Paralleled Transistors

Using multiple bipolar transistors allows not only to distribute temperature between them, but also to lower their collector current. It improves linearity since current gain of transistor drops at larger collector currents (beta droop). Putting transistors in parallel also lowers output impedance, without affecting stability (same emitter resistors). Mosfets don’t have "beta droop" and suffer less stress at high current, while putting them in parallel increases chance for parasitic oscillations, so perhaps it is only good idea for bipolar transistors. Of course you have to use more than one if available transistors don’t deliver required power. Perhaps we have a member who designs amps, to chime in?

If this is true, we've never seen it born out in distortion, square wave or any other measurements. 

To the contrary, ultra simple circuits often have the most "color" added by distortion or high output impedance.

The people over here (including Nelson Pass) would probably have some relevant experience and opinions:  https://www.diyaudio.com/forums/pass-labs/

Pass is a good example, because while it has many fans, their uber simple design does not win everyone over for sound quality.

I think if the idea of transistor smearing was accurate or a convincing win no one would buy any amp that used more than one pair. Instead these types of designs, and tube equivalents, remain a niche sound.

I encourage the OP to find a FirstWatt kit and make his own, see what conclusions he comes to. :)

Best,

Erik

Not really. The more transistors paralleled results in less current through each transistor. The benefits are: less distortion and less of a beta drop, making any inherent beta mismatch not a great factor (and further reduced by emitter resistors that enable each to share current more evenly). The reduction in distortion, especially at 4-ohm loads and lower, probably negates any mismatch.

However, there is no free lunch. More transistors in parallel will make it harder for the amp to control parasitic oscillations. It will also require the voltage amplification stage bias circuitry to source more current for the multiple transistor bases, which may also require inserting a driver stage before the output. All of this drives up the cost rapidly. Considering there are excellent high power transistors available (On-Semi NJW0281G), it doesn’t make economic sense to put a lot of transistors in modest powered amplifiers.

As far as mosfets go, Nelson Pass is the one to ask. Mosfets are less linear, have higher internal capacitance (must be charged/discharged very rapidly at high frequency signals) and they require voltage swings of 5 volts (instead of 0.6) at the gate for them to conduct. Yet Pass makes them work great.

Very interesting and informative posts .....thank you to all. I recently auditioned a Danish Gato amplifier which uses one pair of large N Channel Mosfets per channel.(200/400/700 wpc) The sound was very pleasing, with a natural tonality, similar to a class A Gryphon type sound. Curious!

Short answer: no. Everyone parallels devices. Also paralleling them is not really related to complementary pairs (NPN and PNP) which are used to achieve DC coupling to speakers. Which almost everyone does.
Exceptions: tubes and some really oddball, very low power FET amps.
I’d argue that if you want to go to the trouble and cost to parallel up a gazillion devices, you could get some theoretical benefits. Nelson Pass certainly thinks so.
G
Ps: after posting this i saw that Nelson's name was being bantered about above, but for the opposite reasons i was. Nelson once designed, or tried to, a power amp using hundreds of tiny N-channel JFET TO-92 devices that he really liked.
pps: In sane-land, i recently completed  a design for a headphone amplifier with no loop/global feedback. This adds to the output impedance issue noted above, but is inherently stable into any load (including some crazy impedance 'phones). Among other things, it parallels quite a few to-92 NPN and PNP devices to achieve power handling temperature margin, and linearity. And it just rocks.

There was an extensive discussion on Audiogon about a year ago, when the late great Roger Modjeski hosted a thread entitled something like, "Ask an amplifier designer a question."

We discussed paralleling and matching transistors at some length.

Further to your question, I like epitaxial output transistors from Sanken (beware of counterfeits). Their higher power transistors tend to be slower, which you can absolutely hear as imprecise. Hence, if you want to hear the best your topology can do, you have to parallel.

Parallel transistors, Moffett or tubes is always not a good idea. It is just a compromise - to deliver more power.

I'd just like to reiterate, you might get some really interesting discussion on the pass labs section of diyaudio.  As you can tell above, Nelson's designs are cited both as evidence in support of paralleling devices, and for simplifying and getting down to minimal parts.  Because of that, the crowd over there is actually supportive of both approaches and might give you some additional thoughts once the thread here runs its course.

Thanks again for the helpful information! I guess paralleling transistors is the more favorable approach when intermediate and high power is desired. Brands like Gamut and Gato are the exceptions, using one pair of large N- Mosfets per channel.

This is off topic but I wonder if someone can briefly explain why my high power class A Clayton S-2000 amplifier draws 6.2 amps when first turned on and then drops to 4.2 amp draw after 25-30 minutes....perhaps it is stabilizing its class A bias?  Thank you to all respondents!

I wonder if someone can briefly explain why my high power class A Clayton S-2000 amplifier draws 6.2 amps when first turned on and then drops to 4.2 amp draw after 25-30 minutes....perhaps it is stabilizing its class A bias?

That is kinda strange, i would be normal to be very high for 1-2 seconds as the inrush charges the capacitors int he LPS, but 20 minutes must be some kind of system governance. I would venture a pure guess: i runs very high bias until it hits a thermal level and then the bias is turned down to protect it from excess heat or possibly thermal runaway.
My designs employ a thermal loop as well, but it constantly hunts within a preset range - no big jumps.
G