Hi there!

I just built a small adjustable LM317/337 PSU and forgot the output capacitors. The circuit didn't work at all as intended, with astonishing jumps and breaks in the output voltage. After inserting those caps it worked like a charm, still does.
But that would be my question: I read and heard about unwanted oscillations in a circuit. But I don't really understand it. What is meant if it is said that a circuit oscillates? And I don't mean signal generators or intended oscillators.
Where do those oscillations come from? What does it mean if an opamp oscillates?

Regards
Christian

 

Unwanted oscillation means that the circuit is unstable.

Usually it comes from positive feedback, that keeps the circuit from completely stabilizing.

There can be too much overshoot in the negative feedback, as if you are driving and every time the car goes to the left, you turn the wheel to the right, and then it goes too far right, and then you turn to the left again, but too far... and you look like you are drunk driving down the road.

Did you ever ride on a bicycle with a bag of groceries hanging from your handlebar? If there is too much mass in the bag, you can get into an unwanted oscillation and maybe crash the bicycle.

In a circuit, the unwanted oscillation can sometimes come from too much capacitance on the output of an op amp, or it can be from parasitic capacitance from the output to the non-inverting input...

But this is a really big topic. There are many, many reasons a circuit can contain unwanted oscillations. Sometimes, it can be from EMI (electromagnetic interference) on high impedance signal lines, and so the oscillation is not actually *from* your circuit even though it appears in your circuit.

There are also many strategies to protect against unwanted oscillations, like compensating feedback loops for op amps, and using low-pass filters whenever you do not need high-speed response, etc. There are many, many book documents written on this topic so I will not try to explain any more. I hope this was helpful.

 

Welcome to AAC!

Anytime you have lots of gain you have the potential for oscillation. Voltage regulators have some very high gain internal to themselves to provide quick corrections (with accuracy) to variations in loads on the output. It is a double edged sword though, they have to plan of the user providing the capacitors to kill the probable oscillations, since the capacitors are simply too large to include on the chip die.

 

If we limit discussion to oscillation in amplifiers, as opposed to external interference, there are a few major categories of instability I can think of - I don't suppose this will be comprehensive, but here goes:

Perhaps the most basic problem is excessive delay (phase lag) around a deliberate supposedly negative feedback path. This can result in the feedback suffering 180° phase shift at some frequency, so that it effectively becomes positive feedback. If the loop gain at this frequency is unity or more , we would expect oscillation. This is a very considerable subject of study, and quite a mathematical one. Typically, op-amps are provided with frequency compensation, that is, deliberate tailoring of their frequency responses to facilitate stable operation. Often this compensation is completely internal to the device, but sometimes external components are needed.

Some amplifiers are stable with internal compensation in circuits designed to give down to unity gain, but others are only stable down to some minimum value. These latter types may not be suitable for instance for use as unity-gain buffers, or may require external compensation for this purpose.
These links may be of some interest:
http://en.wikipedia.org/wiki/Frequency_compensation http://www.intersil.com/data/an/an9415.pdf

Amplifier stability can also be conditional on external impedance conditions at the input or (particularly) the output. Voltage output amplifiers and voltage regulators may have restrictions on the permitted load capacitance. Similarly, current outputs may be unstable with inductive loads. Resonant loads, such as tuned circuits and some acoustic transducers, need special care and may require networks to control the load impedance.

In addition, instability may result from unwanted or parasitic feedback. This may result from impedances in the power supplies, and this is one reason that it is usual to connect capacitors across the power supply inputs to devices - to try to minimise supply impedance. The danger is that fluctuating currents drawn by an amplifier can cause supply voltage variations which may affect earlier stages in the amplifier, causing more current fluctuations...and so on.

Signals may also be coupled from output to input circuits by stray capacitances - e.g. between components, wiring or PCB traces. Wiring, or in some cases entire circuits, may require to be surrounded by an electrostatic screen to minimise this. Similarly, unwanted magnetic couplings may exist. These are more likely between wound components like coils and transformers, but may be significant even with wiring at higher frequencies. Unwanted couplings may also be due to common inductances or resistances shared between circuits. Common impedance is a particular issue for ground connections.

You will see that stability is a big subject, in fact many books have been written about it. For a practical hobbyist, it is important to refer to any data-sheet or other information you might have available for the devices you are using. Any requirements for compensation components, minimum closed loop gain, output loading, and supply decoupling should be followed carefully.

Additionally, any more detailed manufacturer's indications about layout should not be ignored. If a manufacturer shows an physical layout in some application notes, quite possibly there are some pitfalls in using the device - like a tendency to unwanted feedback- which this layout is designed to avoid.

Developing good habits for layout is also important, for instance avoiding untidy straggling connections, particularly in critical areas like the inputs or outputs of high gain amplifiers. It is obviously a good idea to keep inputs and outputs apart, as far as possible. Never forget those supply decoupling capacitors!!!

This can be more difficult using things like solderless breadboards or soldered strip-board. It may be useful to avoid using adjacent strips for the input and the output, and perhaps to use grounded strips to separate them.