Tube DIY Asylum

Posted by Henry Pasternack (A) on April 20, 2009 at 10:56:11

Introduction
I said earlier I would write about choke-input power supply design. This is a subject I got very interested in about fifteen years ago. I read as many old books on the subject as I could find and tried to understand the important technical factors. Given the atmosphere around here I feel as though I ought to make a disclaimer. I know from personal experience it's possible to build amplifiers with "conventional" choke-input supplies that sound very, very good. I can't say in any given application that some other configuration might not work better. There are too many interdependencies and personal subjective preferences to make hard and fast rules.

When I start a project, whether it's an audio project or not, it's because I have an idea and I want to see how it works in practice. Usually I spend a lot of time on research and thinking through what I plan to do until I'm comfortable that I understand the project at a thorough nuts-and-bolts level. Then I build it, and often there are surprises and things that need to be rethought. But finally I'm done and I have some results. It's usually pretty good and usable, and I'm satisfied and have no urge to tear the thing apart and tweak it. That's just me, and others may seek different rewards.

My goal here is to give useful information about how power supplies work, perhaps to help people build and evaluate their own designs. I make no claims as to one thing sounding better than another. I know there are many, many different ways to build good-sounding amplifiers.

Overview
What makes a power supply a choke-input supply, evidently, is the filter. The filter is a block of components that sits between the power transformer/rectifier and the amplifier circuit. This is true in general for all power supply filters, regardless of their type. The filter has four main functions, different but related. They are:

1) Provide a load on the rectifier such that it switches on and current flows, in one direction, out of the transformer and into the filter.
2) Store up the charge that comes out of the transformer and deliver it to the load even if the rectifier is not conducting.
3) Provide a low source impedance to the active circuitry at all frequencies from DC to the ultrasonic.
4) Reject hum and noise and keep it from getting through to the power supply output.

So, you could say that a power supply is "good" if it does all of these things well. But everything is a compromise and we may not know for sure which of these factors is most important in a given application. That is where the designer's judgment comes in. I believe the better you understand the problem in theory, the better you are prepared as a designer to make design choices and to evaluate subjective performance results.

What you have here is basically a brain dump of everything I can think of to say about this subject. Because of the amount of material, I will have to do this in parts. It's a snapshot of how I think about the problem, not necessarily a recipe for design. I hope someone finds it interesting or useful.

Rectifier Load
The rectifier is a switch, which is a non-linear component. The rectifier only conducts when there is a positive voltage difference between the transformer secondary and the filter input. It is physically impossible for a single rectifier diode to conduct throughout the entire secondary sinewave cycle. It is also impossible for current to flow backwards out of the filter into the transformer. The diode will always turn off at some point. The current drawn from the rectifier, and the voltage waveform that appears on the filter input terminal, depend strongly on the electrical characteristics seen by the rectifier looking into the filter and on the DC load current drawn from the supply. The DC voltage at the output of the filter, in turn, depends on the shape and the amplitude (that is to say, the average value) of the filter input voltage waveform. If the average DC input voltage changes substantially with load current, the supply will have poor regulation.

In this sense, the transformer, rectifier, and first filter input component are like a kind of complex, ratcheting electrical machine which functions to turn the AC voltage coming out of the transformer into a raw form of DC that can be purified. The complexity of the interaction between these three components is what makes the filter topology and especially the value of the first filter component so important to the performance of the supply.

The main distinction between a choke-input filter and a cap-input filter is that the choke-input filter loads the rectifier so that current flows into the filter almost continuously. If the choke were infinitely large, the current would be constant DC. With a practical choke, the current into the filter consists of a DC component equal to the DC load current with a superimposed AC current approximately equal to the ripple voltage divided by the choke impedance, summed across all the harmonics of the AC supply frequency.

This helps explain the definition of "critical inductance". Ignore the higher order harmonics and only consider the dominant 120Hz (or 100Hz) component of ripple voltage. The first capacitor following the choke, if it is large enough, looks pretty much like a short at this frequency. So the entire ripple voltage appears across the choke and the resulting current is proportional to the ripple voltage and inversely proportional to the choke inductance at the ripple frequency. The higher the voltage or the smaller the choke, the greater the ripple current. To keep the choke conducting, the DC current bias has to be large enough that the negative peaks of the ripple current never reach zero. It is possible to derive a simple, approximate formula for the the minimum inductance to give continuous rectifier conduction as a function of DC load current. The familiar formula is:

Lcrit = Vdc / Ima

where Lcrit is the minimum (critical) choke inductance, Vdc is the supply DC output voltage, and Ima is the load current in milliamperes.

There are three main reasons continuous current conduction is important. The first is that a continuous current with a lower peak value makes more efficient use of the power transformer. This is because power lost to heat is a function of the square of the instantaneous current. You generate less heat conducting a lower current for a longer time than a higher current for a proportionally shorter time.

The second reason is that the smooth, relatively low peak supply current creates less switching noise and electromagnetic interference to pollute the signal circuitry.

The third reason is that continuous current conduction gives much better DC voltage regulation. This is because the duty cycle of the filter charging current doesn't vary as long as the load current exceeds the minimum threshold as calculated by the critical inductance formula. And therefore, the filter input voltage waveform is constant, and the average DC input voltage is constant, too. Contrast this with a cap-input filter where the charging duty cycle is a function of load current. This is a bit of a subtle point and may deserve some follow-up discussion. In the old days, when large filter capacitors were not available, improved regulation was a major advantage of the choke-input filter. This distinction is less important nowadays.

In practice, the diode conduction current is not continuous. This is because diodes do not conduct until their minimum forward voltage drop is exceeded. So even with a critical choke-input filter there is a dead zone in the conduction waveform and the potential for generating switching noise. Solid-state diodes can also conduct backwards momentarily as minority carriers get swept away, creating another source of switching noise. But the noise in general is less than that produced by a cap-input filter, and there are techniques for "snubbing" the noise (e.g., as documented by Jones in "Valve Amplifiers").

To summarize, we can say overall, from the point of view of efficiency, noise generation, and regulation, the choke-input filter provides a superior interface to the rectifiers, and this may be a factor in the subjective perception that choke-input filters give better sonic performance.

(To be continued)

Edits: 04/20/09   05/04/09