3rd order butterworth, proper implementation

I very much respect John Atkinson, his measurements and opinions. You are correct about the vertical polar behaviour of a vertical MTM array, however the power response of MTM with 1st order is exemplary if well designed, and the room responds accordingly. As a consequence of the laws of physics, the corollary of the poor vertical behaviour is better than average behaviour in the horizontal plane, where most humans ears are located. I do not agree that Dunlavy's speakers are highly intolerant of listener position, in fact from personal experience I very strongly disagree.
 
Well, since every such Dunlavy speaker I have heard has, in fact, been highly directional vertically (underlined by JA's measurements above), and given the configuration, as John himself stated, had minimum listening distances, it is not unreasonable to say they are highly intolerant of listener postion. While they were (and are) excellent at what they were designed to do, they were not loudspeakers that work well at a wide range of listening positions.
 
designs.

If you wish to explore the differences in sound quality of different crossover orders and alignments, a digital programable multi-way crossover is invaluable because you can switch rapidly between different designs. Also the active crossover implementation means that each driver is driven from an ideal voltage source, so you can make decisions based on an idealised crossover alignment without have to spend time developing Zobel and conjugate networks to get your drivers to terminate a passive crossover as best as possible.

In my experience DACs and ADCs have huge effects on the quality of sound. Adding two more in the signal path is something I just can't bring myself to do if it isn't absolutely necessary. Clarity and scale are impacted, the image narrows, quality is lost.

The reason I don't buy active speakers is when properly designed passive versions are driven by great amps, they sound better. The highs are always better in passives, and (assuming good design and quality parts) everything else is just as good, with only some exceptions. These exceptions are generally mild. For example, bass is more controlled in an active version: the woofer's passive crossover was made with a high resistance coil. This diminished the damping factor, causing less definition in bass. When a speaker costs $8000, this should not happen. But it does. Even though a low resistance coil only costs another $10, some companies don't bother. They should be overlooked, because what other stupid shortcuts are they taking that save very little money but have large impacts on performance?

Even when well managed, great driver and speaker manufacturers are rarely as capable of making the electronics required to drive them. If you compare a properly designed passive speaker driven by a midrange amplifier to the active version (as is often done to sell the actives), the actives will sound better. But even then the difference isn't huge unless the amp is a real piece of junk or the crossovers aren't built to spec.

With the signal split to high and low by a computer, a great DAC and proper amp hooked up directly to the drivers in their box will sound so, so much better than the active versions made by any company. And that's what I'm trying to do.

Do you know of any PC based programs that would accurately filter the signal - do what it says it does?
If not, could you recommend the make and model of an adjustable digital crossover. I'd use it for practice, if it sounds great I'd keep it.

I should be able to answer question of the change in sound character for individual instruments or sounds, but I have long since forgotten.

The last time I did this was in ~2004 when I was designing a nightclub system with active crossovers. I ended up using first order for sound quality, and making sure the systems were robust enough to achieve the necessary SPL with plenty of headroom.

Damn, forgetting sucks. I'll just have to try it.

Where abouts is this club with first order crossovers? If it's close enough to me I'd like to visit it - it seems like a rare animal for a club. Do they still have the system in use?
 
In my experience DACs and ADCs have huge effects on the quality of sound. Adding two more in the signal path is something I just can't bring myself to do if it isn't absolutely necessary. Clarity and scale are impacted, the image narrows, quality is lost.

I absolutely wholeheartedly agree! But the effects of high order crossovers have nothing to do with whether the implementation is digital or not.

Where abouts is this club with first order crossovers? If it's close enough to me I'd like to visit it - it seems like a rare animal for a club. Do they still have the system in use?

The venue was the Alpine Bar at the Botanic Hotel in Adelaide, Australia. The hotel is still there but unfortunately the night club venue closed 15 years ago or so because of noise complaints from neighbours. Ironically the speakers were sold off for use in home HiFi systems. I did however do a presentation to the Audio Engineering Society - Adelaide Section whilst it was still operating. Years earlier I had done a presentation to the AES in another nightclub about the development of idealised passive 3rd order Butterworth crossovers for some 26" RTR three ways that ran the dance floor, also in Adelaide.
 
Last edited:
Do you know of any PC based programs that would accurately filter the signal - do what it says it does?
If not, could you recommend the make and model of an adjustable digital crossover. I'd use it for practice, if it sounds great I'd keep it.
Sorry I didn't answer these questions.

I do not play with crossover designs other than first order these days and don't have a suggestion for PC software specifically for implementation of filters for crossovers. BTW I am not a night club system engineer per se, but I have been invited to design systems for a few clubs over the years.

The last time I implemented an active crossover I used a Rane RPM 26i, which is a capable signal processor that performs a lot of other functions as well. The one box does all of the functions needed for a simple installation, including an AES input to avoid unnecessary DA and AD conversions for digital sources, system room equalisation, source mixing/switching, interfacing for outside radio broadcasting, crossover, limiting, etc, just add power amplifiers.

https://www.ranecommercial.com/legacy/pdf/old/rpm26idat.pdf

Many professional audio companies made programable DSP based crossovers years ago. In public address systems active crossovers are a necessary evil and digital implementation made them incredibly flexible workhorses. I doubt that many, if any, would rate for high fidelity by today's standards. They have been surpassed by multifunction processors like the RPM 26i above and its much more sophisticated and better specified successors.
 
I absolutely wholeheartedly agree! But the effects of high order crossovers have nothing to do with whether the implementation is digital or not.

My point was just that on top of the unfortunate things that happen to the sound from higher order filtering, is the degradation x2 from conversion as well. It's why I want to do the filtering before the first DAC. Nice how your unit did things in the digital domain whenever possible. I think it's out of my price range for my purpose, unless I'm way overestimating what one would go for. The specs are still decent, very good for 20 years ago

The venue was the Alpine Bar at the Botanic Hotel in Adelaide, Australia. The hotel is still there but unfortunately the night club venue closed 15 years ago or so because of noise complaints from neighbours.

Damn neighbours wrecking good things lol. I wouldn't have made it though, even if it was still around. I don't think I could be further away!

Edit: when you say "the speakers were sold off...", was it individual drivers, or enclosures with multiple drivers covering most of the audio spectrum? I'm trying to imagine, with the especially excellent horizontal dispersion and power response, was it a bunch of smaller enclosures everywhere? (Not literally, just notably more of them and a bit smaller)
 
Last edited:

Attachments

  • Antipodes_LAEA.png
    Antipodes_LAEA.png
    66.1 KB · Views: 162
I was reminding myself about the KEF acoustic butterworth design:

This is the KEF R101 with 5" driver, similar to the LS3/5A:

377516d1382147610-kef-concerto-crossover-rebuild-kef_r101_crossover-jpg


This the KEF 104ab with 8" midbass:

377517d1382147610-kef-concerto-crossover-rebuild-kef-104ab-jpg


I am fairly sure KEF arranged the electrical slopes for a smooth 18dB/octave overall.

486597d1433396292-falcon-acoustics-5ab-crossover-5ab-t27-kef-cs1-electrical-png
 
I've just done a first order design for a KEF transmission system based on B139 and B110. I didn't use the T27 because it is fragile and it's not an outstanding performer anyway.

The circuit does require tweaking to match the individual drives to get it to behave optimally. I stopped tweaking when I could listen to the system in the background for hours without being irritated, which is actually a pretty good test for a speaker.

In fact when I owned a hifi shop and a distributor wanted me to take some super new range of speakers, I would let them demonstrate the system whilst discussing its merits and see how long before the sales person turned the level down or even off. IMHO there is a clear correlation between the ability to talk comfortably and the quality of the speaker system playing in the background LOL.

I can't post the schematic from this computer as I am away from home at the moment; maybe I can put it on the forum in a couple of days. I am pretty happy with how it sounds, given the limitations of the existing drivers and cabinet.
 
If someone wants to implement an analog active solution for a BW crossover a third-order state-variable filter (whose outputs are weighted and summed accordingly) would be an elegant solution for taking the driver responses into account and doing the crossover part in one go.

Regards

Charles
 
The circuit does require tweaking to match the individual drives to get it to behave optimally. I stopped tweaking when I could listen to the system in the background for hours without being irritated, which is actually a pretty good test for a speaker.

So true

I can't post the schematic from this computer as I am away from home at the moment; maybe I can put it on the forum in a couple of days. I am pretty happy with how it sounds, given the limitations of the existing drivers and cabinet.

I'd like to see it
 
Damn, forgetting sucks. I'll just have to try it.

LOL - I have since recalled that although I modelled the nightclub system with active crossovers, in order to meet the budget constraint the system was actually implemented with the speakers original passive first order networks in place, with a few upgraded components for long term power handling. The processing was implemented with a Rane RPM 26z, not 26i.

In order to provide some protection for the driveres, within the DSP I created split-band compressors with crossovers of the same Fcs as the speakers so the power level to each driver could be controlled/limited even though the speakers had passive crossovers.

The outputs of the split-band compressor limiters had to be recombined prior to the amplifier, and listening to this summed output the effect of the filters can be heard. It was necessary to use first order here to keep audio quality integrity going to the amplifier.
 
I'd like to see it
Mike, here's the crossover I referred to. The drivers are from an old KEF B139 / B110 / T27 kit in a TL box I have converted to reflex. I considered the T27 too fragile and having not a low enough fr to use with a first order filter, so I chose a PAE BC25SG16-04 as a reasonable and inexpensive tweeter for matching with the KEFs. The BC25 is ferro fluid damped and cooled.

The crossover is a basic first order with single element low pass (L1) for the woofer, two element bandpass for the mid (C3+7 / L2) and single element high pass (C2) for the tweeter.

The driver's all have far from linear impedance with both fundamental fr resonance blips and rising reactance with frequency. Without addressing these phenomena the crossover would not be terminated correctly and the slopes would be all over the road like a mad person's breakfast, as my mum would say.

The B139/system reflex alignment's resonances are far enough away from fc to be ingnored, but the self inductance of the driver is not. A Zobel network (C1/R1) corrects the impedance making essentially resistive in the octaves above the crossover fc.

The B110 unfortunately has considerable colouration and the impedance bump at resonance exacerbates the low end colouration by preventing the required roll off slope being achieved. A series network (R4/L3/C4) is tuned to exactly null the resonance impedance bump both in amplitude and Q, which corrects the amplitude response of the filter, and reduces the colouration. L3 is adjusted on test by adding or subtracting turns to centre the filter frequency. The mid driver also needs a Zobel (R2/C5) for the the upper fc low pass filter to work as required.

The high frequency driver also needs a series resonant circuit (R5/L4/C6) in parallel with the driver to null the driver resonance to get the desired high pass slope and remove the colouration at resonance. It also is attenuated with series resistor R3 to match the other driver's levels. It doesn't need a Zobel for crossover performance.

To derive the crossover, the on-baffle impedance and amplitude/phase responses were measured and converted to FRD and ZMA files for use in a crossover design program. The crossovers were then physically prototyped and tweaked in situ to optimise the slopes and crossover frequencies to get the desired performance.

The nominal crossover frequencies are 250Hz and 2kHz. The B139 is not very well behaved and the low crossover frequency is needed to mute its behaviour around and above 1kHz. Trying to get the B110 to work so low (at least in these legacy boxes) is also a bit fraught, and the best sounding compromise has a minor suck-out in the mid-bass.

The low pass L1 for the B139 is a 3.5mH P-core with a low 0.27Ω series resistance for good amplifier control of the driver. The high pass C3+C7 for the mid is a film capacitor in parallel with a bipolar for size and economic reasons. All of the other Cs are film, with the exception of C4, where the actual measured values of some nominal 220µF capacitors was used. Apart from L1 and L3 the other coils are all air core. The mid driver low pass L2 has 0.4Ω series resistance which acts as a slight attenuator that helps the balance.

The resistance of L3 and L4 coils in the two series band pass filters is taken into account reducing the values needed for R4 and R5 respectively, so low cost coils can be used.

One might ask: why go to so much trouble? The reality is higher order crossover are also affected by driver impedance anomalies and the crossovers will only work as expected into drivers compensated to appear to be resistive where important.

Many (most?) people don't realise that only a first order crossover (whether passive or active) cannot pass audio without altering how it sounds when the outputs of the filters are recombined - the output sounds different to the input for all orders of crossover higher than 1.

And finally a first order crossover has a flat power response which makes the speaker more room tolerant and the listening position less critical IMHO.

Of course the acoustic outputs of the system will not exhibit true first order phase and amplitude response, however I think this crossover implementation still does the least damage to sound quality, even with less than perfect drivers.
 

Attachments

  • oldkefsX.jpg
    oldkefsX.jpg
    22.2 KB · Views: 209
Last edited:
This is the crossover during prototyping (missing L3) implemented on Parts Express large 3-way crossover PCB Part #260-134. Several tracks were cut to make the layout functional. L3 (12.6mH, 3.45Ω) is an iron core Jantzen #000-2433 made to order from HiFi Collective, and fits on the dot, lower center right.

giFtKSA.jpg
 
Correct to my post #36 above:

Many (most?) people don't realise that only a first order crossover (whether passive or active) can pass audio without altering how it sounds when the outputs of the filters are recombined - the output sounds different to the input for all orders of crossover higher than 1.