dc655321
Major Contributor
- Joined
- Mar 4, 2018
- Messages
- 1,597
- Likes
- 2,235
12th order would be 72 dB/octave, no?
-
WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Active crossover types
- Thread starter dennnic
- Start date
12th order would be 72 dB/octave, no?
Yep. Meant to say 16th order. Thx for the catch.
Post is corrected.
Yep. DSP active. FIR.
Even if such a high order could be done in the analog domain, I'm pretty sure the phase rotation and impulse response destruction make it a non-starter.
It's much easier to get the critical xover region well behaved, the region of summation between drivers (like -20dB used in the examples i gave).
Easier because the driver has less of a frequency range past the xover freq, that needs to be managed.
Most of the time, a driver is already rolling off response near xover. So having to deal with less rolloff by narrowing the range is a plus.
It also helps reduce lobing, simply because the width of the critical region is narrower. The two drivers summing have less of a freq range for polar triangulation issues.
Any FIR based platform...PC, or hardware like some of the miniDSPs, and lot's of other speaker management processors.
I'm using QSC's Qsys platform which has both FIR and IIR. It's a bit unusual in that it allows 16th order IIR, but like said above, such high "analogish" orders are a non-starter.
I've played alot with this. I'm not willing to say higher order linear phase sounds better than lower order IIR, all things equal....and both setups perfectly aligned.
I am willing to say the odds of getting high order linear phase perfectly aligned are way, way, higher...
and for me have resulted in better sound.
Understood. Thanks a lot. I have never dabbled in FIR, because I don't know anything about it, like a lot of novices who enter the DSP domain from the analog side. I'll need to read up I think.
- Thread Starter
- #85
Interesting approach indeed. Slopes of 96 dB/Oct would allow for blending almost any combination of drivers there is.
Since you experimented on the subject... Higher crossover topologies should suffer from worse transient response. Is it also the case with a linear phase crossover? Are there any drawbacks you've noticed?
Just want to point out one common misconception about digital filters and phase.
The misconception is that some people seem to think "if it is digital, there is no phase shift", which is most of the time not true.
The most common digital filter; the Infinite Impulse Response IIR type filter will have the same phase behavior as the corresponding analog filter, be it Bessel, Butterworth or whatever slope you are trying to acheive.
The Finite Impulse Response FIR type digital filter however can be Linear phase, but by design it will add a frequency independent fixed delay which easily can be several milliseconds. This is not necessarily a problem but can be a challenge in multi speaker systems and with sound/picture lip-sync.
Also, about loudspeakers and Transient response. By far the most important design feature if you want to worry about this is how your drivers interact, what type of filter used to achieve this is secondary. If a filter with "bad" impulse response makes them interact better, the "bad" filter is "better" (in regard to transient response). Compare measured step responses from some loudspeakers and you will see that all of them look terrible compared to any filter commonly used in loudspeaker crossovers, digital or analog. This should also give you a pointer about how much you need to worry about this.
I have attached such an example, the otherwise excellent Revel F228BE
The misconception is that some people seem to think "if it is digital, there is no phase shift", which is most of the time not true.
The most common digital filter; the Infinite Impulse Response IIR type filter will have the same phase behavior as the corresponding analog filter, be it Bessel, Butterworth or whatever slope you are trying to acheive.
The Finite Impulse Response FIR type digital filter however can be Linear phase, but by design it will add a frequency independent fixed delay which easily can be several milliseconds. This is not necessarily a problem but can be a challenge in multi speaker systems and with sound/picture lip-sync.
Also, about loudspeakers and Transient response. By far the most important design feature if you want to worry about this is how your drivers interact, what type of filter used to achieve this is secondary. If a filter with "bad" impulse response makes them interact better, the "bad" filter is "better" (in regard to transient response). Compare measured step responses from some loudspeakers and you will see that all of them look terrible compared to any filter commonly used in loudspeaker crossovers, digital or analog. This should also give you a pointer about how much you need to worry about this.
I have attached such an example, the otherwise excellent Revel F228BE
Attachments
Last edited:
KSTR
Major Contributor
That's a somewhat illegal comparision. It has nothing to do with speakers being speakers. The reason for the way most (not all) speaker step responses look like is the allpass (excess phase) contribution to the "default" shape, the latter being governed by magnitude frequency response, notably the roll-offs at either end.
The allpass portion comes from adding two (or more) minimum-phase individual frequency sections, plus sometimes extra excess phase from not having the driver acoustic centers aligned in the same plane. A sum of minphase functions usually does not give a minphase result (with exceptions, like first order acoustic crossover function, among others).
IIR (and analog) filters can be made linear phase, too, in a limited bandwidth (by adding in allpass cascades with peaking group delay that sums up to a mirror image of the main group delay) and these again have, and must have, delay ... but no extra processing delay. Yet this delay of allpass corrected filters is larger than required because of the base delay below the frequencies of interest.
See here, Figure 6, on how this looks like in detail.
The additional processing delay of any FIR filter can be greatly minimised by "partitioning" which splices together short sub-blocks of data, an overlap-and-add type of process. This add some extra CPU load.
The core delay of an FIR is the time span from start to where the main pulse sits, in the convolution kernel, and both delays add together.
Last edited:
Good paper thx, ....and good comments
Yes, it makes tying drivers together much easier. I often have over an octave's width of room to choose xover frequency.
The linear phase xovers have solved (for me) the transient response degradation issue with traditional higher order analog or IIR xovers.
I routinely get measurements with this kind of look for whatever reference axis is chosen to tune to, ie directly on-axis or say 10 deg off-axis.
Of course the measurements don't display the same perfection awayf the tuning axis, but polars are still quite a bit better than most of the on-axis results we see posted. This is 3-way main from 100Hz up without a sub, using 96 dB/oct linear phase xovers.
The transient response is remarkable.
Drawbacks:
First is the processing delay of FIR, when used as linear phase.
The above impulse used about 43ms. If phase linearization extends down into sub territory, i typically use about 170ms of "FIR time".
A 2nd is potential pre-ringing. My experience has been, if fully complementary linear phase xovers are used, they generate no pre-ringing.
I consider fully complementary a must.
It's also best to use a minimum phase high-pass on the sub, if a high pass is needed, to avoid potential pre-ring. So an extra complication, but any processer capable of that much FIR will also have the IIR capability for this task.
Which brings up a third drawback, the cost and complexity. The complexity and cost from an amplification/processing channel count point of view, i think is the same for a normal multi-way active. FIR is a little more costly, in that it takes either a very robust speaker processor, or integrating a PC into the line level signal flow. And perhaps more complicated in that extra knowledge needs to be acquired.
But my experience is, if folks more only knew how much vastly easier xovers are, and speaker tuning in general is, when using steep linear phase xovers,
well....i think they would change their minds about what's complex and what's not.
KSTR
Major Contributor
What do you mean by "complemetary linear phase XO"? Symmetric shapes/slopes (of the final acoustical target)?
Off-axis I've always seen residual pre-ringing as the summation of the individual drivers degrades by time-of-flight differences and non-identical directivities.
As for slopes, I prefer steep (or even infinite) final slopes but a very gradual transition region, like lower order Neville-Theile.
Yes, symmetry of the final acoustic targets. Complementary in the classic style, like describing Linkwitz-Riley mag and phase.
I typically flatten a driver's mag with minimum phase EQ's both in-band, and out-of-band far enough for the steep linear phase xovers to dominate, and hit the acoustic target.
Yes again, although my measurements make it look more of a theoretical problem than one i've observed.
Nonetheless, the potential issue is another reason i like steep...to minimize the freq range that pre-ringing might occur due to ToF differences.
For me the objective has been to minimize the range of the critical summation region between driver sections.
To achieve the protection against pre-ring just discussed, which i see as a function of basic lobing.
Lobing that gets reduced by having a narrower summation range.
So i like steeper in the transition region too.
I've settled in on 96dB/oct LR's as my go-to xover, as it seems to be a good compromise between brickwall, and ineffectively shallow.
ernestcarl
Major Contributor
Do you also use such steep filters when adding in a sub(s)? What do you think about broadly overlapping xo in the bass/sub region? I’m just curious because I’m overlapping mine with the mains.
Yes, I've been using steep between sub and main as well.
I don't like being able to aurally localize a sub when it is the only box playing. Steep at 100-120Hz stops almost all frequency content above xover that can be heard and localized.
Whereas, with shallower, like say 24 dB/oct, i can localize the sub even with a xover freq as low as 80Hz.
2nd reason is trying to keep 1/4 wave length spacing between the sub and main for the entire summation region.
If i may use the same type example i gave earlier, but this time for 100Hz.....
Since 1/4 spacing will only be an issue above xover, the subs low pass will govern.
Red is a 24dB/oct LR @ 100Hz. Blue is 96 dB/oct.
Using a summation region defined by -20dB, Red is summing out to about 175Hz, while Blue sums out to about 115Hz.
1/4 wave for Red is around 19 inches, Blue is around 30 inches.
i often have a hard time keeping center-to-center distances between sub and main drivers within 19", whereas 30" has been pretty easy.
Last reason is most research on phase or group delay audibility that seems modern and credible, says audibility appears to be greatest for low/very low frequencies.
I dunno exactly what to think here, but can say outdoors the bass slam, punch, and rhythm with higher frequency content can at times be stunning.
Not on all tracks though. My guess is tracks that were mixed/mastered to sound best in studios with flatter phase / lower group delay, are the tracks sounding so unusually good.
Indoors is tougher for me to make a call...we all know the low freq soup/mud we hear in a room huh? Hence the whole multiple sub thingy that works for a lot of folks, i guess.
And of course, complementary linear phase doesn't necessarily mean steep.
The sub to main xover order is probably my biggest area of xover order exploration still....despite reasons listed above.
But I do think the xover should be complementary linear phase, whatever order is chosen.
ernestcarl
Major Contributor
I haven't come up with a conclusion on the matter regarding my own small listening room, but my thought process of having a wider overlap was partly influenced by this presentation:
Whichever way -- symmetrically narrow and steep, or widely overlapping bass xo -- I was still able to get good enough results. Wrong localization was not so much a big issue for me given how I've positioned my sub and monitors in the room.
D
Deleted member 46664
Guest
I've designed and built a few active crossovers in my days ... NO OpAmps do not degrade the signal in any audible way. Take a look at the manufacturer's spec sheets ... you will find that with unity gain filters you are looking at frequency response practically from DC to a couple of hundred kilohertz and probably at distortion levels in hundredths of a percent. You won't get it that clean with transistors and absolutely never with tubes.
Still based on your writeup... I would suggest the MiniDSP modules, as the others did. If your needs change, you can simply reprogram and carry on... not so easy with hardware only solutions.
- Thread Starter
- #96
I take it your phrase 'no opamps do not degrade' means they don't make an audible difference.
I went with a mini dsp product, but I really wanted something that doesn't produce a completely new signal. Still in the works, a basic 2way xover.
So far, 7 op amps, but I might ditch last two (output signal 'gain')
D
Deleted member 46664
Guest
No sir ... I meant they don't tend to introduce distortion to a signal.
In a crossover, you are obviously trying to make an audible difference ... change the frequency response.
You should be able to build a pretty good 2way crossover with 2 - 8 pin chips.... Use one chip for each channel. Build one of the opamps on each chip as a low pass filter and the other as a high pass filter... probably half a dozen resistors and a couple of caps for each. The result should be an exceptionally clean sounding crossover.
(My rule is to always keep it simple... minimum parts counts and shortest signal paths always give the best results)
formula 977
Active Member
- Joined
- Apr 8, 2022
- Messages
- 105
- Likes
- 88
You should use these as unity gain output buffer amps.
I would keep the last two op amps.... being able to adjust the output level of an active crossover can come in VERY handy.
Regarding more is less with op amps here is another perspective. This crossover uses op amp input buffers and output gain buffers both of which make it more flexible and stable with any types of input or output devices. I would use it as designed as good op-amps are not going to cause any audible issues and you get the added functionality and op amps are not expensive.
JRS
Major Contributor
I just wanted to second @gnarly 's approach of using brick wall filters as XO's to extend the region of overlap or simply to avoid stressing a tweeter by crossing too low, or bumping into a nasty resonance with a mid or tweeter metallic dome. Take my Aurum Catus G1 ribbons crossed over at 2k to mate with a 6.5" midbass. IIRC the manufacturer suggests running the tweeter at no more than 1750 Hz, Using a traditional LR4, I might be in trouble, wheres with a -96dB/octave XO I'd be well out of the woods. Conventional wisdom would argue that driver "blending" is an issue, but if so, I haven't noticed it. I have had the opportunity to A/B vs a transient perfect XO using the same drivers: while going back and forth took more time than ideal (passive was kept external for the test), the take home was no preference. This was a Bagby design using drivers with a huge overlap--in other words an ideal situation for the time aligned passive to strut its stuff. Given how affordable multichannel amps have become (meanwhile inductor $$ have soared making even a 2 way passive a tidy investment), there really is little to recommend the approach save simplicity IMO).
I do depart from @garly's approach to SW xo's--I usually use fourth order to avoid the long latencies--also seems with the long wavelengths involved, it is less critical. It is also worth having switchable configurations for low latency implementations if it is used for video playback, but if I understand correctly some of the media services can buffer enough video to allow the audio to catch up.
I do depart from @garly's approach to SW xo's--I usually use fourth order to avoid the long latencies--also seems with the long wavelengths involved, it is less critical. It is also worth having switchable configurations for low latency implementations if it is used for video playback, but if I understand correctly some of the media services can buffer enough video to allow the audio to catch up.
