Many individuals have published output stage protection circuits. However, they are usually just one example circuit for a particular transistor. More importantly, they are for a particular value of emitter degeneration resistor. The value of the resistor is important to the protection circuit, changing Re from .22 ohms to .1 ohms means the protection circuit won't protect! The circuit below is a standard dual slope protection circuit, published by Sloane, Self, Duncan and others.
The response above shows the load lines for a 8 ohm and 4 ohm reactive load of 45 degrees, and a 4 ohm resistive load. For the first plot I’ve included both the 3281 and 21193 transistor SOA. Notice that the 21193 is a little bit better than the 3281 transistor (and their associated complementary pair…)
For this circuit R1 = 4.99k, R2 = 4.22k, R3 = 2.21k, R4 = 182, R5 = 665, Re = .22 Ohms…
If you want to go with the 21193 you can move the protection circuit response so that it does not impede as much of the 4 ohm reactive load line.
For this response R1 = 4.99k, R2 = 4.22k, R3 = 1.00k, R4 = 182, R5 = 665, Re = .22 Ohms…
If you want to use the 21193 transistor with an Re of .1 ohms….you can get the following:
For this response R1 = 4.02k, R2 = 2.68k, R3 = 698k, R4 = 182, R5 = 201, Re = .1 Ohms…
I was also curious to see how fast the load line would move outside of the SOA for a single transistor. Below is the diagram with the same MJL21193 transistor SOA and the protection circuit from the example just before this. Instead of plotting 8 ohms, 45 degrees I plotted 4 ohms 45 degrees at just before clipping with an amp with 40V rails, and one with 50V rails. Both load lines (elipse) are CFP amps, with a single pair of output transistors. Notice that with 50V rails a single transistor will NOT be sufficient. In fact for transistor power dissipation, the 40V rails present enough of a challenge…continuous operation into difficult loads with a single pair of transistors requires temperature protection, not just SOA protection.
Remember, the SOA curve is rated for a peak junction temperature. The junctions heat up quickly in a power transistor, there is quite a bit of thermal lag from transistor to case…where you can measure temperature. Also remember that in real life, voltage rails sag somewhat protecting the amplifier. The dynamic nature of music means that the continuous stress on the transistor is considerably less than using tones to test, not enough (in my humble opinion) to get away without SOA protection…but enough to rarely activate the circuit in real use.
I’ve also been asked before if it’s worth-while to worry about load impedance phase angle with amplifiers…and the answer is emphatically yes…
The Avalon Indra, and Sonus Fabor Cremona are just two examples.
http://stereophile.com/floorloudspeakers/1008ava/index4.html
http://stereophile.com/floorloudspeakers/1207sonus/index4.html
I know this list of circuits isn't exhaustive (I've left out Fairchild power BJTs and any type of MOSFET...I'll get around to plotting those one of these days as well.) But I hope this helps.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
The response above shows the load lines for a 8 ohm and 4 ohm reactive load of 45 degrees, and a 4 ohm resistive load. For the first plot I’ve included both the 3281 and 21193 transistor SOA. Notice that the 21193 is a little bit better than the 3281 transistor (and their associated complementary pair…)
For this circuit R1 = 4.99k, R2 = 4.22k, R3 = 2.21k, R4 = 182, R5 = 665, Re = .22 Ohms…
If you want to go with the 21193 you can move the protection circuit response so that it does not impede as much of the 4 ohm reactive load line.
An externally hosted image should be here but it was not working when we last tested it.
For this response R1 = 4.99k, R2 = 4.22k, R3 = 1.00k, R4 = 182, R5 = 665, Re = .22 Ohms…
If you want to use the 21193 transistor with an Re of .1 ohms….you can get the following:
An externally hosted image should be here but it was not working when we last tested it.
For this response R1 = 4.02k, R2 = 2.68k, R3 = 698k, R4 = 182, R5 = 201, Re = .1 Ohms…
I was also curious to see how fast the load line would move outside of the SOA for a single transistor. Below is the diagram with the same MJL21193 transistor SOA and the protection circuit from the example just before this. Instead of plotting 8 ohms, 45 degrees I plotted 4 ohms 45 degrees at just before clipping with an amp with 40V rails, and one with 50V rails. Both load lines (elipse) are CFP amps, with a single pair of output transistors. Notice that with 50V rails a single transistor will NOT be sufficient. In fact for transistor power dissipation, the 40V rails present enough of a challenge…continuous operation into difficult loads with a single pair of transistors requires temperature protection, not just SOA protection.
An externally hosted image should be here but it was not working when we last tested it.
Remember, the SOA curve is rated for a peak junction temperature. The junctions heat up quickly in a power transistor, there is quite a bit of thermal lag from transistor to case…where you can measure temperature. Also remember that in real life, voltage rails sag somewhat protecting the amplifier. The dynamic nature of music means that the continuous stress on the transistor is considerably less than using tones to test, not enough (in my humble opinion) to get away without SOA protection…but enough to rarely activate the circuit in real use.
I’ve also been asked before if it’s worth-while to worry about load impedance phase angle with amplifiers…and the answer is emphatically yes…
The Avalon Indra, and Sonus Fabor Cremona are just two examples.
http://stereophile.com/floorloudspeakers/1008ava/index4.html
http://stereophile.com/floorloudspeakers/1207sonus/index4.html
I know this list of circuits isn't exhaustive (I've left out Fairchild power BJTs and any type of MOSFET...I'll get around to plotting those one of these days as well.) But I hope this helps.
Sure, the junctions heat up pretty fast - but it's not as fast as one might think at first. But of course - using more pairs will give better performance due to less beta droop and such things.
I've played a bit with thermal models because I couldn't understand how commercial equipment gets away with about half the number of output devices many here at diyAudio say are absolute minimum. The thermal models show that the thermal capacity of the transistors is enough to make the short peaks from reactive loads a pretty small problem even at the lowest of frequencies.
For transistors with crappy second breakdown performance like 2N3055, a lot of darlingtons, TIP35 and older transistors you probably do need to stay inside the DC SOA though but for transistors like MJ(L)2119x, MJL3281, 2SC3281, 2SC5200, 2SC3264 and so on it's not a big problem at this low voltage.
The 150 degree max is really set by the packaging material and as this has pretty slow temperature transfer from the chip designing for an average temperature of 150 degrees (using averag power dissipation) during worst-worst-case seems to often be done. This practive seems sensible as long as peak temperature is sensible, say below 200 degrees or so (which it will be). I'd probably derate the calculated thermal resistance by about 1.5 - that would keep almost all peaks below 150 degrees according to models of the transistor time constants.
An example is the QSC RMX1450 PA amp - it has ~75V rails and gives 700W for 2 ohm loads. 4 pairs of 2SC5200/2SA1493. Doesn't have thermal pads for the transistors though so can push them a bit further than otherwise.
But still, nothing wrong with using more pairs - you will often get lower distortion!
I've played a bit with thermal models because I couldn't understand how commercial equipment gets away with about half the number of output devices many here at diyAudio say are absolute minimum. The thermal models show that the thermal capacity of the transistors is enough to make the short peaks from reactive loads a pretty small problem even at the lowest of frequencies.
For transistors with crappy second breakdown performance like 2N3055, a lot of darlingtons, TIP35 and older transistors you probably do need to stay inside the DC SOA though but for transistors like MJ(L)2119x, MJL3281, 2SC3281, 2SC5200, 2SC3264 and so on it's not a big problem at this low voltage.
The 150 degree max is really set by the packaging material and as this has pretty slow temperature transfer from the chip designing for an average temperature of 150 degrees (using averag power dissipation) during worst-worst-case seems to often be done. This practive seems sensible as long as peak temperature is sensible, say below 200 degrees or so (which it will be). I'd probably derate the calculated thermal resistance by about 1.5 - that would keep almost all peaks below 150 degrees according to models of the transistor time constants.
An example is the QSC RMX1450 PA amp - it has ~75V rails and gives 700W for 2 ohm loads. 4 pairs of 2SC5200/2SA1493. Doesn't have thermal pads for the transistors though so can push them a bit further than otherwise.
But still, nothing wrong with using more pairs - you will often get lower distortion!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.