|
|
|
|
The Walt Jung Super-Regulator*Ok so you've all been waiting for this bit, below is the schematic of the Walt Jung Super Regulator, followed by a description and explanation of the various circuit elements and their contribution to the final performance. As will be seen all of the limitations mentioned previously have been addressed, along with many other enhancements, that produce a supply that's as close to perfection as the relative simple topology allows. There's an awful lot of time, thought and listening that's gone into it, so raise a glass to Walt, and say thank you when you're next listening to your music, realising the benefits... Click image, or here, for full-size version. We'll analyse the regulator circuit first, ignoring the 3-terminal regulator (LM317) to the left of the schematic - it's purpose will become clear in time. Voltage ReferenceThe reference chosen is an LM329 sub-surface (or buried) zener (ZD1). Whilst the circuit symbol is that of a zener, it's actually an active device, which results in much lower dynamic impedance than a normal zener, of around 1ohm. It also has low noise relative to it's output voltage (6.95V nom.), due to the sub-surface zener reference. Other devices can be used, depending upon desired output voltage / noise performance. This Zener is biased through a resistor (R5). Note though that the bias voltage is from the regulator +ve sense point, or output, which is the quietest +ve rail of the system. Biasing the device in this manner means that the reference is well isolated from the effects of noise at the regulator input, in effect benefiting from a super-reg to itself! The output of the reference is then filtered, via a simple RC network (R7 / C4). The resistor values are chosen so that the error amplifier (IC2 or IC3) inputs see a low, and matched source impedance of 500 Ohms, thereby facilitating optimum DC stability and dynamic performance and minimising EMC susceptibility. C4 is a very high performance, low E.S.R. device in order to maximise the filtering effect from the low impedances used. The low impedances bring benefits in terms of reducing the inherent noise of the resistors (Johnson noise) and reducing the effects of EMC / EMI on the circuit. Error AmplifierD3 and D4 provide protection against excess differential mode voltage at the op-amp inputs. They are not strictly necessary for the AD797 op-amp, and can be left out for some very minor performance improvement, since the AD797 incorporates these diodes internally. This not the case for many other op-amps though. The reference voltage at the input of the op-amp is compared to a proportion of the output voltage, via the +VSENSE and 0VSENSE inputs and a potential divider (R8 / R9). +VSENSE / 0VSENSE can be connected directly to the relevant regulator outputs, or to the load (Remote Sensing), thereby improving performance by ensuring the regulator performance is not degraded by the interconnecting wiring’s finite impedance. The ratio of these two resistors sets the output voltage, and the combined parallel impedance of these two values should be 500 Ohms, for optimum dynamic and DC stability. To prevent noise being amplified by the gain of the error amplifier, the noise gain is reduced to unity at high frequencies, by the addition of a bypass capacitor (C7). This is matched in value and type to C4, in order to equalize AC impedances at the error amp inputs, thereby improving dynamic performance. The error amplifier also obtains its power from the regulator output (bootstrapping), this brings major performance improvements, and is a fundamental feature of the design. As the error amplifier PSRR (Power Supply Rejection Ratio) degrades, with increasing frequency, the regulator performance will normally suffer. This connection scheme augments the error amplifier performance dramatically, giving a significant, frequency-dependant, improvement in the regulator performance parameters. In particular line rejection is enhanced significantly. Start-up needs careful consideration though, and is covered in more detail later and helps dictate final topology chosen. The output of the regulator is from the series pass device (Q1). This is controlled from an LED-biased current source (T1 / R1 / D1 / R4). At switch on the current source provides current to Q1 base, which turns it on causing the output voltage to rise. This applies power to the error amplifier and reference, which then starts to control the output voltage, by sinking current (via T2 / R6 / D2) away from the base, reaching equilibrium when the correct output voltage is achieved. The level shift Zener diode (D2) is an essential part of the start-up process, ensuring that if the error amp comes on in a low-output condition there is still enough current available at Q1 base to bring the error-amplifier / reference into the linear operating area. It is bypassed by a high performance low ESR capacitor to reduce the dynamic impedance of the device, and further improve dynamic output impedance of the regulator circuit. Input and output decoupling are an essential part of the stability of the regulator (C1 / C5). C6 MUST be left out initially. In rare circumstances when remote sensing is used, the regulator can occasionally oscillate. The latest versions of these regulators, using the AD825's have NEVER oscillated, but earlier revisions using the AD797 error amplifier were very prone to instability. A film capacitor (C6) can be used to decouple the loop at very high frequencies, ensuring stability is maintained. If used the +VSENSE input MUST be connected to the load via a 10R resistor at the load end (to ensure the sense wiring capacitance is decoupled). Failure to do this may cause the regulator to oscillate due to the excess phase shift introduced by the high Q of the film capacitor (C6). The Tracking Pre-RegulatorFinally a tracking pre-regulator can be used to further improve performance. The current source is fed from the raw incoming supply, and is therefore susceptible to noise / degradation of this supply. By using a pre-reg, this degradation can be minimised and the sonic benefit is very large, for minimal cost / complexity. The LM317 is an adjustable monolithic voltage regulator that works by maintaining 1.25V between it's output and adjust pins. By adding the two resistors R2 / R3 we can easily program the regulator to produce any voltage we desire. The regulator (IC1) is set to produce 2.5V via the potential divider connected between the output, adjustment and Kelvin sense points (R2 / R3). A capacitor across R3 further lowers noise, ripple and improves line rejection bringing truly huge musical improvements to the pre-regulator / super regulator combination. The Kelvin sense point TRS1 or TRS2 is connected to the regulator output (OP1 or OP2), the pre-regulator output is then connected to the Super Regulator input (TROUT to SRIN). The raw +ve supply is then fed in on TRIN, the tracking regulator input. By connecting in this way we provide a further enhancement to performance. If the regulator was just set to produce a voltage at the input of the super-reg, with respect to ground, as is the normal configuration, the input to the super-reg will be contaminated by any ground noise present in the PSU. This can be reduced by careful wiring / kelvin sense connections, but is hard to eliminate. By connecting the pre-reg such that it produces a voltage 2.5V above the super-reg output, the input to the super-reg benefits from it's own output performance. This measurably improves the pre-reg's noise performance, by a large margin. It has the additional benefit of not requiring the pre-reg to be adjusted for different output voltages - it tracks the super-reg output. How does it all sound? In a word, stunning! If you doubt the importance of power supplies, try building at least one of the circuits presented and give your ears a treat, I'll place some real-world measurements here soon, but suffice to say that with the Jung topology, and suitable choice of components it really is possible to get a power supply that's very close to being DC, with difficult to measure AC content. No-one really gives a f**k about measurements at the end of the day though. The real benefits are about the music and the listeners enjoyment of it. You'll get a tighter, punchier more rhythmic sound, with a real sense of real musicians and interplay. Noise levels drop massively, revealing more detail, ambience and subtle musical clues such as guitar technique, vocal phrasing / inflection, tempo changes etc. Bass instruments gain real harmonic character, rather than being a low frequency thud, bass guitars differences become obvious; type, manufacture, playing techniques and string types are revealed in detail. Drums become dynamic instruments that leap at you when struck hard. Most importantly though it does not result in musical dissection, but a greater feel for the craft of the musician - in short, it boogies. It gets right into the bit of your brain that makes your feet tap, you're body dance and gives that indefinable connection with the musicians such that you know exactly where the music is heading, because it feels right. I love it :) So can we do even better? There are improvements that can be made to the above circuit, but you're going to have to work those out for yourself. It may be that a different topology is a better route to significant gains, but at this moment in time my music is just so good, I'm not inclined to fiddle. Not sure how long that will last, but I suspect you'll hear about it here first.
*It should be noted that some of the component choices I have made for this regulator are my own, as is the PCB layout, and may not be supported by the Walt himself. I am deeply indebted to Walt for his correspondence during and after the development of this circuit and the massive benefits it has brought to everything I've used it in, from external supplies for portable minidisc players, through to preamps, CD players etc. |
|
Send mail to
webmaster@alw-audio.co.uk with questions or comments about this web
site.
|
