Fullscreen HTML5Med Res (??MB)HTML4

The miniature 100W Hummingbird Amplifier from December 2021 has been popular, but some of the parts used in that design are now obsolete. This improved version features several minor improvements, plus the ability to use a wide range of transistors in several roles. 100W Calliope Amplifier by Phil Prosser T his amplifier is founded on a classic design made famous by Douglas Self as the ‘blameless amplifier’, and is reliable and powerful, especially for its size. As with many things in the electronics field, parts change and go obsolete. Since it is useful and popular, we thought it was worth updating to use currently available devices, with as many alternatives as possible; in case something else becomes hard to get... While at it, we have made a few optimisations to the layout, which at least technically improves the performance of this amplifier. Still, the performance of the original amplifier was very good, so we consider these minor upgrades. Part of the popularity of the Hummingbird amplifier module is that it that packs a surprising punch for its size, while keeping the low-­distortion characteristics of the Ultra-LD Amplifiers from which it takes inspiration. It can achieve up to 60W into 8W or 100W into 4W with distortion below 0.0005% at 1kHz, and less than 0.004% all the way up to 20kHz. That’s way better than “CD quality”. This project is more about the process of dealing with obsolescence, validating the changes and some discussion on the measurements we made. For an in-depth explanation of the design itself, refer to the original article (siliconchip.au/Article/15126). If you are in for a truly deep dive into amplifier design, look up Douglas Self’s books, especially the Power Amplifier Design Handbook that we reviewed in the March 2010 issue (siliconchip.au/Article/89). It is still very much relevant more than 15 years later. Maintaining and supporting this design is a balancing act between making necessary changes and adding improvements while maintaining both physical and performance compatibility with the original design. The part that triggered this update is the KSC3503 transistor (Q14) used in the voltage amplification stage (VAS), between the input differential pair and Scopes 1 & 2: a Calliope amplifier driven into clipping with a KSC3503 transistor as the VAS compared to another Calliope board with an MJE340 for the VAS (right). With the MJE340, it spends quite a long time ‘stuck’ to the negative rail. For high-frequency signals, this can get very ugly. 66 Silicon Chip Australia's electronics magazine siliconchip.com.au the output stage. This does not look that critical, but it provides the bulk of the voltage gain in the amplifier. How this transistor behaves as it enters and leaves saturation, which occurs when the amplifier clips, is an important but unusual requirement. For a transistor to do well as a VAS device, it needs to handle high voltages, and its Cob (output capacitance) needs to be extremely low, ideally only a couple of picofarads. In the old days, this was an easy thing to find, as the video stage in cathode ray tube (CRT) television sets demanded similar specifications. These are broadly: ● Output capacitance (Cob): very low, ideally < 3pF ● Collector-emitter voltage (BVceo): >150V ● Current gain bandwidth product (Ft): ideally in the 50-100MHz region ● Collector current (DC): >50mA ● Power dissipation: >1W When we saw the KSC3503 was marked as ‘end of life’, we went looking for an alternative part. With the passing of the days of CRT monitors, there is little reason for manufacturers to produce this type of device. We knew this would not be a simple change, as we looked for good VAS transistors during the original design. After days of searching suppliers and audio forums, it started looking like the only option was a surface-­ mounting device. Were we being silly? Should we care that much? What happens if we just throw a bog standard MJE340 into the board instead? Well, the amplifier will work, but you take a serious hit on how it behaves around clipping, and in general performance and stability, as shown in Scopes 1 & 2. Less critical but also important for the future, the range of output devices and drivers has changed. We bought a selection of different parts and validated them, so you can confidently build this with one of numerous parts options. Of particular interest to us are new output device options (see Table 1 overleaf) and driver transistor options (Table 2). As can be seen in Table 3, ideal options for the VAS are pretty limited, and there is no ideal part that is readily available in a through-hole package. There are some decent choices in SOT-223 (a fairly large SMD package), though. So we had to bite the bullet and make board changes to accommodate SMD parts for the VAS. To test the VAS devices, we set up a Calliope on our test jig (Photo 1), set the input to 0.9V RMS and connected a 3W load. We then tested each VAS device, one after another, without changing any settings or moving cables. This setup might look agricultural, but has a fair bit of thought in it and allows an objective comparison between modules. The test jig allows rapid changeover between modules and allows us to change components in situ, as the base is narrow enough to access the rear of the boards. It includes enough local filtering to let us run it from a bench DC supply, so there are no mains voltages involved, and we have a 6A current limit on our power supply. Despite that, it is beefy enough to allow us to run the amplifiers hard. The heatsink is undersized, and gets very hot on long tests, but this is handy, as we want to know how the various versions behave when abused. Amplifier Features Accepts a wide range of easy-toget parts (especially transistors) Low distortion and noise Extremely compact PCB Fits vertically on a 75mm heatsink, and can be stacked in a 2RU case Produces the specified output continuously with passive cooling Uses all through-hole parts (or optionally, one SMD for the VAS transistor) Low in cost and simple to build Onboard DC fuses Output over-current and shortcircuit protection Clean overload recovery with minimal ringing Improved tolerance of hum and EMI fields through small size and improved layout Quiescent current adjustment with temperature compensation Amplifier Specifications Output Power (with ±34V DC rails): 100W RMS into 4Ω, 60W RMS into 8Ω Frequency response (-3dB): 1Hz to 150kHz Signal-to-noise ratio (SNR): 118dB with respect to 50W into 4Ω Input Sensitivity: 1.2V RMS for 60W into 8Ω, 1.04V RMS for 100W into 4Ω Input impedance: 22kΩ || 1nF Total harmonic distortion (THD) (32W into 8Ω, ±32V DC): < 0.005%, 20Hz to 20kHz, 50kHz bandwidth Photo 1: our test rig made it really easy to measure the difference in performance as we changed various transistors on the board. You can see the difference between the correct wiring on the right (cyan curve in Fig.2) and the incorrect wiring on the left (mauve curve). siliconchip.com.au Stability: unconditional with any normal speaker load ≥4Ω Power Supply: ±20-40V DC; ideally ±34V DC from a 25-0-25V AC transformer Quiescent current: 53mA nominal Quiescent power: 4W nominal Output offset: typically <20mV (measured) Australia's electronics magazine April 2026  67 Table 1 – output transistor options (available from Mouser, DigiKey & element14) Device Key characteristics Comment MJW21193G/ MJW21194G High gain, optimised for linearity A new version of a ‘bulletproof’ favourite MJW21195G/ MJW21196G High gain, optimised for linearity A new, higher-voltage version of a ‘bulletproof’ favourite MJL21195G/ MJL21196G Same as above but in a slightly different package TTA1943Q/TTC5200Q Targets audio or 2SA1943/2SC5200 applications – linear gain A cost-effective version of another old favourite – higher Hfe and Ft NJW0281G/ NJW0302G Lower power versions of a ‘standard’ Optimised for match and linearity – lower power MJL21193G/ MJL21194G Known good MJL3281A/ MJL1302A Known good NJW21193G/ NJW21194G Known good 2SC5242/2SA1962 Known good Table 2 – driver transistor options (available from Mouser, DigiKey & element14) Device Key characteristics Comment MJF15030G/MJF15031G Insulated-tab versions of standard output drivers Known good MJE15030G/MJE15031G Known good MJE15032G/MJE15033G Known good Table 3 – VAS transistor options (most available from Mouser, DigiKey & element14) Device Package Key characteristics Comment KSC2690A TO-126 A higher Cob than preferred Works, but with some sticking on clipping KSC1845FTA TO-92 On the edge with power Do not use at high handling voltages KSC3503 TO-126 Obsolete/hard to find Known good, if you can get them 2N6517TA TO-92 Higher Cob than preferred Works fine but do not use at elevated voltages BSP19-115 SOT-223 70MHz Ft Prefer BF720/722 BF720 SOT-223 60MHz Ft Works fine BF722 SOT-223 60MHz Ft Works fine (preferred) PZTA42-TP SOT-223 50MHz Ft Assumed OK given DZTA tests DZTA42-13 SOT-223 50MHz Ft Works fine (almost surprisingly well) 2SC2911 TO-126 Obsolete/hard to find A good choice if you can get them 2SC3416 TO-126 Obsolete/hard to find A good choice if you can get them BF469 TO-126 Obsolete/hard to find An outstanding part if you can get them 68 Silicon Chip Australia's electronics magazine We found a handful of BF469 transistors that had been gathering dust. We don’t use obsolete parts in projects, so they are not specified here, but it is an old-school legendary VAS driver. Since we had the test jig up and running, we dropped one in to compare to all the other choices. It did very well, so if you have a few BF469s and are building this amplifier, we suggest you use them. Load line curves, which show the safety margins for the various output transistor options with 4W & 8W reactive loads (simulating typical loudspeakers) are shown in Figs.1(a) & 1(b). You can see that all the options are more than good enough for 8W loads, apart from the venerable TIP35/36, which are marginal but work if you limit the supply rail voltages. The recommended devices for 4W loudspeakers are shown in Fig.1(b) and they are all suitable. Fig.1(c) shows how the single-slope load line protection curves compare to the SOA curves for a selection of output devices. The circuit is designed to limit the devices to stay under the dashed lines, which are fully within the respective SOA curves, providing complete protection. There is no need to actually change the general design, as the ‘blameless’ configuration is known to be good and in use in many amplifiers around the world. However, while we were at the computer shuffling parts around, it gave us the opportunity to make some changes that have been on our ‘to-do list’ for some time. The changes we have made to the PCB layout are: ● Moving the driver transistors from the middle of the board to right next to the output devices, and moving the bias setting potentiometer and associated parts. This reduces routing complexity and provides better thermal coupling of the driver transistors to the output stage, which will improve thermal stability. The better routing should improve high-­ frequency performance, although this is not something we could measure. ● We made room for a VAS transistor in an SMD SOT-223 package, with a modest PCB heatsink area, while keeping the option to use a TO-126 through-hole device. We also tried a few different VAS compensation schemes, in particular, two-pole compensation as used on the Ultra-LD Mk.3 & Mk.4 Amplifiers. siliconchip.com.au Photo 2: the Calliope (Hummingbird Mk2) Amplifier module. We could measure a difference, but with this amplifier being so squished, we saw more effect from wiring layout changes than the compensation change delivered. So we used the KISS principle on this and went back to single-pole compensation. The final board is shown in Photo 2. Performance Fig.2 shows total harmonic distortion plus noise (THD+N) measurements for the new amplifier into a 9W resistive load at 32W. This is what we could generate using our bench supply, but it is representative of what will be achieved at higher supply voltages into normal loudspeaker loads. The performance is essentially the same as the original Hummingbird. If you are comparing this to what we published in the December 2021 issue, it’s important to realise that this plot includes noise, whereas the earlier one was THD only, so that plot showed lower figures. As with the original design, we have tested the distortion of a range of output devices, VAS transistors and output drivers to ensure it behaves well with all of them. An important test for an audio amplifier is how it behaves coming out of clipping, especially with low impedance loads. If you have the wrong VAS device, it will ‘stick’ to the negative rail, and if the output stage has very high-frequency devices, you can find bursts of oscillation near negative rail clipping. Scope 3 shows the behaviour when driving a worst-case 3W load. Figs.1(a)-(c): SOA curves and load lines for the various output device options with 4W & 8W reactive (loudspeakerlike) loads. The dashed lines in Fig.1(c) show the protection lines that the circuit prevents the devices from exceeding, which remain within their safe operating areas (SOA). siliconchip.com.au Australia's electronics magazine April 2026  69 We also have checked the squarewave behaviour. The Calliope is very well behaved with what is essentially a band-limited square wave output. We tested the harmonics generated by the Calliope using a high-quality audio spectrum analyser and found they are very low in level. All harmonics are in the region of -110dBc (0.000316%) to -115dBc (0.000177%). Circuit details Scope 3: clipping with the recommended BF722 VAS. The slight sticking to the rails is normal; the amplifier recovers from saturation at a high output current without ringing or oscillation. Fig.2: three THD+N plots for the Calliope amplifier module at a reasonably high operating power into a resistive dummy load. The cyan curve shows what you can expect if you follow our instructions, while the mauve curve is what you get if you don’t route the output wire as suggested. The red curve shows the result of all the transistors being mismatched. We will not go into a detailed description of the circuit, as the one in the December 2021 issue still applies. This article focuses on building the updated amplifier. You can download or purchase this article from our website if you don’t have a copy and want the detailed design description. The Hummingbird and Calliope amplifiers are physically much smaller than those in the Ultra-LD series, but a review of the circuit (shown in Fig.3) reveals that it is very similar, with the major difference being that Hummingbird/Calliope uses only one pair of output devices (to handle up to 100W) instead of two (up to 200W), and is optimised for operation at the lower voltages that implies. There are three main stages in a ‘blameless amplifier’. These are all described in detail in the original article: 1. The input stage, which uses Q7 & Q8 as a differential pair having a constant-current source (Q3) and a current-­mirror load (Q15 & Q16). 2. The voltage amplifier stage (VAS), comprising Q14 driven by emitter-­ follower Q13 and loaded by constant-­ current source Q2. 3. The output stage, which comprises transistors Q4/Q5 & Q11/Q12, plus protection devices Q6/Q10. This is a conventional complementary output stage. While the Hummingbird Mk1 and Calliope circuits are very similar, if Table 4 – protection resistor values for various output devices NPN device PNP device 22kW W 560W W 220W W Comments MJW21194G MJW21193G 22kW 560W 220W Performs as presented NJW21194G NJW21193G 22kW 560W 220W Performs as presented MJL21194 MJL21193 22kW 560W 220W Performs as presented 2SC5242 2SA1962 15kW 470W 220W Limit to 25V AC transformer if driving difficult 4W loads 2SC5200 2SA1943 12kW 560W 180W Performs as presented TTC5200Q TTA1943Q 12kW 560W 180W Essentially the same performance MJL3281A MJL1302A 15kW 560W 180W Essentially the same performance TIP35B/C TIP36B/C 10kW 680W 180W Limit to 25V AC transformer, prefer 8W load; good performance TIP3055 70 Silicon CTIP2955 hip 12kW 680W Australia's 270W Limit to 25V AC transformer & 8W loads; not verified electronics magazine siliconchip.com.au you compare Figs.4 & 5, you will see how much the layout has changed. Note how the new PCB accommodates either a through-hole or SMD package transistor Q14. We have thoroughly tested the various output transistor, driver transistor and VAS options. We have stuck to the MJE150XX family of driver transistors because they are available, robust and perform well. Any of the devices in this series will do as long as you use the complementary NPN and PNP types. The layout changes have shifted the drivers and resulted in a new layout for the whole top half of the board. This provides better thermal coupling of the drivers to the output devices, and reduces the length of traces with the relatively high base drive current for the output devices, improving stability. Should you be using this amplifier in a very demanding application, there is still room to mount small heatsinks to the output driver transistors. We have kept all mechanical features the same between versions, so if you need to mix and match or replace Hummingbird and Calliope amplifier modules, everything will drop in. We have kept the over-current/safe operating area (SOA) protection for the output devices. This provides protection if you connect a really horrible load or somebody abuses the amplifier. The Hummingbird amplifier delivers the measured performance with the parts specified, but we have checked that it works properly with a range of other parts. For different output devices, change the protection resistor values as per Table 4. An amplifier using a dual 25-30V AC output transformer, diode bridge and capacitor bank will have ±35-42V DC rails, which is safe operating into 4W, 6W and 8W loads. This will deliver 60W into an 8W load and 100W into a 4W load. Component matching Part selection for the Calliope amplifier should be fairly straightforward. We have provided tables of tested parts. Provided you use complementary pairs for the output devices and drivers, and select matched pairs for the input differential amplifier and current mirror, you will be fine. The output pair; for example, NJW21193/NJW21194, and the siliconchip.com.au Challenges with measuring low distortion levels Measuring very low levels of distortion is a lot harder than it might seem. There are a few reasons for this, including: 1. Generating a sinewave test signal that is pure enough to measure distortion at levels below 0.001% reliably is hard, especially if you want to vary the frequency. 2. Knowing if what you are measuring is your measurement system or the device under test is also tricky. 3. Even a tiny bit of EMI pickup can make huge differences in the measurements (simply rotating or moving the DUT can make the readings change massively). 4. Depending on how you Earth the DUT, you can measure voltages induced across ground wires if you are not very careful. For #1, we ended up using a Stanford Research Systems DS360 Ultra Low Distortion Function Generator. To measure the Calliope amplifier’s output harmonics, we are using a high-quality sound card/ADC that requires a line level signal, so we use a 2.2kW/120W resistive divider across the amplifier’s output, in parallel with our dummy load. If we feed the test signal back into the ADC directly, we get a THD reading of 0.00017%, so anything higher than that means we are measuring the amplifier’s distortion. While making amplifier measurements, we got a distortion reading of 0.0018%. While that is still not very high, it’s an order of magnitude higher than we were expecting. After a lot of fiddling, we realised that it depends on which of the two screw terminals of CON4 we make the measurement at! Connecting our measurement system to the unused terminal of the output connector gave a lower distortion reading than the one that is carrying the current (Photo 3). The only differences we see between these two points are: 1. We had to move the measurement probe and cabling slightly, which will pick up different magnetic fields. 2. The output current is going through the PCB-to-connector junction and the connector-to-output wire junction (this is probably the reason, as dissimilar metal junctions can be non-linear). The point of mentioning this is that, when you are aiming for very low distortion, all sorts of second order things start to matter, such as: 1. Earthing and where currents flow (probably the most significant concern) 2. Wiring layout, the magnetic fields the wires produce, and what they can couple into 3. The linearity of loads; we have seen wound Nichrome resistors cause significant problems 4. The types of connector used; our terminals are made of steel, which we think may be a factor 5. The actual ability of measurement equipment Now you know why, in the wiring section, we recommend running the output wire up past inductor L1 and trimpot VR1 to join the supply wires running across the top of the board. Simply running this wire on the other side of L1 has a measurable impact on performance, shown by the difference between the cyan and red plots in Fig.2. The effect is real and repeatable; moving the wire increases the size of the positive rail current loop to the output, which is coupling the positive rail current into the input and VAS stages. This makes the distortion worse across most of the audible frequency range. Photo 3: the two screw terminals on the output connector are both soldered to the output track; the left one is carrying the output current, the right none. We measured 0.0018% distortion on the left, 0.0007% on the right. Australia's electronics magazine April 2026  71 Fig.3: the Calliope circuit is intentionally similar to the original Hummingbird; there are a few subtle tweaks, but most of the improvements are in the PCB layout and expanded transistor choices. While we have nominated NJL21193/4, any of the MJL, MJL, NJL or NJW prefix series with the same numbers will work pretty much identically. Photo 4: the FrankenAmp in all its glory! The input transistors are all different parts, as are the drivers and output pair. Do not do this in your build unless you are truly desperate. drivers, MJE15032/MJE15033 are manufactured to have characteristics such that the NPN and PNP characteristics reflect one another. This reduces distortion when used in an amplifier of this type. The input differential pair, Q7/Q8, does the heavy lifting in making sure that the error in the amplifier output (ie, distortion) is minimised. It also plays a very important role in making sure there is no DC offset. These transistors should be the same as best we can match them, and ideally, thermally coupled. The current mirror, Q15/Q16, keeps the input differential pair in balance and provides gain. These transistors should also ideally be matched and thermally coupled. Feel free to choose pretty much any pairs from the table; match those input transistors and you will be good. At this point, a question arose in my mind: what if you get it really wrong? I couldn’t resist the temptation, so out of my fervid imagination comes the FrankenAmp (Photo 4). In this unit, every single part that can be mismatched was mismatched. It isn’t just a matter of using BC556s from different batches, either; in the FrankenAmp, the input differential pair is a BC558 and a BC556B, the current mirror uses a BC549 and a BC546, and so on. The drivers and output devices are from completely different families. How bad could it be? Fig.4: the original Hummingbird Amplifier PCB layout, shown for comparison to Fig.5. 72 Silicon Chip Australia's electronics magazine Unsurprisingly, the DC offset was terrible, at 140mV. This is because the gain of the input devices is grossly mismatched. Despite this, the amplifier is totally stable and even behaves OK on clipping. The distortion performance is even quite reasonable, as shown in Fig.2! (BD139s are notoriously different between manufacturers; it is likely I used an old Philips one, better than most). So even if you get it really wrong, as long as the DC offset is acceptable, the amp will work quite well. Construction Construction of the Calliope amplifier is pretty easy. It is built on a 63 × 86.5mm double-sided PCB that’s coded 01111212 – see Fig.5. First, based on the output devices you will be using, select the required resistor values from Table 4. These components are shown in red in Fig.5. If you read those same values off Table 4, build it as per our diagrams. siliconchip.com.au Otherwise, substitute the resistors with values shown in red for the different values in the table. After fitting those, install all the other small (¼W resistors). Follow with the 1N4148/1N914 diodes, making sure they are orientated as per Fig.5. Follow by fitting all the capacitors, soldering the smaller ceramic and MKT types first, then the electrolytics. Make sure that the electrolytic capacitors go in the right way around, with the longer (positive) leads to the pads marked +. If you are using an SMD VAS transistor, as recommended, fit it now, as there will be more room. Follow with all remaining transistors, except those that mount to the heatsink. Ensure that driver transistors Q4 and Q12 are installed with their metal tabs facing towards the amplifier input (ie, away from the output transistors). We want transistor pairs Q7/Q8 and Q15/Q16 to be thermally coupled with one another. Our approach is to superglue these face-to-back, then siliconchip.com.au put heatshrink tubing over them. You could glue them together after mounting them, as long as you mount their bodies reasonably close. If you can, select pairs for these devices with similar Hfe by measuring a handful of devices and choosing two that are similar. This can minimise the DC offset of the final amplifier. Now solder the fuse clips, making sure they go in the right way around, with the retention tabs on the outside. After that, solder all the connectors. The wire entries for the power terminal blocks go towards the edges of the board, while the output connector should have its wire entries facing towards nearby diodes D1 & D3. After that, you can mount the larger resistors (0.22W × 2, 4.7W & 15W) and the multi-turn potentiometer, VR1. We need to make sure the potentiometer starts at maximum resistance, so fit it with the screw located as shown, Fig.5: the new Calliope PCB layout moves the driver transistors (Q4 & Q12) closer to the output transistors (Q5 & Q11) and makes room for Q14 to be either in a vertically mounted through-hole (TO-126) package or an SMD (SOT-223) package. Australia's electronics magazine April 2026  73 Parts List – Calliope 100W Amplifier (per module) 1 double-sided PCB coded 01111212, 63 × 86.5mm 1 split rail power supply delivering ±20V to ±40V DC (15-28V AC mains transformer, bridge rectifier, filter capacitors, mains socket, mains-rated wiring, heatshrink etc) 3 2-way 5/5.08mm pitch mini terminal blocks (CON1, CON3, CON4) 1 2-way polarised/locking pin header (CON2) 4 M205 fuse clips (F1, F2) 2 5A M205 fast-blow ceramic fuses (F1, F2) [Altronics S5931] 1 1m length of 0.8mm diameter enamelled copper wire (L1) 1 500W vertical or side-adjust multi-turn trimpot (VR1) 2 TO-3P insulating kits (washers and bushes) 1 TO-126 insulating kit (washer and bush) 3 M3 × 25mm panhead machine screws 3 flat washers to suit M3 screws 3 crinkle washers to suit M3 screws 3 M3 hex nuts 2 blown M205 fuses (for testing, or purposefully blow 100mA fuses) 1 heatsink (we used one Altronics H0545 for six modules) 1 small tube of superglue 1 5cm length of masking tape Semiconductors 5 BC556 65V 100mA PNP transistors, TO-92 (Q1, Q3, Q7, Q8, Q10) 1 MJE350 300V 500mA PNP transistor, TO-126 (Q2) [Altronics Z1127, Jaycar ZT2260] 1 MJE15032G or MJE15034G 250V/350V 8A NPN transistor, TO-220 (Q4) [element14 9556621, DigiKey MJE15034GOS-ND, Mouser 863-MJE15032G] 1 NJW21194G or MJL21194 250V 16A NPN transistor, TO-3P (Q5) [Jaycar ZT2228, element14 2535656, DigiKey NJW21194GOS-ND, Mouser 863-NJW21194G] 3 BC546 65V 100mA NPN transistors, TO-92 (Q6, Q13, Q17) 1 BD139 80V 1A NPN transistor, TO-126 (Q9) [Altronics Z1068, Jaycar ZT2189] 1 NJW21193G or MJL21193 250V 16A PNP transistor, TO-3P (Q11) [Jaycar ZT2227, element14 9555781, DigiKey NJW21193GOS-ND, Mouser 863-NJW21193G] 1 MJE15033G or MJE15035G 250V/350V 8A PNP transistor, TO-220 (Q12) [element14 9556630, DigiKey MJE15035GOS-ND, Mouser 863-MJE15033G] 1 BF722 250V 100mA NPN transistor, SOT-223 (Q14) [element14 1757916, DigiKey BF722,115, Mouser 771-BF722-T/R] 2 BC549 30V 100mA NPN transistors (Q15, Q16) 3 1N4148/1N914 75V 250mA small signal diodes (D1-D3) Capacitors 1 220μF 25V electrolytic [Altronics R5144, Jaycar RE6324] 4 100μF 50V 105°C electrolytic [Altronics R4827, Jaycar RE6346] 1 47μF 50V low-ESR electrolytic [Altronics R6107, Jaycar RE6344] 1 10μF 50V non-polarised electrolytic [Altronics R6560, Jaycar RY6810] 1 220nF 63V MKT [Altronics R3029B, Jaycar RM7145] 5 100nF 63V MKT [Altronics R3025B, Jaycar RM7125] 1 22nF 63V MKT [Altronics R3017B, Jaycar RM7085] 1 1nF 63V MKT [Altronics R3001B, Jaycar RM7010] 1 220pF 100V NP0/C0G ceramic [element14 1600858, DigiKey 56-K221J10C0GH5UH5CT-ND, Mouser 594-K221J15C0GH5TH5] Resistors (all ¼W+ 1% metal film axial unless otherwise stated) 1 220kW 1 82W 4 22kW ♦ 2 68W 2 3.9kW 2 47W ♦ 3 2.2kW 1 39W 1 1.2kW 1 15W 1W 5% 2 560W ♦ 1 10W 1 390W 2 10W 5W 10% (for testing) 4 220W ♦ 1 4.7W 1W 5% 6 100W ♦ 2 0.22W 5W 5% ♦ ♦ two of each may need to change in value depending on the output transistors used ♦ ½W or 0.6W 1% metal film ♦ element14 1735119, DigiKey BC3440CT-ND, Mouser 594-AC050002207JAC00 then rotate the screw anti-clockwise until it clicks. Verify with a multimeter that the resistance between its two outside terminals is below 25W. By the way, side-adjustment pots are better if you’re going to be mounting the amplifiers vertically on the heatsink, while a top-adjustment pot makes most sense if it will be mounted horizontally. Next, make the inductor using 0.8mm diameter enamelled copper wire (ECW) as follows: 1. Find a mandrel that is about 10mm diameter and has a slight chamfer to it so that, once complete, you will be able to slide the inductor off. We chose a large Sharpie marker. 2. Put masking tape around the mandrel with the sticky side faced outwards. 3. Placed a bend in the ECW 30-40mm from the end and wind nine turns onto the tape. 4. Put a few drops of superglue on the ECW; don’t worry if it gets on the masking tape. You do need to be careful not to get glue on your mandrel, though! 5. Give this a minute to set, then wind another layer on top of the first nine turns. You might only be able to get eight in; that is OK. Add more super glue and again, allow it to set. 6. Add a final winding layer and glue it. 7. Push the inductor off the mandrel. 8. Tease the masking tape from inside the inductor; we used needle-­ nosed pliers to do this. 9. Scrape the enamel off the leads and mount it to the PCB above the 4.7W resistor. At this point, the board should be complete bar Q5, Q9 and Q11. From here, you need to mount the output devices to their final heatsink using insulating kits. Then bend the legs of the transistors to match the PCB, as shown in the photos. Slip them into their respective holes on the PCB. The aim here is that these transistors fit reasonably well. Once the three transistors are properly inserted into the PCB, solder them in place. This way, we know that the transistors are mounted with minimal tension on the soldered connections, ensuring a long life of the solder joints (a solder joint under stress has a tendency to crack and go dry). At this point, you should have all Australia's electronics magazine siliconchip.com.au 74 Silicon Chip Photos 5 & 6: The finished Calliope Amplifier, and the test jig used for measurements. The output wire (brown in this case) should be pushed back to run next to the emitter resistor, like it does here. parts mounted to the board, and are ready to test it. Testing & adjustment Your amplifier is probably mounted to the heatsink, but the initial test can be done without it – just make sure that the bias current is set to minimum. This test will check the amplifier is operational: 1. Remove the normal 5A fuses from the board and install blown M205 fuses with 10W 5W resistors soldered across them (‘safety resistors’). 2. Connect a voltmeter between the output and ground, set to a 200V range or similar. 3. Connect a voltmeter across one of the 10W safety resistors, set on a 20V range or similar. If you only have one meter, run an initial check monitoring the output voltage only. 4. With the input to the module disconnected, apply power. Anything over about ±15V is fine. If you can, set the current limit on the power supply to about 100mA. 5. Check that the output voltage settles to 0V ±50mV. We built 14 test units, and all were well within this range. 6. Check that the voltage across the 10W resistor is less than 1V (ie, it’s drawing under 100mA). If either test fails, you need to check for the cause. Do you have VR1 set at Songbird An easy-to-build project that is perfect as a gift. SC6633 ($30 plus postage): Songbird Kit Choose from one of four colours for the PCB (purple, green, yellow or red). The kit includes nearly all parts, plus the piezo buzzer, 3D-printed piezo mount and switched battery box (base/stand not included). See the May 2023 issue for details: siliconchip.au/Article/15785 siliconchip.com.au Australia's electronics magazine April 2026  75 the right end of its travel? Are all the electrolytic capacitors in the right way around? Do you have the input connected? If so, disconnect it. Are all the transistors in the right places and the right way around? Check those output devices are in the right spot! Is your power supply delivering both positive and negative rails, and do you have the ground connected? Assuming it passes the test, it’s time to adjust the quiescent current and run a full operational test. This requires the amplifier to be mounted to a heatsink with appropriate insulators for the output devices and thermal sense transistor. Before powering it up, verify a high resistance between both power supply rails and the heatsink. Apply power and adjust the bias by turning the potentiometer clockwise while watching the voltage across the 10W resistor. Nothing will happen for quite a few turns, then the bias current will rapidly increase. Adjust it to achieve 500mV across the 10W resistor. Allow this to settle and readjust. It can take a while to settle, and should be set with the full rail voltage applied. Power it off, re-install the 5A fuses and you are ready to connect a loudspeaker and run it with an audio signal. You can check the bias (quiescent current) later by measuring the voltage across the 0.22W resistors; you should see close to 10mV across each. Installation Our earlier discussion on measurement pointed out the criticality of Fig.6: when finalising the amplifier wiring in your case, run the supply and output wire to each module like this to get the best performance. layout to get the most of the amplifier. Careful attention to layout and the power supply is required. The power wiring from the main supply capacitors should be delivered on twisted sets of positive, negative and ground wires. The output should fold back toward the output devices, and run parallel to the 0.22W resistors, then follow the power wires – see Fig.6. The output wire should follow the power wires back past the power supply and pick up a ground wire, minimising the loop area created, then run as a pair from there to the speaker terminals. Ensure that the power supply has a ‘star Earth point’ from which you connect to the input ground, the amplifier ground and the speaker output ground, as shown in Fig.7. Also make sure that the way you connect the rectifier and its ground to the capacitors does not inject noise onto your star Earth point. The input cable shields/screens should also be connected to the star Earth point. Make sure all connections are secure and low resistance; poor connections can easily more than double the distortion level. We found this measuring a batch of modules we built to verify our results, having to tighten the connections to achieve consistent results. Is it worth upgrading a Hummingbird to the Calliope? Not really. While the layout is improved, and we provide options for some more recent and ‘optimised’ parts, the Hummingbird performs pretty well. While the Calliope is an improvement overall, its main advantage is that it is more future-proof and easier to source parts for. We have a mix of both in use and are quite comSC fortable with this. Boosting the output power Fig.7: configure your amplifier power supply like this to keep the ripple currents from recharging the capacitor bank out of the amplifier ground lines. 76 Silicon Chip Australia's electronics magazine If you add extra output devices, you can within reason. But watch the ratings of your capacitors and input devices. If you want serious power, you should consider the SC200 (January-March 2017; siliconchip. au/Series/308), which gives roughly double the output power. Otherwise, the Ultra-LD Mk.3/4 Amplifiers (July-September 2011 & August-October 2015) will give you roughly the same power as the SC200 but with lower distortion. siliconchip.com.au