Swift-IIIb (Ultimate)
(High Definition Series)

5 inch 2-way
13 Liters Bass Reflex Satellite

by Michael Chua

This design is free for DIY. Not for Commercial Use.
This article may not be published in part or full without the express permission of AmpsLab.

Difficulties with 4 ohm Tweeters

In Swift-III, the 3rd Order Filter caused the XT25TG30 impedance to dip to 2 ohms.

Surely most power amplifiers' protection circuitry will be activated, causing distortion in the process. Can anything be done to raise the minimum impedance apart from installing a resistor at the front of the tweeter's network?

click picture to Enlarge

Fig 12 - Swift-III Impedance Dips to 2 ohms
click picture to Enlarge

Fig 13 - Swift-III Distortion at 2 ohms

The Case for 1st Order

One option is to use a 1st order network. But this minimalist approach presents it's own set of issues.

When the XT25TG30 was tested for distortion with only a 4uF capacitor in it's path, it registered 4.67% THD at 527Hz (Fig 14). It is by no coincidence that this is exactly the region where the Fs of the tweeter resides. We can safely deduce that the distortion is due to the tweeter resonating.

XT25TG30 Distortion at Fs

Fig 14 - XT25TG30 Distortion with 1st Order Filter (4uF)
(mic 20 ins On Axis with XT25TG30 | 5ms Impulse Window | 1/12 oct smoothing)

Is there a cure for this? Fortunately, yes. It's easy and doesn't cost much. It's what is typically called a Series Notch Filter or LCR. What it does is it damps or "flattens" the impedance peak at Fs (Fig 16). When it's done right, there's no distortion (Fig 17).

click picture to Enlarge

Fig 15 - Vifa XT25TG30 Resonance Frequency (Fs)
click picture to Enlarge

Fig 16 - Vifa XT25TG30 with LCR

XT25TG30 Distortion Eliminated

Fig 17 - XT25TG30 Distortion Eliminated with addition of LCR
(mic 20 ins On Axis with XT25TG30 | 5ms Impulse Window | 1/12 oct smoothing)

At this point, it's an opportune moment to draw a distinction between the types of distortions. In the earlier example, the distortion originates from the power amplifier when the impedance dipped to 2 ohms.

In the latter case, the distortion is not from the power amplifier but from the tweeter resonanting at Fs. This is electro/mechanical in nature whereas the former is from electronics.

XT25TG30 Acoustic Slopes

Fig 18 - XT25TG30 Frequency Response
(mic 1 meter Tweeter Axis | 5ms Impulse Window | Smoothed 1/12 oct)

Fig 18 shows the acoustic responses of the XT25TG30 when swept with the bass removed.

The Black Trace is of the tweeter in it's raw state. The Red Trace is with a 5uF and a L-pad. Lastly, the Brown Trace is with the 5uF, L-pad and LCR. All traces are normalized to Black.

Notice that with the LCR, the acoustic roll-off is about 6dB from 2kHz-1kHz (Brown Trace), conforming to a theoretical 1st order network.

Equally important is the response at 500Hz, where the Fs of the tweeter is. Without the LCR (Red Trace), the SPL is at 67.5dB. With the LCR (Brown Trace), it drops to 52dB. That's a massive attentuation of 15.5dB.

ZA14 Distortion

We shall now turn our attention to the ZA14 midwoofer. Fig 19 is the distortion measurement of the ZA14 in it's "raw" state, ie with no components between the power amplifier and the ZA14. This shows the intrinsic Distortion of the driver. Note the spike in distortion at the cone breakup near 10kHz and the associated elevated distortion all the way down to 1kHz. Of particular interest is a secondary spike one and the half octave down at 3kHz.

Fig 19 - ZA14 Distortion
(mic 20 ins On Axis with ZA14 | 5ms Impulse Window | 1/12 oct smoothing)

Fig 20 below is the Distortion measurement of the ZA14 with a 18dB/oct crossover and a Zobel network. The huge distortion spike at 10kHz vanishes and with it, the havoc it created downstream all the way to 1kHz.

ZA14 Distortion Plots
with 18dB/oct and Zobel networks installed

Fig 20 - Distortion Measurement of ZA14 + 18dB/oct XO + Zobel
(mic 20 ins On Axis with ZA14 | 5ms Impulse Window | 1/12 oct smoothing)

Now that the distortions of the XT25TG30 and the ZA14 have been resolved, it's time to merge the two drivers. In Fig 21, the Brown trace is the ZA14 with a 18dB/oct network and a zobel. The Red trace is of the XT25TG30 with a 1st order filter follwed by an L-pad and finally a LCR. The Blue trace is Swift-IIIb Frequency Response. Ignore the measurements below 500Hz as the room begins to affect the readings.
Swift-IIIb Frequency Response

Fig 21 - SWIFT-IIIb Frequency Response
(mic 1 meter Tweeter Axis | 5ms Impulse Window | 1/12 oct smoothing)

Fig 22 below is the Distortion measurement of Swift-IIIb. Measurement is made with the microphone at 20 ins, on axis with the tweeter. As before, readings below 300Hz should be ignored.

Distortion from 300hz to 10kHz ranges from -32dB to -37dB. Averaging, it's 50dB below the reference level of 85dB. This is a very acceptable figure.

Swift-IIIb Distortion

Fig 22 - Distortion Measurement of Swift-IIIb
(mic 20 ins On Axis with Tweeter | 5ms Impulse Window | 1/12 oct smoothing)

Fig 23 is the Impedance of Swift-IIIb. At 300Hz, the impedance is 8 ohms. The lowest is 6 ohms at 20kHz.

The electrical Phase Angle is not as drastic as Swift-III. At the crossover region, it swings from +18 deg to -36 deg compared to +36 deg to -54 deg in Swift-III.

Overall, Swift-IIIb is a much kinder load for power amplifiers. The minimum of 6 ohms is a good 2 ohms away from activating the amplifiers' protection circuitry.

Swift-IIIb Impedance

Fig 23 - Impedance of Swift-IIIb

The crossover network of Swift-IIIb is shown below (Fig 24). The 3rd order filter has been retained for the ZA14. I do not recommend anything less as metal cone drivers are prone to distortion during cone breakup.

The XT25TG30, on the other hand, has been greatly simplified. The 3rd Order network has been replaced with a 1st Order (5uF). For best sound quality, the 2.2 ohm resistor should be a non-inductive 10W to 20W type. It is totally wasteful to use a Polypropylene capacitor for the 170uF. A cheaper quality E-cap would suffice here.

The rest of the components are air-cored inductors (18 gauge), polypropylene capacitors and 10W resistors.

Swift-IIIb Crossover Network

Fig 24 - Crossover Network of Swift-IIIb

the sound of Swift-IIIb

It is a very relaxing speaker to listen to. No irritation. No "Listening Fatigue". Simply clean music.

Surprisingly, the transparency did not suffer with the 1st Order tweeter. Vocal clarity and focus are as good, if not better, than with a 3rd Order filter. The Swift-IIIb does not "shout" at the listener. Neither does the vocals recede into the background, making it sound like "pipe-in music". It just does it's thing without being intrusive.

After all the design variations, I must admit Swift-IIIb is the best. In the preceding Swift-IIIa, placing a resistor at the front of the tweeter's network to raise the impedance is, in my view, a rather "crude" solution.

The simplicity in Swift-IIIb, in comparison, has a touch of elegance to it. For your musical pleasure, I suggest adding a subwoofer.

Aug 12, 2013


next > Swift | Swift-II Hybrid | Swift-II Passive | Swift-II Music
  Swift-III | Swift-IIIa | Swift-IIIb | Swift-IIIc | Swift Ultima | Swift Ultima-II


60 Downes Street | Calais | ME 04619 | USA