Cause of audio amplifier failure?

[Adrian Tuddenham]

This circuit (except for D53 & D 54)) has been working in a Compton
Melotone cinema organ for about six months and has given no trouble:

http://www.poppyrecords.co.uk/compton/TrebleAmplifier.gif

A few days ago it failed with both O/P transistors (Tr 61 & Tr 63)
totally short circuited on all three terminals. Fuses blew and
everything shut down gracefully.

The connection marked "From Z" is held at a couple of volts & 1
diode-drop below the + 30v rail during normal operation, but rises to
the +30v rail when the amplifier is muted. Until this week, the muting
facility had not been used - the failure occurred shortly after the
amplifier was put into muted condition for the first time.

The instant of failure coincided with the organist pressing a toe piston
which operated several 24 v(dc) solenoids. These are (obviously)
inductive, they draw the best part of an amp each and have unsuppressed
self-break contacts to switch them off as soon as they have mechanically
toggled. Their wiring is independent of the amplifier but runs "all
around the houses" in the vicinity of it.

I replaced the blown O/P transistors and checked for other faults, but
found none, the amplifier is now working again.

My diagnosis was that a transient had occurred on the loudspeaker wiring
when the O/P transistors were muted and effectively open circuit; it had
driven the collector of one of them beyond the power rail and caused a
base-emitter reverse breakdown. The excessive current then propagated
further failure in the opposite transistor of the pair.

To prevent this happening again, I have added D53 & D54


Has anyone experienced this mode of breakdown (or heard of it occurring
elsewhere)? Is there another, more likely, failure mode which I have
overlooked?

 

[Phil Allison]

"Adrian Tuddenham"

This circuit (except for D53 & D 54)) has been working in a Compton
Melotone cinema organ for about six months and has given no trouble:

** Sheer luck.

http://www.poppyrecords.co.uk/compton/TrebleAmplifier.gif

A few days ago it failed with both O/P transistors (Tr 61 & Tr 63)
totally short circuited on all three terminals. Fuses blew and
everything shut down gracefully.

The connection marked "From Z" is held at a couple of volts & 1
diode-drop below the + 30v rail during normal operation, but rises to
the +30v rail when the amplifier is muted. Until this week, the muting
facility had not been used - the failure occurred shortly after the
amplifier was put into muted condition for the first time.

The instant of failure coincided with the organist pressing a toe piston
which operated several 24 v(dc) solenoids. These are (obviously)
inductive, they draw the best part of an amp each and have unsuppressed
self-break contacts to switch them off as soon as they have mechanically
toggled. Their wiring is independent of the amplifier but runs "all
around the houses" in the vicinity of it.

I replaced the blown O/P transistors and checked for other faults, but
found none, the amplifier is now working again.

My diagnosis was that a transient had occurred on the loudspeaker wiring
when the O/P transistors were muted and effectively open circuit; it had
driven the collector of one of them beyond the power rail and caused a
base-emitter reverse breakdown. The excessive current then propagated
further failure in the opposite transistor of the pair.

To prevent this happening again, I have added D53 & D54

** Those two diodes ought to have been included from the start - to
suppress the back emf transient from the speakers when if the mute function
is engaged at the same time as the amplifier is under high drivel levels.
Such transients are well capable of destroying output devices by reverse
breakdown.

Why no output Zobel network ? ( ie 100nF plus 8 ohms in series across the
speaker line ) - I seriously doubt the stage is reliably stable at HF
without one. Excessive heat due to such oscillation will kill the BJT
output devices fast too.

Also, tell me what happens if you remove the negative side DC rail fuse
during operation ??

Fried voice coils??



...... Phil

 

[Eeyore]

This circuit (except for D53 & D 54)) has been working in a Compton
Melotone cinema organ for about six months and has given no trouble:

http://www.poppyrecords.co.uk/compton/TrebleAmplifier.gif

A few days ago it failed with both O/P transistors (Tr 61 & Tr 63)
totally short circuited on all three terminals. Fuses blew and
everything shut down gracefully.

The connection marked "From Z" is held at a couple of volts & 1
diode-drop below the + 30v rail during normal operation, but rises to
the +30v rail when the amplifier is muted. Until this week, the muting
facility had not been used - the failure occurred shortly after the
amplifier was put into muted condition for the first time.

The instant of failure coincided with the organist pressing a toe piston
which operated several 24 v(dc) solenoids. These are (obviously)
inductive, they draw the best part of an amp each and have unsuppressed
self-break contacts to switch them off as soon as they have mechanically
toggled. Their wiring is independent of the amplifier but runs "all
around the houses" in the vicinity of it.

I replaced the blown O/P transistors and checked for other faults, but
found none, the amplifier is now working again.

My diagnosis was that a transient had occurred on the loudspeaker wiring
when the O/P transistors were muted and effectively open circuit; it had
driven the collector of one of them beyond the power rail and caused a
base-emitter reverse breakdown. The excessive current then propagated
further failure in the opposite transistor of the pair.

To prevent this happening again, I have added D53 & D54

Which should have been there to begin with.

Has anyone experienced this mode of breakdown (or heard of it occurring
elsewhere)? Is there another, more likely, failure mode which I have
overlooked?

Nothing obvious I can see.

Graham

 

[Phil Allison]

"Bob Eld"

were not electrically associated with the amp or its power supply.
Furthermore added diodes D53 and D54 don't do anything because the voltage
on the speaker rail will never exceed the power rails and these diodes
will
never forward bias. They're not needed.

** Shame that is 100% wrong.

The vast majority of BJT power amps made have such diodes fitted to the
output - for a damn good reason. That reason is that a dynamic speaker CAN
and WILL deliver a very large back emf voltage spike if ever drive to the
output stage is suddenly disabled.

There are two common scenarios where drive is suddenly disabled:

1: The use of a VI limiting circuit that shunts drive current away from the
output stage when a hazardous ( to the output devices) combination of
voltage and current is sensed in the load.

2. The sort of muting circuit use in the OP's schematic.

In both the above, the usually low output impedance of the amp instantly
becomes very high and allows any stored inductive energy in the load to "
kick back " into the output stage.

You need to read up on the topic:

http://sound.westhost.com/vi.htm



....... Phil

 

[Adrian Tuddenham]

I doubt the toe piston solenoids had anything to do with the problem if they
were not electrically associated with the amp or its power supply.
Furthermore added diodes D53 and D54 don't do anything because the voltage
on the speaker rail will never exceed the power rails and these diodes will
never forward bias. They're not needed.

There are circumstance where they might be needed, but I didn't think
any of those applied here (See my reply to Graham)

I suspect the problem lies in the mute circuit. Transistors TR53 and TR54
are current sources for the input differential pair and the drive circuitry
for the output devices. The two currents are set by the voltage between the
30 volt rail and the "Z" point. It's critical that this voltage be
controlled, known and free from external influences such as noise, spikes,
etc. Muting occurs by removing this voltage causing currents from the
current sources to collapse, removing drive and allowing the output to go to
zero.

Since the "Z" point goes off campus, we have no clue what the external mute
circuit does but it's likely that they problem lies there. For openers, the
current setting voltage should never come from outside the amplifier,
undefined. That voltage should be locally defined by a zener or current
mirror and should simply be turned on and off with a shunting transistor
thus isolating the current sources from the outside world.

I didn't show the rest of the circuit, but the 'Z' wire goes to the
same point on an adjacent card which contains the bass power amplifier
driving stage. The 'Z' wire and the +ve 30v rail both run parallel and
are about 1.5" long with soldered connections both ends; both are
inside the same die-cast box.

Noise, spikes, transients, or wrong voltages on the "Z" point could cause
the problem.

The 'z' wire is well decoupled by C53 and this discharges relatively
slowly when the muting operates. It is difficult to see how a fault on
tha 'z' wire would blow only Tr61 and Tr63.

I have now put up is a full description with circuits at:
http://www.poppyrecords.co.uk/compton/compton_transistorPA352.htm

Also look for the amplifier oscillating especially during the
muting transition or during mute. Unstable high frequency oscillation could
be the culprit.

Before I put the amplifier into service, I switched the circuit in and
out of muting many times on the bench, with and without many types of
load. It withstood all sorts of deliberate abuse and didn't show any
tendency towards instability during transitions. The failure appears to
have occurred after the muting condition had been established for
several seconds (possibly minutes).

Good luck.

Thanks, I feel as though I need it.

 

[Adrian Tuddenham]

Adrian Tuddenham wrote:
[...]

My diagnosis was that a transient had occurred on the loudspeaker wiring
when the O/P transistors were muted and effectively open circuit; it had
driven the collector of one of them beyond the power rail and caused a
base-emitter reverse breakdown. The excessive current then propagated
further failure in the opposite transistor of the pair.

To prevent this happening again, I have added D53 & D54

Which should have been there to begin with.

I have seen designs where they are used and others where they aren't,
but I have never found a formal explanation of the reason why any
particular design does or doesn't need them. I assumed they were to
protect against back EMF from the loudspeaker when:

1) The loudspeaker can be disconnected and reconnected during
operation, with possible inductive sparking.

2) The amplifier can be overloaded and both O/P transistors
fortuitously become open circuit at the same time while the loudspeaker
is passing a hefty current

3) The loudspeaker has a complex crossover network which stores energy
at some point in the cycle and can dump it into the amplifier output.

As none of these situations is possible in this particular application
(each speaker is hard-wired to its own amplifier and the input signal is
from tone generators with fixed amplitudes) I did not incorporate the
diodes. I had not considered what might happen with the output
transistors muted.

Nothing obvious I can see.

It has occurred to me that perhaps the power transistor complimentary
doublet is more vulnerable to reverse voltage than a single transistor -
but I haven't found any evidence to support this supposition, provided
both transistors of the doublet have similar voltage ratings.

 

[Adrian Tuddenham]

"Adrian Tuddenham"


** Sheer luck.



** Those two diodes ought to have been included from the start

I agree that, in hindsight, I should have included them; however none of
the usual reasons for having them seem to apply in this case.

- to
suppress the back emf transient from the speakers when if the mute function
is engaged at the same time as the amplifier is under high drivel levels.
Such transients are well capable of destroying output devices by reverse
breakdown.

The muting action is very slow, taking nearly a second to fully mute.
I'm sorry that this wasn't apparent from the diagram I provided, so I
may have misled you on that point. I have discussed the other possible
reasons for those diodes in my reply to Graham.

The full circuit is at:
http://www.poppyrecords.co.uk/compton/compton_transistorPA352.htm

Why no output Zobel network ? ( ie 100nF plus 8 ohms in series across the
speaker line ) - I seriously doubt the stage is reliably stable at HF
without one. .

I have tested it on various loads and it shows no tendency towards
instability. Some amplifier designs include a Zobel network, others
don't; what is the factor which decides this?

Excessive heat due to such oscillation will kill the BJT
output devices fast too

Under full drive with a shorted output, the calculated junction
temperature never rises to an unsafe value because each transistor is in
close thermal contact with a sensing diode which can trigger a shut-down
through the same muting circuit. I really thrashed this amplifier
(thermally) on the bench before it went into service and it always shut
down long before there was any risk to the output devices.

Also, tell me what happens if you remove the negative side DC rail fuse
during operation ??

Fried voice coils??

There isn't a negative DC fuse, the power fuses are on the AC side of
the rectifier. I take the point that there could be a DC fault, so I
have provided each loudspeaker with its own fuse.

The previous replacement amplifier (a commercial transistorised design),
which this one now replaces again, had failed with a short circuit to
one power rail and had burned out the bass loudspeaker (an 18" Goodmans
from 1954)). Fortunately I was able to make a new former and rewind the
voice coil, so it was back in service with very little delay. Because
of this, I was particularly careful to pre-empt that possibility in my
own design.

 

[Jon Kirwan]

On Sat, 14 Mar 2009 13:09:21 +0000,

I have now put up is a full description with circuits at:
http://www.poppyrecords.co.uk/compton/compton_transistorPA352.htm

Why is the naming of Tr21 and Tr22 in the base amplifier in figure 7
called Tr38 and Tr39 in the text?

Jon

 

[Adrian Tuddenham]

On Sat, 14 Mar 2009 13:09:21 +0000,


Why is the naming of Tr21 and Tr22 in the base amplifier in figure 7
called Tr38 and Tr39 in the text?

Actually they aren't named at all, the numbers 21 and 22 are associated
with their base-emitter diodes. It would be logical to name them Tr21
and Tr22 and then change the text, so that is what I shall do.

Thank you for spotting that anomaly.

 

[Phil Allison]

"Bob Eld"
"Phil Allison"

Thanks.
I agree with your other post that a Zobel R-C network is needed on the
output to insure a high frequency load at all times.

** Zobel ( R-C ) networks fitted across speaker outputs do several things.

1. Provide a known RESISTIVE load to the amp at frequencies up to many
MHz.

2. Swamp unpredictable load impedances and resulting phase shifts found in
all speakers, x-over networks and cables at such frequencies.

3. Provide unambiguous EVIDENCE of high frequency oscillation being the
cause of an otherwise mysterious amp failure - ie the resistor gets
smoked.



....... Phil

 

[Adrian Tuddenham]

** That is only how it seemed to YOU - pal

** All of them stupid and WRONG !!

** My god, what a PIG'S BREAKFAST !!

** Proves nothing except you are an incompetent.

** WRONG.

** Whether the designer was an amateur idiot or not.

** What load of old bollocks.

What the hell does bullshit ...

** More off with the fairies style wishful thinking.

** Damn silly waste of effort.

More dopey, wishful thinking.

[Sigh!] Back into the Kill File.

 

[Phil Allison]

"Jan Panteltje"

Actually a 'Zobel' is not always needed, for example the TDA7294 audio amp
chip
does not need it.

** That is NOT what the maker's data says.

FUCKHEAD




...... Phil

 

[Phil Allison]

"Jan Panteltje"

Actually a 'Zobel' is not always needed, for example the TDA7294 audio amp
chip does not need it.

** That is NOT what the maker's data says.

FUCKHEAD

Try reading ALL of it !!

YOU FUCKING MORON .




....... Phil

 

[Eeyore]

Furthermore added diodes D53 and D54 don't do anything because the voltage
on the speaker rail will never exceed the power rails and these diodes will
never forward bias. They're not needed.

You obviously don't know much about power amps and their typical loads. That is
one of the most crass statements I've seen in all my time on Usenet.

Graham

 

[Phil Allison]

"Eeysore"

You obviously don't know much about power amps and their typical loads.

** Interestingly, essentially same problem ( large back emfs from inductive
loads) occurs with valve amplifiers too and leads to arcing inside the
output transformer or between adjacent pins on the output valve bases and
sockets.

The cure is the same - ie using diodes ( very high voltage ones or a series
string) connected in reverse across each output valve from anode to cathode.

That is one of the most crass statements I've seen in all my time on
Usenet.

** Don't exaggerate.

There have been many that were hundreds of times worse.

Some of them from YOU.



....... Phil

 

[Adrian Tuddenham]

Your output transistor protection circuitry design is weak

Can you elaborate on that please? What are the problems I have not
forseen and what circuit changes would you suggest to overcome them
(apart from adding D53 & D54, which I have now done)?

 

[Eeyore]

This circuit (except for D53 & D 54)) has been working in a Compton
Melotone cinema organ for about six months and has given no trouble:

http://www.poppyrecords.co.uk/compton/TrebleAmplifier.gif

A few days ago it failed with both O/P transistors (Tr 61 & Tr 63)
totally short circuited on all three terminals. Fuses blew and
everything shut down gracefully.

How much heatsinking is there ?

Graham

 

[Adrian Tuddenham]

How much heatsinking is there ?

Each O/P transistor is mounted on an individual 6" x 4" rectangle of
0.125" thick alloy plate with thermal compound and no electrical
insulation (the plates are insulated by nylon spacers). They can be
seen in the picture at:

http://www.poppyrecords.co.uk/compton/compton_transistorPA352.htm

The transistors are mounted just below the centre of each plate, the
screws near the top hold a tagstrip for the emitter resistors.

With a current limit of 4 amps per device and a 30 v supply, that gave a
maximum dissipation of 120 watts per device per half-cycle with a square
wave into a short-circuited load; corresponding to an average of 60
watts (less than that if the PSU impedance and emitter resistor losses
are taken into consideration).

Assuming a total thermal resistance of 2 C degrees per watt between the
semiconductor junction and the cooling plate, setting a limit of 65
degrees C on the plate would correspond to a limit of 185 degrees at
the junction. The thermal sensing diodes are actually in thermal
contact with the transistor mounting tabs, not the plates themselves, so
they will respond to a rapid temperature rise more quickly.

I tried several runs into a short circuit from hot and from cold and
with either or both amplifiers shorted; the shut-down always worked well
before the plates reached a dangerous temperature. From cold, the time
to shut down was around 1 minute.

On sustained full load on both channels, with the amplifier stood on end
(which reduced the convection cooling effect) in an ambient of 25 C, the
system shut down after about 2 hours and re-set after about 5 minutes.

Under actual operating conditions in the organ, I have never found the
plates became hotter than about 50 C, even after more than an hour of
deliberately over-loud and bass-heavy playing.

 

[Phil Allison]

"Jan Panteltje WOG **** "



** That is NOT what the maker's data says.

FUCKHEAD

Try reading ALL of it !!


YOU FUCKING WOG **** MORON





...... Phil

 

[Eeyore]

Each O/P transistor is mounted on an individual 6" x 4" rectangle of
0.125" thick alloy plate with thermal compound and no electrical
insulation (the plates are insulated by nylon spacers). They can be
seen in the picture at:

http://www.poppyrecords.co.uk/compton/compton_transistorPA352.htm

The transistors are mounted just below the centre of each plate, the
screws near the top hold a tagstrip for the emitter resistors.

With a current limit of 4 amps per device and a 30 v supply, that gave a
maximum dissipation of 120 watts per device per half-cycle with a square
wave into a short-circuited load; corresponding to an average of 60
watts (less than that if the PSU impedance and emitter resistor losses
are taken into consideration).

Assuming a total thermal resistance of 2 C degrees per watt between the
semiconductor junction and the cooling plate, setting a limit of 65
degrees C on the plate would correspond to a limit of 185 degrees at
the junction. The thermal sensing diodes are actually in thermal
contact with the transistor mounting tabs, not the plates themselves, so
they will respond to a rapid temperature rise more quickly.

I tried several runs into a short circuit from hot and from cold and
with either or both amplifiers shorted; the shut-down always worked well
before the plates reached a dangerous temperature. From cold, the time
to shut down was around 1 minute.

On sustained full load on both channels, with the amplifier stood on end
(which reduced the convection cooling effect) in an ambient of 25 C, the
system shut down after about 2 hours and re-set after about 5 minutes.

Under actual operating conditions in the organ, I have never found the
plates became hotter than about 50 C, even after more than an hour of
deliberately over-loud and bass-heavy playing.

Adequate for the application but I'd never use that little heatsinking on a
commercial amp. Probably not that then.

Graham