What's new

A Question of Balance (1 Viewer)

BruceD

Screenwriter
Joined
Apr 12, 1999
Messages
1,220
Hoping I can get some discussion on this topic from some of the DIY sub gurus here:

One of the biggest obstacles to good integration between main speakers and subs is the implementation of the digital crossover on an HT processor.

Most of the receivers and prepros on the market were designed with an electrical high-pass @80Hz, 2nd order, 12dB/octave rolloff bass management algorithms to fit THX certified speakers.

If your speakers are anything other than THX certified (not my cup of tea) good crossover integration across the crossover boundary (between those speakers and a sub) is nearly impossible. Great musical sub integration with an HT processor's internal bass management and non-THX speakers is almost an oxymoron.

Personally, I prefer a high quality active crossover with symmetrical 24dB/octave L-R high-pass and low-pass filter slopes.

There are a number of HT processors that allow more sophisticated bass management than what I mentioned, i.e. Theta, Meridian, Lexicon to name a few.

What I'm referring to is the assumption made in the digital bass management algorithms about your speaker's low frequency rolloff characteristics.

I find it very hard to blend a speaker with a 18-36dB/octave low frequency rolloff/slope --that many ported speakers have-- plus the 12dB/octave electrical slope of the HT processor for the high-pass side of this crossover equation (equals 30-48dB/octave slope) with the low-pass side of this equation which the HT processor models on a 24dB/octave slope.

The anomolies (peaks and dips >3dB) at and around the crossover frequency described above make this setup less than desirable for a smooth main-to-sub blending for the best musical enjoyment.

My point is NOT to use a sub's internal crossover, but instead to use an active external standalone crossover (like a Marchand ) to optimize speakers, sub, and room.

To do this means re-directing all bass from Small center and Small surrounds to the main Large L&R preouts and re-directing the LFE to the main Large L&R preouts as well. Then the xover is only done once; from the Large L&R preouts to the active external crossover and then to the L&R mains and sub(s).

I find this provides the best musical optimization of a sub for 2-channel and still works great for HT. For my speakers, room and sub this happens @60Hz.

I've been doing it this way for the last 4 years.
 

Dave Milne

Supporting Actor
Joined
Jul 2, 2001
Messages
568
Bruce,
This is indeed a complex and interesting topic.

I find it very hard to blend a speaker with a 18-36dB/octave low frequency rolloff/slope --that many ported speakers have-- plus the 12dB/octave electrical slope of the HT processor for the high-pass side of this crossover equation (equals 30-48dB/octave slope) with the low-pass side of this equation which the HT processor models on a 24dB/octave slope.
I have struggled with this myself. However, since most receivers and pre-pros have symmetrical slopes, one must assume that they beleive the actual speaker rolloff is far enough away (in frequency) from the crossover to be insignificant. The filter design experts would say you should be a decade apart to fully eliminate interaction. Nobody makes a main speaker with 8Hz rolloff, so this clearly is not the case. In my experience, however, if you're an octave away, it's pretty safe. A lot of main speakers do have 40Hz capability.

The idea of matching the crossover frequency/slope with the speaker rolloff characteristics is not new, but is rarely considered or implemented. It's just not practical for the mass market. As you pointed out, receivers and most pre-pros have limited crossover options... and few main speakers have precise 12dB, 80Hz rolloff. It's hard enough to educate the masses on picking the right crossover frequency...let alone matching asymmetrical slopes to speaker rolloff characteristics. ;)

One of my favorite designs, on paper, is to match a passive RC filter (between pre and main amps) to sealed mains for a true third-order rolloff. This avoids active crossover circuitry in the signal path to the mains and minimizes noise and distortion. Then just use a third-order low-pass on the sub for seamless integration. I've done a few of these designs and they work well. A variation on this is to use 2nd order highpass with sealed mains and 4th order sub lowpass for matched Linkwitz-Reilly responses -but this requires active filters in the main signal path.

However, I've had equal success with symmetrical 24dB LR filters (using my own analog active filter designs or the 24dB LR option in my Lexicon MC-12). In fact, my current theater system has symmetrical 24dB 70Hz slopes with 12db/52Hz speakers :confused:. This measures quite flat through the crossover region in my room. The key here is that rooms often have ragged response this region that easily swamps out a few dB of peak or dip in sub integration. If you have some test gear and are willing to experiment a bit, you can often flatten overall room response with a deliberate crossover mismatch!

Happy experimenting!

My home page with theater and speaker details
 

RichardHOS

Second Unit
Joined
Mar 11, 2003
Messages
454
You can also muck up phase response with a crossover mismatch. ;)

I think the best solution is to use a 6dB/oct symmetrical active crossover that is at least two octaves away from the speakers' natural roll-off points.

Since that isn't practical (except according to Vandersteen, but I don't buy his claim that he has found drivers that allow that to work properly), the next best option IMO is symmetricaly 24dB/oct active crossovers that are at least an octave away from the speakers' natural roll-off points.

It's really the only way to maintain a semblance of phase coherency throughout the full frequency range.

I also don't understand the common discomfort with using active crossovers in the signal path. Passive crossovers are also in the signal path, and they do more nasty things to the signal than an active crossover could ever imagine doing!
 

RichardHOS

Second Unit
Joined
Mar 11, 2003
Messages
454
BTW, I have an interesting idea for a solution to the problem with phase/slope/frequency response that Vandersteen and Dunlavy claim to have conquered. My solution is a lot more complex, but on paper it looks very promising. Hopefully I'll get a chance to build a prototype next summer and make some measurements.
 

BruceD

Screenwriter
Joined
Apr 12, 1999
Messages
1,220
Dave,

However, since most receivers and pre-pros have symmetrical slopes
I think this is False.

Most HT processors have a THX inspired;
high-pass = 12dB/octave slope (depending on speaker for other 12dB)
low-pass = 24dB/octave slope

Now because they implement these in the digital domain, they may be able to avoid phase offsets, but the asymmetrical shapes of the two curves approaching each other causes real problems at and around the crossover frequency in most rooms when trying to blend mains and subs.

I think this is one reason why many are unhappy with their subs when comparing music vs. HT sources (albums sound like they have bass peaks at different frequencies), which is also contributed to by the well known room-mode problem.
 

BruceD

Screenwriter
Joined
Apr 12, 1999
Messages
1,220
Richard,

By the way, I do like 1st order 6dB/octave crossovers.

My Dynaudio Contour 2.8 towers use 6dB/octave crossovers between the drivers.

I selected 24dB/octave for sub(s) primarily for better control over room modes, plus the 360 degree phase offset of a 4th order Linkwitz-Riley was easier to deal with than the 180 degree phase offset of a 2nd order.
 

Jonathan M

Second Unit
Joined
Jul 23, 2002
Messages
267
I agree with most of the comments above.

Many receivers do use the THX inspired 12dB butterworth highpass and 24dB Linkwitz-Riley lowpass option, and in my opinion this is the best possible option for people who don't want big tanky speakers all around.

My speakers were designed as sealed with an F3 of 80Hz for precisely this reason. Not only is it easy on the receiver (Only running from 80Hz up with a 2nd order highpass), it's also easier on the speakers (Most 5-6" speakers don't like too much below 80Hz due to excursion required and the distortion that implies.)

Unfortunately, most speakers on the market (Probably 90% or more) are ported enclosures, so the above isn't so good. Usually though, when room behaviour is taken into account, it works fairly well though - the THX guys got this right IMO.

As for the suggestion to reroute all low bass to the mains then cross to the sub, the only problem there is that you have cascading xovers and more chance of out of phase problems causing nulls and peaks in the signal before it even reaches the speakers - this does depend somewhat on how the processor handles the digital delays applied to the speakers as well. I have yet to measure this on my NAD T752 - I have a method prepared, it's just finding the time to do it. I did notice, however, that when one bundled all the bass and sent it to the mains, then the response was no where near flat - I can't recall if I had any delays on the rears, however.

Feel free to check out my system with comments on my bass integration on my site by either clicking the www link above, or the link below.
 

BruceD

Screenwriter
Joined
Apr 12, 1999
Messages
1,220
As for the suggestion to reroute all low bass to the mains then cross to the sub, the only problem there is that you have cascading xovers and more chance of out of phase problems causing nulls and peaks in the signal before it even reaches the speakers - this does depend somewhat on how the processor handles the digital delays applied to the speakers as well.
Do you mean the DSP chip and bass management algorithm suppliers (approved by Dolby, DTS, and THX) like Cirus and Motorola (who supply most receiver and prepro manufacturers) screw up the bass management to the main L&R speakers when no sub is selected in speaker setup?

Can you explain how they screw it up?

Can you list which receiver's and prepros don't do bass management correctly?

I'm not sure I follow how you calculate a cascading crossover from an HT processor's L&R main preout (with LFE + re-directed bass already digitally integrated and time aligned) to an active crossover with the high-pass to main L&R and the low-pass to sub(s).
 

Jonathan M

Second Unit
Joined
Jul 23, 2002
Messages
267
Hey Bruce,

No, they don't screw it up exactly I don't think. Assuming that any digital delay (To compensate for closer rears or whatever) is applied before the digital xovers, then YES, that is a screw up. The delays should be applied to the highpassed signals only (As that's the only part of the signal that is coming from a speaker that is closer). Many receivers don't have delays for each individual speaker anyway, and even fewer have it for the sub (Which phase control can often remedy to a certain extent).

The problem with the cascading xovers is this:

The pre-pro is digitally crossing over the centre and surrounds (And anything else set to SMALL) and sending it, together with the LFE signal to the mains. If I remember correctly, the LFE gets it's 10dB boost at this stage. There is nothing wrong with this. Note though that the bass from the speakers set to small is crossed over (With a 24dB LR lowpass at say 80Hz or whatever it's set up as) whereas the mains are not lowpassed at all - they're fullrange.

OK, we now send this fullrange+otherchannels lowpassed bass signal offboard to an active xover such as a Marchand or DIY job.

This highpasses this signal and extracts the mains signals. Note that this signal will include some (albeit not much) midbass and bass from the speaker set to small. The xover also lowpasses the signal to send the bass to the sub where it belongs. Let's assume this is a 2nd order xover (For example). The mains' bass is crossed at 12dB/octave to the sub. The speakers set to small, however, is being crossed at 36dB/octave to the sub. Thus one MAY have phasing problems due to the different slopes involved.

Hope this explains things. I personally don't believe this'll be much of a problem when one considers room effects anyway. I do think that keeping things simple with all speakers identical, sealed at 80Hz and using the pre-pro's digital xover is the easiest and best solution. (It's even better than 5 fullrange speakers IMO as it has less bass cancellation issues when one considers a typical home theatre sized room.)

I plan to do some further testing in the future to see how and when the digital delay's are applied on my NAD T752. I've figured out a way to do it - I just have to get the time, and hope I can get sufficient resolution to notice the difference.
 

BruceD

Screenwriter
Joined
Apr 12, 1999
Messages
1,220
I do think that keeping things simple with all speakers identical, sealed at 80Hz and using the pre-pro's digital xover is the easiest and best solution.
I don't disagree with this statement at all.

The problem is how many people do you know that fit this scenario? Me, I don't know many that do, and I sure don't myself. So my suggestions are mostly for people who don't have the ideal speaker setup.

Yes, I think you optimized your speakers to match your HT processor. I didn't want to do DIY speakers, so I'm forced to deal with the speakers I have, Dynaudio Contour 2.8 towers, which I really like by the way.

If your NAD uses the cirrus DSP chips, you can read the cirrus application notes about bass management on their website.
 

RichardHOS

Second Unit
Joined
Mar 11, 2003
Messages
454
I think the THX 12db/oct Butterworth highpass was the "best compromise" they could have made without everyone in the world having standardized equipment (and, in fact, if you get THX certified speakers you do to a degree).

If you have larger mains that can play deeper, I think a 24dB/oct L-R XO at around 80Hz is a better option, so long as you use the same XO design for the sub low-pass (which is what Bruce is doing). If your mains have a 12dB slope with an F3 much above 40Hz though, you're going to get screwy results doing this. You should probably cross over your mains around a full octave above the F3 point if you're using a 24dB/oct slope, and right at the F3 point only if it has a natural 12dB/oct slope and you're using a 12dB/oct butterworth electrical filter.


In addition to the cumulative slope problem Jonathan described, you can also produce phase anomalies when you select different crossover points for surrounds and mains, and sum all low-pass information to a common sub. Although 24dB/oct XO's are phase coherent within a single channel's passbands, if you have two such XO's with different XO points they will not be phase coherent relative to one another.

When you add digital delay into the mix, all bets are off. And, when you consider that the phase response of a typical driver dominates the combined speaker+filter phase response for anything less than a 24dB/oct slope, it becomes excruciatingly difficult to make any progress whatsoever.

I think the system I'm putting together is the best I can do with today's solutions, and that is simply bypassing all bass management in the pre/pro and using a network of outboard active filters with matching XO points and slopes for all channels. Time alignment will be handled by the speaker/listener distance and I'm going to be careful in placement so that no digital delay corrections are needed. The downside is that the common XO point and slope may not be the "best" for each channel (unfortunately, I can't afford identical channels all around, though someday I hope to make that upgrade). That means I'll have some small phase response, and thus frequency response, anomalies within each channel, but this will be minimized by using steep slopes to minimize the bandwith of the off-phase system tracking.

I may venture into parametric equalization for each channel, though that is expensive (not quite as expensive as 7 matching speakers though... of the kind I want). IIRC, equalization not only corrects amplitude response, but phase response as well. It should be possible to adjust with appropriate filter settings such that all channels are phase coherent within their driver elements.
 

Jonathan M

Second Unit
Joined
Jul 23, 2002
Messages
267
Hey Bruce,

The point I was trying to make is that by putting an external xover only on the mains, you ARE using the internal xover in the receiver for the other channels. Here's a quote from your first post:

To do this means re-directing all bass from Small center and Small surrounds to the main Large L&R preouts and re-directing the LFE to the main Large L&R preouts as well. Then the xover is only done once; from the Large L&R preouts to the active external crossover and then to the L&R mains and sub(s).
If you have a 24dB/octave symmetrical Marchand or whatever on the mains, then the bass from the other channels will be filtered twice: Once by the internal pre-pros's 24dB/octave lowpass, and then again by the external 24dB/octave unit. Thus, you'll have a 48dB/octave lowpass on these speakers, but a 24dB/octave lowpass on the others. Seeing as phase is a derivative of frequency response, then quicker the frequency response changes, the quicker the phase changes, thus you'll get phase anomolies and therefore response anomolies. I doubt whether they'd are large enough to be problematic though when one considers the impact in-room.

I really think the key is NOT symmetrical filters. Rather it should be filters that are symmetrical once the acoustic response of the speakers are taken into account. (Just like what I assume is done in your Dynaudio towers. They have first order filters, though the combined acoustic filters definitely won't be first order). Ofcourse, if one is using 4th order slopes, this does tend to mitigate any problems that speaker rolloff induces as the response on either side of the xover is decaying so quickly anyway - especially if the speakers rolloff is kept an octave or so away from the crossover point.
 

BruceD

Screenwriter
Joined
Apr 12, 1999
Messages
1,220
Rather it should be filters that are symmetrical once the acoustic response of the speakers are taken into account.
Using the internal crossover of an HT processor, I would tend to agree with that perspective. The problem is most ported speakers just don't match with the HT processor's 2nd order high-pass (the reason you built sealed with -3dB @80Hz right?).

But, using a 4th order external filter to crossover a full octave above the -3dB for the speakers in question, should reduce the acoustical contribution to the crossover from the speakers to a minimum.
 

Jonathan M

Second Unit
Joined
Jul 23, 2002
Messages
267
I guess it comes down to how the digital filter is actually implemented - measurements would be the only way to know for sure - I'll look to see in the measurements I've done in the past if there is any indication as to the filtering method used.

I think we agree on the best plan of attack in your situation though - the steep 4th order filters do eliminate much of the problem of acoustic roll-off interaction if the F3 point is kept far enough away. I've done up some quick graphs. The first one is a ported speaker with an F3 of 40Hz and a Q of 0.7 (24dB/octave rolloff) coupled with a 4th order symmetrical LR crossover at 80Hz:



The second is a sealed speaker with an F3 of 40Hz and a Q of 0.7 (12dB/octave rolloff) coupled with a 4th order symmetrical LR at 80Hz:



As can be seen, the interaction is fairly minimal at less than 2dB in both cases - I'm sure the room is going to play a bigger role than this!
 

BruceD

Screenwriter
Joined
Apr 12, 1999
Messages
1,220
Jonathan,

Thanks for graphic perspective, that helps (what was used to produce the graphs?). Yes, I agree the room will play a big part, and a good reason for a measurement system to help tune it.

I expect many ported HT speakers are going to be something like F3 between 57Hz to 73Hz with an acoustic rolloff between 18-36dB/octave.
 

Users who are viewing this thread

Forum Sponsors

Latest Articles

Forum statistics

Threads
351,053
Messages
4,940,406
Members
142,973
Latest member
harleyquinn
Recent bookmarks
0
Top