The Gestation of a Regulator

Welcome to Regulator Design School

This page will take you through from the basics of linear series regulator design, to the final circuit showing the Jung Super-Regulator topology. Along the way we will analyse the various circuit elements contribution to performance and look at ways of addressing them. The Sulzer circuit will also be presented as a stop along the road to perfection (or at least somewhere quite close!).

Basic Block Diagram

The circuit above is the basic block diagram of a linear series regulator.

A series control element (imagine it as a variable resistor) is controlled by an error amplifier (X1). This error amp compares a fixed stable voltage from a reference to the regulator's output. If a load is placed at the regulator output the output voltage tends to drop (imagine a potential divider formed by the series control element and the load). This also causes the voltage at the inverting input of the amp to drop (via R1 / R2) which in turn causes X1's output voltage to rise, turning on the series control element harder (imagine it's resistance dropping) causing the voltage at the output to rise - we have regulation!

Practical Circuit

A simple practical example of the basic linear regulator topology is shown above. The incoming raw supply (V1, 12V) is used to feed the pass device (Q1) and to provide a bias current to D1, a 4.7V zener diode, along with X1 the error amplifier. R1 / D1 in fact form a simple shunt regulator giving a reasonably stable reference voltage to the input of X1, the error amplifier. Since the input impedance of X1 is very high, it does not load the reference voltage and hence we have a stable voltage against which we can compare the regulator's output. This function is provided by R2 / R3 which take a fraction (in this case half) of the output voltage to feed to X1's inverting input. The output voltage in this case is twice the reference voltage, or about 9.4V due to the equal values for R2 / R3.

What are the limitations of the above circuit?

Well there are several, the following sections will analyse these and offer possible solutions.

The Reference

The reference voltage produced by R1 / D1 has a number of limitations, the primary ones detailed below: -

Voltage stability.
Zener diodes exhibit a finite dynamic impedance, or to put it another way, if the voltage in changes (V1) the current flowing through R1 also changes. Using Ohms law, the nominal current through D1 is: -

(Input voltage - reference voltage) / R1

This gives

(12-4.7) / 6800, or 1mA.

It's easy to see that if Vin varies, the current varies and as a consequence D1's voltage will vary. If this voltage varies so will the regulator's output voltage.

To address this issue we can do several things, namely: -

1. Reduce the dynamic impedance of the reference, by choosing a different Zener, or maybe an active band gap reference.
2. Reduce the voltage variation, via another Zener, or even linear pre-regulator.
3. Reduce the current variation by using a constant current source in place of R1.

The details of the above solutions are left as an exercise for the reader, for reasons that will become clear later. An elegant solution is always a nice thing to have, so analysis of the other circuit contributions is worthwhile first - sometimes a single cure exists form multiple problems, preventing a lot of additional complexity and cost.

Reference Noise.
Noise generated within the reference (or as a result of variations around it) results in noise at the regulator output. Numerous options are available for addressing this problem, which can be applied individually, or in conjunction with each other, namely: -
1. Choose a quieter reference
2. Filter the reference directly, or via a low-pass filter
3. Feed the reference from a quieter supply

As can be seen at least one of the solutions is common to the solutions above - a 'two birds with one stone' solution, i.e. elegant! In reality several options are likely to be implemented in our search for the ultimate in performance.

Series Control Element

Q1 has a number of parameters that require careful selection for the regulator to work as intended.

It's gain (hFE), in terms of absolute and dynamic variation, affects the amount of current required from X1 in order to control it, if too low X1 cannot provide enough base drive and the regulator's performance will drop. It has a direct effect on regulator output impedance. It should be noted that FET's of any description are a fundamentally bad choice here, since gm is much lower than bipolar parts, by almost an order of magnitude.

The device's AC characteristics impact numerous, sometimes subtle parameters - junction capacitances affect frequency response of the regulator as a whole, line rejection at high frequencies and even stability of the error amp if care is not taken to ensure phase shifts introduced do not affect stability margins.

Sometimes a darlington pair is used here, but ac performance can be severely affected by such choices, so choose wisely.

Error Amplifier

Probably the part that has the single most important effect on performance. If using an op-amp there are a number of parameters that affect performance, some obvious, some more subtle, but all equally important: -

	Operating voltage This needs to be chosen with a view to the raw supply voltage, and the desired output voltage.
	Power supply rejection ratio Affects PSRR (line rejection) of the regulator as a whole - noise on the op-amp rails finds it's way to the regulator output, attenuated by the device's PSRR. Absolute value and bandwidth / response shape are equally important here.
	Open loop (small-signal) response The open loop gain of the op-amp is primarily responsible for the output impedance of the regulator. Whilst precision op-amps with excellent DC performance are readily available, we're interested in the DC and AC performance, so the small signal gain available at high frequencies is also very important. It is here that op-amps start to show their limitations, many have very low corner frequencies for the open-loop response shape, meaning a 6dB / octave rise in output impedance from a very low frequency. This is a direct result of the dominant pole compensation schemes of unity-gain stable op-amps. Look at the open loop response carefully.
	Noise This affects noise at the regulator output. Both current noise and voltage noise can be important, depending upon the circuit impedances present. In most cases impedances are low, and voltage noise dominates. A subtle effect matters here too - the op-amp input stages that are best for low noise are the worst for non-linear behaviour in the presence of high-frequency signals beyond the intended bandwidth of the device. Non-degenerated bipolar inputs are very prone to rectification of high frequency signals (RF, high-speed digital noise) and can at best cause non-linearity, at worst oscillation. Additionally noise corner frequencies vary - make sure noise does not rise significantly at low frequencies.
	Output capability The device must be capable of driving the pass device or it's drive circuitry without non-linearity, or capacitance-induced instability.
	Bandwidth Almost all of the above parameters are affected by bandwidth - for the supply to remain as close to perfect DC as possible, AC errors must be minimised, hence the AC performance of the amp is crucial to the result. You will not actively manage the regulator up to very high frequencies though, this is best done passively, so look at the circuit bandwidth and the audio range of interest, examining performance parameters in that light.

With all the above parameters you MUST look beyond the numbers, to the AC response graphs in the data sheets. The manufacturers only give you the headline best figures, look deeper and things tend to look much worse.

Safety

One thing that should not be overlooked here is potentially we could be powering thousands of pounds worth of precious investment, from a few pennies or pounds worth of components.

What would happen in the above circuit if we drew too much current from the regulator, or one of it's components failed? The answer is it could get nasty, there is no over voltage or over current protection in the above circuit.

Adding these elements is not difficult, but it does raise complexity, and care has to be taken to ensure performance does not suffer. Suffice to say an elegant, simple and reasonably safe solution is available, which we will see, as soon as the schematic for the Jung regulator is presented.

Onto the Sulzer circuit

[ Home ] [ Up ] [ News ] [ Links ] [ Andy's Audio Journeys ] [ Projects ] [ Andy's Guestbook ] [ Press Releases ] [ The Products Page ]