Fullscreen HTML5Med Res (??MB)HTML4

AUDIO OUT AUDIO OUT L R By Jake Rothman Microphone Preamplifier (for Vocoder) – Part 4 The Microphone Preamplifier board, as completed in last month’s column. L ast month, we looked at Adding transformers be needed. These are readily available from Vigortronix in all sizes. Another bonus is that the phantom power could also then be fed via a centre-tap, rather than 6.8kΩ resistors. A further reason transformers can sound better is because there is less voltage drop on the 48V microphone supply due to their lower series DC resistance of around 30Ω. The voltage to many One preamplifier mod it’s often worth doing in electrically noisy environments (such as near lighting rigs), is to add an input transformer to the already balanced input shown in Fig.14 (see Part 2 in Audio Out, June 2021). This significantly increases the CMRR to top-class levels. Also, the CMRR of solid-state inputs is strictly limited to around 5V or whatever level causes the input amps to clip. A transformer, however, can tolerate around 100V common-mode voltage. This double-balancing technique was employed by S. Groner in his microphone-preamplifier in the March 2015 issue of the Journal of the Acoustic Engineering Society (JAES). A 600Ω 1:1 line-level transformer is much cheaper than a step-up microphone transformer. However, a mu-metal screening can to avoid hum pick-up would also Fig.37. The distortion curve of the output transformer at 7V peak-to-peak feeding 600Ω. Note the typical rise at low frequencies. This is emphasised because the transformer is being slightly over-driven at this level. building and testing the Microphone Preamplifier. This month, we conclude with some modifications and design options. A project like this is never truly ‘finished’ – there’s always room for additions and improvements, but you have to draw line at some point! 46 microphones can drop to as low as 35V due to the 6.8kΩ resistors. Output transformer Output transformers offer the unique property of earth isolation, the floating output prevents earth loops however they are connected, even if one side of the balanced output is grounded. (This is the case when feeding unbalanced inputs.) Practical Electronics | August | 2021 D ri ve r a m pl i f i er O utput tra ns f orm er pri m a ry Input R P r1 ( posi ti ve resi s ta nce) n Neg a ti ve output resi st a nce ( – R P r 1 ) g enera tor – ca ncel s th e pri m a ry resi st a nce 0 V All transformers produce a rising distortion as the frequency gets lower, similar to acoustic instruments, magnetic tape, capacitors and loudspeakers. Transformer distortion harmonics tend to be mainly third order plus a bit of second order. Solid-state electronics such as op amps, class AB power amps and digital-to-analogue converters all have the opposite effect, higher-order distortion rising with frequency. Some audio people think these two characteristics complement each other, which is why one finds expensive gold-plated silver-wound transformers installed in expensive digital Hi-Fi systems. There is also a huge demand for class-A Neve-style transformer-coupled preamplifiers, such as the 1073. I take the view that – within reason – low-order harmonic distortion is enhancing on simple audio signals, such as a solo voice and drums, but bad on complex mixes of signals, for example piano and choral music. This is because dissonant intermodulation products arise with multiple tones on systems with high THD. I’ve decided to offer a choice here, normal transformer distortion of about 1% at 20Hz (shown in Fig.37) or a special distortion-cancelling circuit built around IC2b, which lowers LF distortion to around 0.02%, as shown in Fig.38. Distortion minimisation circuit Practical Electronics | August | 2021 Input Zob el netw ork IC 2 a 5 5 3 2 330Ω O utput 2 2 nF – R 2 7 10Ω IC 2 b 5 5 3 2 Lund a h l LL1 5 1 7 DC resistance = 19Ω P ri m a ri es i n se ri es Second a ri es i n se ri es + R 2 6 1.2kΩ C 1 5 4 7 µ F + The BBC originally used huge expensive transformers with 80% nickel 20% iron cores for their own in-house-built equipment, as shown in Fig.39 This was very effective in minimising distortion. When sound mixers began to be bought in from private industry, enterprising engineers developed circuit tricks to get the same performance using cheaper components. Most involved negative feedback from an extra winding, such as some Neve designs. Another method was developed so that ordinary off-the-shelf transformers could be used. It was well known that transformers gave less distortion when driven from zero source impedance. This was because the current drawn by a transformer is distorted because of the hysteresis and saturation effects of the iron core. The saturation effect gets worse as the frequency decreases, since there is more magnetic energy to store for each cycle, accounting for the low-frequency rise. Despite these tricks it is worth noting that some residual distortion is still developed because of the series resistance of the primary wiring, even if driven from a zero-ohm output. Then some bright spark figured that if this resistance could be cancelled out by a negative resistance, distortion should be eliminated, as shown in Fig.40. A circuit developed by Calrec in conjunction with Swedish transformer maker Lundahl is shown in Fig.41. The lower op amp IC2b is the negative-resistance generator. It works by developing a voltage Fig.39. Heft rather than feedback – the BBC used to in anti-phase to the curuse conventionally driven large (43mm high) nickel-core rent flow by using a sense transformers to get low distortion. resistor (R27). If a current goes into the upper end of the sense resistor the voltage developed across it goes down rather than up, as shown in Fig.42. It should be noted this circuit uses a degree of positive feedback and if the negative resistance exceeds the resistance of the primary, the circuit will oscillate. The amount of positive feedback is set by the lower-arm feedback resistor (R25), and this should be adjusted to suit the transformer used. The original Calrec circuit used 2kΩ with a Lundahl LL1517 transformer, which is now very expensive. I use the Vigortronix 49% Ni core VTX 101-3001 with R25 set to 3.9kΩ. Another source of distortion is ground currents. By feeding the transformer return current into the output of IC2b, it is conveniently diverted to the power rails. To achieve the lowest distortion, always drive a transformer directly from the output of the drive amplifier using DC coupling. Adding a coupling capacitor causes the LF distortion to rise again because of the – Transformer distortion Fig.40. Basic principle of reducing the transformer distortion. The negative resistance cancels out the transformer’s primary DC resistance. + Fig.38. The effect of the negative-resistance distortion-reduction circuit (Fig.42). Same conditions as in Fig.37, but with an almost 10-times reduction in distortion. R 2 5 2.0kΩ 0 V Fig.41. The distortion-reduction circuit used on late 1980s Calrec broadcast desks using a Lundahl output transformer. 47 V+ 1 N4 0 0 1 + Input + IC 2 a 2kΩ V– R sense R 2 7 10Ω – 1 N4 0 0 1 – IC 2 b 1 N4 0 0 1 0 V 1 + M eta l w ork 1kΩ r n 1 0 nF it V– t t ta e r rti na t trans r er ri ar c rrent t anti ase c are t C2a + R 2 6 1.2kΩ 2 0 V + F 8 . 2 nF a ance t t s cket e terna ie Vi g ortroni x 101 3001 – R 3 30Ω DC resi st a nce V+ R 2 5 3.9kΩ Zob el netw ork 1 N4 0 0 1 O pti ona l cl a m p d i od es Neg a ti ve resi st a nce g enera tor an arti icia g round – DC ckin C 1 5 3 3 0 µ F C a pa ci tor 6 .3 V T a nta l um Fig.42. The negative resistance is developed by producing an anti-phase voltage drop from the current passing though sense resistor R27. The effect is boosted by positive feedback network R26 and R25. DC has to be blocked by C15 to prevent the op amp going off-set. siti e ee ack net rk ee s t e ta e e e e acr ss 2 int C2 s n n in ertin in t an sts it s icient t ake t t resistance ne ati e capacitor’s reactance, as shown in Fig.43. One problem with driving transformers directly is the low DC resistance of the winding, which can give rise to high fault currents if an offset voltage develops. The 5532 seems to be able to survive such situations, but its dissipation rating maybe exceeded in the worst-case situation where both op amp outputs are latched to opposite power rails. It’s a worthwhile safety precaution to add clamping diodes (shown dotted in Fig.42) from the outputs to the rails to deal with possible voltage spikes caused by the transformer driving faulty leads. Capacitor distortion In Rupert Neve’s obituary (JAES, March 2021), Malcolm Atkin, MD of Audio Consultants Ltd, recalled Neve describing what caused the unique sound quality of his mixing console installed at Sir George Martin’s Air recording studios in Montserrat. He said it was not just down to the transformers, but tantalum capacitors. Using tantalum capacitors for signal coupling (especially when unbiased and driving low impedances) gives a similar low-frequency distortion rise to transformers. Fig.44 shows this capacitor distortion on the microphone-preamplifier. Ordinary wet aluminium electrolytics also make this distortion, but around five times less. Going bipolar Fig.43. The output transformer has to be directly driven / DC coupled. Adding a 47µF bi-polar capacitor in series with the output of IC2a spoilt the distortion cancelling effect. I found enclosing this capacitor in a negative-feedback loop around IC2a eliminated the distortion, but then caused low-frequency ringing around a few hertz. Negative resistance is a good way of making an oscillator. This is a work in progress. Fig.44. Distortion effect of a single 100µF 20V tantalum capacitor driving 13.6Vpk-pk into 600Ω unbalanced output (ie, no transformer). It is reverse polarised each cycle, the voltage increasing as the frequency decreases, causing a rise in distortion at low frequencies. It is unlikely the capacitor will normally be driven at this level. 48 Bipolar wet aluminium electrolytics are almost distortion free and provide full protection from hard offsets of either polarity. This approach was used by Clive Green in his Cadac mixers used in most West End theatres. Connecting two polarised capacitors back-to-back effectively makes a bipolar capacitor, but some distortion is still present. To eliminate this, a few volts of bias at the junction is needed. Inactivity can kill One problem little known about these wet bipolar capacitors is their short shelf life. Fig.45. Bipolar electrolytic capacitors have negligible distortion, but they can degrade if not used. All of these Suntan capacitors almost exploded in storage. Practical Electronics | August | 2021 Fig.46. The Hitano capacitor brand was no better than Suntan in terms of component degradation. I’ve just dumped a load of Suntan capacitors from Rapid and some Hitano types from JPR which exploded in storage after a few years. When I complained, they pointed out the data sheet specified they need a polarity reversal every 250 hours. Those from Nichicon and Panasonic have the same requirements. They have to be used or they die – see the photos in Fig.45. and Fig. 46. A sensible check is to measure the capacitance. If it has markedly risen from its stated value then it’s a sign that the rot has already started, and the dielectric film is thinning. One manufacturer whose products do not suffer from this defect is the yellow Nitai type shown in Fig.47. The Midas touch The clever way round this problem is a technique developed by Midas who made the mixers and sound systems used by Pink Floyd. Aged 13, I remember being taken to their factory near Euston in 1975 by my uncle who worked there as a test engineer. I saw the best audio design, construction and components. They C 1 6 .8 µ F * * P l a st i c f i l m H i g h - pa s O n F rom i nput D C f i l ter O n T oT R 1 /2 b a se s of P rea m p D a m pi ng 1 .2 H 12 Ω resi st a nce C 2 6 .8 µ F * R D 0Ω Fig.50. A high-pass filter can be placed directly on the input to the preamplifier. For better headroom, remove low-frequency signal rubbish before amplification. had Bourns conductive plastic pots and Philips TDA1034 op amps, (later renamed the NE5534) before anyone else. A quirk of the 5534 is that sinking a small current into the output of the 5534 reduces its distortion. The engineers at Midas (who I suspect discovered this technique) also combined it with a bipolar capacitor biasing circuit, as shown in Fig.48. This worked wonders with tantalum capacitors. The resulting curve is shown in Fig.49. We now have a virtually distortionless coupling capacitor that lasts a lifetime. One of the Midas desks is now in the UK’s National Science and Media Museum’s mixer collection. Low-cut filter Fig.47. These yellow Nitai capacitors gave no storage degradation problems, but I’m not sure if you can still get them. V+ 1 2 V 330kΩ Wind noise, vocal pops and handling cause microphones to generate a lot of low-frequency transients which cause all sorts of problems, such as clipping, intermodulation and transformer saturation. These are best removed as soon as possible in the signal chain. A simple compromise is to put a smaller capacitor in series with C8, C9 and VR1 (the gain control). This can consist of two 22µF tantalum bead capacitors in parallel back-to-back. These are then shorted out by a switch to turn the filter off. Unfortunately, the cut-off frequency varies with gain setting. A selection of capacitors could be wired to an extra gang in a gain 3 5 µ A kΩ 1 5 0 µ F 1 6 V T a nta l um + 5 5 3 2 /4 + + – Input 1 .5 V Isink 1 5 0 µ F 1 6 V T a nta l um D ecrea se resi st ors b y f a ctor of 5 f or w et a l um i ni um el ectrol yt i c ca pa ci tors d ue to h i g h er l ea ka g e. Fig.48. The biased bi-polar tantalum capacitor circuit used in Midas mixers. Low distortion, lasts for ever. Midas also biased the 5534 output stage into displaced class A for lower distortion. (Although in this case the current sunk will have little effect – around 5mA is needed). Practical Electronics | August | 2021 Fig.49. The effect of the Midas circuit on a 7Vpk-pk signal into 600Ω (no transformer). 49 steepens the slope and raises the cut off frequency to that shown in Fig.51. The inductor must be well shielded to avoid hum pick-up, such as a ferrite pot core. Leaving the 750Ω damping resistor off gives a resonant hump which can be useful for adding ‘warmth’ to voices, as shown in Fig.52. This can be made switchable. Measuring distortion Fig.51. The second-order high-pass filter response from Fig.50 with a cut-off frequency of 95Hz, which should be changed to suit different voices. Distortion is a measurement requiring expensive equipment. I started off with a homemade ETI Linsley-Hood distortion analyser (see Fig.53), and then moved to a Lindos, one of the best examples of degenerate digital design with a grey-on-grey LCD. I recently moved to the Audio Precision instrument shown in Fig.54 – having waited 15 years for the price to drop like a good car from £14,000 to £1500. The software is horrid, like Windows Vista, but the hardware (apart from the noisy fan) is unsurpassed. PNP transistors There is provision on the PCB to allow PNP transistors to be used to take advantage of their slightly better noise and Rbb figures compared to their NPN complements. The polarity of supplies to the collector (R7 and R8) and emitter resistors (R9 and R10) can be reversed by means of links. Notice that D5 and D6, C8 and C9 also have to be reversed. Fig.55 shows how centre-collector Japanese transistors, such as the 2SD655 are installed. Fig.52. Effect of removing the damping resistor RD in Fig.50. The humped characteristic is enhancing in some speech applications. switch. A better solution is a 1200mH inductor strapped across the input after the coupling capacitors C1 and C2, as shown in Fig.50. This does not cause a significant distortion- or impedance-induced noise problem. Reducing the capacitors to 6.8µF 48V transistor power Groner found running the transistor collectors off the 48V phantom power rail reduced the noise at minimum gain by 5.3dB. This mod can only be done with NPN transistors. The resistors R7 and R8 have to be increased to 9.1kΩ 0.5W. This can be done by simply putting in 6.2kΩ resistors into the positions marked REX1 and REX2 on the PCB, and putting a diode in the link to the positive 15V rail. PNP transistors may benefit by taking the emitter resistors to +48V which will provide a more constant current, giving better matching. (There is no provision on the board for this – this is for experimentation only, so implementing it will be a bit messy!) Battery version Fig.53. This Linsley-Hood distortion analyser gave 25 years of solid service. 50 The second op amp stage can be eliminated by feeding the outputs of the gain stages (IC2) into an output transformer or balanced output directly. This will reduce current consumption by 4mA and avoid the differential amplifier noise. I’ve found the Vigortronix VTX-101-007 to be an excellent transformer for this job. It is possible to then use the preamplifier with a couple, or even four PP3 batteries as a remote gain booster. Current consumption Practical Electronics | August | 2021 Fig.54. My Linsley-Hood unit has now been replaced with this Audio Precision unit, so graphs can be more easily produced. Fig.55. Most low-noise Japanese transistors have their collectors in the centre. These have to be inserted in TO5 sockets, as shown here. Fig.56. A low-impedance phantom-powered gain block; this had to be inserted between the RE20 low-impedance microphone and the Ted Fletcher Brick to get a decent signal-to-noise ratio. into a complex piece of studio-quality equipment taking a month of work. It won’t fit into the vocoder box now, so it will have to stand alone. So far it has compared favourably with the £700 Ted Fletcher Designs Brick microphone-preamplifier shown in Fig.57. Leading on There’s no point using a low-noise amp with a noisy lead, so next month in Practically Speaking I’ll be showing how to make a top-quality microphone lead. 1966 Practical Electronics Transistor Guide – now complete! Fig.57. The Ted Fletcher Brick is an amazing microphone preamplifier plus processor for studio condenser microphones, but not suitable for low impedances, hence the replacement design given here. on ±9V was 13.8mA. Amazingly, I found the unit works down to ±4.5V. Distortion will of course be much higher for the battery version. Large rail decoupling capacitors of say 470µF will reduce it. I’m thinking of making a phantom-powered Practical Electronics | August | 2021 single-rail version, like the gain booster shown in Fig.56. Final thoughts What started out as a simple microphone-preamplifier has now morphed You may remember a few issues back we included part of a handy Transistor Guide from the May 1966 issue of Practical Electronics. It was incomplete because our well-thumbed copy had lost a few pages. Well, a couple of readers were kind enough to help us out, and this month’s download from the August 2021 page of the PE website includes the full Guide. Thank you David Allan and Bob Ashby. 51