mmt or mtm give less lobing?

[obiwan]

I have a DEQX and am thinking of making just a "practice" pair of bookshelf speakers to go with my Velodyne DD15 sub.

As DEQX offers up to 300dB per octave crossovers, if I want to use 2 mid bass drivers, am I theoretically better off doing an MTM or an MMT. I'm thinking that in the MMT with the midbass drivers closer together you'd have less lobing. Would the benefits of this outweigh the benefits of the theoretical point source you get with an MTM arrangement?

I know you'll say why not just have one driver for the midbass duties, but that's what I have now with my Dynaudio Special 25's that I've biamped via DEQX. I just want something more articulate in the midrange and I feel the 8 inch Special 25 midbass driver just isn't refined enough in the upper midrange. But if I go down to a 5.5 or 6.5 inch say Scan Speak Revelator driver, I'm worried about the loss in sensitivity and output that a single smaller driver would deliver.

Any thoughts?

Thanks

Obiwan

 

[augerpro]

2 way TMM might reduce combing between woofers but the tweeter and outside woofer combing would get worse. This would cause more unpredictable lobing I would think than the MTM.

 

[Dag]

Why not try a MTB 2.5-way??
With M for both mid and bass and B only for bass. M&B can be the same woofer. => God bass and no strange loobing in the midrange.
Peerless HDS 5.5" or 6.5" could be a good woofer (and MUCH cheaper than the SS)

 

[augerpro]

Why not try a MTB 2.5-way??
With M for both mid and bass and B only for bass. M&B can be the same woofer. => God bass and no strange loobing in the midrange.
Peerless HDS 5.5" or 6.5" could be a good woofer (and MUCH cheaper than the SS)

That reminds me of something I've wanted to try. It's an attempt to make a MTM without the combing problems. Use the 5.25" HDS Nomex woofers and a Vifa D26NC55 with the flat edge of the flanges touching. Push the XO as low as possible and you *may* not have the combing between the woofers that you normally would.

 

[obiwan]

Thanks for the replies guys,

The problem with going to MTB means that I'll use up my 3 frequency bands on the DEQX leaving me without the ability to add the sub with the DEQX. If I want DEQX to blend the sub, then the main speakers have to be 2 way.

Augerpro, I think with 300dB per octave crossover there would be virtually no lobing between the tweeter and the woofers as they are operating in totally different frequency ranges.

Obiwan

 

[sreten]

Hi,

You could go active / passive 0.5 way TMM, using 8" unit for BSC.
That is use that driver with a series inductor, other full range.

/sreten.

 

[Dag]

If you have the amps you can use a very simple 1st order RC filter (at ~500Hz) for the amp to the 2nd woofer.
You can start with an MTM 2nd order and if you don't like the mid you can add another amp + filter later. This way you will avoid the crappy inductor in series with the woofer.

Streten:
If the two woofers are different will you not have problems with the phase between the two woofers?

 

[sreten]

Streten:
If the two woofers are different will you not have problems with the phase between the two woofers?

Hi,

I'm not sure I understand the question, but to use different sized
woofers they have to match to a degree, moreso if they share the
same box volume, moreso if they also share a port.

Sensitivities need to be very similar, also cone mass per unit area,
or another way Vas x cone area similar etc ....

You are right about the input filter if a spare amplifier is available.

In a 0.5 way the phase digressions should be well above the 0.5 way point.

/sreten.

 

[DSP_Geek]

Augerpro, I think with 300dB per octave crossover there would be virtually no lobing between the tweeter and the woofers as they are operating in totally different frequency ranges.

You don't necessarily want to cross over that sharply.

DEQX looks as if it uses FIR filters, so the sum on-axis should be unity. Off-axis could be a problem, especially if woofer and tweeter dispersions are different, because at the (smallish) overlap there will be a disparate contribution from the two drivers. Since FIRs of that steepness have fairly long impulse responses, the room contribution could well be ringing at a narrow band around the crossover frequency, so I'd advise using a shallower slope even with a linear-phase crossover.

 

[obiwan]

Thanks for responses guys.

DSP_Geek, I've heard people comment on ringing with steep crossovers, but I don't know what it is. I remember trying to search about it, but couldn't find much.

Are you able to explain it in layman's terms for me?

I think the shallowist DEQX will let me select is 48dB per octave.

Obiwan

 

[augerpro]

I think there is a topic on this at Linkwitz's site.

 

[DSP_Geek]

DSP_Geek, I've heard people comment on ringing with steep crossovers, but I don't know what it is. I remember trying to search about it, but couldn't find much.

Are you able to explain it in layman's terms for me?

I think the shallowist DEQX will let me select is 48dB per octave.

48 dB/octave on a FIR filter is perfectly reasonable, actually a bit shallower than I'd have expected. FIRs can have sharper cutoffs than IIRs while still maintaining reasonable impulse responses (this is a gross oversimplification, by the way).

As to why steep crossovers ring, the answer involves some hefty dollops of filter theory. To simplify grotesquely once again, you trade off sharpness in the frequency domain for more length in the time domain: sharper filters will have longer impulse responses. Filters typically have some damped oscillation centered at the corner frequency, so you can see that a longer impulse response will have more oscillation, or in other words ringing.

This can be surprisingly audible. For DAYJOB I designed some lowpass filters for program material, with cutoffs around 11 kHz. The original spec called for a 1 kHz transition band, but that was completely obnoxious: it felt like someone was scraping sandpaper against my eardrums. After discussions with the lead technologist we relaxed the transition bands to 2 kHz. That sounded a lot more relaxed, with much less sandpaper effect. The 2 kHz filters had about half the ringing of the 1 kHz filters, which convinced me that relatively low-slope filters are the way to go even in digital form.

You might want to listen to the bands individually and figure whether you like the sound without the other bands helping out - it'll sound strange because you're only getting part of the information, but you should be able to find your threshold of tolerance for filter steepness. I suspect the 300 dB/octave filters might whistle like a bird near the crossover point.

 

[obiwan]

Very interesting reading. Thanks for that DSP_Geek. I'll have a go at shallower filters.

 

[obiwan]

I've copied the following from the DEQX user groups forum about ringing. Somewhere else though in the forum though they recommended 60dB per octave slopes between a midwoofer and a dome tweeter.



7. Can you comment on high slope filter ringing?

If you look at the time response of higher order crossovers, it is true
that ringing exists, but this is due to a physical reality known as
Gibbs phenomena, this is where you cannot create perfect transitions (in
a square wave for example), for a bandlimited system. In fact, as the
system is less and less bandlimited (as in the case of the higher order
crossovers), more 'overshoot' exists. All this aside, the ringing is at
a specific frequency (the crossover frequency), and it is cancelled very
cleanly by the other loudspeaker driver (because they sum to a flat
response). In fact, because the crossovers are so sharp, the crossover
region is actually unperceptable, even if they are totally out of phase.
It is well known that steep dips in the response are not perceptable.

There are a number of things that make a good crossover, but we can say
with total certainty that the effects gained by removing the interaction
between the two drivers and avoiding the lobing effects caused by that
interaction, is the single most noticable improvement in this design.
There are additional advantages such as avoiding the use of passive
components, which are non-ideal and degrade in time, and the fact that
it is not possible to design an analogue crossover that is linear phase.
All analogue crossovers have phase distortion! Other factors regarding
transients are totally negligble with regard to these gains, which is
why we highly recommend the use of higher-order crossovers exclusively.
Please note that you can vary the slope of the linear-phase crossovers
with our system, and even simulate analogue crossovers (in the PDC
configuration screen), so you can compare the results for yourself.


That's about it for now. Please let's have your views on the above
points. I will now proceed to listening sessions of various settings and
crossover slopes.

 

[DSP_Geek]

7. Can you comment on high slope filter ringing?

If you look at the time response of higher order crossovers, it is true that ringing exists, but this is due to a physical reality known as Gibbs phenomena, this is where you cannot create perfect transitions (in a square wave for example), for a bandlimited system. In fact, as the system is less and less bandlimited (as in the case of the higher order crossovers), more 'overshoot' exists. All this aside, the ringing is at a specific frequency (the crossover frequency), and it is cancelled very cleanly by the other loudspeaker driver (because they sum to a flat response). In fact, because the crossovers are so sharp, the crossover region is actually unperceptable, even if they are totally out of phase. It is well known that steep dips in the response are not perceptable.

The ringing will very well be cancelled out when the two drivers are exactly matched in amplitude and delay, but things change somewhat when you go off-axis. Listening above or below the design axis will cause different path lengths from the drivers, and at some angles [1] the ringing will get worse than from one filter alone. Going off-axis, whether horizontally or vertically, also gets into the spatial response of the individual drivers: tweeters are OK, but most mid-woofers start rolling off off-axis near the crossover point - that's partly why they're crossed over in the first place! - and once again the sum isn't a perfect cancellation.

Much of the energy hitting the ear at the average listening position is from room reflections, so off-axis behaviour, including lobing and ringing, does have an effect on the perceived spectrum. Just because something is OK on-axis doesn't give a design a free pass to act strangely elsewhere.

[1] where the path length difference corresponds to half the ringing wavelength, because the sum goes from cancellation from reinforcement.

One more thing: the drivers need to be perfectly aligned to cancel out ringing. Imagine a 2.2 kHz crossover: in air the wavelength is about 6 inches, so a half wavelength is 3 inches and a quarter wavelength is 1.5 inches. Why do I wonder about a quarter wavelength? That's where the cancellation begins to fail: adding two waves in quadrature gets you a wave at 45 degrees, of the same amplitude as either of the originals. Some folks might want to use a FIR crossover with steep slopes to avoid worrying about time aligning the tweeter to the woofer, but the ringing would still be present if the woofer acoustic center is 1.5 inches behind the tweeter. Steep digital crossovers are not a panacea which will cure flaws elsewhere in the design, and in some cases they can aggravate the problem by introducing artifacts of their own.

 

[obiwan]

Thanks for that DSP_Geek. I think I need to get a new university degree. Optometry's getting boring anyway!

 

[DSP_Geek]

Thanks for that DSP_Geek. I think I need to get a new university degree. Optometry's getting boring anyway!

An engineering degree will give you the tools to analyse this stuff, but not the exact solutions for this particular case. Truth be told, there aren't a lot of jobs in audio: my first few years out of school were with graphics and cell phone companies. I've been fortunate to work alongside some fairly eminent authorities in applied signal processing where I picked up a number of useful techniques. Finding good mentors is half the battle, and from what I recall Oz has some pretty good audio people.

Just as an aside, I screwed up the phasor addition for the cancellation calculation [1]: 90 degrees difference in path length yields ringing at +3 dB relative to the individual filters, 60 degrees has ringing at the same level, and even at 30 degrees difference the ringing is only 6 dB down. 30 degrees, by the way, is half an inch at 2.2 kHz crossover point. Sounds like you'd need to have your head in a clamp to have the highpass and lowpass filters cancel ringing.

[1] before my first cuppa, it was, so that's an excuse. Honest.

 

[obiwan]

"[1] before my first cuppa, it was, so that's an excuse. Honest."

I have moments like that with increasing regularity. With my mind, if it wasn't for emails telling me that someone had replied to this thread, I'd either never find this thread again, or worse still, forget I even started such a thread. I'm hoping this has nothing to do with me turning 40 next week.

 

[DSP_Geek]

"[1] before my first cuppa, it was, so that's an excuse. Honest."

I have moments like that with increasing regularity. With my mind, if it wasn't for emails telling me that someone had replied to this thread, I'd either never find this thread again, or worse still, forget I even started such a thread. I'm hoping this has nothing to do with me turning 40 next week.

Jeez, thanks for making me feel good. I just hit 50. On the other hand, exercising the grey matter does keep the mind young, and figuring out what the hell went wrong with the latest DIY project is awfully good for that (I tend to examine my work a tad critically).

Anyway, thanks for a productive question. I hadn't looked at FIRs in crossovers since I ran a simulation a few years ago which convinced me that lobing problems with relatively shallow filters was mostly inconsequential. I'd more or less ignored the time domain problems of steep filters.


Cheers,
Francois.