Your output transistors have ample margin on voltage and current. On power dissipation you monitored the temperature and did not see thermal run away. Typically BJT transistor fail short collector emitter (or excessive leakage between them). However if short B-E junction is seen (without C-E short) it may point to exceeding reverse B-E voltage. Your output stage is dual stage emitter follower (unity gain). The only way to exceed reverse B-E voltage is if the emitters are held by capacitors while the base is declining fast. This does not happen during listening music, where no high level high frequency is applied. This circuit has no current limiting and shorted load may be a risk. Can you examine your speaker cable/connectors for potential for shorts.

From LTspice, I should be around 2v for VB-E:

 

The TIPs failed with Collect to Emitter shorted this time. One of them actually split in half when it let the smoke out.
There isn't a short in the cable or connectors. I verified nothing was touching or shorted, and the cables and speakers are fine when connected to a different amp.

If it adds any info, R16 was burned and failed open at the same time.

Thanks

Petkan:
If R16 is damaged - check the small transistor Q8, Q9. After the repair temporary short C3 (zero bias to eliminate thermal run away) and let it run a sine into a resistive load - say a 12Vcar bulb (to eliminate inductive kicks). Is your dual rail power source solid? (transformer/rectifier? or switching mode). Check the layout in respect of too long high current lines from power source to output stage. The output stage should take its power directly from the bipolar rectifier caps. No harm adding local ceramic decoupling caps (potential inductive kick in power wires). If ringing is suspected - adding a damping series RC circuits across each rail may help (say 2R, 22uF). With bias at zero only high level sine will be correctly propagated. After reaching a settling in temperature, remove the short from C3 and connect volt meter across the 0.68R resistor and evaluate indirectly the bias current. check the others for symmetry. Lower the bias, see the distortion at low level and then gradually increase the bias until low level sine is undistorted. The voltmeter reading divided by 0.68 will tell you the bias current. Resistor may have grown in value (having been stressed). Two transistor in parallel form each half and these resistor help to balance their current split. If one particular transistor happens to fail each time - one more reason to check the resistors. Low Ohmic resistors are best measured by a lab power supply in constant current (limiting) mode - by driving a known current - say 1A into the resistor while measuring the voltage across it. This could also be done in pairs i.e. driving current into two in series (same current, more accurate resistance evaluation).

 

Petkan:
If R16 is damaged - check the small transistor Q8, Q9. After the repair temporary short C3 (zero bias to eliminate thermal run away) and let it run a sine into a resistive load - say a 12Vcar bulb (to eliminate inductive kicks). Is your dual rail power source solid? (transformer/rectifier? or switching mode). Check the layout in respect of too long high current lines from power source to output stage. The output stage should take its power directly from the bipolar rectifier caps. No harm adding local ceramic decoupling caps (potential inductive kick in power wires). If ringing is suspected - adding a damping series RC circuits across each rail may help (say 2R, 22uF). With bias at zero only high level sine will be correctly propagated. After reaching a settling in temperature, remove the short from C3 and connect volt meter across the 0.68R resistor and evaluate indirectly the bias current. check the others for symmetry. Lower the bias, see the distortion at low level and then gradually increase the bias until low level sine is undistorted. The voltmeter reading divided by 0.68 will tell you the bias current. Resistor may have grown in value (having been stressed). Two transistor in parallel form each half and these resistor help to balance their current split. If one particular transistor happens to fail each time - one more reason to check the resistors. Low Ohmic resistors are best measured by a lab power supply in constant current (limiting) mode - by driving a known current - say 1A into the resistor while measuring the voltage across it. This could also be done in pairs i.e. driving current into two in series (same current, more accurate resistance evaluation).

Testing the failed unit, Q8 and Q9 are shorted Base-Emitter, and both measure 0V on diode test mode on the DMM testing Base-Emitter. Base-Collector measures around 0.7v. I'm guessing they fried too.

Power supply uses a toroidal transformer and rectifier.

All 0.68 ohm emitter resistors measured 0.7 ohms on DMM using relative mode.

 

Testing the failed unit, Q8 and Q9 are shorted Base-Emitter, and both measure 0V on diode test mode on the DMM testing Base-Emitter. Base-Collector measures around 0.7v. I'm guessing they fried too.

Power supply uses a toroidal transformer and rectifier.

All 0.68 ohm emitter resistors measured 0.7 ohms on DMM using relative mode.

Petkan:
What is the distance from rectifier caps and output stage? (Wires have inductance. Change in current produces spikes)
My first power amp used Germanium PNP transistors. I struggled with hum due to 2" long thick wire between the two rectifier caps. I had the rectifier wire on one end and the wire to the output stage on the other. Filtering caps have heavy current spikes and produce respective voltage drops in wires. It is important to separate the wire from rectifier to cap and from cap to output stage. Ideally a star point connection should be made at the caps. Apart of the interconnecting GND wire between them ho harm having two separate wires from the rectifier return to respective caps. Now the middle the this cap to cap interconnecting wire is the place to take a wire from for the output stage GND.
Note that the heavy filtering currents flow through their respective wires, leaving alone the cap to cap GND interconnect i.e. no voltage drops reflecting the filtering currents. Is the power source at the same PCB or at the transformer?

 

Petkan:
What is the distance from rectifier caps and output stage? (Wires have inductance. Change in current produces spikes)
My first power amp used Germanium PNP transistors. I struggled with hum due to 2" long thick wire between the two rectifier caps. I had the rectifier wire on one end and the wire to the output stage on the other. Filtering caps have heavy current spikes and produce respective voltage drops in wires. It is important to separate the wire from rectifier to cap and from cap to output stage. Ideally a star point connection should be made at the caps. Apart of the interconnecting GND wire between them ho harm having two separate wires from the rectifier return to respective caps. Now the middle the this cap to cap interconnecting wire is the place to take a wire from for the output stage GND.
Note that the heavy filtering currents flow through their respective wires, leaving alone the cap to cap GND interconnect i.e. no voltage drops reflecting the filtering currents. Is the power source at the same PCB or at the transformer?

The power is fed from a transformer to a PCB that has the filter caps and rectifiers. From there, it feeds the amplifier boards through maybe 10" of 18AWG wire. This photo will give you an idea (I removed all the TIP transistors that blew up):

 

So should I mount Q8 and Q9 off-board and maybe epoxy them right to the output transistor heat sinks? Would that provide the thermal feedback you mentioned?

Q11 needs to be on the output heatsink.

Here's http://www.ece.drexel.edu/courses/ECE-E352/ClassABAmp.pdf a blub on amplifier design including the Vbe multiplier.

But definitely use an AGX fuse on the output of the amp.

 

Although, probably not a 1st order issue, layout should have taken into account the high current and reference grounds. This is where the "star ground" really prevails.

Isolated the input ground with say a 2.2. Ohm resistor can help DC ground loop issues. i.e. Use isolated phone plugs and take a 2.2 ohm resistor to whatever you call the reference ground.



I've attached a pic of my Leach Amp that was built in the 1980's. the transformer is a custom 4x35 Vac windings providing 2 separate +-50 Vdc supplies. The heatinnks in the center are huge and if you look closely, there is one of the 3 diodes mounted for thermal feedback. The amp is unusual because it has a ground plane construction, because it is a high frequency design (0 to 800 kHz) without being intentionally rolled off.

Tne the front is the transformer with a protection board on the top. It logartihmically ramps the audio thorough an opto-coupler. There are various LEDs on the PCB which can be seen through the perforated screen top.
the entire top is perforated.

The rail fuses are mounted on a L-bracket on top of the transformer. There is also a 12 VDC supply that powers the start-up circuit. There'ss a flameproof resistor in line with the mains that is shorted out at when all supplies are about 2/3 of 50 V. AT that time, the audio starts ramping and the speakers connect.

I never added a clipping indicator. The PCB accommodates LED drivers for L clip, R clip, High temp and protect, but they are unused,

The top-left corner of the pic has the relay and resistor.

The AMP boards have an integral phono connector. The high frequency network is on the binding posts.
An Amp channel can be removed for troubleshooting with an extension harness. When removed, the opto is disconnected.

Near the center at the top of the picture are two optocouplers with an LED indicator.

The right top, is the speaker output relay and THE WIRING MESS is a real STAR ground, Everything is returned to a single LUG.

The AMP does not have a power switch.

I just never got around to

Building a clipping/tru-clip or distortion indicator, but I did build a really cool one for work. The work one used a single b-icolor LEDand would flash red or green depending on the polarity of the clip for 1 second if the voltage as >10 or <10 volts.

Upgrading the power supply, It;s actually fed by a 500 W AC regulator.

Making a much better slow turn-on circuit that doesn't pop the resistor.

Adding a hi-temp or DC protect circuit.

 

Although, probably not a 1st order issue, layout should have taken into account the high current and reference grounds. This is where the "star ground" really prevails.

Isolated the input ground with say a 2.2. Ohm resistor can help DC ground loop issues. i.e. Use isolated phone plugs and take a 2.2 ohm resistor to whatever you call the reference ground.

View attachment 119234

I've attached a pic of my Leach Amp that was built in the 1980's. the transformer is a custom 4x35 Vac windings providing 2 separate +-50 Vdc supplies. The heatinnks in the center are huge and if you look closely, there is one of the 3 diodes mounted for thermal feedback. The amp is unusual because it has a ground plane construction, because it is a high frequency design (0 to 800 kHz) without being intentionally rolled off.

Tne the front is the transformer with a protection board on the top. It logartihmically ramps the audio thorough an opto-coupler. There are various LEDs on the PCB which can be seen through the perforated screen top.
the entire top is perforated.

The rail fuses are mounted on a L-bracket on top of the transformer. There is also a 12 VDC supply that powers the start-up circuit. There'ss a flameproof resistor in line with the mains that is shorted out at when all supplies are about 2/3 of 50 V. AT that time, the audio starts ramping and the speakers connect.

I never added a clipping indicator. The PCB accommodates LED drivers for L clip, R clip, High temp and protect, but they are unused,

The top-left corner of the pic has the relay and resistor.

The AMP boards have an integral phono connector. The high frequency network is on the binding posts.
An Amp channel can be removed for troubleshooting with an extension harness. When removed, the opto is disconnected.

Near the center at the top of the picture are two optocouplers with an LED indicator.

The right top, is the speaker output relay and THE WIRING MESS is a real STAR ground, Everything is returned to a single LUG.

The AMP does not have a power switch.

I just never got around to

Building a clipping/tru-clip or distortion indicator, but I did build a really cool one for work. The work one used a single b-icolor LEDand would flash red or green depending on the polarity of the clip for 1 second if the voltage as >10 or <10 volts.

Upgrading the power supply, It;s actually fed by a 500 W AC regulator.

Making a much better slow turn-on circuit that doesn't pop the resistor.

Adding a hi-temp or DC protect circuit.

Thank you for all the great info. I'm probably decades away from attempting anything as complicated as that!

I'm just trying to salvage this project, which is my first. I'm still scratching my head as to what parameter I might be exceeding that keeps blowing up the output transistors, and in this occurrence, the driver transistors as well.

Could having the VBE multiplier transistor NOT mounted on the heatsink possibly cause the repeated failures I'm having? Or am I exceeding some other voltage/current value of either the TIPs or the 2N5551/2N5401?

Thanks

 

Could having the VBE multiplier transistor NOT mounted on the heatsink possibly cause the repeated failures I'm having? Or am I exceeding some other voltage/current value of either the TIPs or the 2N5551/2N5401?

Yes, It could cause your DC bias to run away. The typical thermal runaway problem. When it heat up. It conducts more. When it conducts more, it heats up and repeat.

Stick a voltmeter across on of the emitter resistors and monitor it. For a test, no input, just set the bias like you think it should be an watch the voltmeter across on of the output emitter resistors. The current should not creep up.

Q10. Q11, Q12 and Q13 should really share the same heatsink. You could see if the amp does better with continuous fan cooling, but fix C3.

CHANGE C3 from 47 uf to a metal film cap around 0.1 uF. This could be the bigger problem.
You want that regulator to respond vary fast, 47 uF is just asking for trouble.

==

Why do I see an air inductor on the board?

The the note here http://leachlegacy.ece.gatech.edu/lowtim/part3.html concerning R49/R50 or your L1/R19. You chose a 22 ohm resistor. The Leach amp chose 10 ohms. Wind a solid insulated wire (18-22 AWG) around a 5 Watt CARBON composition resistor for that component.

quote=[From part 3]
L1 - 10 to 12 turns #22 solid insulated wire wound tightly around R49 and soldered to the leads of R49 where they emerge from the resistor body. Click here to see an illustration. Solder one end of a piece of #22 solid wire to one end of R49. Wind the wire tightly around R49 to form L1. Clamp the windings to R49 with a small bench vise. Strip and solder the other end of L1 to R49. Do not use stranded wire.[/quote]

PS:

I know the pain your going through, In the amp I made, I mirrored the boards when I had them made, It took me forever to figure that out. Once I did, I found out how I could use the mirrored boards. I blew up stuff too.
This is where a metered/variac (particularly current) comes in real handy. Totally initial power ups basically disable the Vbe regulator, use an input signal and look at the output and the AC power draw ammeter.

At one point, I tested just the Vbe regulator section with an external supply until I understood it. Since you have two darlingtons like the Leach amps has you have 0.6(2)+0.6(2) or about 2.4 V. So, it' basically a 3 V shunt regulator that reduces it's output by about 10 mV/deg C, I think as the transistor heat up.

The cool part, is that you can make the regulator output 0 V with a jumper. It's a shunt regulator, so it can easily be shorted, right? When you have no regulator, the amp doesn't get warm when idleing.

The Leach Vbe multiplier is here: http://leachlegacy.ece.gatech.edu/lowtim/2ndstage.html So, you could add the 1-4 diodes to the large heatsink containing the output transistors, but keep the transistor you have independent.

The mounting takes some finesse, but you only have to do it once. Two plastic body diode in a force fit hole, then lay another diode on it's side and bend up the leads. They need to be mounted, so they see an average temperature.

That 47 uF capacitor in your amp has to be changed to something like 0.1 to 1 uF.

the TIP42 is probably (I didn't look) pretty slow and not considered a good audio transistor. I didn't look at the other selections either, The TIP42 datasheet here: http://www.mouser.com/ds/2/149/TIP42-890119.pdf says I think you out of the safe operating area of Vce. So, look at the SOA for those last 4 transistors.

The finals can be totally overrated. In the Leach amp they are rated for a 30 Amp collector current and don't ever get close.

==

You might look here: http://audiojudgement.com/speaker-impedance-curve-explained/ and get the idea that the speaker a very dynamic load and since it's inductive, the current through an inductor doesn't change instantaneously and you want to make it go where it doesn't want to go. Hence, an overrated output stage is a "good thing".

Your not ready yet, but looking at the rise time of the amp is another thing to eventually look at.

 

Yes, It could cause your DC bias to run away. The typical thermal runaway problem. When it heat up. It conducts more. When it conducts more, it heats up and repeat.

Stick a voltmeter across on of the emitter resistors and monitor it. For a test, no input, just set the bias like you think it should be an watch the voltmeter across on of the output emitter resistors. The current should not creep up.

Q10. Q11, Q12 and Q13 should really share the same heatsink. You could see if the amp does better with continuous fan cooling, but fix C3.

CHANGE C3 from 47 uf to a metal film cap around 0.1 uF. This could be the bigger problem.
You want that regulator to respond vary fast, 47 uF is just asking for trouble.

Thank you again for the excellent information. I'm going to do some reading on that Leach amp site.

I'll definitely try changing C3 to 0.1uF and see what happens. I will also make an new output inductor using the method you described. I was under the impression that air-core was good for audio.

What type of diodes would be required for the Vbe multiplier shown in the Leach amp design? Could I use 1N400X type diodes?

The transistors I was using were TIP41C and TIP42C, which are rated to 100V and 6A for C-E. I'll swap those out for something faster as well.

Would you happen to have any idea why the driver transistors blew up in this case? (2N5551/2N5401) Could that have happened when the output transistors shorted?

 

Also, would there be any way to test my amp up to the driver stage? Can I determine if there's a problem with anything before the output transistors if I run the amp into a 100ohm load for an extended period of time before testing at 8ohms ?

 

What type of diodes would be required for the Vbe multiplier shown in the Leach amp design? Could I use 1N400X type diodes?

yes, epoxy packages.

The transistors I was using were TIP41C and TIP42C, which are rated to 100V and 6A for C-E. I'll swap those out for something faster as well.

You can't look at Ic being 6A. You have to look at the SOA graph or Safe Operating Area. That's where I think it flunks.

Would you happen to have any idea why the driver transistors blew up in this case? (2N5551/2N5401) Could that have happened when the output transistors shorted?

If you don;t have any thermal feedback, you can easily be in a thermal run-away situation. External diodes can move the temperature sensing to the heatsink. Separate heatsinks for the each transistor is NOT a good idea for the all current gain output stage. the NPN and PNP need to share a heatsink.

Devices in parallel should really be matched. Vbe would be a first order match and so would Hfe. That could be another reason.

You should be able to see idle current (no signal) issues by looking at the voltage across an emitter resistor over time. Time being 1 to 2 hours.

The output of the Vbe regulator is another quick figure of merit. It's probably going to be around 2.4 to 3.x volts and it should change as the transistors heat up. I think it's supposed to go down about 10 mV/deg C. Make sure the potentiometer varies that voltage.

Another way to see the effects of the quiescent current is to monitor the AC current. For testing you just might want to put a low value fuse in the AC input, but it might pop on turn on.

I need to go.

 

Would it be a bad idea to mount one diode on each of the output transistor heatsinks wired in series to help the Vbe multiplier?
Or would that add too much capacitance due to being wired off-board?

 

Don't know. The diode is basically a temperature sensor.

The two output NPN's have to have their DC current gains matched, say within 10-15% anyway. The same goes for the PNP transistors. The PNP and the NPN don't have to have similar gains, just the paralleled pairs.

I'd think you would do vastly better with the 4 output transistors on a single heatsink and a diode string, to basically read the average temperature of the heat sink.

You could also look for a single transistor replacement for the pairs in parallel and mount external from the PCB with thermal feedback.

 

You could also look for a single transistor replacement for the pairs in parallel and mount external from the PCB with thermal feedback.

I was going to ask about that. I had wondered if I could buy a new single heatsink for each channel of the amp and mount one transistor in the PCB to it, and the other off-board using soldered wires back to the PCB. At least that way the NPN and PNP would both be on the same heatsink. I could also mount the Vbe transistor off-board onto this heatsink as well. Think that's a step forward?

Also, would this be OK to replace the 47uF cap:
http://ca.mouser.com/ProductDetail/...GAEpiMZZMv1cc3ydrPrF7l45uRd9dVhc1EB6i%2bmP0o=

 

I think it would be better to place all 4 of the output transistors on the heatsink. I think you could do he diode sensor on the heatsink or mount the Vbe transistor on the heatsink, but NOT both. In all likelyness, you could mount the current sharing resistors there too creat

It's actually been a while since I sized heatsinks

If you happen to have a fan, any fan even a house fan and force cool the amp, you can see if it behaves better, but do get rid of the 47 uF cap and IF you use a speaker use an AGX fuse or fast acting fuse in the speaker output.

With forced cooling, the idle test should be no sweat. Use a low value fuse in the mains.

The main problems, I think are:
1) Lack of thermal considerations and feedback
2) The 47 uF cap on the Vbe multiplier.
3) Not matching Hfe of the outputs

Getting ahead after things don;t blow up: Do a frequency response test using a sine wave. Go out to when the output is down 70.7% for the same input signal. Do a rise time test with a square wave 10 to 90%. Another test you might want to do is to measure the damping factor. See: http://www.bcae1.com/dampfact.htm

 

Last night I re-mounted a single pair of TIP41C/TIP42C onto a different heatsink along with the Vbe multiplier transistor, and replaced the 47uF cap with a 0.1uF cap. I was hoping to test running like that but I blew up both TIP transistors immediately upon power up twice. I don't have a variac to test with unfortunately.

I need to re-check the following things tonight that can push too much current through output transistors:

VAS emitter resistor (100ohm) shorted
VAS constant current diodes open
VAS constant current transistor shorted Base-Collector

Vbe multiplier 1.2k resistor open
Vbe multiplier open from 1.2k resistor to Base
Vbe multiplier shorted Base-Emitter

Driver transistor shorted Base-Collector

This time I'm going to hook up the board to a lab power supply so I can limit the current and slowly bring up the voltage.

This is not the same issue as it usually lasted for a few hours, so something else must have been damaged when the output stage originally blew up, and I'll start with checking the Vbe multiplier stuff.

Anything else to check?

Thanks!

 

I hope you didn't overlook something simple, like the lack of a transistor mounting kit (some need thermal grease, others do not) which keeps the transistors electrically insulated, but thermally not?

You have the ability to isolate with R8 and R27. I think you can check the Vbe regulator with a voltmeter. You want the variable resistor such that the regulator is at minimum when you first power it up,

You can jumper the collector and emitter of the Vbe regulator together when first powering up. This disables the Vbe regulator. the amp should work, but with crossover distortion.

Remember to match hfe for output NPN's and PNP's. You can breadboard a circuit that injects Ib and measure Ic. Two resistors and a fixed and variable power supply. Set the voltage across the bse resistor and measure the voltage across the collector resistor. Convert to currents and eventually Ic/Ib or hfe.

 

I used a connector to plug the transistors into (since I'm exploding them so often) and it turns out the Vbe multiplier wasn't making proper connection. With that transistor out of the circuit my power transistors draw maximum current.

I fixed it up today and got one set of output transistors mounted temporarily to the other heatsink and it's passing a waveform again! Yeah!

I also replaced the 47uF cap with a 0.22uF cap and mounted the Vbe multiplier to the heatsink as you suggested.
I'm going to continue testing tomorrow starting from setting idle current and then leaving it for a few hours without a signal input and watch the voltage across the Emitter resistors. Here's the setup:

Remember to match hfe for output NPN's and PNP's. You can breadboard a circuit that injects Ib and measure Ic. Two resistors and a fixed and variable power supply. Set the voltage across the bse resistor and measure the voltage across the collector resistor. Convert to currents and eventually Ic/Ib or hfe.

That's an exellent tip! I'll definitely start doing that. I currently have been just testing HFe with a cheap DMM.

Thanks again for all your time and help, it is greatly appreciated!
I'll report back after some more testing.

 

That's an exellent tip! I'll definitely start doing that. I currently have been just testing HFe with a cheap DMM.

I didn't know you had one. That works.

The setup looks good.