2 or 4 Output Transistors per Channel

[gort69]

Probably a dumb question. Speaking of TO-3 devices, curious what is the determining factor(s) for number of outputs per channel. Was it voltage, watts output, or something else? Most of what I own is 80WPC or less, and all use 2 per side. The most powerful - 130WPC - uses 4 per side. Thanks.

 

[DRecovery]

Some devices and heatsinks are better than others and some designs are more aggressive than others, but ultimately multiple parallel outputs are used to share the power dissipation and keep any one device from becoming too hot.

There is an argument to be made that the answer could be output current, with paralleled devices exhibiting less gain sag at high current, but ultimately those devices could still deliver the desired current if driven hard enough. Nothing will stop them overheating, though, so I say it's still power.

 

[gadget73]

Largely its current related. Four per side will handle basically double the current that two per side will, assuming there is enough heat sink. Double the current at the same voltage is double the wattage. You could get double the wattage with double the voltage too, but there are limits to what a transistor will deal with.

 

[BinaryMike]

Not a dumb question. Beta droop (gain sag) at high current causes distortion to increase by placing a heavier load on the driver and voltage amplifier stage, which reduces its gain and therefore loop gain. Also, paralleled output transistors yield reduced crossover distortion even with light loads. This effect by itself could convince an amplifier designer to double up on output devices in the high end of a product line, even if the extra power dissipation capability wasn't a factor (but it usually is).

 

[Alan0354]

How many transistors used depends on the (1)power handling, (2)distortion and (3)beta droop at high current.

(1)Power handling is very obvious, you need more transistors to handle more power. Don't just look at the 200W rating of a lot of transistors, you have to look at the SOA ( safe operating area) curve of the transistor. At higher temperature, the 200W transistor cannot handle even close to that.

(2)The following ONLY applies to the most common emitter follower output stage(EF) in class AB power amp. Main problem of class AB is the crossover distortion. Crossover distortion of output stage depends on the load it's driving. Ball park is distortion double when the load it is driving is halfed. That is distortion almost doubled from the spec of driving 8ohm load if you use it to drive 4ohm load. Or look at it in another way, the distortion halved from driving 8ohm if the load doubled to 16ohm. This is very important to know and use it as guideline.

Suppose your amp has only one pair of EF as output stage ( complementary output pair is NPN and PNP). So the one output pair is driving 8ohm load and produce a certain amount of distortion. Now if you double the output pair to two pairs. So two pairs are driving the same 8ohm load. BUT NOW, each pair is driving half of the load.......That is as if each pair is driving a 16ohm load ( half the amount of load of 8ohm). So, the distortion is almost half.

Now the same theory applies further, If you amp has 4 pairs of EF stages, each pair drive 1/4 of the load or 32ohm. Distortion lower another half from two pairs or 1/4 of a single pair. Same theory applies to 8 pairs...........and so on. Of cause there are law of diminish return. Anything over 32ohm load is so low it's not significant anymore. But why people use 8 pairs? this is because impedance of the speakers are not constant with frequency. Some dip as low as 2ohm or under. If you look at the more exotic speakers, they are mostly lower than 8ohms, more like 4ohm. It is reasonable they dip even below 2ohm. That's when you need more pairs. That's the reason I design with 9 pairs of EF pairs in my current design. My first design was 5 pairs, result was good, I just want to improve more. Power section distortion is the major component of the distortion of class AB power amps.

3) All transistors have beta droop at high current, just look at the graph in data sheets. The best transistor pair are MJW1302/3281 by On Semi, it still starting to droop above 6 or 7A. So if you have an amp with 50V supply rail, and if you can swing to 45V. If you drive a 4ohm speaker, peak current is 45/4=11.25A. You definitely need at least 2 pairs of output transistors for output stage. Beta droop put extra load on the driver transistors which reflect back to the VAS stage ( voltage amplifier stage) of the power amp. That's the reason good amp design use 3EF instead of 2EF stage. 3EF meaning emitter follower pre-driver driving the driver stage, then the driver stage drives the big output transistors. With all the betas, the effect to the VAS stage is minimized even though the beta droop happens at the big transistors.

Hope this help. This is why my amp has 9 pairs of output transistors and 3EF, all to lower distortion. And I am designing for 120W into 4ohm only. It's not the amount of watts, it's the quality of the watts. I get 8W of class A into 4ohm and 16W into 8 ohm. It's going to be a class A at my listening level.

Now, the most important (2) does not apply to class A amplifiers. By definition of class A, both NPN and PNP transistor pairs are ALWAYS ON, NEVER TURNS OFF, unlike the class AB where one is off while the other one is driving in one direction. There is no crossover distortion, you don't need to worry about (2), as you will notice, class A amp usually have less pairs of transistors. (1) and (3) still applies. There is no way out of (1) and (3) but more pairs. BUT when you see class A, you better ask what load. Amp that guaranty class A for 8ohm speaker might not be class A driving 4ohm speaker. Result might be worst than a well designed class AB when a class A amp goes out of class A into class B.

 

[rcs16]

This is why my amp has 9 pairs of output transistors and 3EF, all to lower distortion. And I am designing for 120W into 4ohm only. It's not the amount of watts, it's the quality of the watts. I get 8W of class A into 4ohm and 16W into 8 ohm. It's going to be a class A at my listening level.

It is the OPS bias current through each device that determines the Class "A" region, not the amount of devices in parallel.
Some designers prefer MOSFETS as OPS devices, since they have no beta droop, SOA secondary breakdown, for lateral MOSFETS, no theraml run away or bias temperature compensation circuitry. MOSFETs do not load down the VAS with changes in load Z.
MOSFETS usually are bias higher than a bjt for the gm doubling aspect of a Class AB design.
All stuff that is pointed out in Bob Cordell's excellent Power Amp design book.

 

[Alan0354]

It is the OPS bias current through each device that determines the Class "A" region, not the amount of devices in parallel.

Of cause, that is the most basic fundamental thing about class A that you need the bias current to cover the full swing regardless of how many pairs. For example, if you max swing is 40V and you declare it's class A for 8 ohm, your peak current is 40V/8=5A. So your bias current needs to be 2.5A at idle in order to be a class A amp for 8ohm speaker. If you use MJW1302/3281, You might get away with using only 2 to 3 pairs of output transistors.

I got into a lot more detail about why you need more pairs for class AB to lower the crossover distortion as I explained in detail in the last post. For class A, you don't have crossover distortion, so you don't need a lot of pairs as described in (2) in my last post. You just need to observe (1) and (3) for class A.

Class A amp is in a way easier to design, you don't need to worry about Oliver's condition. More so, you can use higher value emitter resistor to prevent current hogging. You can use 0.33 to 0.5ohm and drop 100mV across the resistor and completely null out the Vbe difference between the transistors. Only big problem of Class A amp is the heat sink size and the heat it generates.