Fullscreen HTML5Med Res (??MB)HTML4

AUDIO OUT AUDIO OUT L R By Jake Rothman Back to the buffers – Part 4 T his month, we’ll wrap up the series on discrete op amp/buffer designs. The complexity of a full discrete op amp is not really required for a buffer and the simple three-transistor circuits given (PE, April 2024, Fig.24, Fig.27) are usually sufficient. However, sometimes op amp characteristics, such as a low output offset are occasionally needed. The high open-loop gain of a full op amp (discrete or monolithic) also enables a very low output impedance to be obtained. There are some expensive dedicated buffer chips available, such as the LM302, which employs 18 transistors. So, I thought a ‘halfwayhouse’ discrete solution based on an op amp input stage with the voltage amplifier stage (VAS) omitted would be worthy of some R&D. Op amp buffer design My resulting design is a two-stage op amp, sometimes called a ‘Schlotzaur circuit’. It has an open-loop gain of around 6000 if constant current loads are used. To start, the discrete op amp board from the October 2023 article was used with a long-tailed pair (LTP) stage followed by a class-A emitter follower. A possible problem with this topology using 100% negative feedback, is that the operating level of the input stage is the same as the output. Indeed, this was a problem, with the distortion level increasing in proportion to the output voltage, as shown later. Development trials and tribulations Originally, the output stage of the first discrete op amp which I used as a buffer just consisted of an emitter follower (TR3) and current sink running at 10mA (TR7) – see Fig.35. This was adequate for high impedance loads, but not for full output into 600Ω. The solution was a three-transistor follower and a modulated current source. The first circuit tried was to simply connect the three-transistor buffer to the op amp input stage. A successful circuit design 58 strategy is to start by combining circuit blocks you already know work. If it doesn’t work, it’s easier to find the fault. There was a hard offset upon turn-on because I forgot that the omitted voltageamplifier stage was inverting. To get round this, the negative feedback had to be applied to the non-inverting terminal OA+ on the input stage. Guess what? It still didn’t work. Then I remembered the current mirror had to be turned round as well. One side of the mirror is output and the other, the diode-connected transistor side (TR5 on the original op amp circuit (PE, October 2023, Fig.15)), is for current sensing. Having done this necessary ‘flipping’, the circuit worked. Next, I needed to do some optimisation. Oscillation subtle low-frequency rise in THD that occurs with capacitors, for example. To fix the oscillation, I placed a capacitor (CComp) across the collectors of the long-tailed pair. I used 470pF out of the ‘mixed cap’ drawer which fixed the oscillation problem. I then checked for slewing distortion which occurred at a too low a frequency of 40kHz, so I then reduced the capacitor to 150pF – but then some other unwanted highfrequency signal appeared. Fortunately, this was not an oscillation, but 470kHz RF noise emitted from a new Metcal soldering iron I had recently started using. Unfortunately, this new problem also had me going round in circles for a couple of days after sudden rises in distortion on various amplifiers I was testing. The problem was solved by moving the soldering iron away from the test circuit, at which point I was relieved to determine that the 150pF As is usual with new circuits with negative feedback, it oscillated when first turned on. It was barely visible on the scope as a thicker trace, but +25V very apparent on the 19mA R3 R4 R7 distortion analyser, PR1 200Ω 200Ω 10kΩ 5kΩ with a reading of 0.02% THD+N 4.8mA a t 1 V rms ( t o t a l DC offset harmonic distortion TR5 TR4 BC556 BC556 TR3 plus noise). I knew BC546 from before that *Compensation R22 both circuit blocks capacitor 47Ω CComp* were capable of VO VIN 150pF at least ten-times TR2 TR1 BC546 BC546 better distortion. 100% R18 22kΩ Negative R1 R2 One reason I pursue feedback 110Ω 110Ω vanishingly low 0V 2mA 2mA distortion figures TR7 BC546 4mA in my circuits, R6 2.2kΩ even though the TR6 BC546 improvements may 10mA be inaudible, is to LED1 R8 1.7V R5 Red clear the view for 100Ω 240Ω other problems and –25V non-linearities that 19mA can hide behind a h i g h g e n e r a l Fig.35. Initial discrete op-amp-style buffer. This is reconfigured THD floor. These to compensate for the omission of the inverting VAS stage. can be low-level Note component numbering refer to the discrete op amp PCB oscillations or the (PE, November, 2023). Practical Electronics | May | 2024 Fig.36. The effect of using an output MOSFET for the JFET discrete op-amp-based buffer circuit was to reduce the distortion compared to a Darlington (TR3) in Fig.37. R2 200Ω PR1 5kΩ 2mA DC offset 2mA 5mA TR4 *Compensation capacitor VIN R7 10kΩ 19mA R8 180Ω 1.78V TR5 TR4/5 HN1A01F TR1/2 2SK2145 R3 200Ω +25V D C1* 150pF TR1 R1 1MΩ + D S R4 240Ω R10 47Ω VO 100% Negative feedback 4mA TR3 BC546 TR6 BS170 D TR2 S 0V C3 10µF 50V R5 2.2kΩ C2 100nF LED1 Red R6 2.2kΩ TR7 BC546 10mA 1.7V R9 100Ω 19mA –25V Fig.37. Final discrete MOSFET op-amp-based buffer circuit with JFET input and MOSFET output. (Note: components renumbered, coupling/decoupling capacitors omitted.) Fig.38. Building the MOSFET op-ampbased buffer on the discrete op amp PCB (PE, November, 2023). This is how prototypes can look. Note that individual devices can be plugged in to sockets for changing devices during testing. capacitor was sufficient to fix the oscillation problem. Another cause of ‘raised distortion’ was a bench power supply placed too close to the Audio Precision audio analyser. I found magnetic coupling between the power supply’s large, laminated transformer and the audio analyser’s output transformers. About 200mm clearance was required. Since the output stage was enclosed in a negative feedback loop, I found the value of the sense resistor (R8, Fig.37) was much less critical for minimum distortion than in the three transistor circuits. I then set it for maximum output swing into a 600Ω load. Next, I decided to run the input stage at double the current (2mA per each transistor) to increase the slewing, give greater drive to the output stage and to get more transconductance out of JFETs if these are used for the long-tailed pair. Component selection Following on from my experience with parts for the discrete op amp, I strongly suspcted dual transistors would give the best results – and they did. Using JFETs for the long-tailed pair revealed them to be more sensitive to loading from the output stage. I tried an MPSA29 Darlington transistor for the emitter follower (TR3) and the distortion dropped by 75% compared to a BC546B. Fig.39. Increasing the output level for the FET (JFET input, MOSFET output) op amp buffer in Fig.37 in 1Vrms steps to 12Vrms into a 600Ω load increases the distortion as expected. Note the rapid rise of distortion at 20Hz to 1.3% at 12V due to the 10µF 50V tantalum bead modulation capacitor (C3) being stressed. Changing this device to a 22µF 50V metal-cased tantalum reduced it to 0.025%. Practical Electronics | May | 2024 59 However, some sort of curvature cancellation magic happened when I put in a BS170 MOSFET with currentsink modulation. Using this device the circuit achieved the excellent figure of 0.001% THD+N, as shown in Fig.36. My final circuit is shown in Fig.37, and the messy prototype construction of the discrete op amp PCB in Fig.38. Distortion depression Fig.40. Using the original full discrete op amp as a buffer suffers from rising highfrequency distortion. Note that the top curve is for the output just before clipping at 17Vrms. Fig.41. Distortion for three-transistor buffer just before clipping (Fig.24). These curves are taken in steps from 1V to 15Vrms into 600Ω. Clipping is just beginning to occur at 15Vrms, hence the sudden jump in distortion. Fig.42. Distortion plot for JFET single-ended buffer in Fig.27 at 1Vrms, 4Vrms and 8Vrms respectively. 60 However, this low distortion was only for an output signal of 1Vrms (2.8Vpk-pk) into 600Ω. I needed to check the distortion at different output levels all the way up to clipping. I increased the output level in 1Vrms steps on the Fig.37 FET op amp buffer giving the curves shown in Fig.39. What looked good at 1Vrms didn’t look so good at higher test levels up to clipping. There was a rise in distortion at the low-frequency end at 12Vrms due to the modulation capacitor C8 getting stressed. I then tested the original full discrete op amp (PE, October 2023, Fig.15) wired as a buffer with 100% negative feedback and this gave a large high-frequency (10kHz) distortion hump of 0.026% at 16Vrms, shown in Fig.40. This is because a lot of compensation capacitance is required to make this topology unitygain stable. This causes a problem with slewing when driving the capacitance and there is little negative feedback at high frequencies left for output stage distortion reduction. It did give a large output of 17Vrms (48Vpk-pk) just beginning to clip with ±25V rails, 2Vrms more than the single-ended three transistor circuit for the same rail voltage. Back to the threetransistor buffer I then checked the three-transistor buffer circuit shown in Fig.24 which gave better performance than the op amp version at the higher test levels (above 1Vrms), as shown in Fig.41. I can’t leave things alone and decided to replace R7 with a more stable constant-current source using a 1mA CRD (current-regulator diode). The distortion dropped by a third to 0.0015% at 8Vrms and the low-frequency distortion increase vanished. At 1Vrms the distortion was flat over the whole band at 0.0005%, a useful upgrade. Sadly, this was not so effective on the JFET version which also needed the sense resistor R8 adjusting to 470Ω. I was hoping this constant-current mod would reduce the distortion variation between different individual JFETs. It didn’t. If I was a total obsessive, I would make R8 a preset adjustment tweaked for each individual JFET. Fig.42 shows the distortion at different levels produced Practical Electronics | May | 2024 Fig.43. Spectrum of distortion harmonics produced by the single-ended JFET buffer in Fig.27 driven at 1kHz 0dB. Note the relatively large amount of second harmonic (2kHz) which is 90dB down relative to 1kHz. The third harmonic is 122dB down. Since it is a simple circuit with low feedback the higher harmonics are invisible, buried in noise. by the JFET single-ended three-transistor buffer. However, the bulk of distortion generated is second harmonic, as shown in the spectrum analysis plot in Fig.43 produced using fast Fourier transform analysis. FFT is a useful feature of the Audio Precision analyser, TiePie scopes and various software packages. This type of distortion at this level is inaudible, but at high levels can even sound nice, so the variation between JFETs is nothing to worry about. C5 100nF It’s for real 4.8mA R3 10kΩ VIN C7 100nF C1 470nF R2 1kΩ R1 100kΩ C2 100pF R4 130Ω Ib Ib TR2 BC456B TR5 ZRX651 R8 100Ω 4.5mA TR1 BC556B Iq 4.2mA R6* 6.8Ω al rm he T link *Set Iq 0V Th erm al link R10 22Ω + C3 22µF 10V R11 22Ω 92mV C4 47µF Bipolar R12 47Ω VO R7* 6.8Ω R9 100Ω 0V D3 1N4148 C6 100nF D4 1N4148 TR4 BC456B R5 130Ω TR6 ZTX751 4.2mA 13.5mA –25V Fig.45. ‘Diamond’ buffer. An unusual circuit with two parallel, but opposite polarity input emitter follower stages. These then drive a complementary class-AB emitter-follower output. Fig.44. ‘Cupboard based programming language’ for analogue circuits. This shelf of passive components comprises 650 Farnell plastic drawers alone. (These cost more than the parts in some cases!) Practical Electronics | May | 2024 D2 1N4148 TR3 BC556B +25V + It is possible to use simulation for this type of R&D and it is great for checking whether a circuit will basically work, avoiding the latch-up scenario I had earlier. However, I dislike simulation for distortion analysis because it tends to underestimate problems. Also, I’ve been doing practical circuits since I was 12 and I’m now 61, so I stick to what I’m good at! It takes me longer to input the schematic into LTSpice than it does to build it. I can pull out any component from my vast ‘CupboardBased Programming Language’ (CBPL, shown in Fig.44) quicker than downloading any component model on-line. When I need a simulation I can always get a friend with the necessary 10,000-hours practice time to do a skills swap with me. The catch with my CBPL method is that the file size is 1500 square feet. You need to have bought a house before 1995 or have had a big inheritance to install it. 0V D1 1N4148 13.5mA Fig.46. This Avondale Audio active crossover board uses four diamond buffers with a similar proprietary circuit to Fig.45. The transistors are thermally coupled to ensure a stable quiescent current. Note there are lots of SMT components under the board. 61 Diamond buffer C9 C5 LED1 C7 +VE R5 R6 R12 GND -VE C4 R7 C6 R10 ZD1 C3 R3 I/P GND SGL SUPPLY LINK The Diamond buffer is an interesting, if poorly named circuit, which I first came across in an Elektor oscillator design (February 1979). It is a classAB push-pull mirror-image circuit that has curvature cancellation, giving very low distortion of 0.0018% at 1Vrms into 600Ω with no overall feedback. The transistor input bias currents also cancel out, assuming good Hfe matching. Despite its name I can’t see a ‘diamond shape’ in it; rather it has a ‘crossover’ like the classic multi-vibrator. I don’t know what the ‘diamond’ refers to, its name seems to derive from Burr Brown blingy marketing. The circuit in Fig.45 needs six transistors, which is rather a lot for a buffer. The Elektor version was simplified down to four transistors by using bootstrapping for the input transistors’ collector loads. Note the thermal linking of the transistors to reduce drift. A practical application is for the buffers in active filters, such as the active speaker crossover shown in Fig.46. C1 R8 TR1 R9 O/P GND R1 R4 TR2 R11 TR3 C10 R2 C8 R13 C2 R14 Fig.47 Held over from last month, the overlay diagram for the PCB for last month’s three-transistor discrete buffer circuit. Software sorrow… and a little serendipity During the test phase for the discrete buffer I thought I had a fault with my AP audio analyser – the output drive level seemed to be fixed at 1Vrms, but I was wrong. Buried in the manual (six A4 ring binders), I found – after a year of use – that the analyser was running a fixed test routine (macro) each time I clicked on the little THD icon. If I went to the sweep menu and clicked ‘start with append’ I could then change the output drive level with at least three more entries and clicks. I was amazed to find the unit could actually deliver 26Vrms (36.9Vpk-pk) at 0.0002% THD into a 600Ω load. This is very useful for performing component stress tests. I was even able to produce distortion in resistors! But that’s for another column. I’m sure there are many other settings I could use better, does anyone out there know how to use an AP SYS-2712 to its full potential? PCB A PCB for the discrete buffer is available from the PE PCB Service – layout, Fig.47. STEWART OF READING Fluke/Philips PM3092 Oscilloscope 2+2 Channel 200MHz Delay TB, Autoset etc – £250 LAMBDA GENESYS LAMBDA GENESYS IFR 2025 IFR 2948B IFR 6843 R&S APN62 Agilent 8712ET HP8903A/B HP8757D HP3325A HP3561A HP6032A HP6622A HP6624A HP6632B HP6644A HP6654A HP8341A HP83630A HP83624A HP8484A HP8560E HP8563A HP8566B HP8662A Marconi 2022E Marconi 2024 Marconi 2030 Marconi 2023A HP 54600B Oscilloscope Analogue/Digital Dual Trace 100MHz Only £75, with accessories £125 (ALL PRICES PLUS CARRIAGE & VAT) Please check availability before ordering or calling in PSU GEN100-15 100V 15A Boxed As New £400 PSU GEN50-30 50V 30A £400 Signal Generator 9kHz – 2.51GHz Opt 04/11 £900 Communication Service Monitor Opts 03/25 Avionics POA Microwave Systems Analyser 10MHz – 20GHz POA Syn Function Generator 1Hz – 260kHz £295 RF Network Analyser 300kHz – 1300MHz POA Audio Analyser £750 – £950 Scaler Network Analyser POA Synthesised Function Generator £195 Dynamic Signal Analyser £650 PSU 0-60V 0-50A 1000W £750 PSU 0-20V 4A Twice or 0-50V 2A Twice £350 PSU 4 Outputs £400 PSU 0-20V 0-5A £195 PSU 0-60V 3.5A £400 PSU 0-60V 0-9A £500 Synthesised Sweep Generator 10MHz – 20GHz £2,000 Synthesised Sweeper 10MHz – 26.5 GHz POA Synthesised Sweeper 2 – 20GHz POA Power Sensor 0.01-18GHz 3nW-10µW £75 Spectrum Analyser Synthesised 30Hz – 2.9GHz £1,750 Spectrum Analyser Synthesised 9kHz – 22GHz £2,250 Spectrum Analsyer 100Hz – 22GHz £1,200 RF Generator 10kHz – 1280MHz £750 Synthesised AM/FM Signal Generator 10kHz – 1.01GHz £325 Synthesised Signal Generator 9kHz – 2.4GHz £800 Synthesised Signal Generator 10kHz – 1.35GHz £750 Signal Generator 9kHz – 1.2GHz £700 HP/Agilent HP 34401A Digital Multimeter 6½ Digit £325 – £375 62 17A King Street, Mortimer, near Reading, RG7 3RS Telephone: 0118 933 1111 Fax: 0118 933 2375 USED ELECTRONIC TEST EQUIPMENT Check website www.stewart-of-reading.co.uk HP33120A HP53131A HP53131A Audio Precision Datron 4708 Druck DPI 515 Datron 1081 ENI 325LA Keithley 228 Time 9818 Marconi 2305 Marconi 2440 Marconi 2945/A/B Marconi 2955 Marconi 2955A Marconi 2955B Marconi 6200 Marconi 6200A Marconi 6200B Marconi 6960B Tektronix TDS3052B Tektronix TDS3032 Tektronix TDS3012 Tektronix 2430A Tektronix 2465B Farnell AP60/50 Farnell XA35/2T Farnell AP100-90 Farnell LF1 Racal 1991 Racal 2101 Racal 9300 Racal 9300B Solartron 7150/PLUS Solatron 1253 Solartron SI 1255 Tasakago TM035-2 Thurlby PL320QMD Thurlby TG210 Modulation Meter £250 Counter 20GHz £295 Communications Test Set Various Options POA Radio Communications Test Set £595 Radio Communications Test Set £725 Radio Communications Test Set £800 Microwave Test Set £1,500 Microwave Test Set 10MHz – 20GHz £1,950 Microwave Test Set £2,300 Power Meter with 6910 sensor £295 Oscilloscope 500MHz 2.5GS/s £1,250 Oscilloscope 300MHz 2.5GS/s £995 Oscilloscope 2 Channel 100MHz 1.25GS/s £450 Oscilloscope Dual Trace 150MHz 100MS/s £350 Oscilloscope 4 Channel 400MHz £600 PSU 0-60V 0-50A 1kW Switch Mode £300 PSU 0-35V 0-2A Twice Digital £75 Power Supply 100V 90A £900 Sine/Sq Oscillator 10Hz – 1MHz £45 Counter/Timer 160MHz 9 Digit £150 Counter 20GHz LED £295 True RMS Millivoltmeter 5Hz – 20MHz etc £45 As 9300 £75 6½ Digit DMM True RMS IEEE £65/£75 Gain Phase Analyser 1mHz – 20kHz £600 HF Frequency Response Analyser POA PSU 0-35V 0-2A 2 Meters £30 PSU 0-30V 0-2A Twice £160 – £200 Function Generator 0.002-2MHz TTL etc Kenwood Badged £65 Function Generator 100 microHz – 15MHz Universal Counter 3GHz Boxed unused Universal Counter 225MHz SYS2712 Audio Analyser – in original box Autocal Multifunction Standard Pressure Calibrator/Controller Autocal Standards Multimeter RF Power Amplifier 250kHz – 150MHz 25W 50dB Voltage/Current Source DC Current & Voltage Calibrator £350 £600 £350 POA POA £400 POA POA POA POA Marconi 2955B Radio Communications Test Set – £800 Practical Electronics | May | 2024