19-05-2011, 05:29 PM
(18-05-2011, 03:58 PM)ThePillenwerfer Wrote: Thanks for your suggestions, Mark. I'll await the breadboard as messing about on veroboard is courting disaster. I know you know your onions with regards to PSUs so how about telling us (sans maths!) why things like de-coupling are needed. I've never seen the point of putting a cap across a battery as it's as smooth as it gets and there won't be any mains-borne interference coming from it. In fact it would contribute to turn-on transients by taking time to charge. Obviously there are good reasons but I don't know what they are.
Hi Joe,
Sorry for the delay; busy day!
The first concept is that, to a signal, a supply rail looks like ground (or at least, it should). I accept this might need some thinking about if you've never heard it before, but I promise it's true. If you need convincing, think about the model of a "perfect" voltage source: an absolutely constant voltage with no internal resistance. In other words, if a signal current is applied to this voltage source, no voltage change would be observed.
Now, I also accept that the last point might need some thinking about as well. Surely a signal is a voltage, right? No, that ain't necessarily so (Gershwin's Law). If you apply a signal voltage to a resistor, the current flowing in that resistor must vary in sympathy with the voltage. So there is a signal current in that resistor.
Funnily enough, transistors (bi-polar and FET) and valves are Voltage In, Current Out. The resistor in the collector (or drain or anode) circuit is used to convert the current back into a voltage...
Anyway, I digress. In an amplifier, there are varying voltages, and varying currents. Now, an amplifier designer will make assumptions about the power supply - (s)he will assume that the supply rail(s) are "ground" as far as the AC signal is concerned. Clearly, they are not "ground" as far as DC is concerned, but it's important to separate out AC and DC conditions in any circuit...
The easiest way to make a DC supply rail "look" like ground is to connect a capacitor to ground. We know that capacitors block DC but allow AC to pass; job done
You mention a battery; they can be quite good, but they do depart from the model of the perfect voltage source because they have some source impedance. In other words, think of your PP9 battery containing a "perfect" 9V voltage source, and a resistor of a few ohms between it and the terminals. So now, as you deliver audio into your loudspeaker, the varying current demand results in a varying voltage across the internal source impedance - or, to put it another way, the voltage observed at the supply rail will vary according to the audio content.
Of course, as the battery runs down, its internal resistance rises, hence the variations get bigger. As a result, the poor amplifier gets horribly confused by the supply rail waggling about all over the place...
Naturally, the wire between the power supply and the amplifier will have some resistance, and possibly inductance too. The latter is a problem, and likely to cause issues at higher frequencies - hence it's common for amplifiers to spuriously oscillate at high frequencies when not properly decoupled.
Pick any amplifier IC at random, and the datasheet will make pretty specific recommendations about decoupling. I was surprised that the LM386 datasheet doesn't mention this, but I would always include it...
Next, you have an LM317 regulator. These always need decoupling (although plenty of people forget and often get away with it - or seem to, at least). What's inside a voltage regulator? Primarily, an amplifier. Yes, there is a voltage reference and a big hairy transistor to deal with the power, but the amplifier is the key element. And, as a rule, amplifiers always need decoupling. A gross simplification, but well worth remembering
Here, you won't need to decouple the two ICs separately - the recommendations in my previous post will do the job for both of them.
Oh well - I tried to be brief! Hope this helps, and sorry if I didn't get the "level" right. At least there was no math
Mark