|
|
|
 |
Amplifier design for dummies It's not rocket science, you just have to think things through.
Most amplifiers today consist of receiving some magnetically or digitally derived signal and drivin a magnetically operated device - a speaker.
You get variously linear and non-linear devices employed from BJTs (bipolar transistors) to MOSFETs, Valves or Tubes to do this in class A, AB, B and D. Class C is reserved for tuned circuits - i.e. radio transmitters.
Amplifier biasing: class You should probably already know about Class A, AB and B by now. You may not however realise something about class A and class B techniques. Single-ended amplifiers have only one active device in the circuit, which gives an asymetrical drive - it is however class A - always conducting.
To view the classes you have to imagine a +ve rail on top, an active element (transistor or tube) followed by the output followed by another one connected to a -ve rail, like a totem pole. the top device swithing on then swings the output up, the lower device swings it down.
| Class | Description | Pros | Cons
|
|---|
| A | Both devices conduct, wasting all the power. Power is merely diverted to the output instead of the other device, so power consumption is constant. | This takes most of the power-supply (PSU) out of the sonic equation, simplifying the sonic path considerably. | It uses a lot of power
|
| AB | A mixture of class A and B. Will be class A at low levels - almost ideal. | Loses dynamics as the PSU hops in and out. Class H is an extension of this. | Not many cons, but the class is vaguely defines
|
| B | One device is usually off. With no output, both are. | Low power | For BJTs you have switching effects which cause distortion of low level signals. Curing these effects can make things worse. The PSU is also completely involved - so you listen to that too.
|
THD Total Harmonic Distortion When you listen to Jimmy Hendrix etc. on your uber-fi, you may be interested to know the huge amounts of distortion that have been added to the guitar and vocals even before it gets to the mixing desk. Just like radioactive dirt is not the same as fine compost, all distortions are not the same. The THD (total harmonic distortion) measurement is completely pointless as by adding all the different distortions together destroys the information in the measurement.
For example, 2nd harmonic sounds lovely, but 3rd harmomic sounds awful. So a lovely and awful sound can have the same THD figure - this figure tells you nothing, except if it is low the amount of NFB used may be too high.
Feedback A long time ago a person/company called Williamson discovered that by feeding the output back to the input of a (tube) amplifier could reduce noise and distortion, and flatten the frequency response, by suppressing/controlling the lower frequency gain.
Some people think that adding NFB widens the frequency band, which technicly it does, but it does not extend the high frequency gain, it merely flattens the low frequencies - i.e. the amplifier is not intrisically 'faster', it just behaves at a lower gain.
Doing this the wrong way turns amplifiers into oscilators and kills them, plus any attached speakers. Early UK and US transistor amplifiers did the same, but they reckoned without the effects of phase change delay.
DIY Audio discussion about NFB
Delay There is a small delay between a signal at the front of an amp and the signal that comes out of the end. This is much bigger than the speed of light, as it is governed by various parasitic capacitances. In effect it means that if you feed a negative (-ve) signal into an amplifier from it's output then at a certain tone (frequency) the amplifier will see that as positive (reinforcing) feedback and turn into an oscillator. Tubes (valves) are very forgiving of this, MOSFETs aren't to bad either. The quality of the PSU is also important, a class B BJT amplifier with a pithy PSU and good amount of negative feedback is therefore most likely to self destruct, as many US and UK ones did. The Japanese amps didn't, which is why their retro hi-fi is still around over 40 years later, still going strong.
Smoothness When a designer uses a lot of feedback he is demonstrating how people fail to learn from lessons gained years ago. BJT transistor amplifiers commonly sound brittle, scratchy, screechy, unpleastant to listen to loud, agressive and just plain wrong. To eliminate the crossover distortion introduced by using class B (and a lesser extent class A), and to try and hide any PSU issues, feedback is added. Negative feedback (NFB) however does not cure the problem, it merely converts it into something else that is rarely measured, because people rarely look for it.
This distortion is a temporal distortion, cause by echos of delayed signals circulating round and round the feedback loop many times. At high frequencies and big impedance mismatches (i.e. a BJT directly driving a speaker) these will become very audible as described above. That is why IM (intermodulation distortion) is a good measurement as it can see this a little, but it cannot see the whole effect.
There is a paper about this here and here
although the best paper I've read is this one from Royal Device: http://www.ossaudio.com/WhitePaper-RoyalDevice.htm
The Best Amplifier An amplifier needs to fill the following criteria if it going to try to be any good
- Class A or heavy Class AB
-
A Good PSU
-
Low/no negative - certainly no long loops
-
Impedance matched speaker outputs is useful (usually transformers)
-
Good devices
-
Hook it up to easier speakers
Basically the intrinsic design needs to be right, as you cannot correct with NFB because that ruins the sound. This causes you to build a better amplifier - not always more expensive to do, and a better amplifier will have a smaller NFB signal anyway, because the original output is more accurate.
The Usher R1.5 sounds good because it uses a good PSU, good devices (paralleled Motorola BJTs), Class A/AB and therefore needs very little NFB. An old Sansui 1000A running at 10watts bias will also blow most modern amps away because it has most of the essential ingredients.
The last point is that some speakers are easier to drive than others, so the ultimate match would be for example an output transformer connected to an alnico and/or piezo driver... in fact just like the new Fender Tweed Deville guitar amps, which sounds awesome and uses many of my points.
| | | | | |
|
|
