This was my first preamplifier design, conceived in 1975, the first year that I worked in the audio industry. The two gain-control solution to the dynamic-range problem was inspired by what was thought of as advanced industry practice at the time—notably in products by Radford and Cambridge Audio. I thought the rumble-gate concept was rather clever, but I seem to have been in a minority of one, as it was never even mentioned in the correspondence that followed. That was rather a disappointment.

The major part of the circuitry consisted of two discrete opamps with differential inputs, which performed the amplification and tone-control functions. The ±24V supply rails, which permitted a maximum signal level of about 16 Vrms, were provided by two 7824 IC regulators. The Moving Magnet (MM) input stage, however was a simple three–transistor stage running between +24 V and 0V. The reasoning was that in the MM stage low noise was the priority, the low signal levels making distortion much less of an issue, and so it was better to have a single input transistor rather than a differential pair; I had a vague notion that the pair would be 3 dB noisier. This is not actually the case, as the transistor of the pair that handles the negative feedback works under much better conditions for low noise than the input transistor, which is faced with the highly inductive source impedance of an MM cartridge. I have not yet got around to quantifying the difference, but clearly two devices must be noisier than one to some extent; having said that, the 5534 opamp with its differential inputs gives a very acceptable noise performance as an MM input stage. I later developed a more advanced discrete MM stage that had still only one input transistor, but could have been run between the +24V and −24V rails, giving a theoretical 6 increase in headroom. This approach is fully described in Chapter 2.

As shown in the block diagram (Preface Figure 1.2) the RIAA equalisation was divided into two halves. The HF roll-off at 2.1 kHz was implemented in the first stage, and the feedback around the normalisation amplifier implemented the boost starting at 500 Hz and levelling off at 50 Hz. Doing it this way round avoids a massive loss of headroom in the first stage [1], though at the cost on some compromise on noise. Splitting up the equalisation of course makes the RIAA component calculations very much easier, and unless I’m very much mistaken that is the inglorious and only reason that the idea has the popularity it does. That was not the reason I did it here. The main motivation was to make life easier for the MM input stage, which was a relatively simple three–transistor Walker configuration [2] with my addition of second-transistor collector load bootstrapping, what I called a BootWalker at the time, though it is perhaps not the most euphonious of terms. I did not want to implement the LF boost in this stage as its rather limited open-loop gain would have led to excessive distortion.

Preface Figure 1.1 The Advanced Preamplifier neatly boxed: 1976

During the development period, the need for an HF correction pole to counteract the levelling-out of the response as unity gain is approached in the first stage was appreciated; this is implemented by R1, C1 in Preface Figure 1.2. The extra 6 dB/octave roll-off exactly compensates in level and phase for the levelling-out in the first stage so long as you get its frequency right; this is not a case of an approximate bodge.

While giving excellent results considering its simplicity, the BootWalker did not have the capabilities of a discrete opamp; therefore it was not required to implement the boost in the LF region, where it might have had inadequate negative-feedback to give good linearity. The normalisation amplifier is a discrete opamp with nine transistors rather than three, and better fitted to giving high gain at low distortion. The components implementing the LF part of the RIAA were switched out of the feedback network of the normalisation amplifier when a line input was in use; all my later opamp designs do the whole of the RIAA equalisation in one stage, thanks to the Lipshitz equations. I have heard the discrete opamp approach disrespected on the grounds that ultra-low distortion occurs almost automatically; such statements show woeful ignorance, as there are many design issues to that need to be dealt with properly. These are described in [3]. Be aware that the discrete opamps used in the article here are more start-of-art than state-of-art and could be much improved in the light of today’s knowledge; for example, now I would put current mirrors in the collectors of the input pairs to enforce equality of collector current, and so prevent the generation of second-order distortion. For full details of how to make a state-of-art discrete opamp, see [3].

Absolute phase was an unheard of concept in 1976, and the phase-inversion produced here by the tone-control was ignored. It is difficult to see how it could have been fixed without adding another stage, which did nothing but invert to get the phase correct again; not exactly an economical design approach.

You may be wondering why the coupling capacitors are in most cases rather small non-electrolytics. This is because electrolytics were not as reliable then as they are now, and so I wanted to use as few as possible. Looking back, this was not my brightest idea, although I was right in the sense that the lifetime of a piece of audio equipment is still primarily limited by electrolytics drying out and falling in value. In top-end mixing consoles where internal temperatures are relatively high and 24/7 operation is the norm, the electrolytics need replacing (an extremely specialised job) long before the pots and switches are worn out.

You will note the 25 kΩ level pots and the 100 kΩ pot in the tone-control network; these values were typical of the era. In succeeding designs, I used electrolytic coupling capacitors, often of 47 uF, because this allowed the circuit impedances to be an order of magnitude lower, with corresponding reductions in Johnson noise and the effects of device input current noise. Pot impedances were reduced to 10 kΩ; for an example see Chapters 5 and 6. Since there were always electrolytic reservoir capacitors in the power supply, it seemed pointless to worry too much about a few coupling capacitors working under rather more benign conditions from the point of view of reliability.

As my thoughts on preamp design progressed, I reduced the circuit impedances in my high-end designs by another order of magnitude, so that pots were 1 kΩ, and coupling capacitors of 470 uF or 1000 uF were used whenever DC had to be blocked. The limits on this process are to an extent the increased drive requirements, though this can be easily overcome by using multiple opamps in parallel, which in itself also reduces noise by partial cancellation; see Chapters 14 and 15. A more final limit is the unavailability of two- and four-gang pots in very low values, but this too can be circumvented by using multiway switches and resistor ladders.

Let us see how far this idea can be pushed. If the pots/switched-resistor-networks are reduced to 100 Ω then driving them with paralleled opamps starts to get clumsy, as you would probably need to use eight of them for a good distortion performance. Using opamps combined with discrete output transistors may cut the component count but is likely to impact the THD for the worse. If you decide to get really radical and replace the opamps with power amplifiers, I suppose the ultimate limit would be overall power dissipation; imagine driving a 10 Ω volume control at 10 Vrms (roughly maximum opamp level). You would be dissipating 10 W in the pot, or 20 W for stereo; if you insist on a pot rather than switched-resistors, it’s going to be a wirewound rheostat (I actually have some of them in my component collection, but unfortunately none are double gang for stereo.) Good luck getting a rheostat with a log law. Admittedly 10 Vrms is a high level, and for a more likely 2 Vrms, the pot dissipation is 400 mW, and more doable, but this will be repeated at many places in the circuitry. Apart from the problems of sourcing dual-gang log rheostats, there are likely to be crosstalk and offness problems if the ground connections are not equally reduced in impedance, which won’t be easy.

Pushing the idea a little further, the use of power amplifiers removes the level limitations of opamps. With ±45 V rails, we could sustain 28 Vrms, 9 dB above 10 Vrms, which could be used to either increase headroom or signal/noise ratio (by raising the internal level), but we would then be dissipating 80 W in our 10 Ω volume control. This is obviously well beyond the line of sensible, but in hi-fi, that is a very blurred line. If you want to take the idea even further, I will merely observe that there is no difficulty in designing an amplifier to drive a 2 Ω load, and leave you to it.

If you’re wondering if this sort of thing has been tried for real, the answer is yes, sort of. I recall a microphone preamplifier with a very low impedance negative feedback network intended to reduce noise; this had to be driven by a power amplifier.

The LED biasing of current-sources (to minimise current changes with temperature) was an elegant method but, as I strongly suspected at the time, quite unnecessary, as the discrete op-amps used are really not that sensitive to their internal operating conditions. Nevertheless, as LEDs were pretty new at the time, it was generally regarded as well cool. A PCB was available from Wireless World for the princely sum of £3.50.

The level monitoring circuitry was perhaps a bit over the top, with a big pair of VU meters originally intended for a mixing desk. There were also ’1V Peak’ LEDs to indicate nominal level, and bipolar ‘Clip’ LEDs to alert the user that the ±24V rails had been hit. I have to admit that I never once saw the ‘Clip’ LEDs come on in real use. The bipolar clip-detect circuit proved extremely useful in other applications however, and I have been using it ever since. There is also a bit of an infelicity in the ‘1V Peak’ driver circuit, where the full output swing of one of the discrete op-amps is potentially applied to the trigger input of a 555 timer. Still, it never failed.

The Advanced Preamplifier was inspired by an article written by Daniel Meyer in Wireless World for July 1972. It was called: ‘Audio Pre-amplifier Using Operational Amplifier Techniques’, and still makes interesting reading, though some of the statements made are questionable. I have not been able to confirm it, but this is almost certainly the same Daniel Meyer that founded SWTP and designed the ‘Tiger’ series of power amplifiers [4].

When I measured the distortion performance of Meyer’s version of a discrete opamp in April 1975, I found the distortion performance was less than inspiring, because the output is connected directly to the VAS collector, and loading on this point is very bad for linearity; I measured 0.008% at 1 Vrms out, with a flat closed-loop gain of 16 times and ±15V supply rails (The supply rails used by Meyer). Although I was but a lad, it still seemed to me that this was rather high for what was supposed to be an advanced configuration, and I set out to improve it. In the course of this, I taught myself some of the rules of good amplifier design, such as do not connect your main gain stage directly to the load, but instead, buffer it with some kind of emitter-follower output stage. I used a constant-current emitter follower, increased the supply rails to ±30V, and voila! Less than 0.003% at 20 Vrms out. This pleasurably straightforward process started off an interest in reducing distortion which has stayed with me ever since.

A follow-up article was published called ‘Additions to the Advanced Preamplifier’ in October 1977. This added a second-order subsonic filter and a switched-frequency second-order scratch filter to the signal path. Also added was a virtual earth summing amp to give simultaneous mono and stereo outputs. (The mono output was to drive a sound-to-light converter, or light show, as we called them in those heady days.) Rather than add another complete discrete opamp, I used a simpler four-transistor circuit with the gain provided by a cascode stage using current-injection (this technique is described in the preface to Chapter 2) and a constant-current Class-A output. Other features that were described were meter suppression (to stop the VU needles smacking the end-stops at switch-on) and a remote muting facility. The latter was a switch on the end of a piece of cable rather than an IR system.

In designing these later add-ons, I was deliberately following the path of other preamp designers in Wireless World. It was definitely the done thing to come up extra features at a later date, and somehow gave the project an aura of greater import. I was danged to heck if I was going to be left out of that.

This was the first preamplifier that I developed into a completely worked-out and neatly boxed entity. The initial design work on the discrete opamps was done with the Sound Technology 1700A in 1975, using one of the first examples to enter Britain. In its day it was a major advance in audio measurement. The prototype Advanced Preamplifier was built on stripboard, but it could have been laid out as a PCB and put into production. This is therefore preamplifier No. 1 in my personal numbering system, and I described it as the MRP1. The RP naturally stands for Reference Preamplifier, and the M comes from Meadow House in Swaffham Prior, Cambridgeshire, where I was living when I did the design work.

References

1. Self, D., Small-Signal Audio Design, 2nd edition, Chapter 9, pp. 252–256. Focal Press (Taylor & Francis) 2014. ISBN: 978-0-415-70974-3 Hardback. 978-0-415-70973-6 Paperback. 978-0-315-88537-7 ebook.

2. Ibid., pp. 301–303.

3. Ibid., pp. 110–115.

4. http://en.wikipedia.org/wiki/Daniel_Meyer_(engineer) (accessed Jan. 2015)
(The Wikipedia entry claims that Meyer published an article in Electronics World in May 1960: this must refer to a USA magazine of that name as the British journal was called Wireless World at the time, and its May 1960 issue does not contain an article by Meyer.)