This variant , they've sorted out the bad thermal air path "design" for
the main heatsink, but introduced another.
Zobel network involving 2x 4.7R 5W in parallel litterally back to back
and also hemmed in, touching, by the caps, relay and the copious white
gunk.
A pathetic looking flat baffle plate that presumably was someone's idea
that the inlet air would then magically divert over the tops of these
Rs, unlikely.
So instead of 10W rating , presumably effectively derated to about 3W.
Etiology looks as though PbF failure of one R solder joint (burnt pcb
hole in the pcb around the pin ) leading to that R actually cracking up
and going o/c and eventually the other one going high, and along the way
the heat cooked one of the paralleled caps to short circuit . Also
cooked and melted relay case , so the contact is permanently closed,
explains the switch on and off thumps/chirps , over a few weeks, that
did not used to be there.

2 exactly the same size and shape yellow block encased caps, in parallel.
Surviving one marked 104K 400V and measures 94nF but the other one
marked 33 ,then possibly 4, then the rest is burnt away.
So amp , as a combo ,built for a specific speaker fair enough, but would
zobel using 330nF instead of 430nF +/- 20% or whatever , be that
significant?

N_Cook wrote:

This variant , they've sorted out the bad thermal air path "design" for
the main heatsink, but introduced another.
Zobel network involving 2x 4.7R 5W in parallel litterally back to back
and also hemmed in, touching, by the caps, relay and the copious white
gunk.
A pathetic looking flat baffle plate that presumably was someone's idea
that the inlet air would then magically divert over the tops of these
Rs, unlikely.
So instead of 10W rating , presumably effectively derated to about 3W.
Etiology looks as though PbF failure of one R solder joint (burnt pcb
hole in the pcb around the pin ) leading to that R actually cracking up
and going o/c and eventually the other one going high, and along the way
the heat cooked one of the paralleled caps to short circuit . Also
cooked and melted relay case , so the contact is permanently closed,
explains the switch on and off thumps/chirps , over a few weeks, that
did not used to be there.



** Typical scenario when an amp has been made to oscillate at a supersonic frequency. The WW resistors in the Zobel overheat, destroy whatever is next to them and then detach from the PCB.

If the output devices have survived, repair is simple enough. What event caused it is not a job for a tech but requires a detective with police powers.

Had a DJ customer one who blew his power amp up twice because he did not appreciate the difference between a speaker lead and a signal lead. When he came to me the second time, trying to claim warranty on my repair work, he left very disappointed.


..... Phil

On 17/11/2016 14:21, N_Cook wrote:
This variant , they've sorted out the bad thermal air path "design" for
the main heatsink, but introduced another.
Zobel network involving 2x 4.7R 5W in parallel litterally back to back
and also hemmed in, touching, by the caps, relay and the copious white
gunk.
A pathetic looking flat baffle plate that presumably was someone's idea
that the inlet air would then magically divert over the tops of these
Rs, unlikely.
So instead of 10W rating , presumably effectively derated to about 3W.
Etiology looks as though PbF failure of one R solder joint (burnt pcb
hole in the pcb around the pin ) leading to that R actually cracking up
and going o/c and eventually the other one going high, and along the way
the heat cooked one of the paralleled caps to short circuit . Also
cooked and melted relay case , so the contact is permanently closed,
explains the switch on and off thumps/chirps , over a few weeks, that
did not used to be there.

Worse air problem than that. The "baffle" is actually mu-metal plate as
the tip of the input jack is about an inch from the speaker line.
So another air flow mod for this, a full baffle directing air towards
the Zobell, which should also rebalance air inlet to bias towards the
main h/s inlet, a bit like the usual MB problem.
So perhaps internal HF howl-round was the problem. So add capacitance in
the Prea, to downgrade from 20KHz amp, no one has complained before
doing this, a bass guitar amp anyway.
The relay , now desoldered, was seriously buckled and burnt casing so
the contacts were jammed on. As the click/chirp noise had been like that
for a couple of months, 3 times a week use, difficult to believe the
problem was full on ultrasonic instability driven into 10 W of resitors
over the speaker line, all that time.
There is a hole behind the front fascia, for a telltale LED to drive off
the speaker line. What sort of dropper and cap and zener over the LED+ a
rectifier , to come on at elevated amplitude 10KHz and not have the LED
fail at full on ultrasonic "full-rail" oscillation

Nutcase Kook is at it again




Worse air problem than that. The "baffle" is actually mu-metal plate



** The Kook see Mu-metal everywhere in his dreams.



the tip of the input jack is about an inch from the speaker line.
So another air flow mod for this, a full baffle directing air towards
the Zobell,

** Air cooling for the Zobel - must be a real first.


So perhaps internal HF howl-round was the problem.


** No **** Shylock.


The relay , now desoldered, was seriously buckled and burnt casing so
the contacts were jammed on. As the click/chirp noise had been like that
for a couple of months, 3 times a week use, difficult to believe the
problem was full on ultrasonic instability driven into 10 W of resitors
over the speaker line, all that time.


** Likely only took a minute or so to do the damage, following a short mishap with cables or plugs.

The Kook is so pig ignorant there is no explanation he can comprehend or accept.


..... Phil

Put in replacements and fan and sp relay came on , then a bang and loads
of smoke. 680nF 400 V black cap in series with large toroid inductor,
split open, I'd not noticed the bulge from in-use heating previously.
Seems to be across the sp line also. Replaced that and running wiht no
hotspots and amplifying perfectly happily. That L-C filter and amp being
2x power Darlingtons it would seem (under those sprung clips , you have
to prize all 8 of them back with stirrups and remove the h/s, to see the
idents though).
A lot of differences to the 1 known Mark Bass schematic out there,
marked CMD 102P , as they often seem to be, M3003 is leading digits of
the S/N if the model type is encoded in that.
Anyway putting a scope on the sp output, there is a constant 200KHz
1.5V "modulated" on the audio, nothing of that heard on speaker of
course . Not present on either +/-80V lines.
The first thing to try is de-looming the unscreened ribbon connections
that go across 2mm from the inverter, with rear send/rec lines to prea
go right over the top of the SMPS Tx. Not investigated S/R bypasses, if
opening, then possible more switching noise is injected then.
So reloom with extra sleeving away from the Tx looks possible. Going by
the rest of these amps that loom path was probably "designed" to go on
the worst possible routing.
Failing that , what else to look for. ?
I'm assuming that 1.5V 200 KHz flares up
at times in-use. No idea how much in-use local heating around the Zobel
before the relay got jammed closed (so owner noticed bangs and whoops at
sw on/off) ,but about 300 hours of use since the relay jam until final
overheating

for previous black cap - internally blackened yellow cap

Repositioned that loom and no change.
Did not think previously about disconnecting the PREA, but doing so and
the same level of 200KHz on the zero audio out. OK unterminated input to
the PA but why is this bass PA amplifying 200KHz anyway. No schematic ,
so looks like a small cap going in before the assumed Darlingtons (no
predriver size transistors and cold DVM-D would suggest Darlingtons).

So this amp is class D and that L-C filter should remove the 200KHz
switching modulation, but is 1.5 pk-pk acceptable, time to visit s.e.d I
think.
Uses Si8244BB soic, at least there is full data on that, PWM input to
that and so presumably driving a pair of powerFETs under the h/s clips ,
not Darlingtons

The god folk on s.e.d reckon 1.5V at 200KHz is what there should be as
remnant signal after the L-C filter.
So this is what I reckon happened. The 680nF of the L-C went ohmic
(still measures variously 150 to 250ohm) and lost C so more 200KHz was
passed to the zobel, not made for that level of 200KHz , one of the
paired 4.7R solder points failed with all the heat , then the remaining
4.7R heated up even more.

L'esprit d'escalier. Next time I'd remove the speaker connector on the
main pcb, and solder in the sp wires and add solderable/desolderable
tags on the rear outlet pcb, to give some space for the replacement
components. So can get some air through the replacement components
rather than just around, as well as the same simple baffle to direct air
into that backwater area. Did use long W/W Rs rather than the squat ones
that were in there, so at least the tops will be in moving air.
Added a simple LED monitor , requiring 25V pk-pk 4KHz or 10 V pk-pk
40KHz and about 4V of 200KHz before the LED is visible in a dark room,
wiht instructions to owner to ignore any pulsing with notes, just a
continuous LED would mean a problem in the making.

Unfortunately I did not think to take throughput power measurements
before handing back. Owner immediately realised it was putting out more
audio for his familiar settings and guitar.

Another Mark Bass amp - symptom fluttery rustling noise .
Just a wooden cocktail stick was enough to lever the "solder" joint from
the central pin of the rear elbow jack on the speaker lead. Must have
been,in effect, just the resin/flux holding the wire on there, ball of
solder resting in the cup of the pin with no sign of solder on the pin.