaker systems allow for higher power handling, but high-order filters create higher degrees of phase errors and require additional networks to correct for phase errors, etc.  The more devices in a crossover, the more power is wasted creating heat and the lower the system's sensitivity (which seems to matter very little in line-level filters).

Designing audio products that use filters is an art as much as a science.  You're hearing the differing results of your art now.

[Davey]

I'm a little surprised that Rick and a few others are embracing (is that the right word?) this latest trend because it represents a complete 180 degree change in the traditional speaker design approach.  Normally, a designer would select appropriate drivers for the application, design a box/enclosure/baffle and then take a methodical approach with a series of measurements (usually at 1 meter on and off axis) to evaluate where the crossover points should be, how steep the slopes should be, how they'll effect polar response, power handling issues, etc, etc.  This is designing for essentially anechoic conditions.  Once implemented this would be the end of the process for the designer (and most users,) but there are some users that would apply room EQ for their particular environment to further enhance the presentation.  However, the basic design of the speaker is still the same.

The DSP-based units like the DEQX (used as crossovers) have the capability to complicate this process greatly in my view since the typical anechoic or quasi-anechoic design "rules" are bypassed and the "design" becomes one based on the listening position location vice anechoic measurements.  Intuitively that might seem like a good thing, (and I'm not sure it isn't,) but if there are no constraints placed on the DSP unit then the crossover slopes it generates (maybe very steep,) boosts and dips in response, and various other parameters can certainly have consequences.  Possibly abrupt changes in the polar response of the speakers....An overstressed tweeter because of a too low xover frequency....A strong resonant peak in a magnesium driver that is not fully identified because of the microphone at the listening position....etc, etc.  It's interesting...but possibly very complicated.

I think what Rick is advocating for his speakers is more of a "hybrid" approach wherein the slopes and frequencies of the crossovers are constrained/defined to his specifications and the DSP unit would then adjust time alignment, room EQ boosts/dips, relative gains, etc, to even the response at the listening location.  This seems like a prudent approach however the advocates of these units point to the linear-phase and extremely high roll-off rate crossovers as one of the prime advantages of the devices.  To me, it seems like taking maximum advantage of these "features" would require a back-to-square-one approach to the design and not simply a retrofitted hybrid approach.  Of course, ultimately the user would end up experimenting with various aspects and find "improvements" with possibly unknown consequences.  At that point in time it's really not a Selah, or VMPS, or whatever design anymore.  

It seems like it could be a real can of worms to me.....But what the heck do I know.  

Cheers,

Davey.

[woodsyi]

Rick,

I don't have a DSP and I may not go that route.  I am using a Marchand XM44 to triamp with tubes on the ribbons and ss on the woofers and a larger sub.  I love the sound of these ribbons driven by tubes.  I am, however, intriqued by a digital x-over, driver and room correction capability of these units.  What I would like to find out is what kind of MUSIC this will reproduce and how it would mate with tube amps.  It may turn out that all these digital processing will kill even the beneficial bacteria, yeast, and fungi of music to present nothing more than a synchronous symphony of dynamic sterility.   It would be like a Bauhaus home -- impressive and showy but hard to live in.  But that is my taste and some may love it.  I am also devoted to vinyl because some of the older recordings are not available in CD format.  I would like to hear from a person who has played a vinyl disc through ADC/DSP/DAC to find out how it sounds. How does Meridian deal with Phono?

[Rick Craig]

I'm a little surprised that Rick and a few others are embracing (is that the right word?) this latest trend because it represents a complete 180 degree change in the traditional speaker design approach.  Normally, a designer would select appropriate drivers for the application, design a box/enclosure/baffle and then take a methodical approach with a series of measurements (usually at 1 meter on and off axis) to evaluate where the crossover points should be, how steep the slopes should be, how they'll effect ...

You made some excellent points. My idea is that the manufacturer (Selah, VMPS, etc) has a vested interest that the end result be optimal, both for the owner and anyone else who listens to the owner's speakers. I envision situations where the owner has not optimized the system properly and then everyone places the blame on the speaker designer.

The idea I had was to establish the correct crossover points / slopes and gain settings with a the right amount of equalization correction. I would also use my measurement system to verify the results. The end user could then add the room equalization to optimize the system for the listening environment. Of course they would be free to change the voicing to suit their preferences.

It will open a can of worms but hopefully there won't be as many nightcrawlers in the container after I'm done with it  

[woodsyi]

Jeffrey,

I was referring to the digital filters with 132 dB slope or something llike that.  Unless the filter pole is staggered, I don't see much overlap there. And if they are staggered at that slope, it would be interesting to hear!

[Rick Craig]

woodsyi:  "  if different drivers (especially if they are different type) have different sonic "texture" at the same frequency, wouldn't it be beneficial to blend the two sounds?"

None of these or any analog filter ('crossover') is 'brick-wall' in operation; there's ALWAYS overlap.  A filter 'slope' is just that--a line heading down (or up) as frequency changes.  That line is NEVER vertical.   The 'slowest' (ie first-order) filters have the most overlap between adjacent drivers, the steep ones (ie 3rd- a ...

Just curious as to what listening experiment you did to hear something minus 50db down from the main level?

[mac]

My idea is that the manufacturer (Selah, VMPS, etc) has a vested interest that the end result be optimal, both for the owner and anyone else who listens to the owner's speakers. I envision situations where the owner has not optimized the system properly and then everyone places the blame on the speaker designer.

The idea I had was to establish the correct crossover points / slopes and gain settings with a the right amount of equalization correction. I would also use my mea ...

I totally concur with you.  There really is a conflict between you (the designer) and the DEQX customer.  I think the DEQX appeals to people who already own a set of reasonable speakers and wants to achieve performance that non-DSP turn-key solitions are capable of offering.

With increased power comes the potential for bad things to happen.  I could not imagine trying to "voice" or configure crossover points & slopes without the aid of a good measurement system (I'm not talking about a Radio Shack SPL meter and CD).  Further, even having the right tools doesn't guarantee success.  As Davey points out, tools enable someone with speaker design knowledge to generate desired results.

Fortunately, the DEQX PDC takes a lot of the design knowledge out of the equation.  Sure, you can still screw things up if you try to cross your woofers at 10kHz or even worse, your tweeters at 100Hz.  However, their software is wizard driven and well integrated into their measurement routine.  It's actually quite difficult to end up with bad results using the PDC.

As far as your assertion about driver integration goes, you obviously don’t have enough experience with the PDC.  Old school thinking doesn’t work with DSP technology.  Driver integration is actually enhanced by the use of “super high order” slopes.  There is far less lobbing effects between the drivers and this dramatically improves off-axis frequency response.  The PDC also uses linear-phase FIR filters which don’t have the phase shift that all analog filters have.  In addition to all of this, the PDC actually corrects for phase response and allows one to realize the benefits that some associate with conventional 1st order “time-aligned” designs (w/o all the drawbacks).

You should really listen to a DEQX calibrated system if you have the opportunity (and I don't mean at your friends house).  It may change your way of thinking.

[mac]

None of these or any analog filter ('crossover') is 'brick-wall' in operation; there's ALWAYS overlap. A filter 'slope' is just that--a line heading down (or up) as frequency changes. That line is NEVER vertical. The 'slowest' (ie first-order) filters have the most overlap between adjacent drivers, the steep ones (ie 3rd- and 4th-order) filters the least. Even a 4th-order filter is down only 27dB one octave beyond and 51dB two octaves beyond. Fifty-one dB down is quite audible.

This is what a 96db/octave filter looks like (note the scales):

High-order filters in multidriver speaker systems allow for higher power handling, but high-order filters create higher degrees of phase errors and require additional networks to correct for phase errors, etc.

If you're referring to conventional filters, then this is correct.  However, the  DEQX PDC uses linear-phase filters.  These filters achieve very high-order slopes without any introduced phase errors.

[mac]

Intuitively that might seem like a good thing, (and I'm not sure it isn't,) but if there are no constraints placed on the DSP unit then the crossover slopes it generates (maybe very steep,) boosts and dips in response, and various other parameters can certainly have consequences. Possibly abrupt changes in the polar response of the speakers...

The user of a DEQX PDC can place constraints on how it performs optimization.  This gives the user some control if he thinks he knows better than the engineers at DEQX.  

[jeffreybehr]

Hmm...seems most of the comments addressed to me have to do with analog v. digital filters.  I THINK I qualified my comments adequately, and I'll stand by them.  Woodsyi didn't say he was using a super-steep slope digital filter, just that he's seen manufacturers offering speakers for them.  I think it's wonderful that digital filters can do phase-errorless filtering; didn't know that.  I hope the music sounds as good after 2 conversions as it did before...(not that this tinned-eared audiofool could tell).  

My basis for saying that sound 'only' 50dB down is audible is based on humans' dynamic range of AROUND 100dB (pls don't quibble with this number--it's only a generalization).  

I think I'll bow out of this digital-filter thread, now that I know what it is.  

[Davey]

Mac,

It's still not clear to me how the use of "super high order" slopes enhances driver integration.  It seems to me that this concept taken to the ultimate results in "no" driver integration.....if you see what I mean.    Blending of the drivers response over a finite range is the essence of good crossover/speaker design IMHO, and can result in a "seamless" transition between xducers.  Also, if you think about this further how can there not be abrupt changes in the polar response of the system when the filters are this steep?  (I'm making the assumption that abrupt changes are probably a bad thing and have a negative effect on the "spacial integration" of the system.)  Of course, if the adjacent drivers are selected such that their raw polar responses are essentially similar in the desired crossover range then this scenario seems like it might work okay.  However, I don't think there are many driver combinations used by the typical designer that would satisfy that requirement.  In any case, I don't see how the DEQX can know anything about the polar response of the speaker system when the microphone is so far away at the listening position.

The whole linear-phase crossover concept has been discussed at length on various forums.  I'm not sure how a person can isolate the "benefits" of linear-phase crossovers in the context of an entire system because there are other variables that play into the equation and make it an apples/oranges comparison.  The only concept that I've seen that makes sense is processing a full range signal with an all-pass filter function equivalent to say a typical 4th-order crossover and then A/Bing it with an original copy using headphones or a high-quality speaker system.  I've done this myself using a 100Hz LR4 function and I'll be darned if I can tell the difference between the two.  This would indicate to me that "linear-phase" crossovers have no audible advantage (all other things being equal) over a typical/conventional crossover system.

It would be nice if the DEQX was easily auditionable in ones own home, but currently this is not possible.  The DEQX discussion forum is restricted to owners only and although there is increasing discussion by advocates (like you) on other forums it still requires a major leap of faith for a possible customer who doesn't have access to an audition.  There's also the price, and you know my feelings on that.  

Cheers,

Davey.

[mac]

Hmm...seems most of the comments addressed to me have to do with analog v. digital filters.  I THINK I qualified my comments adequately, and I'll stand by them.  Woodsyi didn't say he was using a super-steep slope digital filter, just that he's seen manufacturers offering speakers for them.  I think it's wonderful that digital filters can do phase-errorless filtering; didn't know that.  I hope the music sounds as good after 2 conversions as it did before...(not that this tinned-eared audiofool could tell).  ...

Yes, you did adequately qualify your statement.  However, I thought you'd find it interesting to see how much overlap there really is with a "super-steep filter".  And yes, it does sound quite excellent after the conversion.  The DEQX PDC does accept a digital input, so if you're already in the digital domain there's really only one conversion done.

Cheers,
mac.

[mac]

It's still not clear to me how the use of "super high order" slopes enhances driver integration...

Dave, I believe we're both basically in agreement on this.  WRT polar response differences between drivers... It wouldn't make much sense to try to cross a 15" woofer over to a 3/4" dome tweeter at 2khz for (I hope) obvious reasons.  I would however think that for the vast majority of applications, one will find that the radiation patterns of two reasonable drivers to be similar (enough) that when a high-order filter is used at a reasonable crossover point the results will be quite seamless.  

Both of us know how two (or more) non-coincident drivers cause lobbing patterns and how the filter slopes effect the interaction.  Isn't it true that the steeper the slope the less interaction/lobbing there will be?

"The whole linear-phase crossover concept has been discussed at length on various forums..."

Yes, it sure has.  There are lots of people on both sides of the fence.  One experiment that I plan to conduct with the DEQX is to setup and optimize my system using both linear-phase filters and conventional LR filters.  With the flip of a switch I'll be able to compare the differences.  A wile ago you offered to make me a "phase scrambled" recording to perform a DBT against a regular recording.  I'm willing to take you up on that offer...

It would be nice if the DEQX was easily auditionable in ones own home, but currently this is not possible...

It might be possible depending on the dealer.  I don't know anyone currently making that offer but that may change soon.  Having been burned badly by over-hyped hifi products  on the internet, I can appreciate your apprehension.

[John Casler]

While I don't speak for Brian, I might add a word about VMPS being made available for Digital Signal type processing.

VMPS speakers "are" designed by B, to be what he considers the VMPS sound, based on maximum performance to his design philosophy and preferences.

It is clear that DSP type applications are in their infancy, and "WILL" grow in sophistication and accuracy in the future.

At present, all who use this type of system are at the mercy of the limitations and restrictions of the technology, as well as the "assumed" algorythms and measurment analysis of the various manufacturers.

I know acoustic engineers "Are Not" speaker designers and I, for one, don't expect that speakers used in this way will sound as they were intended.  I do think that there "are" some very interesting abilities, that can provide benefits, once the total comprehesnion of their integration becomes known.

Phase, frequency, dynamics, coherancy,  and other issues which are very sensitive, yet not well known will become the playground of audiophiles.  Much of the "magic" of Designer Speakers will "NOT" transfer to this technology, until the designers themselves see it as a viable appliction for them to use and offer.

VMPS (to my way of thinking) is not "embracing" this technology as much as "addressing" a client/customer group that may find exploring this new path with our products, fun, useful and beneficial.

I don't anticpate that in the begininning that the overall VMPS sonic personality can be reproduced anywhere true to B's intentions, but, that does not mean that in the future, early adaptors might find additional information and applications that can or may find their way into the sound.

That said, we are looking at two areas:

1) Speaker Correction
2) Room Correction

At this juncture I see the room correction as being the most valuable for those wishing to maintain the sound we represent.

Again these are my thoughts, simply observing the potential and the direction of what is currently available.

[mac]

While I don't speak for Brian, I might add a word about VMPS being made available for Digital Signal type processing....

John, This may be a good opportunity for you to discuss some of the technical aspects of the VMPS speakers.  Can you tell us about the accuracy of their measured performance?  Yeah, I know a lot of audiophiles poo poo measurements, but some of us find them fascinating when interpreted properly.  

DEQX claims that they can correct the amplitude response of a speaker to within a decibel.  They also make a similar claim for phase response.   In addition, they also provide room correction because not all rooms are perfect.  (Check out my gallery here if you need confirmation.)

It is clear that DSP type applications are in their infancy, and "WILL" grow in sophistication and accuracy in the future.

If this is considered emerging technology, I can hardly wait until it matures.     What's cool about that aspect is it's easily user upgradeable!  DEQX makes firmware and software changes available to all of their users for free.  

I won't debate you about people wanting "house sound" from their speakers.  Plain and simple, those people will probably not be interested in a product like the DEQX PDC.

[John Casler]

John, This may be a good opportunity for you to discuss some of the technical aspects of the VMPS speakers. Can you tell us about the accuracy of their measured performance? Yeah, I know a lot of audiophiles poo poo measurements, but some of us find them fascinating when interpreted properly.  

Hi MAC,

I am not qualified to speak to the more technical aspects of Brian's drivers.  Only he knows what they do, and are intended to do.

I generally don't put too much stock in measurements since it is sometimes debatable as to how they translate to the real world of sound, but as measuring gear becomes more accurate and measuring applications become more meaningful, I will adjust.

If this is considered emerging technology, I can hardly wait until it matures.     What's cool about that aspect is it's easily user upgradeable!  DEQX makes firmware and software changes available to all of their users for free.  

I won't debate you about people wanting "house sound" from their speakers.  Plain and simple, those people will probably not be interested in a product like the DEQX PDC.

What I mean by infancy, is that the programs are set up using current knowledge bases and associations.  The science of acoustics is far from complete, due to interpretation and application differences.

Programs must be focused on your acosutic goals and I am not knowledgable about the current hardware/software to know how well it does address these issues.

The hardware/software to ear/brain chain is very intricate and it will take many minds to open the layers needed to really reach a large percentage of its potential.

We're just unzipping the zipper right now.  We have a ways to go.  I can't wait to get to "first base" (or is that bass?? )

[ted_b]

Mac,

  Also, if you think about this further how can there not be abrupt changes in the polar response of the system when the microphone is so far away at the listening postion

Just a clarification, then I'll bow out:
The DEQX processes are two-fold.  The requirement is that speaker measurement and correction be done near-field (i.e 1 meter or so) and as anechoic as reasonably possible (yes, they even ask if the speaker is able to be transported outside!).
The room measurement process, however, is indeed done from the listening position, or for more accurate and smooth response, an averaged set of listening postions, if possible.

Ted_B

[Davey]

Ted,

Well, you didn't quote me correctly, but no matter.  Thanks for the clarification on the DEQX methodology.  The two step approach seems like a much better way of doing it....I wasn't aware this is how the unit operates.  Good info.  Thanks.

Davey.