This page discusses power supply issues and describes the operation and performance of series and shunt regulators. A guide to the location of examples to be found on this site is given at the end of this article.
Most people tend to be a bit skeptical about supply regulation, not without reason. However; it is not generally recognized just how complex the behavior of a passive power supply actually is. The combination of transformers, chokes and capacitors usually results in a resonant circuit. The first curve shows the modeled Z vs F performance of a typical capacitor input PSU. The rectifier and transformer DC resistances are lumped as 100 Ohms (which is on the low side if using a tube rectifier). R1 (40 Ohms) is the choke DCR. The LF impedance spike is typical of such a supply. Furthermore, the supply impedance does not fall even to the equivalent series resistance of the capacitor until 200Hz: Removal of the reservoir capacitor C5 to create the popular choke input arrangement makes little difference; the resonance of the L1/C1 combination prevails. IF C1 is truly non-inductive then the impedance over the flat portion of the curve is governed by the equivalent series resistance (which I have included in the model as 2 Ohms). However, 1mH of inductance is sufficient to cause the impedance to curve back up to 130-140 Ohms at 20kHz. LF resonance will remain until (in this example) the capacitor value reaches more than 500µF. The LF resonance which exists in many if not most passive power supplies, may explain why some people feel that tube amps have soft or "flabby" bass when compared with solid state. It does not have to be so. The take home message is: While passive supplies may appear simple, the LF operation and performance is actually very tricky while the HF performance is governed by the quality of the bypass capacitor (in this case C1).
Please note that you can download free Duncan Monroe's very useful passive PSU designer. Look for Duncan's link on my tube links page.
By contrast, a regulated PSU can be designed to have a constant source impedance of below 1 Ohm from below 10Hz to 100kHz or more.
There are three aspects to power supply regulation:
A: The stability of the supplied voltage: A well regulated power supply exhibits very little change in output voltage as a function of ‘normal’ variations in line voltage.
B: The stiffness of the supplied voltage: The ‘stiffer’ the power supply, the less the supplied voltage varies with load current.
C: The quality of the supplied voltage: A well regulated power supply will not exhibit noise on the supplied voltage above – at the most – 5 millivolts pk. Even using tubes, it is possible to reduce noise level to the microvolt region.
There are two basic regulation techniques:
Series regulation - See Figure 1
This method samples the output voltage via the network R2 and R3. The sample is fed to the grid of an error amplifier, X2. The error voltage is the difference between the grid voltage and the cathode voltage (Vgk). Any changes in the error voltage from the set-point due to variations in the output voltage are amplified by X2; the resulting amplified error voltage is used to modulate the grid of the series pass element X1, such that if the output voltage tries to fall, X1 will turn on more and vice-versa. This method can provide extremely high stability, stiffness and noise rejection. The problem with the series regulator from an audio stand-point is that if an output bypass capacitor is not used, the signal current draw due to the signal applied to the audio amplifier will flow in real-time through the series element X1 and around the loop formed by the Vin filter capacitor. This is fairly tortuous: Thus it is hard to design a series regulator having a level of simplicity commensurate with that of the best audio circuits. Thus, not only is a series regulator bypass capacitor essential but to be effective, it must also exhibit much lower impedance than the regulator across a very wide bandwidth. This is a tough requirement; such regulators can readily be designed to accomplish source impedance in the milliohm range.
Shunt Regulation - See Figure 2
This method samples the output voltage via the network R2 and R3. The sample is fed to the grid of a combined shunt element and error amplifier, X1. (In some high-performance designs, the shunt and error functions are performed by separate devices.) The error voltage is the difference between the grid voltage and the cathode voltage (Vgk). Modulation of the error voltage by the sample is converted by the transconductance of the shunt device into modulation of the shunt current such that as the load current rises, the shunt current falls and vice-versa. Since the total current remains constant, the voltage drop in the feed resistor R1 also remains constant. Thus, for a constant input voltage Vin, the output voltage also remains constant. If Vin changes, Vout out will also try to change causing a deviation from the error voltage set-point resulting in a commensurate change in the total current such that the voltage drop in R1 changes to restore the output voltage. Due to the simplicity of the circuit, these corrections occur extremely quickly. The slew rate of such a circuit may be better than 20v/µsec and so if the output voltage tries to shift by say 0.1V, a correction will occur in less than 5nanosec. The design of a separate error amplifier for a shunt regulator is a little tricky, many people use op-amps. Without the use of a separate error amplifier, the ‘specification’ performance of the shunt regulator is inferior to that of the series regulator. However the signal current draw due to the signal current applied to the audio circuit will flow directly via the parallel shunt current path: this is much ‘cleaner’ than the path formed by a series regulator. This attribute combined with the speed of a shunt regulator means that it is possible to design a good sounding shunt regulator that tracks the signal current demands extremely accurately such that it does not require a bypass capacitor ‘band-aid’.
There is another issue - I often hear about the notion of the sound of the rectifier: If you hear a difference when you change rectifiers then (assuming the supply voltage does not change) the bypass capacitors are not effective in isolating the amplifier from the non-linearity of the rectifier circuit. A supply regulator will improve this situation. In the case of a series regulator, the input voltage is greater than the output voltage by the amount required for proper operation of the series pass element. This additional voltage creates an energy buffer in the reservoir capacitor: That is, the reservoir voltage can be drawn down some (say on transients) without the output voltage of the regulator falling. A well designed shunt regulator will almost completely isolate the circuit from the rectifier since the current demand on the raw supply remains constant.
It would appear then, that the shunt regulator is almost a no-brainer if you are looking to surpass the limitations of a passive PSU. However, there is a big sticking point: If the load is removed from a shunt regulator (say a heater or filament in the supplied circuit fails), the shunt device must be capable of handling the resulting current. In the case of signal tubes, this is usually not a problem but for with power tubes, the current increase may be significant. In fact, the shunt device must be capable of handing more current than the largest tube supplied which for most is not a practical proposition. My current thought is to continue to use series regulators for heavy currents but to deliberately increase the source impedance so as to force the bypass capacitors to work. This way combines stability and freedom from resonance (if the design is right) and very low supply noise with the ability to voice the sound of the amplifier by playing around with bypass capacitor types and combinations.
Examples of regulators to be found on this site:
All tube series type: 845 SE and KT88 PPP amplifiers; also MK1 416 phono stage. (The KT88 design includes a 650V regulator with feed-forward to improve line regulation.)
Mosfet / tube hybrid series type: 300B SE and EL34 PP amplifiers. (The EL34 design is not one I would repeat but is included since it exists.)
Mosfet slow turn-on series type without error amplifier*: 6AS7 PP and Calrad amplifiers; also 2A3 line stage.
Shunt type with error amplifier - all tube, MK11 416 phono stage.
Please Note: clicking on the project names above will open the schematic pdf file. If you want more detail then go to the vacuum tube audio page.
All these designs have had regular use by me and others over several years. It is possible to build reliable high-voltage regulators!
The 2A3 line stage, 300B SE and 845 SE also include examples of filament regulation.
* This design includes a slow turn-on feature for use where a tube rectifier or a slow turn-on rectifier (such as a 5AR4) is not used. Note, I do not include filamentary rectifiers in this category; they turn on much more quickly than is desirable for the longevity of expensive signal tubes.