Transistors in audio curcuits

[MrZwing]

Hello

I have a simple question with probably not so simple answer.

when i look around on google about this circuit people just recommend transistors and not any reason to why except "these are best" witch i fond annoying and uninformative. and since i am a person who likes to experiment rather than just "follow the standard recipe" i would like to know what is it that identifies a transistor as suitable for audio and dive into the jungle myself. simply put what should i look at in the datasheet to see what makes it more och less suitable for a audio circuit and why?

/MrZ

[Benta]

Well, the first hint is, that the supplier himself calls them "Audio Transistors" in the datasheet.
This implies that the device is well characterized for linear operation (as opposed to switching) and  that care has been taken to focus on parameters relevant for audio, eg, linearity, noise, frequency response, power capability, SOA etc.
Another good hint is, that both NPN and PNP devices with the same characteristics are available.

[BrianHG]

What's your driving application.  Headphones? 5-15 watts? 100 watts?  Class AB, or class D?

For headphones, any small signal transistor with an above 200ma should do.

For 5-15 Watts, any medium sized transistor, something in a TO-220 will do.

Now, for 100 watts class AB, there are some transistors better than others.  It depends on the quality, SNR and your base driving circuitry, but, for full power at at full 20Khz bandwidth, the simpler and properly specked the transistor die with proper driving will work better than going for the all brute force approach.  Example, super high current mosfets have really high gate capacitance and they are designed to operate as a switch, not to be operated in the linear domain.  This also may make them noisy.  There also wont exist well matched complementary pair N-Type and P-Types.  When online if you are told to use specifically, if your source is good, usually it's a chosen complementary designed by the manufacturers for linear audio and servo applications.

The final main reason you may be told to use specific devices is that the circuit you are looking at may rely on specific transistor beta, or, VGS specifications to operate without going through thermal runaway, or, leave you with crossover distortion on your output.

[shteii01]

Just had something similar at AAC.
 
1.  Audio is defined as frequencies from 20 Hz to 20 kHz.  So.  Like someone already said, you start looking in the datasheet at frequency bandwidth/gain bandwidth product/frequency response.
2.  Someone else at AAC pointed out that while we amplify the voltage of audio signal, we drive the speaker with current.  So.  How much current can transistor send to the speaker becomes important.
3.  One interesting option is to use Power transistors, they do both, voltage and current.  In one of the Amp Hour episodes the guest was building some crazy power supplies and said they often use audio components (audio amplifiers?) because they can provide voltage and high current that normal/regular amplifies don't provide.

[T3sl4co1l]

when i look around on google about this circuit people just recommend transistors and not any reason to why except "these are best" witch i fond annoying and uninformative.

Well, there you go.  Because people aren't smart enough to figure things out for themselves, they much prefer to copy something, say "me too!" and pat themselves on the back.

That's why people pay me to do things right...

...That said, what do I look for in an "audio transistor"?

There isn't too much to choose from, between transistors.  MOSFETs are MOSFET-ey, and BJTs are BJT-ey.  They all have nonlinearities that the designer must address: base current, transfer (output current vs. input voltage), output characteristic (MOSFETs have a more exaggerated saturation (ohmic) region), and terminal capacitances.

So, the aspects we can vary and choose, mainly come down to:
- BJT hFE: flat.  Does not drop off significantly at low currents, and stays reasonable up to the peak current needed.
- Capacitance: smaller is better, as always.  Although, do avoid RF power transistors: they'll be more hassle than they're worth (which is a lot, anyway; RF transistors always seem to be 2-10 times more expensive (and more!) than comparable general purpose power transistors).
- Less variation in capacitance, as well.  SuperJunction MOSFETs are fantastic for switching, but may not be all that great for linear applications.  (I've been meaning to try this -- I expect the result is wide frequency response in the middle linear range, but terrible saturation recovery as you approach maximum power output.)
- Linearity of transfer function: MOSFETs win in this regard, as they have a quadratic transfer curve.  It's not linear, but it's a lot closer than the exponential of a BJT.
- High gain: that said, BJTs have more gain, so that by the time you reduce the gain to the same level (by application of negative feedback, usually by emitter degeneration resistors, emitter follower connection, or global feedback like an op-amp circuit), the distortion is lower again.
- Wide SOA: all linear amps have one thing in common, lots of power dissipation.  Both BJTs and MOSFETs exhibit 2nd breakdown (where hotspots form on the die, causing runaway failure).  You need to pick parts that have a high enough limit.

Historical references proclaim MOSFETs as "free from breakdown", which should be read as: "we're too lazy to make a good MOSFET, so the low power density is a feature, not a bug!"  MOSFETs, from the last decade or two, are more than dense enough to require reduced power ratings at high voltage.

BJTs and MOSFETs made for linear applications, are either made physically larger (like the old ones were by accident), or are made with positive-tempco emitter/source ballasting to minimize hotspot formation.  Which really means, 2nd breakdown is still there -- it's just at such a high voltage or power level that you've already melted the package, so don't worry about it.

(For the same reason, high-resistance packaged parts are "free from 2nd breakdown", because they simply can't be cooled enough to begin with, to get into the operating range where hotspot formation is fatal.  Small air-cooled packages, like SMTs and TO-92s, and cheap insulated (full pack) packages, fall into this range.  It's the same as buying the regular metal-tab version and simply not running it at full ratings. )

So, BJT or MOSFET?  An open design choice, really.  MOSFETs tend to be more difficult to use, because of their lower gain and higher G-S voltage drop.  BJTs have a low enough B-E voltage drop that using an emitter follower, directly, with no special attention to drive voltage, gets you very close to the supply rails (within 1V).  MOSFET amps usually address this by powering the drive circuitry from slightly more voltage (say, +/-55V rails, where the output power is +/-50V).

Tim