This is super interesting. Thank you, @perrymarshall, for describing this concept in detail. And thank you, @AllenB, for adding in the series tweeter idea. I will play with this for my OB three ways, which use a passive parallel crossover and DSP tweaking.
Simplicity is a HUGE draw here.
Never a dull day in diyAudio. Always a tweak and an experiment lurking around the corner.
Simplicity is a HUGE draw here.
Never a dull day in diyAudio. Always a tweak and an experiment lurking around the corner.
OTOH, in the modern era (IMO, DSP-based parametric EQ is a given), then often the simplest xover that gets DI and phase right while keeping the drivers from harm is very useful approach. You can apply EQ to fix the frequency response (and thus snap the horizontal polar into line too, because DI is good), and still drive the thing from a single amp channel. I've done that with parallel crossovers before as well.
If you have a driver with a bad out-of-band breakup (e.g. metal cone midwoofer) can you still apply a notch filter within the series crossover? (In case it's not obvious; I've never tried one.) I'd expect it would take some juggling of parts values in V-CAD etc. to make it work, but is that a non-starter?
Absolutely you can do that.
Many of the same techniques you'd use in a parallel configuration can be applied in a series configuration. (In fact, that is the norm in most real-world series implementations.) The only speakers where you might get away without extra manipulation are systems with drivers that are resistive and have no large resonances to tame. Magnepan's are a good example. And, of course, the project right here worked well with a simple config.
Regards global equalization.....yes that's certainly an option as well should you find two drivers that hump in the crossover range. It's the easiest option to solve a situation like that.
Although in parallel crossover you have the option to fix it in the network and not resort to that.
Dave.
Many of the same techniques you'd use in a parallel configuration can be applied in a series configuration. (In fact, that is the norm in most real-world series implementations.) The only speakers where you might get away without extra manipulation are systems with drivers that are resistive and have no large resonances to tame. Magnepan's are a good example. And, of course, the project right here worked well with a simple config.
Regards global equalization.....yes that's certainly an option as well should you find two drivers that hump in the crossover range. It's the easiest option to solve a situation like that.
Although in parallel crossover you have the option to fix it in the network and not resort to that.
Dave.
Last edited:
You can but why would you do that?
And let's clarify from the start, here we are talking about 1st order series xo, not 2nd, not 3rd.
There is a huge difference and I would recommend reading Rod Elliott's articles on the topic:
https://sound-au.com/parallel-series.htm
https://sound-au.com/articles/6db-xover.htm#s3
1st order series xo does not need either Zobels or notch filters as the very nature of the filter takes care of all this.
The issue usually comes down to careful selection of drivers.
If you insist on using such networks you'll be better off with a parallel filter.
Ignore the parallel filter crowd who'll tell you parallel and series filters are the same and you can use the same techniques in both and achieve the same result. That's rubbish.
Those people crave control in the belief that somehow it will produce a better sounding speaker.
Ever thought why the resurgence in popularity of Full Range drivers and single driver speakers?
Many reasons but an important one is because they sound natural.
A 1st order series filter with carefully selected drivers creates sort of a quasi-full range speaker. It is not the same but comes very close.
We can keep talking and I'm sure this discussion will continue but all it takes is a Open baffle of some sort because it's simple, two drivers, a woofer and a small full range like Vifa, TB, or the more expensive Mark Audio, an inductor and a capacitor with values calculated for your xo point and maybe a resistor or two for level balancing.
Wire it up, measure the impedance to be sure you won't stress your amp, fire it up and listen.
Later if you like the result play with XSim and/or VituixCAD, optimise it and listen again.
BTW, the late John Dunlavy used to evaluate his speakers with square wave signals. As it's known 1st order filters are the only filters that can reproduce a square wave relatively well. Of course Mr. Dunlavy was using quite complex 1st order filters trying to produce the perfect phase coherent speaker.
Now try the same exercise with a series one (6dB) and be amazed.
I built a system using the Ciare HWG165-4 which is an excellent driver but has a peak at 5K that must be dealt with.
Here are the models for a parallel crossover that I modeled and actually built. It's worth pointing out that this circuit is not a traditional low pass, it's a notch filter specifically targeted at the 5k peak but which also compensates for Baffle Step:
Here are the models for a series crossover that I ran just to see how good I could get it (I did not build it FYI):
You can see in the series crossover the LC trap circuit on the Ciare woofer which removes the 5K peak. .56mH || 1.8uF+5.6 ohms. @Davey is right, the series has more interdependencies so when you actually build it and inevitably have to tweak stuff to fine tune the sound, the series config may be more challenging.
Last edited:
Yes, that's a very typical real-world situation. Folks who have built systems with Seas Excel drivers (for example) would find that familiar looking. 
I note that you also had to reverse the (conventional) relative polarity on the drivers in both instances.
So right away you found yourself in a trade-off situation to achieve your goal. (This is the sort of real-world challenge thing I'm talking about.)
Nice project there.
Dave.
I note that you also had to reverse the (conventional) relative polarity on the drivers in both instances.
So right away you found yourself in a trade-off situation to achieve your goal. (This is the sort of real-world challenge thing I'm talking about.)
Nice project there.
Dave.
Last edited:
I agree with all that but with one caveat: Actually achieving real world 6dB slopes is nearly impossible (even with standard IIR DSP!) without really exceptional wideband drivers; and as @Davey points out, a series circuit with a woofer and FR driver will most likely require reversing phase. The nice steep rolloff of the FR driver with a series crossover, will not be a 6dB slope, it will be more like a 15-30dB shelf filter with a Q between 1 and 2.
The slopes are still gentle and the sound is still natural but it will take some serious filter design chops to achieve what John Dunlavy achieved with his designs. Even with DSP it's difficult unless you use FIR filters - and then it becomes very easy of course.
In the Bitches Brew and Live Edge Dipoles, the phase shift of the crossover from woofer to mid is pretty similar to a 12dB/octave slope requiring polarity reversal on the midrange. I correct that in the FIR DSP file by adding an all pass filter that shifts phase 180 degrees and brings the system back into linear phase. Part of my strategy with crossover slopes was to make this easy to do. If you read the Live Edge Dipole article in AudioXpress carefully, you see the layers of filtering and I don't fully explain in the article that I'm trying to minimize the amount of gymnastics that the FIR filter has to do.
In the Bitches Brew design, the MiniDSP Flex8 only has 2 FIR channels and I use both of those channels to rotate the phase 360 degrees between bottom and top and that renders the phase response flat and the impulse response close to perfect. This goes back to a priority that I learned by studying John Dunlavy and his viewpoint that a square wave is a huge litmus test of a speaker. 99.9% of designers don't have the chops for that but he did.
Last edited:
This is the FIR filter I made using Eclipse software. More details at https://evo2.org/wp-content/uploads/2023/09/Live_Edge_Dipoles_2023_Beryllium_Update.pdf
I just posted a VituixCad model of the Live Edge Dipole 150Hz series crossover for the Eminence Kappa 18LF woofer and the Radian 5208 mid-woofer here: https://www.diyaudio.com/community/...ompetition-updated-design.405104/post-7500040
B&C's FB4648 dedicated passive crossover for the DCX462 and DCX464 8 ohm HF coaxial drivers appears to work fine in their measurements, the polarity of the HF is reversed, "a huge wide dip" would result if it wasn't.
Great thread !
Years ago I designed series XO's for two projects with great results. (always 2way)
On was for a disgruntled Marantz customer that wasn't happy with his speakers sound.
As we were importing spectrum analyzers at the time, I measured the mod. result >
and it was flat, I mean really flat.
Depending on the drivers & simple component values, you can actually 'FLUKE' amazing results.
Years ago I designed series XO's for two projects with great results. (always 2way)
On was for a disgruntled Marantz customer that wasn't happy with his speakers sound.
As we were importing spectrum analyzers at the time, I measured the mod. result >
and it was flat, I mean really flat.
Depending on the drivers & simple component values, you can actually 'FLUKE' amazing results.