Fullscreen HTML5Med Res (??MB)HTML4

Restoring the QUAD 303 Power Amplifier and Preamplifier A vintage hifi article by Jim Greig The QUAD 303 amplifier and associated QUAD 33 preamplifier (that they call a “control unit”) were introduced in 1967 and sold until 1985. These units belong to fellow HRSA member Ray Thomas. I was very happy to exchange some time refurbishing for a chance to listen to a classic QUAD amplifier. T he specifications of this equipment are ordinary by today’s standards but compare very well with the valve amplifiers of the time. While the output impedance is specified as the emitter resistance (0.3W), the negative feedback across these resistors will reduce it. However, the output filter and series capacitor will increase it somewhat. Power to the amplifier is from a single-ended, regulated 67V supply. Amplifier circuitry The amplifier circuit is broadly similar to a modern ‘blameless’ amplifier circuit in many ways, with a complementary emitter-follower output buffer and a ‘voltage amplification stage’ or VAS based on NPN transistor TR102 – see Fig.1. The main difference is in the input stage and feedback system, which doesn’t use the balanced, symmetrical two-transistor input that’s common today. The amplifier has quasi complementary symmetry output with transistor 92 Silicon Chip triples (TR104, TR106, TR2) to simulate a PNP transistor and provide a linear (through local feedback R120) transistor equivalent for both the PNP and NPN ‘transistors’. With this technique, ordinary 2N3055 transistors act as superior PNP and NPN devices. Diodes MR105 (NPN) and MR106 (PNP) protect the output transistors from overcurrent. On the NPN (upper) side, when the current through R123 approaches 4.3A, MR105 conducts, driving the base of TR103 more positive. As a result, the current through it decreases, cutting off TR105 and decreasing the drive to the output transistor, TR1. The VAS transistor, TR103, has a quasi constant-current load based on resistors R116 and R117 plus capacitor C106. The current available to the base of TR103 would decrease as the Any work on this unit should be done with the mains disconnected, as there are exposed mains connections when the outside cover is removed. Australia's electronics magazine output moved towards the supply voltage if not for C106. As the output moves positive, C106 also takes the junction of R116 and R117 positive, ensuring there is voltage across R117 to drive TR103. This is known as bootstrapping. The output signal from the junction of R124 and R125 idles at half the supply voltage, so a coupling capacitor (C1) is required for the speaker output. That is somewhat frowned upon today as capacitors constitute a significant source of distortion. Still, it simplifies the design and would have resulted in a relatively large cost saving at the time. The amplifier is stabilised with a Zobel network (R128/C108) and series filter (R129/L100). This must have been a very early appearance of the Zobel network in a hifi amplifier to ensure a primarily resistive load to the amplifier, regardless of loudspeaker impedance changes with frequency. The driver stages are DC-­coupled common-emitter singles where siliconchip.com.au Fig.1: one channel of the power amplifier circuit from the QUAD 303 Power Amplifier Service Data manual. differential or long-tailed pairs would be utilised today. Overall DC negative feedback is through R113 and R108, with R130, R110 and RV100 forming a voltage divider for setting the output to half the supply voltage. AC feedback follows the same path, but the gain is limited by C104 shunting R111 to ground. The overall AC gain is 82kW/2.2kW or 37 times. So 0.5V RMS at the input is amplified to 18.5V RMS at the output, giving 43W into 8W. Trimmers allow the output idle DC voltage (RV100) and standby current (RV101) to be adjusted. The standby/ quiescent current is set using the same Vbe multiplier circuit still in use today, based around TR107. Power supply circuitry The power supply (Fig.2) is interesting because the regulator is in the negative rail. The cans of filter capacitors C2 and C3 must not touch ground. The supply is referenced to zener diode MR201 (16V). The zener and associated resistor R204 are connected across the stable 67V output to keep the current through it constant, for a more stable reference voltage. A fraction of the output from the divider formed by RV200, R202 and siliconchip.com.au R203 is compared with the reference voltage. TR200 and TR201 amplify the difference; the result is applied to the emitter-follower regulator, TR3. Trimmer RV200 adjusts the output voltage. R201/MR200 ensure that TR200 is conducting at switch-on, while R200 ensures that MR200 is back-biased and not active during regular operation. The power supply can be configured for 110/120/220/240V AC mains supplies using the external selector switch. A neon indicator glows when power is applied. The chassis The amplifier is elegantly crafted with a pressed steel chassis that has the Power, Input and Output connectors on one end and a heatsink on the other. The two amplifier and power supply circuit boards fit across the bottom of the chassis, held in place with QUAD 303 amplifier weighing 8.2kg: » » » » » » » » Power output: 2 × 45W into 8Ω Frequency response: 30Hz to 35kHz +0,-1dB Total harmonic distortion at 45W: 0.03% at 70Hz and 700 Hz; 0.1% at 10kHz Output source impedance 0.3Ω (+ output capacitor & Zobel network reactance) Hum and noise: 100dB below full output Inter-channel crosstalk: better than -60dB from 30Hz to 10kHz Input sensitivity: 0.5V RMS Speaker load impedance: 4-25Ω QUAD 33 preamplifier (“control unit”) weighing 3kg: » » » » » » » » Frequency response: 30Hz to 20kHz, ±0.5dB Total harmonic distortion: 0.02%, 30Hz-10kHz at all controls level; 0.5V RMS out Input sensitivity (RMS): 2mV (moving magnet), 100mV (ceramic), 100mV (line) Signal-to-noise ratio: 70dB (moving magnet), 85dB (line) Tone control: approximately ±16dB at 30Hz and 20kHz Filter: flat to -20dB per octave at 5kHz, 7.5kHz and 10kHz Inter-channel crosstalk: better than -40dB, 30Hz to 10kHz Output level (RMS): 100mV (line), 0.5V (Pre out) Australia's electronics magazine January 2024  93 Fig.2: the regulated DC power supply circuit, again from the QUAD 303 Power Amplifier Service Data manual. plastic clips. These are easily opened to allow the boards to be removed for service, and still function without breaking. This layout is tidy but necessitates long leads from the PCBs to the output transistors. They are neatly bundled and laced, giving the amplifier a professional appearance (see Photo 1). However, compared with today’s short leads following similar paths, the layout will limit performance. Still, we are in 1967, where 0.1% distortion is considered very good. QUAD 33 preamplifier The QUAD 33 preamplifier complements the appearance of the 303 amplifier. Appearance and construction are clearly design inputs. The preamplifier is built on a steel chassis and implemented on five modules plugged into a passive motherboard and filter board (see Photos 2 & 4). The modules are the Disc Adaptor, two Preamplifiers, Tape Adaptor and Right/Left Hand Amplifier. The Disc Adaptor provides matching for Low Output Magnet (M1), High Output Magnetic (M2), Ceramic (C1) or Spare (S1) inputs by connecting different components in the preamplifier input and feedback paths. The card that plugs in from the back of the unit provides the four functions, as it is square and can plug in one of four ways – see Photo 3. The preamplifier (one channel shown in Fig.3) has two DC-coupled BC109s with R313/R314, R310 and R302 providing DC feedback to stabilise the operating point. RIAA equalisation is provided by the Disc Adaptor using connector M2 and components C104 and R110/C101 from the output to the emitter of TR301. Capacitor C308 connected to the emitter of T301 ensures that the emitter side of R302 closely follows the AC input to the base. The signal current through R302 is then minimal, greatly increasing its apparent resistance (another form of bootstrapping). The amplified disc signal may be selected along with Radio 1, Radio 2 and Tape as inputs to the Tape Adaptor board. It has an emitter follower stage, with the full output passed to the volume control and jumper selectable full or partial output to the Tape Record connector. The outputs from the volume control feed the Left and Right Amplifiers, which drive the tone controls, balance control and filters. Fig.4 shows the tone control circuit, with the output level control circuitry at lower right. The input stage is an emitter follower driving the Baxandall tone control circuit. The output passes through the filter network. There is a top cut switch with -3db points around 5kHz, 7.5kHz, or 10kHz and a slope control (RV8), varying the response from flat to a steep cut at the selected frequency. The filters can be independently set on/off, and a cancel switch bypasses the tone controls and filters. The power supply is a simple zener-­ regulated 12V configuration. A second supply connected to the output plug Photo 1: the internals of the QUAD 303 amplifier are very neat, with multiple modules built on small PCBs, wired together very neatly with loomed wiring. 94 Silicon Chip Australia's electronics magazine siliconchip.com.au Fig.3: the preamplifier front-end circuit for one channel, an extract from the QUAD 33 Control Unit Service Data manual. The Disc Adaptor shown on the left can be inserted in one of four ways, effectively acting like a four-way switch to select the S1, C1, M1 or M2 connections to suit different signal sources. is not used, other than to power the indicator. The QUAD 33’s construction is of the same high standard as the 303 amplifier. For example, the balance control is a standard potentiometer operated by a mechanical link from a slider on the front panel. The outer cover slides off on rails attached to the chassis, and the wiring is neatly loomed. Restoration These units are over 50 years old and, if untouched, require attention. Several companies offer “upgrade kits” that contain new electrolytic capacitors, resistors and transistors. There are three versions of the power amplifier, and the upgrade kit must match your model. I used a kit from Dada Electronics (https://dadaelectronics.com.au & https://dadaelectronics.eu) to restore the amplifier, while I purchased parts separately for the preamplifier. The following steps are listed in the Dada documentation for QUAD 303 power amplifiers with serial numbers above 11500, to replace: 1. The two 2000μF filter and the 2000μF output coupling capacitors with three 4700μF capacitors. 2. All electrolytic capacitors on the power supply and amplifier boards with new electrolytic capacitors (plus one 0.68μF foil type). 3. All trimmer resistors. 4. All resistors on the power supply board. 5. Both diodes on the power supply board. 6. Some power supply cabling. The existing wires from the power transformer to the rectifier and on the filter capacitors are solid 24 gauge (about 0.5mm diameter & 0.25mm2). They look good, neatly bent to follow the components, but are inadequate by today’s standards. So I replaced them with 0.7mm diameter (0.5mm2) multicore cable. Photo 2 (right): on the rear panel of the QUAD 33 preamplifier, you can see the Disc Adaptor board that plugs in at lower right, the Tape Adaptor card to its left, plus the various inputs and output connectors. Photo 3 (above): the Disc Adaptor can be plugged in on any of its four sides, setting the preamp up for one of four different input signal types. siliconchip.com.au Australia's electronics magazine 95 Fig.4: the tone control, output level control and top-cut circuitry plus the final amplification stage of the preamplifier; another extract from the QUAD 33 Control Unit Service Data manual. The power-on indicator (Photo 5) is a neon bulb connected to the incoming mains supply through a 100kW resistor. This may be faulty; the recommended replacement is a square LED with a 12kW series resistor across the DC power supply; still, a new neon indicator could be used. Along with the indicator, there are other exposed mains conductors on the fuse and voltage selector on the rear panel, near the input connector, so care must be taken to avoid contact with them. I unclipped the PCBs but did not disconnect them from the cabling. I replaced the components following Photo 4: the QUAD 33 preamplifier internals are very neatly organised and laid out, with highly organised cable routing. 96 Silicon Chip standard procedures, carefully observing the polarity of diodes and electrolytic capacitors. The PCBs are old phenolic types; care must be taken when desoldering and soldering to avoid lifting tracks. For testing, the boards must be kept away from the chassis to avoid short circuits as the chassis is Earthed. As the history of this unit was unknown, I disconnected the positive rail from both amplifier boards before power-on and added a 3.3kW 5W resistor as a load. I then used a variac to gradually apply AC voltage while monitoring the DC output volts. The power supply’s regulated DC output voltage increased slowly and stabilised at around 70V DC. Adjusting the trimmer (RV200) reduced it to the desired 67V. However, the voltage decreased further; at least 10 minutes passed before it was stable. Power amplifier testing I switched the power off and connected the first amplifier board to the supply through a 100W ½W resistor as a fuse. I also connected 10W highpower resistors across the amplifier outputs as loads. I re-applied power and monitored the amplifier DC output voltage (5 on the PCB). Fig.5: the 1kHz square wave response of the QUAD 303 amplifier is very clean. Figs.6(a) & 6(b): the leading edge (left) of the QUAD 303 amp output with a 1kHz square wave fed in. The rise time allows us to calculate the time constant of the high-pass filter formed by the coupling stages throughout the amp. The trailing edge (right) of the same waveform indicates that the response is symmetrical. It steadily increased to 29V and nothing was getting hot. I then adjusted trimmer RV100 until the output was at 33.5V. The output current can be monitored by checking the voltage across both 0.3W emitter resistors (4-6 on the PCB). I adjusted VR101 to get 8mV (allowed range 6-9mV), corresponding to 13mA. This time, both the supply voltage and the amplifier settings were drifting, so I repeated the adjustments after 20 minutes. I removed power and connected the amplifier directly to the 67V supply, then wired up the second board via the 100W resistor. After verifying that it worked, I removed that resistor and I repeated the adjustment procedure for the second board, aiming to have the standby current in both channels the same. The upgrade instructions state, “use for some hours at normal volume and repeat the calibration”, so I followed that recommendation. It is interesting to compare the procedure with the setup for the Silicon Chip Class-A 20W Amplifier that was initially designed 25 years ago (in 1998, as a 15W version) and improved to 20W in the May-August 2007 issues (siliconchip.au/Series/58). While the Class-A amplifier power supply is unregulated, the separate positive and negative supplies with the input referenced to ground ensure the output voltage is close to zero regardless of voltage fluctuations. Also, the quiescent current (1A) remained stable after setup. After monitoring voltages for a while, I ran a few simple tests. The frequency response (-3dB) was from 10Hz to over 60kHz, measured with a digital oscilloscope. It achieved 19.9V RMS output just before clipping, and with a magnified trace, crossover distortion could not be seen on an oscilloscope. A 1kHz square wave output looked good with a reasonably flat response, no overshoot and rise and fall times around 10μs (see Figs.5 & 6). A 32Hz square wave showed significant low-frequency roll-off. However, a good frequency response does not necessarily translate to a good square wave response. I measured the time constant (time for the waveform to drop to 63% of the original value) on the CRO as around 15ms. The most likely cause of this time constant is the 0.68μF capacitor 22kW resistor in series at the amplifier input. The calculated time constant is T = RC = 68μF × 22kW = 15ms, in agreement with the measurement. I connected a 22μF electrolytic capacitor across the 0.68μF capacitor, and the response improved significantly, but it still was not flat. Rather than attempt to redesign the amplifier, I left the input coupling capacitor at 0.68μF. Restoring the preamplifier Photo 5: there are exposed mains connections on the front panel, including for the neon indicator. That indicator can be replaced with a modern LED powered by the DC supply. siliconchip.com.au Australia's electronics magazine It’s important that the preamplifier has decent performance since the power amplifier will amplify any noise and distortion it introduces. In the Dada procedure, the modifications are more extensive than for the power amplifier. The following steps are recommended: 1. Replace all electrolytic capacitors. 2. Replace all BC109 transistors with lower noise BC550 types. 3. Replace some resistors with metal film types for lower noise. 4. Change some resistors to alter the gain so CD players do not overdrive it. 5. Increase the supply voltage from 12V to 16V for more headroom and lower distortion. 6. Remove the secondary power supply as it is not used. January 2024  97 Fig.7: the 1kHz square wave response of the QUAD 33 preamp (bottom) is not as good as the QUAD 303 amp, with a noticeable shift in the level during what should be flat portions. Fig.8: the 100Hz square waves response of the QUAD 33 preamplifier is noticeably triangular. Fig.9: the preamp’s 32Hz square wave response degenerates into something barely recognisable as a square wave. I changed the components as recommended and rebuilt the secondary 8V power supply that powers the indicator. The indicator bulb showed signs of heat damage, so I replaced it with a high-intensity LED soldered to the old bulb metalwork. After changing the lamp type, I moved the power supply’s yellow (indicator) lead from position 2 (AC) to 5 (8V DC). I also changed R502 to 330W to limit the LED current to 18mA. I powered the preamp on without the modules connected and measured the DC supply voltage as 15.9V. As with the power amplifier, there is exposed mains wiring that must be covered while the lid is off. I reconnected the modules and commenced testing with 1kHz sinewaves into the Radio 2 input. As the output reached around 0.5V RMS, it dropped to almost zero. The output resistances of both channels had dropped from 4.7kW to around 80W and stayed there. After some checking, I determined that the low resistance was from the metal frame of the filter switches to the output. Powering it off and pushing switches cleared the problem, but it returned as soon as the output reached the critical value. Several articles mentioned that these switches can cause problems, so I cleaned them without removing them from the PCB. Removing them would be challenging, as the spring-loaded contacts are tiny and would pop out much more easily than they would go back in (see Photo 6). Unfortunately, the problem returned after cleaning, but only in one channel this time. As the low resistance path was to the switch frame, I sprayed the gap under the switches well with contact cleaner and then washed them out with isopropyl alcohol. After that, both channels worked correctly. This would likely have cleared the faults in both channels if applied earlier. What the substance was and why it was triggered into a low resistance state depending on the signal level, I do not know. I checked the signals through the preamp from the Radio input with the filter switch in Cancel (no filter and no tone controls). The sinewave response was -3dB from 15Hz to 220kHz. The square wave response at 1kHz shows signs of poor low-frequency response (Fig.7). At 100Hz, it is obvious (Fig.8), and at 32Hz, the response is horrible (Fig.9). It is the same on both channels, so presumably the original release had similar performance as I did not reduce any capacitor values in the preamp. Examining the amplifier more closely, the signal is losing shape as it arrives at the base of the first transistor, TR400. With the preamplifier in “Cancel” mode, the output from TR400 emitter is coupled via a 2.2μF capacitor and 5.6kW resistor in series (time constant = 6ms) to the base of TR401. TR401 and TR402 constitute a high-gain amplifier with the output returned to the inverting input, making the base of TR401 a ‘virtual earth’. Investigating further would involve breaking the feedback loop. The output has another 2.2μF coupling capacitor connected to a 4.7kW load resistor. Increasing the value of capacitors in this circuit would help the square wave response, but they are not the only factor. I was reluctant to change any of those capacitor values as such changes could have a flow-on effect elsewhere. After all, this is a refit to make the best of the existing unit with minimal changes, not a redesign. Note that the measurements and comments above apply to these modified units alone. Photo 6: one of the switches that caused so many problems by intermittently shorting the signal to the case. Presumably, some kind of conductive gunk had built up; a thorough cleaning finally sorted it out. 98 Silicon Chip Australia's electronics magazine Listening tests Any comments on the sound must acknowledge that my 70+ year old ears are not in great shape. The ‘test’ was listening to Fleetwood Mac’s Rumours on a Thorens/Ortofon Blue combination played through home-built threeway Vifa Speakers. For comparison, I used the 20W Class-A amplifier I mentioned earlier, the Magnetic Cartridge Preamplifier (August 2006; siliconchip.au/ Article/2740) and a two-linear-IC tone control network. This combination is at least 30 years younger, so the comparison is unfair, but it is my reference. I found the QUAD system acceptable but not as clear as my existing system. References (www.dadaelectronics. eu/downloads): QUAD 303 Power Amplifier Service Data and Instruction Book QUAD 33 Control Unit Service Data and Instruction Book QUAD 33-303 Service Supplement QUAD 303 all versions illustrated upgrade guidelines V2.0 QUAD 33 Revision – Illustrated Guidelines V2.7 SC siliconchip.com.au