The most frequent visitors
to the workshop nowadays are printed circuit boards for a variety
of applications. Mostly lift controllers, but also chillers,
central heating, electric fences, and fuse panels for motorcars;
all sorts of odd things arrive each week.
The faults are often not dissimilar to those found in old CRT
TV sets, being cracked solder joints and their consequences.
One of the types of fault however that is not found in domestic
appliances is a damaged varistor. With a thing like a lift controller
there's always the chance the installation will get struck by
lightning and, to combat consequential problems, the designers
generally fit a varistor across each vulnerable input circuit.
Varistors come in various shapes and sizes dependent on the amount
of surge protection likely to be needed.
When a varistor has done its job it may be still in good order,
however if a significant amount of energy has to be dealt with
it can burn up and sometimes catch alight. When this happens
the thing will turn into a very low value resistor and blow any
fuse in the circuit.
Most times the device does its job perfectly, protecting hundreds
of integrated circuits, but will require changing before the
circuit will work. Because of the dratted harmonization of European
mains supplies varistors will go pop because the designers didn't
realise what "harmonization" really meant... but that's
another story.. |
 |
Replacing a varistor without
a circuit diagram or a components list for the board is a bit
of a black art as usually little is left of the original markings
and often there's no firm indication of the circuit parameters.
If there's a row of the things, clearly protecting identical
circuits, one can see the rating of the device on the side of
one that's still nice and blue and shiny.
Unlike resistors and capacitors, varistors from different manufacturers,
but nevertheless having the same rating, can be marked in different
ways and some background knowledge is necessary when purchasing
replacements.
Fixing things like lift controllers (pictured above) is not easy
as sometimes the lift is a hundred miles away and it's not a
good idea to have the circuit board winging its way backwards
and forwards until the fault is fixed. Not just a pain for the
engineer having to drive backwards and forwards, but certainly
not nice for the lift users, who may be old ladies in a nursing
home, having to walk up several flights of stairs while their
lift is waiting to be fixed.
One lift I fixed was the one that went up to a main operating
theatre in a hospital on the Isle of Wight. As the hospital is
judged on the number of operations it can perform the chief executive
was rather upset when his lift failed. Such was the panic that
I was presented with, not one circuit board, but a huge box full
of the things. I had to look at all of them even though it was
most unlikely there were more than one or two faults in all.
It's easy to diagnose a fault when there's a blackened and burnt
diode but very tricky when faced with a faultless board full
of microprocessors and logic devices. Finding that there's no
fault present always takes a lot longer than changing a burnt
diode.
Thankfully most faults are due to the failure of common components
such as relays, diodes and varistors and very few due to failure
of complex chips.
A lot of failures can be blamed on the original designer. These
generally fall into the category of excess heating causing the
solder to fail or the board material or the component to burn.
Many faults are due to finger trouble such as connecting a wire
wrongly or short-circuiting connections and some are due to lightning
or surges on the mains supply. Rarely does a component fail when
it is run within its ratings.
I recall that when I worked in Defence, many millions of pounds
was spent calculating reliability figures so that spare parts
could be made available to keep equipment running. Looking back
I think that the figures that were turned out were absolute rubbish,
as the most common reasons for failure were never considered.
I bet that there are MoD warehouses stacked full of components,
costing many billions of pounds, that will never be used. No
one will ever admit this, as goodness knows how many little empires
are dependent on the calculations, supply and handling (and disposal)
of the stuff. |
|
|
|
Just horsing around!
The latest arrival (March 2016)
is a large circuit board. It's come from a nursing home where
someone thought it would be nice if they introduced a horse to
an inmate unable to leave the first floor. The doors closed and
the lift moved a few inches then just stuck and wouldn't go up
or down. An hour later the people were rescued from the lift
complete with horse, having burnt out the motor contactors which
I duly replaced. |
Typical drive unit repairs
|
|
Below: a picture of the
chassis interior of a damaged lift motor drive unit. This one
is known as a "Vacon" unit. |
 |
|
The circuit boards and
the IGBT module have been removed and you can see where a failure
has occurred by the soot that's been ejected from the module.
In this instance the module
was made by Semikron and contained all the major high power semiconductors
necessary to control the 3-phase motor.
There's a 3-phase bridge rectifier,
a set of IGBTs (insulated gate bipolar transistors) for driving
the motor and operating the brake.
The circuit board carries the
IGBT drive components, power supplies and interconnections for
the module etc.
Unfortunately, when an IGBT
module fails some 600 volts or so is placed on the drive circuits
because the insulated gate often breaks down. This causes lots
and lots of damage to the low voltage circuitry which is very
difficult to repair because the parts are surface mounted and
carry microscopic codes instead of part numbers.
Generally, one can reckon on
a dozen replacement parts in addition to the expensive IGBT module.
This type of unit can provide
typically from 7KWatts to 12 KWatts of power. |
 |
|
Above: a view inside
a typical blown-up IGBT module used in a Kone drive unit. It's
about100mm x 50mm or in sensible units: 4 inches x 2 inches.
The plastic lid has been removed
so you can see the inside.
The square white areas are power
transistors and diodes.
The black smudges are made by
soot which is centred on parts which have failed. The soot is
underneath a jelly-like substance which is used to encapsulate
the assembly.
Amazingly, a module like this
can supply 60 or 100 amps at 600 volts or more to drive a lift
motor.
A failure can occur for several
reasons. For example if a motor bearing is seizing, excessive
current flows and this results in the module getting too hot.
The hotter the module gets the less is it able to provide the
high currents taken by the motor and it fails catastrophically
often taking out a large fuse rated at 100 amps.
A pair of IGBTs for each of
the 3 phases supplying the motor is usually connected in series
(a "totem pole" push-pull circuit). If the top transistor
collector carrying 600 volt HT breaks down to its insulated gate,
a considerable amount of damage will be caused to the control
circuit board.
Early drive units used SCRs
(Silicon controlled rectifiers) but more modern types use a set
of very high power IGBT transistors (IGBT=Insulated Gate Bipolar
Transistor).
An IGBT can control huge currents
at the expense of almost zero power input, much like a high power
thermionic valve used in broadcast transmitters. |
 |
|
This is the top
side of the board that connected to the module shown above.
The underside of the board is
covered with tiny surface mounted parts and the IGBT module is
soldered by over 30 tags which are located around the 4 blue
capacitors left of centre.
The parts which are usually
damaged when the module fails are located in 4 places. Below
the 3 orange blocks (pulse transformers) top left; to the left
of the pulse transformer at the centre, and underneath both these
areas.
The damaged parts will be primarily
the optical couplers (the white i/cs) the capacitors, diodes
and zeners which feed the insulated gates in the module.
In addition the pair of contactors
(bottom right) which are mounted on a second control board may
have burnt or welded contacts.
These contactors govern the
direction of rotation of the motor.
In this model of drive unit
the contactors are not mechanically coupled together so, if both
operate and connect the motor to simultaneously go forward and
backwards (perhaps due to welded contacts) the IGBT module will
fail catastrophically.
The six large black capacitors
smooth the rectified 3-phase mains producing up to 600VDC or
so; the HT supply to the IGBT module. |
|
Not all jobs are straightforward
and clinical. Take this circuit board for example where the designer
forgot to consider localised heating from feed resistors.
Two areas were affected similarly, and
both sides of the board needed fixing.
Why not chuck it away? Well sometimes
a replacement circuit board is no longer available and there's
no option but to make repairs, no matter how messy, as the only
viable alternative is to replace major parts of the lift system
costing tens of thousands of pounds.
If the burning is too serious a new
section of board material has to be grafted in place once the
burned area has been cut out.
In this case the board carries a pair
of numeric indicators showing the floor of the building. Sitting
at ground "G" most of the time resulted in the pull
up resistors for the relevant LEDs in the matrix getting very
hot. Eventually the solder melted and the joints began to go
intermittent. One or more resistors then got even hotter and
one eventually failed. The only warning something was amiss was
when part of the "G" disappeared. |
 |
|
Here's an instance of
what happens when a lift engineer connects a high voltage to
a low voltage circuit. Because a connector was poorly numbered
Pin 4 wasn't Pin 4 carrying a ground connection of zero volts
it was Pin 21 carrying 24 Volts.
Unfortunately, in this case
24 Volts was connected to the local intercommunications bus network
and damaged 3 boards.
On the left is a bus chip (an
NXP PCA82C251T/N3) and on the right four zener diodes and a termination
resistor. Clues to the mishap are the small holes in the top
of the zener diodes and the burn mark on the 220 ohm resistor.
Sometimes, one of the hardest jobs is to identify the damaged
parts. Here the letter "J" lying on its side is a good
clue as it's a date marker used in this particular orientation
by one specific manufacturer. For example, the same marking "Y4"
is used for completely different devices by 13 different makers
on an SOT-32 or similar 3-pin package. Checking data sheets for
each likely candidate reveals only one maker who marks the date
of manufacture with a sideways "J". If this method
fails to identify a likely candidate, or the markings are obliterated,
the only option is to reverse engineer the circuit because circuit
board schematics are not available. Some manufacturers even go
to the length of filing off chip markings so board repairs are
virtually impossible outside the manufacturers repair department. |
 |
Now a Magnetek Drive
Unit, an HPV900
|
 |
|
Before fault diagnosis
these units generally need to be stripped down to constituent
parts as above. Drives usually require 3-phase mains because
they can consume lots of power. This model isn't particularly
big, rated at only 5KWatt. As you can see the thing is not very
old and all the parts are very clean, looking quite new. |
 |
|
First to be removed is this input/output
interface board which sits on top of the processor board.
Bothe relays were OK. In other makes
similar relays can be under-rated causing contact burning and
drive failure. I wonder how many units worth around £5,000
have been scrapped for want of a £2 relay?
Below, the processor board, connecting
to the main board via a multi-way flat cable. The cooling fan
for the low voltage power supply carried on the main board is
screwed to the processor board mounting plate. |
 |
|
 |
|
Before the drive was dismantled,
I checked the various test points and found that no low voltages
were present. The 3-phase was connected and the HT of around
600 volts was present at that 25 Amp fuse below. The likely problem
is failure of the low voltage power supply which is fitted on
the main board shown below.
There are several possible reasons
for the failure. Checking circuit components is problematic because
the board carries conformal coating to protect it from damp,
but I was able to confirm the fuses were intact and all the rectifier
diodes and the chopper transistor were OK. This leaves two possibilities.
The feed resistor between the high voltage line and the chopper
chip (the resistor needs to drop not far from 550 to 600 volts
and only one resistor is fitted instead of the usual chain of
three which are necessary to divide the operating voltage to
be within the rating of the resistor's ). The resistor was intact
leaving only the decoupling capacitor which stabilizes the chopper
supply voltage. Checking in-circuit gave doubtful results but
removing it showed it was open-circuit. |
 |
|
Above, just below that orange
relay, you can see where I've removed the dud capacitor. It has
been carefully sited adjacent to a really hot component (that
pink resistor) because the designer forgot that capacitor lifetime
is inversely proportional to temperature.
Below, the underside of the
same main board. That large component is the IGBT module plus
3-phase bridge rectifier. Low edge, centre you can see markings
"+ and -" and resoldering where the new 100uF capacitor
was fitted. |
 |
|
Here's the old 100uF 25volt
electrolytic capacitor. |
 |
|
 |
|
After reassembly the unit will
be as good as new. The user manual had 180 pages and manufacturer's
fault diagnosis is based on the information given on the front
display. Alas, because this was dead, fault diagnosis needed
a different approach. I did in fact find another version of the
manual, containing an extra 40 pages. Two of these showed test
points on the processor board. Because none of these carried
any voltage led me to the chopper power supply.
Electrolytic capacitors have a finite
lifetime... a bit reminiscent of valves in old radio sets. In
this particular application (chopper chip stabilization) there
are now a multitude of electronic items subject to its failure
including TV sets, Sky boxes, light bulbs etc etc and it's about
time manufacturers introduced something new and more reliable
at a cost which doesn't force manufacturers to specify a cheap
pathetically poor alternative. Many, if not all, designers specify
ludicrous MTBF figures for reliability, totally ignoring this
weak link in their products. |
Yaskawa 31KVA Drive Unit
|
 |
|
This drive arrived from
Gatwick recently. Oddly I'd seen an identical drive from an adjacent
lift from the same site back in 2015 so had a good idea what
the fault was. The reported problem was the lift would only move
a couple of inches instead of zooming up to the next floor. You
can see below the sort of power levels involved. Note the 75Amp
fuse and the sizeable thyristor modules... This drive is particularly
well specified as it tolerates 380 to 460 volts and can draw
up to 46Amps from the mains supply. Many continental designed
drives have a much smaller voltage tolerance and frequently bite
the dust if a 3-phase mains imbalance occurs. This particular
model has a weakness in a smallish relay mounted on the chassis.
To accommodate a high current the designers made the mistake
of running relay contacts in parallel. In this case three paralleled
sets. I sometimes see this mistake in small relays, perhaps when
the designer cut costs by using a common relay for both double
pole and single pole applications. Sorting out the bad relay
put the lift back in operation. Not an easy thing on which to
work as its covered in soot, runs 31KVA, and weighs nearly 65
pounds. |
 |
Unidrive
|
|
I mentioned that some
equipments, particularly those designed originally to work from
"220 volt mains" might be vulnerable to damage if used
in the UK (without a power conditioner). Of course there are
other reasons a drive unit can fail, not just from a higher than
average mains supply. I examined this "Unidrive" example
recently. It was unrepairable as are many like this one. Below,
the first picture shows the IGBT module detached from the circuit
board and with its cover removed. The black stuff is soot and
shows up like this because of the jellylike substance covering
the circuit. Because this drive was beyond economic repair a
new one will be fitted. Hopefully, the new one will have a better
mains tolerance. |
 |
|
Not just the IGBT module,
but the circuit board is also knackered. The difficulty met by
designers is the size of the protecting fuse. This needs to be
50 to 100 Amps depending on the drive's ratings. The black soot
is a mixture of residue from copper tracks used to carry 3-phase
mains to the circuitry and burning relays, capacitors and resistors.
The fuse was intact.
The method of driving a lift
motor with all this fancy circuitry is jolly clever but has a
drawback of needing a method of reversing the motor otherwise
people would have to descend by the stairs. The usual method
for reversing a motor is switching the control voltage using
two large contactors. A contactor is just another name for a
relay and contactors can fail just like their smaller relatives.
Some contactor pairs use a metal pin as an interlock so that
they can switch correctly. Unfortunately some pairs do not have
an interlock because the designer didn't think about failure
modes. The result can be seen above and below. Some contactor
pairs are external to the drive unit (like this example) and
its to be hoped that when this drive was replaced the new example
didn't go bang. |
 |
Just in case you like looking at these kind of things,
here's a couple more...
|
 |
 |
Big trouble..tiny cause
|
|
Schindler circuit boards are of very high quality,
at least compared with those from lots of other manufacturers,
but I suspect their designers go to town on the design as they
seem to be over-complex to me. Also I've seen a few with pronounced
weaknesses such as the SMIC range which end up rather charred
after a few years use. Today's example (below) is a PEBO pcb
sent to me some time ago and languished unrepaired until just
now. |
 |
|
Many times I see over-stressed components looking
very sorry for themselves but, in this example, all looked pristine.
To make matters tricky many pcbs such as this have conformal
coating for protection from corrosion due to damp conditions
and making simple resistance measurements needs some patience
to ensure the meter probes penetrate the coating. Looking at
the picture above suggests a fault-finding exercise based on
comparisons on stressed parts is not too difficult because there
are two indentical power circuits.
In fact two 24 volt batteries connect to two battery
chargers and two DC-DC converters so comparing the pair of circuits
should reveal any discrepancy and hence pinpoint the faulty component.
Nothing seemed amiss however and pressure of other
jobs resulted in the PEBO pcb being set aside. Yesterday I picked
it up and was determined to discover the problem perhaps by powering
the thing up? Below is a drawing (click it to see a clearer image
although this drawing is very much a simplification of actual
circuitry) which might indicate how to get the pcb to work as
intended. Alas this proved to be too difficult as it meant a
virtual reconstruction of the entire lift wiring on my work bench.
I initially tried by using a pair of 24-volt power supplies for
batteries and another 24 volt PSU for powering the logic area
but the need to add other wiring in too much complexity made
me suspend operations, after at least proving some circuits were
good (the pcb would detect the battery voltages as good or bad)
and the microprocessor was alive and well.
I suppose I did also discover that there were no
critical problems, in that supply currents were minimal, which
led me to believe the fault was going to be less obvious than
the run of the mill types. Of course I'd already tested all the
electrolytic capacitors for poor ESR, diodes and transistors
for shorts etc so time to don my magnifying goggles. Years ago
when testing obscure and complex TV chips a trick was to compare
a suspect chip with a known good chip by measuring across its
pins with a diode test meter and an ohm-meter. Often one found
a short between two pins and a new chip could be fitted with
fingers crossed that it didn't suffer the same fate. |
 |
|
I eventually spotted an anomaly and then identified
the chip in the centre of the picture below (Converter 2) and,
after removing the conformal coating I could see it was marked
SG3524P. I compared the resistances of its 16 pins to ground
against the same chip in the other converter. The clear anomaly
is indicated in red below. |
|
Chip-1 Pin |
Ohms |
Chip-1 Pin |
Ohms |
|
Chip-2 Pin |
Ohms |
Chip-2 Pin |
Ohms |
|
1 |
197K |
9 |
4.3M |
|
1 |
199K |
9 |
4.2M |
|
2 |
3.6K |
10 |
6.9K |
|
2 |
3.6K |
10 |
6.9K |
|
3 |
2.8K |
11 |
4.7K |
|
3 |
2.8K |
11 |
4.6K |
|
4 |
47 |
12 |
6.5K |
|
4 |
47 |
12 |
6.5K |
|
5 |
0.1 |
13 |
6.5K |
|
5 |
0.1 |
13 |
6.5K |
|
6 |
5.6K |
14 |
4.7K |
|
6 |
4.9M |
14 |
4.7K |
|
7 |
4.7M |
15 |
1.8K |
|
7 |
4.5M |
15 |
1.8K |
|
8 |
0.1 |
16 |
4.4K |
|
8 |
0.1 |
16 |
4.4K |
|
 |
|
It had been mentioned that Converter 2 output
was being reported by the system as being outside its correct
limits and it was Converter 2 that seemed to have the anomalous
reading. Of course this might have been a red herring and Converter
1 had an incorrect reading but further analysis showed otherwise.
It was virtually impossible to accurately trace the PEBO circuitry
because it's a mult-layer board and the conformal coating made
it awkward to check connections, but comparing the layout of
components on the PEBO pcb indicated that the SG3524P manufacturer's
recommended circuit below came pretty close. As you might notice
above the designers have used extra components to allow control
and monitoring and it's these which make testing quite complex. |
 |
|
In fact the chip requires two timing components
which set its operating frequency, being those at Pins 6 and
7. So there it is... Pin 6 needs a resistor to ground and in
the "good" circuit this is 5.6K. In both cases each
SG3524P has a chip resistor marked 5.61K but in one case this
chip resistor read open circuit. I removed the bad chip and fitted
a new resistor. To make things easier I used a slightly larger
resistor* (shown in the pcb picture
marked 5.6Kohm) which for sake of authenticity I selected from
several marked 5.6K as reading exactly 5.61K. In fact the value
is likely to be uncritical as the associated 10nF timing capacitor
is unlikely to be closely toleranced.
* original 0603 @ 1.6mm
x 0.8mm (see below) compared with new 0605 @ 2mm x 1.2mm |
 |
|
Why did that resistor fail? I might look at
it with my new USB microscope but I can't think of a reason other
than a manufacturing defect. Once the pcb has been returned to
the customer we'll see if it works.... |
