Amp output transistor short circuit protection needed

[uslrs]

Hi, this StageLine PMX100 amp has blown the output transistors for the second time when the speaker output was inadvertently shorted.  Both 4 amp fuses in the ac transformer output have blown, but this has not prevented the destruction of the output transistors. 

There is a thermal cutout attached to the heatsink which is in the ac supply line to the transformer primary and Q4 is also mounted in the heatsink (see pics).  The thermal cutout is obviously to prevent overheating and I assume Q4 is for thermal compensation in the bias circuit, but I’m struggling to understand how the stages between the op-amp and the driver transistors Q5 and Q6 function.

I’d like to build in some short circuit protection for the output transistors – ideally without adding too much extra complexity to the circuit.  I’m wondering if adding fuses between the dc supply B+ and Q7 collector and B- and Q8 collector might provide a fix and if so what fuse type and rating would be appropriate, given that Q7 & Q8 are rated at 20A continuous collector current and 40A peak.  Many thanks in advance.

[strawberry]

looks like low end amplifier

MOSFET output stage? original transistors? MOSFET's generally are made smaller but there are expensive grades for more power dissipation
if transistors are weak then protection wont help it

fuse per channel for multiple channel amplifier

I made amplifier with 2SC5200N pair with +/-25V and is not afraid of short circuit(will get hot at full power)

https://www.tindie.com/products/sakurawizard/speaker-protection-module-ta7317-upc1237/ with two sense  transistors can current protection

[uslrs]

looks like low end amplifier

MOSFET output stage? original transistors? MOSFET's generally are made smaller but there are expensive grades for more power dissipation
if transistors are weak then protection wont help it

fuse per channel for multiple channel amplifier

I made amplifier with 2SC5200N pair with +/-25V and is not afraid of short circuit(will get hot at full power)

https://www.tindie.com/products/sakurawizard/speaker-protection-module-ta7317-upc1237/ with two sense  transistors can current protection

Not MOSFETs - output transistors are TIP35 and TIP36 BJT's. These are as shown in the schematic and supplied by RS, so were good quality components. 

[magic]

It looks like Q9,Q10 already implement output current limit. They seem to trigger at about 1.75V across R25,R26, which corresponds to 8A current.

[uslrs]

It looks like Q9,Q10 already implement output current limit. They seem to trigger at about 1.75V across R25,R26, which corresponds to 8A current.

I have tested both Q9 and Q10 and both are good, as are R21,22,23,24,25 and 26.  I assume their function is to cut off the base currents of Q5 and Q6 when the voltage across R25/26 is sufficient to turn on Q9/10, so as all the resistors R21-24 are good (and D2/3, Q5/6), I'm wondering why this didn't protect the output transistors, Q7/8.

[strawberry]

I think not current cut off but constant current circuit it is
I think TIP35C more like 3A would be maximum average current
about 5A TIP35 gain starts roll-off
100W rating is when heatsink is kept constant @25degC
8A * ~25Vavg = 200W
reason why good quality amplifiers use parallel transistors for same power level

[marekm]

This current limit may be above the secondary breakdown SOA limit.  It could be reduced at high Vce by adding resistors from Q9/Q10 bases to B+/B-.  About 33k should give about half the current for output shorted to ground, while still allowing full current limit at output close to supply rails (full power at resistive load).

[elecdonia]

I looked at a schematic for this amplifier. It is a “minimalist” circuit design. Some of it resembles audio power amplifiers from the 1960’s or early 1970’s.

It might be feasible to replace the output transistors with something more robust. That “20A” current rating for the original output transistors only applies when the E-C voltage is <5V. However, if the audio output is shorted to ground the E-C voltage will be about 35V. In this case the output transistors can only handle about 3A.

The current limiting transistors (Q9, Q10) aren’t reducing the overload sufficiently to save the output transistors. They only limit the maximum current on a cycle-by-cycle basis.

What is actually needed is to either disconnect the speakers or mute the audio within about 100 msec of the current limiting getting activated. The existing circuit is incapable of doing this. This amplifier needs to have a muting circuit added to it which will entirely shut off the audio signal for several seconds whenever the current limiting transistors begin to conduct. Most solid-state amplifiers made in the 1970’s and later employ a “speaker relay” to disconnect the output for several seconds when overcurrent occurs. Alternately the line level audio signal going into the first opamp stage can be clamped to ground during overload. The concept is to prevent the output transistors from experiencing overcurrent for >100 msec.

Another issue with this circuit design is the “boot strapped” collector load (R15, R16, C14) for the voltage amplifier transistor Q3. This generates a large DC offset at turn-on/turn-off which will cause a thump or pop from the speakers. Since about 1980 nearly all amplifier circuits have an active current source instead of this bootstrap circuit. It can be switched off during turn-on/turn-off to greatly reduce the thump/pop issue.

Overall I would rate this amplifier circuit design as “marginally functional.” Of course that’s what keeps the price low.

[uslrs]

I looked at a schematic for this amplifier. It is a “minimalist” circuit design. Some of it resembles audio power amplifiers from the 1960’s or early 1970’s.

It might be feasible to replace the output transistors with something more robust. That “20A” current rating for the original output transistors only applies when the E-C voltage is <5V. However, if the audio output is shorted to ground the E-C voltage will be about 35V. In this case the output transistors can only handle about 3A.

The current limiting transistors (Q9, Q10) aren’t reducing the overload sufficiently to save the output transistors. They only limit the maximum current on a cycle-by-cycle basis.

What is actually needed is to either disconnect the speakers or mute the audio within about 100 msec of the current limiting getting activated. The existing circuit is incapable of doing this. This amplifier needs to have a muting circuit added to it which will entirely shut off the audio signal for several seconds whenever the current limiting transistors begin to conduct. Most solid-state amplifiers made in the 1970’s and later employ a “speaker relay” to disconnect the output for several seconds when overcurrent occurs. Alternately the line level audio signal going into the first opamp stage can be clamped to ground during overload. The concept is to prevent the output transistors from experiencing overcurrent for >100 msec.

Another issue with this circuit design is the “boot strapped” collector load (R15, R16, C14) for the voltage amplifier transistor Q3. This generates a large DC offset at turn-on/turn-off which will cause a thump or pop from the speakers. Since about 1980 nearly all amplifier circuits have an active current source instead of this bootstrap circuit. It can be switched off during turn-on/turn-off to greatly reduce the thump/pop issue.

Overall I would rate this amplifier circuit design as “marginally functional.” Of course that’s what keeps the price low.

Many thanks for all your responses on this - much appreciated!  I have seen a number of protection circuits employing a relay in the output and I think this is probably the way to go if re-designing this power amp. I also recognize that this is a pretty low end product, so am really looking for a minimal intervention to overcome this problem.  Given the inclusion of a (basic and not very effective!) current limiting circuit (Q9/10, etc), as elecdonia has indicated, the basic issue appears to be the current capacity of the TIP35/36 output transistors, so would it be feasible to increase this current handling by means of adding an additional TIP35/36 in parallel with the existing components? There is adequate space on the heatsink to accommodate this addition in close proximity to the existing o/p transistors (which should prevent runaway because of performance differences) and could be accomplished without modifying the pcb or adding additional circuitry.  Would this be a goer do you think?

[uslrs]

In relation to adding output transistors in parallel, I'm wondering if this circuit would be appropriate.  The additional transistors would be mounted on the same heatsink in close proximity to the originals and both bases are connected directly to the emitters of Q5/6.  I'm proposing adding a 0.15ohm, 5W ballast resistor in the emitter paths of each o/p transistor, as per the attached schematic.  Does this look like a practical solution? I'm assuming the additional load on the drivers Q5/6 would not be problematic.  Many thanks in advance.

[elecdonia]

In relation to adding output transistors in parallel, I'm wondering if this circuit would be appropriate.  The additional transistors would be mounted on the same heatsink in close proximity to the originals and both bases are connected directly to the emitters of Q5/6.  I'm proposing adding a 0.15ohm, 5W ballast resistor in the emitter paths of each o/p transistor, as per the attached schematic.

Only one additional resistor is needed per + or - side of output stage. But the values will be different, so you will need to get a total of 4 pieces of 0R39 5W resistors. Change original R25 (and R26) to 0R39. Connect 2 more 0R39 resistors from E of each added output transistor to the output signal rail. Current limit circuit only needs to sense voltage across one 0R39 resistor (per + or - side). There is no need to obtain an average for both output transistors. The schematic will resemble your drawing but without the original R25 and R26. The 4 0R39 resistors will be in the locations you labeled Rx and Ry.

Does this look like a practical solution? I'm assuming the additional load on the drivers Q5/6 would not be problematic.

I've seen amplifiers where one driver transistor fed as many as 4 output transistors in parallel. Some designers add a small value series resistor to the B of each output transistor (often 4R7 ohms). If these are 1/4W "flameproof fusible" resistors then they tend to protect the remaining output transistors from damage caused by a single output transistor shorting out.

[uslrs]

Thank you Elecdonia - that's really helpful - I think this mod will do the trick. Even if the current limiting is not perfect the output transistors only need to survive long enough for the fuses in the transformer secondaries to blow.  Thanks again for your help.😁😁

[elecdonia]

I found mfg. datasheets for TIP35C/TIP36C.
The ON Semiconductor datasheet is complete, but the ST datasheet lacks a vitally important parameter for audio amplifiers: "SOA" (safe operating area). This is a chart of maximum allowable collector current vs. E-C voltage. (ST "fudges" by omitting SOA from their data sheet. This might cause some designers to overload the transistor.)

It is easy to think "everything will be OK" because this transistor is rated at 25A, 100V, and 125W.
However, it cannot handle all 3 conditions at the same time.
If one tried to put 25A through it with 100V E-C, this is 2,500W, (25 times larger than the rated 125W).

This is what the SOA chart tells us (see attached chart):
  25A of collector current is permitted up to 5V E-C voltage.
  But maximum collector current drops to 12.5A at 10V E-C (limited by requirement to keep dissipation <125W)
  Another limitation comes into play when E-C voltage is >30V. This is known as "secondary breakdown."
  At 38V E-C, collector current must be <2A. In other words this transistor can only dissipate ~75W when E-C voltage is 38V. (This represents the fault condition where the output is shorted to ground.)
  Maximum allowable collector current continues to drop rapidly for E-C voltages >40V.  Only ~100mA can be delivered when the C-E voltage approaches 100V. This is only 10W

Higher current can be delivered for short pulses ( <10ms). But 10ms isn't very long. This is why overload muting circuits are essential. When a direct short-circuit occurs the input audio signal needs to muted within ~10ms. There should then be ~1- 2 second "dead time" before the audio input signal is allowed to get through again. In fact I prefer designs where a long-term direct short causes the power supply to switch off and remain off until the user cycles the power button. This alerts the user of the need to troubleshoot the fault in their wiring (or inside the speaker). It is also informative if the amplifier has an "overload indicator LED" which switches on (and stays on for several seconds) after a fault occurs.
 
 

[uslrs]

That explains it all! I was struggling to understand why the transistors failed given the transformer secondaries are fused at 4A.  From the chart it's clear that at 38v they're near the limit, even paralleled. It looks like 4A fast blow fuses are required!

[uslrs]

In relation to adding output transistors in parallel, I'm wondering if this circuit would be appropriate.  The additional transistors would be mounted on the same heatsink in close proximity to the originals and both bases are connected directly to the emitters of Q5/6.  I'm proposing adding a 0.15ohm, 5W ballast resistor in the emitter paths of each o/p transistor, as per the attached schematic.

Only one additional resistor is needed per + or - side of output stage. But the values will be different, so you will need to get a total of 4 pieces of 0R39 5W resistors. Change original R25 (and R26) to 0R39. Connect 2 more 0R39 resistors from E of each added output transistor to the output signal rail. Current limit circuit only needs to sense voltage across one 0R39 resistor (per + or - side). There is no need to obtain an average for both output transistors. The schematic will resemble your drawing but without the original R25 and R26. The 4 0R39 resistors will be in the locations you labeled Rx and Ry.

Does this look like a practical solution? I'm assuming the additional load on the drivers Q5/6 would not be problematic.

I've seen amplifiers where one driver transistor fed as many as 4 output transistors in parallel. Some designers add a small value series resistor to the B of each output transistor (often 4R7 ohms). If these are 1/4W "flameproof fusible" resistors then they tend to protect the remaining output transistors from damage caused by a single output transistor shorting out.

Elecdonia, I just want to check I've interpreted your circuit description correctly - please see attached amended schematic.  Many thanks.

[uslrs]

Schematic attached....

[marekm]

Even with two parallel transistors, it still makes sense to add foldback to keep current limit within the SOA at high Vce.  Adding more complex protection circuits will probably be more work than replacing the complete power stage with an IC with better on-chip protection, like LM3886 or TDA7294.

[elecdonia]

That explains it all! I was struggling to understand why the transistors failed given the transformer secondaries are fused at 4A. From the chart it's clear that at 38v they're near the limit, even paralleled. It looks like 4A fast blow fuses are required!

There are two possible places for these fuses:
 
1)  The original circuit:
Fuses located between the two ends of the transformer secondary and the bridge rectifier.
   (Curiously the published schematic omits the center tap of the transformer secondary, which is connected to the ground rail near the two large filter capacitors.).
With this fuse placement there is no protection from high current surges coming from the significant amount of energy stored inside the two large filter capacitors. Therefore these fuses aren't very effective for protecting the output transistors. (However these fuses will protect the power transformer in case the user "fixes" a blown AC mains line fuse by wrapping it with tinfoil and putting it back into the fuse holder). Also when located in this position the fuses are usually “slow blow” types. There is a large surge of current at power-on to charge up the two large power supply capacitors. Fast blow fuses might blow simply by switching on the power.

2)  Alternate circuit:
Place series fuses in the +38V and -38V DC supply rails, after the filter capacitors. Fuses in this location provide much better protection for the output transistors compared to circuit #1. This is because fuses located here will interrupt excessive current flowing not only from the transformer/bridge rectifier, but they will also interrupt current coming from the two large filter capacitors. These should be “fast-blow” fuses.

Early high-power solid-state amplifiers commonly had “fast blow” B+ and B- fuses located in this circuit position.  Examples: Phase Linear 400 & 700, Dynaco ST400 & ST416. Fuse values were typically 5A to 8A and the fuses were "fast blow" types. For stereo amplifiers each channel had a B+ fuse and a B- fuse.

[uslrs]

Many thanks. I'll incorporate fuses in B+ and B- as part of the circuit modification. I'll install 5A fast blow types and see how they hold up under test.