Active crossover types

[dennnic]

I was hoping for one octave on both ends. Two 12 inch woofers per speaker allow for reproduction down to 40-45 Hz - so, there's one octave missing.
Tweeter range is as it is, I'll toe them in to extend the sweet spot coverage. Open baffle woofers should beam at a little higher frequency, that should help as well.

 

[dguillor]

I do like their stuff, but shipping from USA with taxes and everything else usually costs more than a product itself.

On the other hand, I do like the idea of DIY and that's the main problem!


Thanks for the explanation. It seems that the most elegant solution would be to buy a 15-0-15 transformer and a power cable and put it besides Rod's power supply and crossover in the same box.


Good to hear. I was about to go with the 4th order, in order to extend the full range as low as possible. Although, it struggles under 250 Hz.
A lot of chip op amps were already mentioned. What would be your pick?

This is an incredible op amp that costs around $2 for a dual: LM4562NA-NOPB

 

[dennnic]

Update: Just ordered Rod's PCBs, both for the crossover and it's power supply.

What is your take on the crossover frequency?? My goal is to push it as low as possible, ideally to avoid the vocal region. Full range driver measures flat down to around 250 Hz. It's a 24 dB/octave filter.

I found capacitor/resistor combination that would give me around 430 Hz crossing frequency. Would it be comfortable for the FR and dare I go lower?

 

[tcpip]

On the matter of analog vs digital active xo, one point is often missed by newbies.

When newbies build analog active crossovers, they almost always build "stock" crossovers, i.e. textbook "knee" and slope, or textbook values of Fc and Q. These are also the builders who get attracted to Rod Elliott's PCBs or Marchand Electronics boxes. These builders never measure their speakers, or even bother to take the manufacturer-published SPL graphs, digitise them with graph tracers, and then put the FRD data into something like VituixCAD with the active crossover schematic.

Needless to say, these builds show problems later because of phase mismatch between drivers at crossover points, since drivers are rarely perfectly linear at the crossover frequencies. This means that when you measure the acoustic output of your crossover + drivers, using a mic, you'll see dips and peaks, and the sound will be coloured. This requires correction with parametric eq.

This is where analog active xo is worse than digital. A single channel of parametric eq can take 4 opamps, and even an analog notch filter can take between 1 and 3 opamps, depending on the Q you need. The tenacious builder will add all this extra circuit to the inputs or outputs of his analog active xo. Most newbies will simply not bother, and will marvel at the clarity of their new speakers and talk about veils being lifted. In reality, the veils being lifted are colourations caused by the peaks in the SPL graph.

In this area, I find digital active xo to be superior, because you can simply throw in a few channels of PEQ as and when you want to "tune" the sound, either with your own ears or with the help of a techie friend who may actually know how to measure speakers.

Rod Elliott (who I respect hugely, I feel he hasn't gotten the respect he deserves) too leaves the newbie speaker builder high and dry here. In his entire website, I've never seen any mention of actual speaker response SPL graphs, and how the xo modifies those graphs. Since I have some experience in speaker design, I am very happy to use his lovely 3-way active xo PCB, but I'll throw away all component-value-calculator spreadsheets, plug his schematic into VituixCAD and tune the R and C values after adding the speaker driver FRD data and optimising the simulated output.

Newbie speaker builders need to understand that stock crossovers, whether active or passive, almost always make mediocre speakers. And PEQ is the great blessing of digital active xo for such builders.

 

[dennnic]

Thank you for answering, it gives more material to think about.

I am not sure if Rod's p09 crossover allows for any change of variables, except for a frequency selection. I will measure my drivers again, but I think I'm stuck with what I bought.
However, I am considering another alternative and it would be interesting to read comments on that, as well.
Although the crossover function will be strictly on Rod's PCB, since only digital media will be used, I will be able to use DSP within a streamer.

The idea is to measure the whole system, with crossover and two amplifiers in place and correct it with a convolution file. Something like a rephase could could alter both the phase and add linear EQ and even some kind of room correction could be implemented as a separate parametric EQ.
What is your opinion on using such a DSP with a 'stock' crossover I can get from Rod's work?

 

[tcpip]

I am not sure if Rod's p09 crossover allows for any change of variables, except for a frequency selection.

Rod's crossover can't prevent you from changing values of components. You can change the frequency and Q-setting component values freely (they are all R and C) without having to change the circuit schematic. But in order to decide the new values, you'll need to use a crossover simulator which lets you enter arbitrary active circuits and see the effect of component value changes. Enter stage left, VituixCAD. It's the most awesome free crossover simulator, the only one which supports analog active crossovers.

The idea is to measure the whole system, with crossover and two amplifiers in place and correct it with a convolution file. Something like a rephase could could alter both the phase and add linear EQ and even some kind of room correction could be implemented as a separate parametric EQ.
What is your opinion on using such a DSP with a 'stock' crossover I can get from Rod's work?

Yes you can do that. Essentially, you want to use an analog active crossover, and then use a digital filter to implement equalisation. That's great. This is what most non-DIY music lovers do when they set up room correction. In your case, the room correction will not just correct for the room but will also even out any unevenness in the SPL which will be introduced by your less-than-optimal crossover. It'll work.

 

[DonH56]

When I designed my original subwoofer crossover, ca. 1980 or so, I included all-pass filters to allow me to adjust the relative phase among speakers to facilitate integration. Some subs do include phase or delay control, e.g. my Rythmik subs have a "phase" knob.

 

[levimax]

Thank you for answering, it gives more material to think about.

I am not sure if Rod's p09 crossover allows for any change of variables, except for a frequency selection. I will measure my drivers again, but I think I'm stuck with what I bought.
However, I am considering another alternative and it would be interesting to read comments on that, as well.
Although the crossover function will be strictly on Rod's PCB, since only digital media will be used, I will be able to use DSP within a streamer.

The idea is to measure the whole system, with crossover and two amplifiers in place and correct it with a convolution file. Something like a rephase could could alter both the phase and add linear EQ and even some kind of room correction could be implemented as a separate parametric EQ.
What is your opinion on using such a DSP with a 'stock' crossover I can get from Rod's work?

This is exactly what I do. I use Rod's crossovers, adjust the crossover gain on each driver to get as close as possible to the desired FR and then use REW and Rephase to fine tune the speaker / room response with a FIR filter and Foobar's convolution plug in. I even add an RIAA filter and Brickwall subsonic filter for LP playback. I am sure it is not perfect but it is certainly way easier than trying to fine tune an analog crossover with more op amps and filters. If I was starting over with nothing I would certainly look at full digital crossovers but not sure it is worth tearing apart a working system and buying a lot of new hardware at this point.... although I probably will just to learn and have fun.

 

[levimax]

When I designed my original subwoofer crossover, ca. 1980 or so, I included all-pass filters to allow me to adjust the relative phase among speakers to facilitate integration. Some subs do include phase or delay control, e.g. my Rythmik subs have a "phase" knob.

You can now to the same with Rephase and FIR filters which is amazing and much easier.

 

[KSTR]

This is exactly what I do. I use Rod's crossovers, adjust the crossover gain on each driver to get as close as possible to the desired FR and then use REW and Rephase to fine tune the speaker / room response with a FIR filter and Foobar's convolution plug in. I even add an RIAA filter and Brickwall subsonic filter for LP playback. I am sure it is not perfect but it is certainly way easier than trying to fine tune an analog crossover with more op amps and filters. If I was starting over with nothing I would certainly look at full digital crossovers but not sure it is worth tearing apart a working system and buying a lot of new hardware at this point.... although I probably will just to learn and have fun.

IME this is a very efficient approach and I did similar things in the past. Sometimes, with a two-way, you just need to get the drivers's responses right at the XO point (read: make the phases track as best as possible, and match level) with simple passive crossover and then do all the gross and fine EQ digital upstream (and even full DRC), on the input signal (no need to actually go multi-way true active).

When driver offset have to be corrected for you're out of luck and the either an active XO with allpass features is called for or do all digital with multi-channel DACs. Electronic allpasses get hard to work with when you have to delay a tweeter significantly (or a main speaker vs. the sub).

---------:---------

For the circuit tinkerers, here is the so-called Steffen allpass circuit I've learned about some 4 years back which is very simple and robust and uses only a single OpAmp (copying from a post of mine at DIYaudio.com):

FWIW, I'm using this single OpAmp allpass design (not covered by D.Self?) which is unity gain and has rather low noise (see noise gain plot) as well as reasonably low component value sensitivity. Component values are a bit tricky to find, though.

Note that E1 is for simplicity of simulation (value finding), actually R2 is a divider from output to GND with output resistance of R2 and gain factor g.
Use R3 or R4 to fine-trim gain flatness.

This is just the general idea.

For exact dimensioning and analysis I have a more detailed LTspice sheet for this. I could post and comment it if there is any interest.

 

[dguillor]

On the matter of analog vs digital active xo, one point is often missed by newbies.

When newbies build analog active crossovers, they almost always build "stock" crossovers, i.e. textbook "knee" and slope, or textbook values of Fc and Q. These are also the builders who get attracted to Rod Elliott's PCBs or Marchand Electronics boxes. These builders never measure their speakers, or even bother to take the manufacturer-published SPL graphs, digitise them with graph tracers, and then put the FRD data into something like VituixCAD with the active crossover schematic.

Needless to say, these builds show problems later because of phase mismatch between drivers at crossover points, since drivers are rarely perfectly linear at the crossover frequencies. This means that when you measure the acoustic output of your crossover + drivers, using a mic, you'll see dips and peaks, and the sound will be coloured. This requires correction with parametric eq.

This is where analog active xo is worse than digital. A single channel of parametric eq can take 4 opamps, and even an analog notch filter can take between 1 and 3 opamps, depending on the Q you need. The tenacious builder will add all this extra circuit to the inputs or outputs of his analog active xo. Most newbies will simply not bother, and will marvel at the clarity of their new speakers and talk about veils being lifted. In reality, the veils being lifted are colourations caused by the peaks in the SPL graph.

In this area, I find digital active xo to be superior, because you can simply throw in a few channels of PEQ as and when you want to "tune" the sound, either with your own ears or with the help of a techie friend who may actually know how to measure speakers.

Rod Elliott (who I respect hugely, I feel he hasn't gotten the respect he deserves) too leaves the newbie speaker builder high and dry here. In his entire website, I've never seen any mention of actual speaker response SPL graphs, and how the xo modifies those graphs. Since I have some experience in speaker design, I am very happy to use his lovely 3-way active xo PCB, but I'll throw away all component-value-calculator spreadsheets, plug his schematic into VituixCAD and tune the R and C values after adding the speaker driver FRD data and optimising the simulated output.

Newbie speaker builders need to understand that stock crossovers, whether active or passive, almost always make mediocre speakers. And PEQ is the great blessing of digital active xo for such builders.

I think that you have overstated this problem. If you choose drivers with flat response and bandwidth considerably wider than the crossover frequency the problem at the crossover point caused by phase mismatch will be small. If this were not true, none of the passive crossover speaker reviewed on this sight, almost all speakers reviewed would get a recommendation.

 

[mdsimon2]

I think that you have overstated this problem. If you choose drivers with flat response and bandwidth considerably wider than the crossover frequency the problem at the crossover point caused by phase mismatch will be small. If this were not true, none of the passive crossover speaker reviewed on this sight, almost all speakers reviewed would get a recommendation.

@tcpip was talking about newbie DIYers, not competent designers. I think his point is very relevant as it is very clear that many hobbyists have no idea how electrical and acoustical responses interact.

You see this all the time with sub integration where someone will take a speaker that has some natural roll-off, apply an electrical high pass filter to it and then apply a low pass filter to the sub exactly matching the electrical high pass filter of the speaker.

Michael

 

[dennnic]

.... If you choose drivers with flat response and bandwidth considerably wider than the crossover frequency the problem at the crossover point caused by phase mismatch will be small...

This is exactly the info I was trying to obtain. By using a high order crossover (4th), how much is considerably wider, or just wide enough? One octave below/above crossover frequency, two?

By the way, if crossover is not right at / near FS of one driver, what phase shift are we talking about? Each crossover order will have a phase shift of its own, not related to the Q value. At least as far as I know.

 

[dguillor]

I am not sure right now. I believe it was one octave for an active crossover with fairly steep slope, 3rd or 4th order. More is needed for a passive crossover because a passive crossover will be affected by the impedance of the driver whereas an active crossover system will not. I would recommend the book Loudspeaker Design Cookbook by Dickason.

 

[tcpip]

This is exactly the info I was trying to obtain. By using a high order crossover (4th), how much is considerably wider, or just wide enough? One octave below/above crossover frequency, two?

By the way, if crossover is not right at / near FS of one driver, what phase shift are we talking about? Each crossover order will have a phase shift of its own, not related to the Q value. At least as far as I know.

Good question. How much is wide enough? For most "well behaved" drivers with no huge resonance distortion amplification or ugly cone breakups, you are ok if your xo pushes the driver's output down by about 30dB or more. This means that if your driver's output is well behaved till the xo pushes its output down by about 30dB, then that slope is good enough. So if the xo takes two octaves to reach -30dB and the driver is well behaved till that point, the xo is steep enough for this driver. This means a second-order acoustic rolloff. For another driver, you may get only one octave to reach this rolloff, in which case you need steeper slopes. So you use 4th order acoustic.

For drivers which have "badly behaved" regions, your xo needs to push down the badly behaved regions much further. How much further depends on the speaker designer and the "character" of sound he's aiming for. Me, I hate these metal-cone break-ups with a passion, so I like to push them down by at least 60dB.

The toughest design I've built till date is a 2-way MTM with very good but very hard-to-tame midbass drivers: Dayton RS180 metal-cones. I used a Seas TDFC tweeter, well known for low distortion and low Fc support, and crossed over at 1.5KHz, 6th order acoustic. See http://tech.dhandanought.org/asawari5

At the other end are unbelievably benign drivers, which can be used with 1st order xo slopes even by conservative and careful designers. I've seen Jon Marsh at HTguide DIY doing a 1st order 3-way design using Accuton drivers. For such designs, he demands that the drivers be well behaved at least for 3 octaves beyond each xo point. Very few drivers can deliver such even-tempered behaviour.

 

[tcpip]

I think that you have overstated this problem. If you choose drivers with flat response and bandwidth considerably wider than the crossover frequency the problem at the crossover point caused by phase mismatch will be small. If this were not true, none of the passive crossover speaker reviewed on this sight, almost all speakers reviewed would get a recommendation.

Whether I have overstated the problem is a matter of opinion. If I have to adopt your point of view, 80% or more of the drivers I find excellent will become unusable because their even and flat response region will not be wide enough for my use. I would hate to let go of so much choice, in the modern world where excellent free simulation software like VituixCAD lets me tune the acoustic response and match it to a target acoustic slope with little effort. If I adopt your approach, I'll become much less effective as an engineer.

 

[gnarly]

This is exactly the info I was trying to obtain. By using a high order crossover (4th), how much is considerably wider, or just wide enough? One octave below/above crossover frequency, two?

Maybe this will help.
Using rePhase to plot electrical xovers of different orders, using electrical as a surrogate for the purpose of determining how wide is needed acoustically.

Here's a 2nd order LR 12 dB/oct @ 300Hz. (ignore the horiz dashed-red phase trace, xover is lin phase)
If we want smooth summation to say -20bB,
that looks like the low-passed side needs good performance down to 100Hz,
and the high-passed side to 900Hz.

Here's 4th order LR 24/dB/oct @ 300Hz.
If we want smooth summation to the same -20dB,
that looks like the low-passed side needs good performance down to 175Hz,
and the high-passed side to 515Hz.

I use mainly 12th 16th order LR 96 dB/oct (linear phase)
Same -20bB,
which gives low-passed side needs good performance down to 350Hz,
and the high-passed side to 250Hz.
(Achieving steep complementary linear phase acoustic offers considerable latitude in choosing xover points.)

Hope that helped.

 

[gnarly]

For drivers which have "badly behaved" regions, your xo needs to push down the badly behaved regions much further. How much further depends on the speaker designer and the "character" of sound he's aiming for. Me, I hate these metal-cone break-ups with a passion, so I like to push them down by at least 60dB.

Thanks for some good points in your reply.



With regard to badly behaved regions, i take it you mean out-of-intended passband regions that a xover doesn't quite squelch.
I've become a big fan of out-of-band parametric EQs to kill those regions, making the xover's task easier.

Likewise, out-of-band shelving EQ's can sometimes make the summation region easier to get smooth.

 

[gnarly]

If you choose drivers with flat response and bandwidth considerably wider than the crossover frequency the problem at the crossover point caused by phase mismatch will be small. If this were not true, none of the passive crossover speaker reviewed on this sight, almost all speakers reviewed would get a recommendation.

Agreed.
Biggest problem is that drivers with flat response across wide bandwidth, inevitably have limited SPL.
I mean heck, if it weren't for that, i guess multi-ways wouldn't even exist.

 

[tcpip]

I use mainly 12th order LR 96 dB/oct (linear phase)
Same -20bB,
which gives low-passed side needs good performance down to 350Hz,
and the high-passed side to 250Hz.
(Achieving steep complementary linear phase acoustic offers considerable latitude in choosing xover points.)
View attachment 170143

Hope that helped.

Wow, this is the steepest I've heard any designer go. I'm guessing that this is all DSP active? Hard to imagine building such slopes in the analog domain.

Why do you need to do such steep slopes?

Which tools give you such slopes? I've seen support only for about 8th order in most tools.

Do you have any thoughts about lower order xo sounding different from higher order?