A common procedure to select drivers for a loudspeaker project is by examination of its parameters, mainly its response chart or its FRD file. This makes it easy to match 2 drivers before purchasing it. But the selection for this project didn’t follow that procedure. I wanted high end drivers to build a high-end loudspeaker, and I believed any two drivers can always be matched when time and crossover requirements have no limit.
I have always wanted a Seas Excel but I couldn’t source it so I choose Focal 5W4211 over an ordinary Scan-Speak. For tweeter, I choose the smoothest sounding available, a Dynaudio D21/2. This tweeter is one of the best. With resonant frequency of 1.3 kHz this tweeter is more suitable for a three way loudspeaker, but 5W4211 is only a 5-inch woofer so I thought it must be capable of producing high octaves to match with the tweeter.
Only later I found out that these 2 drivers are not really a good match. The woofer has terrible peaks in the midrange frequencies that are very difficult to tame. And especially when your ears are very sensitive to peaks, you may want to have a rather low crossover frequency for the woofer, and also a rather over-damped response, something that can hardly be matched by the tweeter.
Strangely I had in my occupation a Vifa XT25 tweeter that in my opinion was a perfect match for the woofer. But I regarded D21 a better tweeter than XT25 so I wanted the better drivers to work together, at any cost!
[5w4211] [D21-2] [XT25]
The published response chart of the woofer didn’t show any serious breakup. The specified Fc of 46Hz does look promising as well. But from examination it was found that the bass response of the woofer is not good enough and the breakups pattern in the midrange frequency is not so good either. Worse, the major breakup is in the mid-band!
BSC [5W4211 SPL]
Parameters suggest that 5W4211 should go with 8lt vented or horn enclosures. I’ve always wanted a horn but I don’t know anything about its design, let alone the size and construction difficulties.
5W4211 has a free-air resonance frequency of 46Hz. Such is usually low enough for a mid-woofer to produce critical lows but I found it useless to build the suggested 8lt bookshelf with this woofer.
Anyhow, a quick examination suggests that with both drivers it is perfect to clone or to build something like the LS3/5A. This is a famous mini monitor with sealed enclosure having volume less than 5lt. In smallish LS3A clone it is possible to get the critical lows at the price of reduced sensitivity. This and with the nature of sealed enclosure, ensure the loudspeaker can be driven with high power amplifier even though the woofer is only rated at 50W.
The FocalAudio#1 is a realization of this LS3/5A clone and is indeed better than the LS3/5A. But with the high price of the drivers I urge something better than a mini monitor. I didn’t like bass reproduction that is un-natural from a smallish bookshelf like that. The frequency is not low at all. It is just a mid-bass frequency that is pronounced through a baffle step compensation scheme. My friend liked it a lot but I didn’t, so found another solution that would become the FocalAudio#2.
Smallish woofer is usually not suitable for a floor-standing speaker. The resulting bass is unnatural as it is generated by the box not from direct cone movement. So going after this kind of bass and loosing bookshelf strength which is the coherency of its midrange is usually not a good idea. But 5W4211 is extremely dynamic and is very detail (as the stiff material would suggest) so going floor-standing or Transmission Line with FocalAudio#2 is a good option to get the critical low frequencies while taming the midrange breakups!
The transmission line enclosure used is almost identical with the one used in Daline 3.1 loudspeaker by Orca Design which is using different Focal woofer with properties that are quite similar with 5W4211.
I was stuck with Martin J. King’s theory and MathCad program and was not sure with what I was doing so the design of FocalAudio#2 and FocalAudio#3 were actually done with vented bookshelf enclosure (as can be seen from the impedance curve). The crossover then be transported to the transmission line enclosure and comparison will be done to select which one is better, the vented bookshelf or the transmission line enclosure.
It was a hard time for me to select which enclosure is better, the vented bookshelf or the transmission line enclosure. Of course, all listeners preferred the transmission line, but you know I have different ears. Basically, I don’t like the sound of a TL speaker. I wouldn’t put the blame into the design as I have heard several TL speakers and still didn’t like the sound.
Though dynamic, soundstage and resolution are better with bookshelf speaker, the low frequency reproduction of TL speaker gives more charm, especially with recording materials where low frequency content is dominant.
Crossover Type and Frequency
D21 tweeter has a free-air resonance frequency (Fs) of 1.3 kHz. From tweeter point of view it is probably wise to have a crossover frequency of no lower than 2.6 kHz (twice of the Fs). In the FocalAudio#1 (The LS3/5A Clone) this realization is easy as the woofer (and also the tweeter) was damped. In FocalAudio#2 (The Concertino) where over-damping was the last thing expected from the design, this would be a big challenge for 5W4211 as it has nasty peaks in the midrange frequency. But strangely I decided that 5W4211 should be crossed higher than anyone might suggest. I thought that if it is crossed low, I will loose the incredible mid-bass properties, the woofer’s strength point. And the critical breakup is right after 1 kHz, so you cannot avoid it by crossing it low anyway.
On the other hand, D21 roll-off is smooth and shallow. It is easy to tailor the slope into any type of crossover slope at crossover frequency around 3 kHz. So the constraint here is the woofer. The woofer’s natural roll-off suggest that 4th order electrical should be used (Actually, when crossed around 2.6 kHz, its natural roll-off is steeper than 4th order Butterworth).
A simple solution as is learnt from FocalAudio#1 (The LS3/5A Clone) is to pad down 5W4211 from 84dB to 81dB right after 1KHz where the breakup starts, then going flat and pad it down again at around 2.6 KHz with 4th-order Linkwiz-Riley or Bessel slope. The extra 3dB before 1KHz will be acting as the baffle step compensation for the speaker. It is an effective solution to solve the breakup problems of 5W4211 and a high Fs of D21 simultaneously. But I wanted to get rid of the sound of an over-damped in-efficient mini-monitor. I want something that reveals the strength of both drivers, something that sounds so real and lively like a real concert.
Honestly I forget the working crossover for FocalAudio#2. This is because the two drivers are not a good match so that computer simulation is hardly the optimum filter for both drivers. And I relied on my ears more than on computer simulation result. In short, I have to change the components values during tweaking and forget to measure them. But that was for FocalAudio#2 (The Concertino) that I have given up.
The enjoyment was pure with FocalAudio#2, and I have often dreamed of having it back, hoping I could find a way to tame the woofer breakup without having to bring the response at 1 kHz down to 81dB. It was the only speaker that sounded so real to me. D21 vocal is just incredible. Unfortunately I experienced a bit of fatigue while listening to some recording materials for a long time so I decided to upgrade the loudspeaker by bringing the crossover frequency lower and thus “sacrifice” the tweeter.
With FocalAudio#3 the crossover filter is completely different. The objective of this upgrade is to reduce the glare or peaks of the woofer by bringing the sensitivity down and having the crossover frequency lower than before. Of course the tweeter will have more tension so notch filter is mandatory for the tweeter
So here is the 4th order Linkwiz-Riley crossover filter for FocalAudio#3:
The chart doesn’t look too good, I know. But of course I can show you a good looking chart if you want me to. But that should not be necessary as we’re after a good sound not a good chart anyway.
What I’m trying to say is, the crossover was designed more by ears than by computer simulation. Computer did help for speeding up the design, especially to find possible network with good phase overlap.
Looking at the chart you can see a gentle dip at mid frequency. To this I’d say that most speakers have midrange that is even more pushed to background. The phase overlap around crossover point is the most important design goal (if not the only). The rest is to put the drivers to work in their optimum network which is done by ears.
I know, big inductors and series notch filter at 1 kHz will kill the woofer dynamics. I have tried to find options where the notch can be left without serious exposure of the peak but I couldn’t get something as clean as the one with notch filter. I also compared the series notch filter with parallel one. While I can get better dynamics with parallel notch filter, it is not as effective as series one to tame the peaks out of the woofer.
Changing the notch filter from series to parallel means a completely different crossover. And the resulted phase pattern may lead to the need of costly and harmful delay network for the tweeter. These suggest that further experiment with parallel notch filter wouldn’t be necessary.
Baffle Step Compensation
As can bee seen from the chart, there is no special BSC circuit implemented. This might be forgiven as this is a small power loudspeaker that is used in a small room. I was reluctant to provide the BSC mainly because the D21 tweeter is already too sensitive. Padding it down too much wouldn’t give me the charm it produced as in FocalAudio#2. I realized that the objective of the loudspeaker upgrade was to reduce woofer peaks by bringing the sensitivity and crossover frequency down and thus sacrificing the tweeter. But I didn’t intend to sacrifice the tweeter to a degree where my XT25 tweeter would do the job better! A 10 Ohm resistance is the maximum I intended to put in series with the tweeter.
The Sound Quality
The sound of FocalAudio#3 is quite okay. I realize that there’s no perfect loudspeaker. There are so many trade-offs in loudspeaker design. You just cannot have all the best for cymbal airiness, midrange sweetness, bass tightness, dynamic transient, etc in one design. Often, it is the designer preference or his ability to judge good sound that determines the final result.
For me, I know exactly what I’m after. I want no more than a loudspeaker that makes me feel absorbed into the music without getting tired. Getting only these two qualities is not easy because clean sounding loudspeaker usually not interesting due to bad dynamic transient. And dynamic loudspeaker often has overexposed midrange that leads to fatigue. And even worse, good sounding loudspeaker cannot be seen from its response chart or other related computer-aided design outputs.
In term of resolution, vocal sweetness, high airiness, bass tightness, this FocalAudio#3 is a bit worse than FocalAudio#2 but overall I like FocalAudio#3 just better. It has a magic in its sound
Knowing that I have worked with both drivers for quite sometime I think I’m running out of idea on how I could possibly improve the crossover design. Although both drivers are not a good match, the network pattern has been carefully chosen to put both drivers working in their best possible condition.
In FocalAudio#3 the woofer peaks are sufficiently reduced through series notch filter and steep slope as well as a relatively low crossover point. Due to fine blend of both drivers, this necessarily complex crossover filter doesn’t smack the speaker dynamics down.
The tweeter can be brought low to match with the woofer at 2300 Hz without sounding harsh and still maintain its natural vocal and airy highs (worse than it is in FocalAudio#2 but in my opinion is still too good to compare with what can be achieved with XT25).
It is probably the bass response that is left for upgrade through vented floor-standing or big bookshelf enclosures. Nothing really wrong with transmission line but I must say that I’m not a fan of it. I still prefer the sound of the ordinary floor-standing vented enclosure. Not going as low but a lot faster and provide more dynamics.
Last but not the least, below is a picture of a preliminary design using 5W4211 woofer and XT25 tweeter in an effort to upgrade the FocalAudio#2. Here parallel notch filter is used to replace the series one, and Vifa XT25 is used to replace the Dynaudio D21/2 tweeter.
Listening to the prototype I couldn’t see any promising future with either the parallel notch filter or the XT25 tweeter so I didn’t look for more possibilities with the network and went straight to the FocalAudio#3. The tweeter sounded okay but I know exactly that it could do better in third order electrical slope without the delay network. Even so, to my preference it wouldn’t be as good as the Dynaudio tweeter!
Yes, there is a very simple way to make this speaker just sound better! Now that I know it, I hope I could revisit all the previous designs, especially the FocalAudio#2. FocalAudio#2 is very awesome, incredible sounding design. The only problem is the peak created by the woofer cone because I had to bring it up in the frequency ladder to match up with the Dynaudio tweeter.
Is it really an easy upgrade? Yes, because the problem can be solved by changing only the notch filter! The previous notch filter was found in many other designs using the same woofer (such as The Ego of the Dillon Acoustics). That notch filter just doesn’t work! It has to be changed!