Fullscreen HTML5Med Res (??MB)HTML4

AUDIO OUT AUDIO OUT L By Jake Rothman PE Mini-monitor crossover for Wavecor drivers – Part 2 Basic circuit A pair of PE Mini-monitor prototypes in action – note the anti-reflection foam on the desk! I started last month by extolling the virtues of the famous LS3/5A, describing it as, ‘arguably the best mini-monitor speaker’, but I qualified that with, ‘it is very expensive and getting the parts can be tricky’. The purpose of the PE Mini-monitor project is to get L2 2mH Ferrite core R9 56Ω 2W C2a 6.8µF C2b 10µF (bipolar) 0V Woofer WF120BD06 + C5 5.6µF C6 7.5µF L3 0.22mH Air core Tweeter TW022WA04 + Fig.16. Complete provisional crossover from last month, plus damping network (R9, C2b). 50Ω 50Ω 40Ω 40Ω 30Ω 30Ω 20Ω 20Ω 10Ω 0Ω 20 Hz I got lucky here, phase is often difficult to get right in crossovers, requiring all-pass filters or moving the driver’s positions around, with deep horizontal steps or undesirably wide vertical spacings. In this case, the difference in the horizontal radiation plane between the two drivers was only around 15mm, with the large hard woofer dome and recessed tweeter dome helping to 10Ω 50 100 200 500 1k 2k 5k 10k Fig.17. Impedance curve of first crossover in Fig.16 without damping network. 48 For our crossover design, we arrived at the circuit shown in Fig.16 in last month’s article (repeated below). As explained, it is not the fi nal design, but it is a satisfactory foundation on which to build. I noticed there was a bit of a response hump in the crossover region and more worryingly, a dip (at 2kHz) below 8Ω in the impedance curve (Fig.17). It is always important to keep a check on the system impedance curve when designing crossovers. The speaker designer should not make life difficult for the amplifier designer by having low dips and very rapid (reactive) changes in the impedance. This problem was fixed by adding a 56Ω damping resistor (R9) across the lower arm low-pass filter capacitor C2a. Of course, we don’t want 14% of our precious bass power being wasted in this resistor, which is effectively in parallel with the bass unit. Since we only needed its effect around 2kHz, a 10µF bass-blocker capacitor (C2b) was placed in series. This capacitor does not need a low ESR, so a cheap bi-polar electrolytic can be used if you are a penny-pincher. The resulting impedance curve with the damping network is shown in Fig.18. Phase serendipity R4 10Ω + Input from amplifier most of the true LS3/5A’s bang for only a fraction of the buck! We have spent some time going through speaker parameters and crossover designs and we are nearly at the point where we can actually build a complete speaker, but first we need to complete our new crossover. R2 22Ω L1 1.2mH Ferrite core R 20k 0Ω 20 Hz 50 100 200 500 1k 2k 5k 10k 20k Fig.18. Impedance curve with damping network (9, C2b). The impedance curve for the final circuit in Fig.21 is the same. Practical Electronics | March | 2020 n +10 0dB –10 –20 –30 –40 20 Hz 50 100 200 500 1k 2k 5k 10k 20k Fig.19. Improvement to the low-pass attenuation at 10kHz curve by adding C3a and C4. minimise this. Also, the low crossover frequency made the phase shift in wavelength terms quite low; so low it cancelled out with the electrical phase shift from the filters. I didn’t even have to reverse the phase of the tweeter, which is normally the case with crossovers of this topology. Listening to white noise, it is possible to identify phase notches around the crossover point by moving up and down in front of the speaker. (Do this when nobody’s watching!) The white noise should ‘gel’ on-axis to the cabinet at a point between the two drive units. An objective test with any crossover is to deliberately connect the tweeter out of phase. A deep symmetrical null should develop (as shown later in Fig.32). Little tweaks Once the basic crossover has been sorted out some refinements can be made, often involving listening tests, called ‘voicing’. The expensive components are the inductors (usually about aB bridge: dip at 850Hz C7 220nF C5 5.6µF R5 270Ω 0.5W High-frequency boost above 6kHz (actually a low-treble cut) Input from amplifier L3 0.22mH Ferrite or air core Tweeter TW022WA04 The 10kHz peak is still slightly audible on the woofer output on-axis with the tweeter disconnected. Further attenuation can be obtained by notch-tuning the final inductor with a parallel capacitor C3a. A capacitor C4 in parallel with the woofer also helps. The difference to the low-pass electrical curve is shown in Fig.19. It just reduces the 10kHz by about 6dB. Since there is now no treble radiated from the woofer, the overall treble response sounds a bit smoother because there are no high-frequency cancellation notches. Softer treble The crossover as it stands may sound a little ‘hard’ or ‘mid-rangey’ to older listeners (50+) with a degree of age-related high-frequency loss or presbycusis, which I have. This can be ameliorated with a bit of treble lift on the amplifier or by inserting the network consisting of C8 and R7 in series with the tweeter after the high-pass filter. This decreases output in the crossover region a little and boosts it a few dB at 10kHz. This gentle rise also compensates for reducing the high-frequency output off-axis from the tweeter. C6 is also reduced to the more common value of 6.8µF because of the increased impedance. Here’s another subtle tweak, one considered by KEF and others to smooth the high-pass curve at the lower end. It suppresses the tweeter’s 850Hz resonant peak using an acoustic-Butterworth bridge consisting of C7 and R5. I could only hear it on white noise with the woofer disconnected, but it costs less R7 3.3Ω + Peak suppression Tweeter aB section C8 10µF C6 6.8µF £4 to £6 each) so we avoid adding more. Most tweaks just involve relatively cheap resistors and capacitors, generally adding extra refinement to the upper-mid and treble area. Some tweaks do not show up on the acoustic response curve because they are low down in the stop-band area of the crossover curves; but they can still be heard. Because of this, some electrical curves (Fig.19 and Fig.20b) are plotted on a 50dB scale. + 0V R5 270Ω 1.0W Fig.20a. Adding a tweeter anti-resonance dip using an aB network (C7, R5). Note also added a treble boost network (C8, R7). R4 10Ω 3W C5 5.6µF 0dB C8 10µF C6 6.8µF R7 3.3Ω + L3 0.22mH (Air core or ferrite) Input from amplifier +10 C7 220nF Wavecor Tweeter TW022WA04 4Ω + 0V –10 Note: tweeter in phase with woofer –20 R2 22Ω –30 –40 20 Hz 50 100 200 500 1k 2k 5k 10k 20k Fig.20b. Resulting tweeter anti-resonance dip using an aB network (C7, R5). L1 1.2mH 0.23Ω Ferrite core C2a 6.8µF 0dB –5 –10 L2 2mH R9 56Ω 0.25Ω 2W Ferrite core C2b 10µF (bipolar) + C4c 3.3µF Wavecor Woofer WF120BD06 8Ω –15 –20 20 Hz C3a 680nF 50 100 200 500 1k 2k 5k 10k 20k Fig.20c. Adding a treble boost network (C8, R7) cuts the low treble signal slightly. Practical Electronics | March | 2020 Fig.21. Final crossover circuit with all the tweaks. These did not alter the impedance curve. 49 R3 C4a C4b C7 C3b C3a R2 R7 C4c L2 C8 R5 C2a R9 C1b Link C6 C2b C1a Input + LF – R1 C9 R8 R6 L1 +HF – R4 L3 + LF – +HF – Output C5 Fig.22. (above left) Component overlay for PE Mini-Monitor crossover on the Universal Passive Crossover PCB. (Note: the tweeter phase marking on the PCB is incorrect and should be reversed – ie, bottom right, swap ‘+’ and ‘–’ either side of ‘HF’ output. (It is correct for the LS3/5A version, which has an than £1, so it’s worth it. The modified circuit and resulting curve are shown in Fig.20a and Fig.20b. This feature is more important for third-order high-pass filters where the tweeter has no inductor across it or other means to damp the resonance. There is an option for the tweeter to have ferrofluid in its magnetic gap, which provides high damping and power handling. I don’t like it, because the oil dries/thickens after a few years and sounds bad at low levels. Final crossover circuit This is shown in Fig.21. There are a lot of ‘Rs and Cs’, but the minimum of inductors. I asked Volt Loudspeakers in Devon to wind these and was gratified to find their DC resistance was lower than they originally specified and their tolerance was within a few percent. Final frequency response Building the crossover on the Universal PCB makes a refreshing change from the normal microscopic components of today. Fig.22 shows the overlay and Fig.23 a photo of the completed unit. Plated-through-holes with leaded components gives maximum strength, ideal for a crossover being shaken about inside a loudspeaker. It’s a good idea to wind a piece of insulation tape round the inductor where the termination wire is clamped under the cable tie. This prevents a shorted turn forming from abrasion of the enamel arising from vibration. Hot glue, foam and electronic grade (acid-free) silicone can reduce possible buzzes. 50 inverted tweeter connection, but the tweeter for the PE MiniMonitor is wired in-phase.) Also note the two central links need to be soldered to the board. Fig.23a. (above right) shows the completed PE Mini-Monitor crossover (do include the R3 wire link!). 23b (below) Note the vertical mounting for R9. Parts list Notes n T he following list is per crossover – two sets needed for a pair! n A ll components are ±5% tolerance or better. The most important thing is that the components on left and right crossovers are matched for good stereo imaging. So if, for example, you have two 10µF capacitors that are 10% too high, put them in the same position on each board. n ‘WW’ means wire-wound. Resistors R2 22Ω 6W WW R4 5.6 to 12Ω 2.5W WW, select value according to taste to set tweeter level R5 270Ω 0.5W carbon-film R7 3.3Ω 2.5W WW R9 56Ω 2W metal-oxide or WW Capacitors Note the Universal Passive Crossover PCB has multiple holes, so constructos can use either radial or axial types For plastic film capacitors, use polypropylene (MKP), polycarbonate (MKC) or polyester (MKT) – in that order of preference – rate at 63V minimum. C2a, C6 6.8µF C2b, C8 10µF C3a 680nF C4c 3.3µF C5 5.6µF from Blue Aran – part no. CVMPC25560 (or parallel combination: 4.7µF and 820nF) C7 220nF Inductors L1 1.2mH 0.23Ω 1mm WW on 12.5 x 50mm core (Neosid F6 manganese ferrite rod part number 36-702-26). Volt part number F002. L2 2mH 0.25Ω same construction as L1. Volt part number F020 L3 0.22mH 0.3Ω 9mm or 12mm x 25mm long ferrite. Alternatively, use an aircore type, 0.21mH 0.4Ω (Volt type 72). L1, L2, L3 (ferrite core) coils made for PE by Volt Loudspeakers are available as a complete set, with the resistors and capacitors from the PE shop – Part WAVXO. Miscellaneous Cable ties 3mm by 100mm, 3off Universal Passive Crossover PCB from the PE Shop – Part UPC0320 2mm turret tags, 12 off 22swg tinned copper wire for links, 100mm 19mm-wide PVC insulation tape, 150mm Practical Electronics | March | 2020 R5 270Ω 1W Input from amplifier + R4 10Ω 3W Notes *C5 and C7 are unchanged Low pass section is unchanged PCB needs a cut to use autotransformer C7* 120nF C5* 3.9µF Outer winding finshes at 0.4mH C8 10µF R7 3.3Ω 2.5 or 3W C6 6.8µF Bespoke ferrite autotransformer inductor from Volt, 0.3Ω 0.22mH tap Inner winding start Wavecor Tweeter TW022WA04 4Ω + 0V Fig.24. New high-pass section using auto transformer. 50Ω 40Ω 30Ω 20Ω 10Ω 0Ω 20 Hz 50 100 200 500 1k 2k 5k 10k 20k Fig.25. Impedance of the auto-transformer version, significantly higher in the top-end. Fig.27. The wire from the outer winding of the tapped coil is fed into the isolated track using the spare capacitor hole. Autotransformer – future upgrade I have commissioned Volt to wind some tapped inductors to implement the higher-impedance approach used in the LS3/5A crossover. This helps avoid the loss of efficiency in matching a 4Ω tweeter to an 8Ω woofer. These are 0.4mH tapped at 0.22mH. The new high-pass section is shown in Fig.24 and because the input impedance is higher, the first capacitor C5 has to be reduced to 3.9µF (mouser.co.uk part ECWFD2W395K) The aB capacitor value C7 also has to be altered to 120nF (available from Tayda). The response curve is the same, Fig.26. To use the auto transformer the PCB needs the track to be isolated by cutting as shown. This will be fully covered in a future article. R3 C4a C3a R2 Final edit One of the ‘perils’ of CAD PCB design is that it is easy to change and modify Fig.28. (below-left) Overlay for the new board layout with the low frequency coils (L1 and L2) at right angles. Also note that the positions of capacitors C1a and C2b have changed. The (Fig.22) note about the tweeter output also applies here. Fig.29. (below-right) Photo of the new board with the auto-transformer. C4b C7 C3b however the impedance curve shown in Fig.25 is now 4Ω higher in the treble region, reducing amplifier distortion. There is an important construction detail to take note of when this part becomes available: the PCB will need to be cut, see Fig.26. The autotransformer wiring is shown in Fig.27: feeding the coil from one of the spare capacitor (C5) pads. This will be fully covered in a later article. R7 C4c L2 C8 R5 C2a R9 Link C6 C1a C2b C1b R1 Input + LF – C9 R8 R6 L1 +HF – Practical Electronics | March | 2020 Tap R4 Outer coil end L3 Inner coil end + LF – +HF – Output C5 51 0dB –5 –10 –15 –20 20 Hz 50 100 200 500 1k 2k 5k 10k 20k Fig.31. Final frequency response using the tweaked crossover on a two foot open stand in a semi-anechoic living room. A bit of BBC style ‘voicing’ to compensate for the small size has been applied. Notice the slight bass hump to compensate for the lack of deep bass. There is a slight dip at the crossover to compensate for the sound being too ‘forward’ for close monitoring. Also, there is a high-frequency gentle rise on axis to compensate for the reduction in dispersion off axis. This response is not as smooth as the original curve (Fig.16b, see last month) but in fact, it ‘sounds’ smoother. Fig.30. The multi-pad connections can be used to parallel components for tolerance adjustment or making up values. In this case C5 (5.6µF) has been made up of a ‘high’ 4.7µF and a 680nF. things very easily. I felt uneasy with the two inductors (L1 and L2) in the low-frequency section being in-line because it says in every text book ‘don’t do it, put them at right angles’. This advice mainly applies at high frequencies, not the low frequencies involved here. Also, the overall magnetic circuit path length was quite long. The end result being I could measure no difference, but I changed the board anyway. The only other things that had to change were the positions of C1a and C2b. Fig.28 shows the overlay of the revised board and Fig.29 a photo of the completed board with autotransformer L3 installed. 0dB –5 –10 –15 –20 20 Hz 50 100 200 500 1k 2k 5k 10k 20k Fig.32. Crossover curve showing the woofer and tweeter outputs separately. The woofer slope is around −15dB per octave and the tweeter nearer −24dB/oct. The slopes are monotonic, that is they go down continuously without jumping up again. The different rates of attenuation are desirable because of the difference in directivity for each driver at the crossover point. (The woofer is beginning to ‘beam’ and the tweeter is almost omni-directional). 0dB Your best bet since MAPLIN Chock-a-Block with Stock Visit: www.cricklewoodelectronics.com Or phone our friendly knowledgeable staff on 020 8452 0161 Components • Audio • Video • Connectors • Cables Arduino • Test Equipment etc, etc –5 –10 –15 –20 20 Hz 50 100 200 500 1k 2k 5k 10k 20k Fig.33. The response with the tweeter connected anti-phase relative to the woofer. It is deep and symmetrical, indicative of good phase control through the crossover region. If you can’t hear this I suggest you find an audiologist. Another interesting component insertion technique is optimisation of capacitor values by paralleling components: mounting them one on top of the other, as shown in Fig.30. This is particularly handy if either you need to match a pair of values on – for example a pair of crossover boards – or, if you want create an unobtainable component that lies between two E-series values; for example, 4µF is within 0.5% of 3.3µF + 680nF. Final listening Visit our Shop, Call or Buy online at: www.cricklewoodelectronics.com 020 8452 0161 52 Visit our shop at: 40-42 Cricklewood Broadway London NW2 3ET The final frequency response, incorporating the tweaks is shown in Fig.31. The separate curves for both drivers are shown in Fig.32. The anti-phase connection is shown in Fig.33. I may be biased, but I have just had a good session with the PE Mini-Monitors and I think they give superb clarity and spatial imagery for the price (around £300). I was listening to Goldfrapp and it brought tears to my eyes, a true indicator of effective audio design. (Here I am even more biased – Alison Goldfrapp has bought my theremins!) Practical Electronics | March | 2020