Using driver -12db/oct rolloff in xover calculations

[deandob]

Hi,

I'm looking to make a cheap but high performing speaker system for my computer using TangBand W3-871S full range drivers and a single 8" peerless 830491 XLS woofer as a subwoofer.

There is a great article for using the Tangband drivers at
http://home.new.rr.com/zaph/audio/audio-speaker11.html that will form the basis of the satellites either side of the computer screen, for near field listening.

For the subwoofer, Rod Elliot's project 71 will be used as an active equaliser to improve the peerless woofer's low end in a sealed 20l box. It needs about 15db of gain to make -3db at 25hz which is quite a bit, however this woofer does have an "extra long stroke" (XLS) and I wont be listening to it at high SPL levels so should be OK.

The difficult part is working out the crossover between the subwoofer and the almost full range TangBand drivers. I will use active equalisation here as well, and I want to keep the crossover point as low as possible as I only have a single subwoofer so dont want it to sound like the subwoofer box is under my desk (ie. the higher the crossover, the more the ear can pick the directionality of the sound). However, unlike the Peerless woofer, the small TangBands will hit their -3b point at about 140hz and cannot shift much air below this frequency without distorting (ie. "extra short stroke"!), using a closed box of 2.2 litres.

I'd like to use a 24db/oct linkwitz-riley low pass filter for the subwoofer at 140hz, no problems, Rod Elliot has a Visual Basic program that will design the filter for you. For the TangBand, as the speaker is starting its natural -12db/oct rolloff at 140Hz, I was wondering if it is feasible to use a -12db/oct linkwitz-riley high pass filter at 140hz, which will add up to a -24db/oct high pass filter (driver rolloff + filter rolloff) to hopefully match the -24db/oct low pass for the subwoofer. My questions about this approach:

1) Is it valid to use the natural driver low frequency rolloff to add to a filter rolloff? ie. will I end up with a high pass match for the subwoofer -24db/oct low pass filter?

2) Is -12db/oct of the high pass filter sufficient to reduce the bass energy going to the Tangbands to avoid overloading them with bass that they can't reproduce? I think it should be (ie. power to the driver with the filter will be 1/16th at 70hz compared to without a filter).

3) What is the practical impact of the phase differences between the -24db/oct low pass filter for the subwoofer and the -12db/oct filter + speaker rolloff for the Tangband high pass? The -24db/oct filter is pretty phase neutral, while the -12db/oct filter will invert the phase, and around the crossover frequency as its also the -3db point for the tangband natural rolloff, the driver phase at this freq will be -90, reducing the absolute phase to +90 (180-90). At 140hz the woofer will have a phase change of +45. Is this OK?

4) I'm assuming that if I use a closed box for the subwoofer then I should also stick to a closed box for the full range drivers, so that the sound integrates better?

4) Should I use a larger box for the TangBands and cross them over at a lower frequency (eg. 100hz), ie. is 140Hz just too high for a crossover to a single subwoofer, for a high quality integrated near field listening experience?

I appreciate that I will have to build this setup, listen, measure and tweak to get it right, but I was hoping that someone more knowledgeable than I can tell me if I'm on the right track or just plain nuts!

Regards,
Dean

[deandob]

Thinking about it more, the answer to the last question is that it should be OK to have the subwoofer cross over at a frequency higher than normal (140Hz) as the -24db/oct linkwitz-riley filter has a pretty sharp drop off, leaving very little signal going to the subwoofer above 200Hz. This should make for reasonable integration with the TangBand full range speakers.

Regards,
Dean

[Bill Fitzpatrick]

The technique you describe works very well and is how I do it on my system where the cross is at 90Hz to my 5 1/2" mid. Cone excursion on the mid is quite low even at high volume levels and would be lower still if you cross at 140Hz.

The only way this works correctly is if you have the mid in a closed box with the Q at .7.

I believe you will be quite pleased with the results.

I have tried to push this technique on this forum but nobody seems to want to listen.

[deandob]

Bill, it does look like a good method, and according to the theory should work well. The only reservation is what it happens to the phase plot where the interaction of the filter and the turning phase of the speaker at the -3db point. If you include the ability to vary the subwoofer phase or at least invert the subwoofer output you should be OK.

As you have said, it will only work with closed boxes to get the -12db/oct rolloff, a vented box gets a much steeper rolloff past the box resonant frequency, looks like about -36db/oct, and the small tangband driver will certainly be out of puff below resonance.

A fellow forum member has a Tripath 2 x 100W amp kit for sale at a reasonable price that I can drive the subwoofer with, and I have a pair of gainclone chips which will be just right for driving the tangband.

I'm hoping that this combination will work well as I spend a lot of time listening to music on my PC and want a better experience than what even expensive PC speakers can give.

Regards,
Dean

[Bill Fitzpatrick]

Originally posted by deandob
If you include the ability to vary the subwoofer phase or at least invert the subwoofer output you should be OK.

You don't want to do either. As is provides a 4th order L/R network, exactly what you need.

Originally posted by deandob

As you have said, it will only work with closed boxes to get the -12db/oct rolloff, a vented box gets a much steeper rolloff past the box resonant frequency, looks like about -36db/oct,

Actually the vented box rolloff is 24 db/o

Originally posted by deandob

I'm hoping that this combination will work well as I spend a lot of time listening to music on my PC and want a better experience than what even expensive PC speakers can give.

Like I said, I'm sure you will be quite happy.

[deandob]

Hmm. WinISD was showing a steeper rolloff, more like 36db/oct.

Thanks for your feedback Bill, I'll start building this system and report back what it sounds like.

Regards,
Dean

[DSP_Geek]

Originally posted by Bill Fitzpatrick
The technique you describe works very well and is how I do it on my system where the cross is at 90Hz to my 5 1/2" mid. Cone excursion on the mid is quite low even at high volume levels and would be lower still if you cross at 140Hz.

The only way this works correctly is if you have the mid in a closed box with the Q at .7.

I believe you will be quite pleased with the results.

I have tried to push this technique on this forum but nobody seems to want to listen.

I'm about to use that very technique on my speakers (12" woof, 15" PR, pile of 6" mids, 1" tweet). You actually can fudge a bit on the mid Q if you can change the filter Q. If you have a mid box with Q = 0.807, for example, you can set the crossover filter Q to 0.607 and you get substantially the same polynomial, and ergo the same response.

The algebraic reason for this is left as an exercise to the student.


Cheerio,
Francois.

[phase_accurate]

What Dean proposes is nothing else than good solid engineering practice: taking the response of a driver into account when calculating a crossover !
This is the first step to a system that will need the least amount of tweaking to get right !

Your proposed system should basically work well. There is just one thing that you should keep in mind: Depending on how close you listen, how close to near boundaries the speakers are placed (like monitors or back-wall etc) and the speaker's size, you might have some baffle-step (or even the contrary like a bump) at the lower midrange.
But since this is an active system it can be easily taken care of afterwards (if needed at all) !

Regards

Charles

[Bill Fitzpatrick]

Originally posted by DSP_Geek
You actually can fudge a bit on the mid Q if you can change the filter Q. If you have a mid box with Q = 0.807, for example, you can set the crossover filter Q to 0.607 and you get substantially the same polynomial, and ergo the same response.

True. I happen to have both Qs set to .707 (or close to it) but there is room for variations as long at the final Q is .5.

I'm glad that others see this. Hopefully the technique will now begin to spread. Note that this works only as long as the woofer and midrange driver are at substantially the same distance from the listener. I'd keep the difference at no more than 4" when the crossover is at 100Hz. That's about 12 degrees phase shift. Aim for 0 degrees if your room will let you!

[brummkreisel]

Hi Dean!


1. Yes. However, as Francois said before, when you combine a "natural" roll-off (i.e. due to the sealed box), you will have to have an eye on the filter parameters of the additional HP to match the properties of the LR4 LP filter from the woofer. (Although, given the very long wavelengths involved, phase accuracy should be not even a minor concern.)

2. Whether the 12dB/oct additional HP is "enough" is mostly determined by the max SPL you want from your TangBand. It's only a 3", but he fact you do use it as a near-field monitor certainly helps. If you somehow had a very large baffle, you could profit from the baffle-step gain (rather: absence of loss of 6dB). Do, as Charles said, not forget that for a NFM, you will have to compensate for the baffle step, so take it into account when designing your HP filter. After all, you want to take full advantage of the near-field case!

3. The bandpass properties of the woofer will somewhat modify the pure LR4LP behaviour, but you should not take that too seriously (just be aware of it when designing the all-pass). If the TangBand's total (box+HP) filter function it is something like a LR4, it matches the LR4 of the woofer sufficiently well, and you're fine. After all, no need to overstate phase errors at these low frequencies. You'll sleep much better.

4. The choice of "all sealed" (or "all ported") will not influence the "integrity" of the sound or whatever to an important degree. However, a vented fullrange driver (as opposed to: drone-cone or passive radiator) will likely display something like port resonances and transmission of mid- and high frequency signals. Also, the rather high tuning frequency might be audible in the sense that every hi-bass note produced by the TangBand will be followed by a long decay on this frequency. Thus some say, and this is a old-fashioned reason not to use high-tuned (>50-60Hz) ported constructions. In short: you're best off with closed box for the TangBand. For the woofer, it might be worth thinking of a ported version (you're electronically assisted anyways) to allow for somewhat higher SPLs at low frquency (25-40Hz or so). But then, again, you're sitting very close to the speakers, so that it might be a moot point.

5. (or 4. again) The combination of a woofer plus a FR driver is indeed a very good approach, most certainly the best one if small FR drivers (<6") are to be used. Personally, though, I'd go for two woofers (one per TangBand) crossed over around 250Hz or above, and using aa 2nd-order filter (phase subtractive or so) instead.


Cheers to hot Brisbane,
bk