Custom power transformer post-mortem analysis

[P E Schoen]

We have a circuit breaker test set PI-2500, which can operate on nominal
480, 240, or 208 VAC mains, with an output current of about 10 volts and
2,000 amps continuous. It provides that current and up to 15,000 amps to
trip breakers with no problems.

But... We need a source of 24 VDC at up to 5 amps for some relays, and a
source of 104-240 VAC for a 25 watt 12V switching supply, and a
series/parallel source of 120 VAC for AC relays and some instrumentation.
And also a low power voltage source that reads 1/4 the input voltage for a
meter and a small voltage relay, drawing only a few milliamps.

Originally I used a 250 VA 240x480/120x240 transformer with the input wired
in series to the mains, which could be 480, 240, or 208. The secondary was
connected in series or parallel to get the 120 VAC at 1 amp for the control
circuitry. I also put a 120 watt 24 VDC switching supply on the center tap
of the primary, so it would see 240, 120, or 104 volts. But we had problems
with the transformer overheating and burning up.

We had a custom transformer built, which was specified to meet these
requirements. It seemed to work fine, but we have had a couple of failures,
so I decided to investigate. We were unable to get our next order of these
transformers so we had to revert to the original design which actually used
three transformers. I took some measurements with 208 VAC input, and I found
that the current from the center tap of the input to the 120W power supply
was at most 0.67 amps, and the output of the control winding to the 25 watt
switcher was also 0.67 amps.

Since this is a 250 VA transformer, the input windings should be rated at
480V and 0.5A, so already there is some overload at 0.67A. The 240 VAC
output should be rated at 1 amp, so it should be OK. I think there will be
no problem at 480 VAC and it may be marginal on 240V, but at 208, especially
if the line voltage is low, there is about a 34% overload.

Now for the post-mortem of the custom transformer, which was supposed to
have fixed this problem.

The top layer, X5-X6, was not damaged. It had a DC resistance of 16.25 ohms.
It was 0.0145†dia wire, or #27 AWG, which appears to be rated at 0.59A.
That seems fine.

The next layer, X3-X4, was not damaged. It had a DC resistance of 2.72 ohms.
I think it was 0.0275†dia, or #22 AWG, which is rated 1.88A. That seems
fine for 1.2A, and copper losses of 4 watts seems reasonable.

The next layer, X1-X2, undamaged, had a DC resistance of 0.66 ohms. I had
thought this was because there was a short, but I did not find any damage.
In any case it was the same gauge and should be fine.

H1-H2/H3 measured 4.15 ohms, and was also 0.0275†dia. It was undamaged, and
it should provide current for the entire transformer, less the current for
the load on the tap, so it is only about 200mA.

The H2/H3-H4 winding showed evidence of overheating and the enamel had
disintegrated in some areas to expose bare wire, and there was one spot
where the wire had definitely arced and burned. This primary winding carries
current for the entire transformer as well as the current for the tap load.
It is 1.54A according to a simulation. All currents are as follows:

Supply 1.54A * 200V = 308VA

H3-H4 (winding) 1.54A * 103V = 159VA
H1-H2 (winding) 0.21A * 97V = 20VA
=====
179VA

H1-H2/H3(load) 1.38A * 97V = 134VA
X5-X6 0.36A * 91V = 33VA
X1-X4 1.10A * 110V = 121VA
=====
288VA

The apparent discrepancies are probably due to the series resistance I added
to the model, as follows:

H1-H2 0.21A at 4.0 ohms = 0.2W
H3-H4 1.54A at 1.8 ohms = 4.3W
X1-X4 1.10A at 3.4 ohms = 4.1W
X5-X6 0.36A at 16 ohms = 2.1W

The simulation may not be totally accurate, but it does seem to identify the
problem to be the H3-H4 winding. Although the current appears to be within
the rating of the wire, it is possible that the power supply could draw
enough current to cause a voltage drop and a runaway condition, if it does
not have an undervoltage cutout.

Another factor may be that the output of the 24 VDC supply is connected to a
large (30,000 uf or so) capacitor. The inrush current was so high that the
power supply would often "motorboat" and keep cycling from its overcurrent
shutdown, so I added a simple linear current limiter which basically burns
off 120 watts (24 volts at 5 amps) long enough for the capacitor to charge,
and then is just a 1 or 2 volt drop. But I think it is possible that an
unstable mains voltage supply might cause the switcher to draw higher than
normal current, or its power factor may be such that the RMS current draw is
much higher than expected. But the 34% overload may be the reason for the
eventual failure. Actually the transformer had a thermal switch under the
last winding, but it was somewhat insulated from the overloaded winding so
it probably did not get hot enough to open. And even if it did, it would
only have removed some of the output load, and not the 120W supply which was
the reason for the failure.

We are now in the process of getting a new transformer with a higher current
rating, and probably a separate isolated output for the 120W supply, rather
than using the center tap. But it appears that a much heavier wire could
have been used for the high side, and very small wire for the low side.
Essentially, the top winding was supplying power for everything at a current
twice what it was rated for.

If you want to look at the simulation I did, it is at:
http://www.enginuitysystems.com/pix/PI2500_Transformer.asc

Thanks,

Paul

 

[Spehro Pefhany]

We are now in the process of getting a new transformer with a higher =
current=20
rating, and probably a separate isolated output for the 120W supply, =
rather=20
than using the center tap. But it appears that a much heavier wire could =

have been used for the high side, and very small wire for the low side.=20
Essentially, the top winding was supplying power for everything at a =
current=20
twice what it was rated for.

If you want to look at the simulation I did, it is at:
http://www.enginuitysystems.com/pix/PI2500_Transformer.asc

Thanks,

Paul=20

Does your SMPS have PFC? Usually you need around double the xfmr VA
compared to the watts you get out of the DC output of a FW bridge
rectifier+capacitor.



Best regards,
Spehro Pefhany

 

[Spehro Pefhany]

There's a noted absence of temperature measurements recorded in your
tale. A transformer that gets hot enough to damage it's mindings can
usually be smelled out, particularly if the varnish is air-drying and
not baked. There are more reliable ways of monitoring and indicating
this overstress.

For 50/60Hz power transformers, a decent way is to measure the
resistance change in the windings-- shortly after power is removed, of
course. ;-)

 

[legg]

We have a circuit breaker test set PI-2500, which can operate on nominal
480, 240, or 208 VAC mains, with an output current of about 10 volts and
2,000 amps continuous. It provides that current and up to 15,000 amps to
trip breakers with no problems.

But... We need a source of 24 VDC at up to 5 amps for some relays, and a
source of 104-240 VAC for a 25 watt 12V switching supply, and a
series/parallel source of 120 VAC for AC relays and some instrumentation.
And also a low power voltage source that reads 1/4 the input voltage for a
meter and a small voltage relay, drawing only a few milliamps.

Originally I used a 250 VA 240x480/120x240 transformer with the input wired
in series to the mains, which could be 480, 240, or 208. The secondary was
connected in series or parallel to get the 120 VAC at 1 amp for the control
circuitry. I also put a 120 watt 24 VDC switching supply on the center tap
of the primary, so it would see 240, 120, or 104 volts. But we had problems
with the transformer overheating and burning up.

We had a custom transformer built, which was specified to meet these
requirements. It seemed to work fine, but we have had a couple of failures,
so I decided to investigate. We were unable to get our next order of these
transformers so we had to revert to the original design which actually used
three transformers. I took some measurements with 208 VAC input, and I found
that the current from the center tap of the input to the 120W power supply
was at most 0.67 amps, and the output of the control winding to the 25 watt
switcher was also 0.67 amps.

Since this is a 250 VA transformer, the input windings should be rated at
480V and 0.5A, so already there is some overload at 0.67A. The 240 VAC
output should be rated at 1 amp, so it should be OK. I think there will be
no problem at 480 VAC and it may be marginal on 240V, but at 208, especially
if the line voltage is low, there is about a 34% overload.

Now for the post-mortem of the custom transformer, which was supposed to
have fixed this problem.

The top layer, X5-X6, was not damaged. It had a DC resistance of 16.25 ohms.
It was 0.0145” dia wire, or #27 AWG, which appears to be rated at 0.59A.
That seems fine.

The next layer, X3-X4, was not damaged. It had a DC resistance of 2.72 ohms.
I think it was 0.0275” dia, or #22 AWG, which is rated 1.88A. That seems
fine for 1.2A, and copper losses of 4 watts seems reasonable.

The next layer, X1-X2, undamaged, had a DC resistance of 0.66 ohms. I had
thought this was because there was a short, but I did not find any damage.
In any case it was the same gauge and should be fine.

H1-H2/H3 measured 4.15 ohms, and was also 0.0275” dia. It was undamaged, and
it should provide current for the entire transformer, less the current for
the load on the tap, so it is only about 200mA.

The H2/H3-H4 winding showed evidence of overheating and the enamel had
disintegrated in some areas to expose bare wire, and there was one spot
where the wire had definitely arced and burned. This primary winding carries
current for the entire transformer as well as the current for the tap load.

Do you have a build sheet for the failing part? This should indicate
stack build, former thickness, layering order, turns, magnet wire
gauge and type. You seem rather uncertain about the materials used.
Who wound it and what method was used to fix the core stack? Did you
check for bolt current or the specified mag current?

With all your Xs and Hs, I couldn't follow how the thing is actually
used, or the variations in hookup that could produce differing stress.
These variations in use are not illustrated in your simulation, nor do
the Xs and Hs show up to assist.

Connections that do not exist in the real circuit, but are required
for ease of simulation, are best represented by megohm connections,
not short circuits.

As this is test hardware; what's the duty cycle; how many times is the
circuit turned on and off in a five minute period and how are
interruptions/reconfigurations achieved? Did you witness the actual
application/environment/test sequence that produced the failure?

There's a noted absence of temperature measurements recorded in your
tale. A transformer that gets hot enough to damage it's mindings can
usually be smelled out, particularly if the varnish is air-drying and
not baked. There are more reliable ways of monitoring and indicating
this overstress.

RL

 

[P E Schoen]

"legg" wrote in message

Do you have a build sheet for the failing part? This should indicate
stack build, former thickness, layering order, turns, magnet wire
gauge and type. You seem rather uncertain about the materials used.
Who wound it and what method was used to fix the core stack? Did you
check for bolt current or the specified mag current?

The transformers were built by Magnum Transformer, but unfortunately the
owner died and they are now out of business, which we learned when we tried
to place an order. We never had details of the build and they only supplied
the spec drawing below. We did not do exhaustive testing and just ASSumed it
would work as needed. I have been working with the customer part-time and I
am usually called in when something goes wrong, but not always, and there
are several versions of the product in the field, used under different
conditions.

With all your Xs and Hs, I couldn't follow how the thing is actually
used, or the variations in hookup that could produce differing stress.
These variations in use are not illustrated in your simulation, nor do
the Xs and Hs show up to assist.

I have now posted our specification drawings which may help:
http://www.enginuitysystems.com/pix/M-D357_PI2500_Control_Transformer.pdf
(our original spec drawing)
http://www.enginuitysystems.com/pix/M-D357_PI2500CTL Transformer.pdf (from
original supplier - Magnum)
http://www.enginuitysystems.com/pix/M-D357_Signal_Spec_Dwg_B-2409SD_P1.pdf
(from new supplier - Signal - before redesign)

Connections that do not exist in the real circuit, but are required
for ease of simulation, are best represented by megohm connections,
not short circuits.

As this is test hardware; what's the duty cycle; how many times is the
circuit turned on and off in a five minute period and how are
interruptions/reconfigurations achieved? Did you witness the actual
application/environment/test sequence that produced the failure?

The control transformer and its loads are generally kept on for hours. The
operator will change relay taps several times a minute, perhaps, and the
relays pull a high current from the DC supply but then a reduced holding
current. The large capacitor should keep the PSU from seeing very much of
this, and the continuous draw is much less than its rated 120 watts
(although PF may be a problem).

There's a noted absence of temperature measurements recorded in
your tale. A transformer that gets hot enough to damage its windings
can usually be smelled out, particularly if the varnish is air-drying
and not baked. There are more reliable ways of monitoring and
indicating this overstress.

These failures have generally occurred in the field, under unknown
conditions. I think we had one or two failures during early production, but
there were other problems we addressed at the time and we thought everything
was OK. But we usually do the most comprehensive testing on 480 VAC input,
where the stress on this transformer is minimal, so once it passes the short
low-voltage testing phase we would not encounter the conditions that
probably destroyed the transformer. There are other components that get
quite hot so we are used to smelling that, and although this transformer was
running hot in some earlier testing, I think we made changes so that it ran
fairly cool. But I am not there to supervise testing, and the technicians
often overlook or ignore some of the warning signs.

Another item I just realized is that the clamp-on meter I used for the
readings may not have been properly calibrated for AC+DC true RMS. The first
meter I used would not work at all on DC, so I was given two others to use.
One of them had a note that the DC offset could not be adjusted, but the
other one I think did not read any DC current at all (which is what I
expected). But it could have been a problem with metering. We are going to
put together another unit and I'll repeat the measurements using a shunt and
a millivolt DMM, as well as a scope. But I think the new design will take
care of the problem completely.

Thanks,

Paul

 

[Jamie]

"legg" wrote in message


The transformers were built by Magnum Transformer, but unfortunately the
owner died and they are now out of business, which we learned when we
tried to place an order. We never had details of the build and they only
supplied the spec drawing below. We did not do exhaustive testing and
just ASSumed it would work as needed. I have been working with the
customer part-time and I am usually called in when something goes wrong,
but not always, and there are several versions of the product in the
field, used under different conditions.



I have now posted our specification drawings which may help:
http://www.enginuitysystems.com/pix/M-D357_PI2500_Control_Transformer.pdf
(our original spec drawing)
http://www.enginuitysystems.com/pix/M-D357_PI2500CTL Transformer.pdf
(from original supplier - Magnum)
http://www.enginuitysystems.com/pix/M-D357_Signal_Spec_Dwg_B-2409SD_P1.pdf
(from new supplier - Signal - before redesign)



The control transformer and its loads are generally kept on for hours.
The operator will change relay taps several times a minute, perhaps, and
the relays pull a high current from the DC supply but then a reduced
holding current. The large capacitor should keep the PSU from seeing
very much of this, and the continuous draw is much less than its rated
120 watts (although PF may be a problem).



These failures have generally occurred in the field, under unknown
conditions. I think we had one or two failures during early production,
but there were other problems we addressed at the time and we thought
everything was OK. But we usually do the most comprehensive testing on
480 VAC input, where the stress on this transformer is minimal, so once
it passes the short low-voltage testing phase we would not encounter the
conditions that probably destroyed the transformer. There are other
components that get quite hot so we are used to smelling that, and
although this transformer was running hot in some earlier testing, I
think we made changes so that it ran fairly cool. But I am not there to
supervise testing, and the technicians often overlook or ignore some of
the warning signs.

Another item I just realized is that the clamp-on meter I used for the
readings may not have been properly calibrated for AC+DC true RMS. The
first meter I used would not work at all on DC, so I was given two
others to use. One of them had a note that the DC offset could not be
adjusted, but the other one I think did not read any DC current at all
(which is what I expected). But it could have been a problem with
metering. We are going to put together another unit and I'll repeat the
measurements using a shunt and a millivolt DMM, as well as a scope. But
I think the new design will take care of the problem completely.

Thanks,

Paul

Common failures on some large transformers, depending on winding
design are caused by arcs, arcs that occur at times when the connected
load or input may have intermitting connections while the transformer is
(fluxing). High voltage is generated and depending on the circuit can
breach the insulation.

We've seen unexplainable transformer failures in systems where in
others the same transformer operates just fine. We identified this issue
some years ago with Hi-pot equipment and the fix is to put large MOV's
or some kind of fast suppressers on the transformer.

You are not so much protecting against the out side, it's what it can
generate on its own that can be damaging.

This may not be your issue but it is a good note to keep.

Now and then, the electrical guys replace the 3 phase MOV's
(large ones) on one of our large HI-POT cage. I've done it once myself.

This usually happens when they have up around 10kV DC with a large
cable that has a lot of dielectric material in it, and turning it off
instead of snubbing or ramping it down is damaging.

Jamie

 

[legg]

"legg" wrote in message

The transformers were built by Magnum Transformer, but unfortunately the
owner died and they are now out of business, which we learned when we tried
to place an order. We never had details of the build and they only supplied
the spec drawing below. We did not do exhaustive testing and just ASSumed it
would work as needed. I have been working with the customer part-time and I
am usually called in when something goes wrong, but not always, and there
are several versions of the product in the field, used under different
conditions.

It sounds like the end-user is using the H1H2 winding to develop 240V.
The winding that the autotransformed load is placed on actually sees
less current than the 'drive' winding. Or perhaps you misnamed or
misidentified the terminals, on deconstruction.

If the autotransformed load was present on H3H4 (as expected or
suggested in the schematic's orientation), the current in H1H2 could
be expected to be double that on H3H4. The designer would have to be
made aware of the load configuration and the '0V' terminal identified,
to allow placement of more generous copper in the driving layer - the
layer NOT connected to '0V'. H3H4 current actually reduces with
autotransformer loading on it's terminals, when loads on isolated
outputs are present.

This should be demonstrable in your simulation. Just be sure to
display actual winding currents, rather tha assuming these currents
from load values.

It's possible that your internal testing applies loads correctly, but
end users don't, or do so randomly, without the common autotransformer
terminal being identified. Correct application is made more apparent
if the actual terminals of the transformer are assigned to permit a
dual contact to the '0V' functional contact. Obviously a double
connection on H2H3 is only needed if the input voltage is actually
expected to be either 480 or 240. Quite frankly, I still find the
intended application to be somewhat cloudy, and wonder if the end-user
isn't placed in somewhat the same state.

Your own drawings on file should also include phasing dots on all
windings, particularly those expected to function in series/parallel.

RL

 

[Fred Abse]

I have now posted our specification drawings which may help:
http://www.enginuitysystems.com/pix/M-D357_PI2500_Control_Transformer.pdf
(our original spec drawing)
http://www.enginuitysystems.com/pix/M-D357_PI2500CTL Transformer.pdf
(from original supplier - Magnum)
http://www.enginuitysystems.com/pix/M-D357_Signal_Spec_Dwg_B-2409SD_P1.pdf
(from new supplier - Signal - before redesign)

They appear to be just mechanical drawings, with a basic schematic.
What *would* help is a proper winding plan, order of winding, interleaving,
insulation, etc.

There's more to a transformer than just number of turns, wire gage, and iron.
Sometimes you even need to specify wire tension.

 

[P E Schoen]

"Fred Abse" wrote in message

They appear to be just mechanical drawings, with a basic schematic.
What *would* help is a proper winding plan, order of winding,
interleaving,
insulation, etc.

There's more to a transformer than just number of turns, wire gage, and
iron.
Sometimes you even need to specify wire tension.

Yes, but I would expect a competent custom transformer manufacturer would
know how to meet the customer's electrical specifications and take care of
those details.

We had problems in the past with transformers that were specified for
240/480 volts in series and parallel. But we were driving the primaries
through two Powerstats and there was a big problem with current imbalance.
This was caused by the fact that the two 240 VAC windings were stacked as
two layers, so the impedances were different. So we had to work with the
manufacturer and we suggested a bifilar winding, but that was expensive and
they thought it might lack reliability because in series mode the two
windings would always have 240 VAC potential which could break down. I
suggested they use alternating layers, and we came up with an ABBA winding
scheme which solved the problems.

Who would have thought a Swedish singing group would help with transformer
design?

Paul

 

[Fred Abse]

Yes, but I would expect a competent custom transformer manufacturer would
know how to meet the customer's electrical specifications and take care of
those details.

Don't bet your life on it...

We had problems in the past with transformers that were specified for
240/480 volts in series and parallel. But we were driving the primaries
through two Powerstats and there was a big problem with current imbalance.
This was caused by the fact that the two 240 VAC windings were stacked as
two layers, so the impedances were different.

How do you work that out? Same number of turns on the same core should
have *substantially* the same impedance, What your real problem is, IMHO
is using the primary as an autotransformer, supplying substantial current.
Effectively, there is little or no current in the "auto-secondary" half,
which is also half the main primary. Maybe excusable to drive a 110- volt
fan in an 11/240 instrument, but bad practice the way you're doing
it. Use a fully double-wound transformer.

So we had to work with the
manufacturer and we suggested a bifilar winding, but that was expensive
and they thought it might lack reliability because in series mode the two
windings would always have 240 VAC potential which could break down. I
suggested they use alternating layers, and we came up with an ABBA winding
scheme which solved the problems.

Any time I use a subcontractor to wind anything, they are free to make
suggestions, but the specs are always ours, and we get a C of C for
everything. I insist they make a couple of prototypes, which we then
extensively test, often destructively. Only then do I sign off. We've got
better test equipment than most winding companies I've seen.

 

[P E Schoen]

"Fred Abse" wrote in message

Don't bet your life on it...

We have learned that we must perform fairly exhaustive tests on
transformers. And we have seen quality control problems with some companies
when they have moved to other locations or have been bought by other
companies.

How do you work that out? Same number of turns on the same core should
have *substantially* the same impedance, What your real problem is,
IMHO is using the primary as an autotransformer, supplying substantial
current. Effectively, there is little or no current in the
"auto-secondary"
half, which is also half the main primary. Maybe excusable to drive a
110- volt fan in an 11/240 instrument, but bad practice the way you're
doing it. Use a fully double-wound transformer.

In this case we did not use the center tap as an autotransformer. This was a
totally different animal, having 5 primary cores connected with bus bars as
the secondary, so that individual sections could have the primaries either
shorted or supplied with power to obtain variable output voltage (or
current) into various loads. It has one section rated 5 kVA and four
sections rated 10 kVA. The 5 kVA section is driven by Powerstats. There is
some interplay since each winding can act as a source as well as a sink for
applied power, so the problem may be better described as an imbalance of
coupling between primary and secondary.

Any time I use a subcontractor to wind anything, they are free to
make suggestions, but the specs are always ours, and we get a C
of C for everything. I insist they make a couple of prototypes, which
we then extensively test, often destructively. Only then do I sign off.
We've got better test equipment than most winding companies I've seen.

I have been involved with several companies where we needed custom
transformers for very high current output, and we have found that very few
application engineers really understand our needs. And unfortunately I do
not have the skills to define all of the design criteria. Quite while back,
I designed a high current transformer using toroidal primary cores and bus
bar secondaries, and once I realized the problem with aluminum bus and used
copper, and made sure the connections were absolutely clean, smooth, and
tight, it performed very well.

But then we designed a much larger version, and had a toroid manufacturer
make 16 cores of one specification and four more with a different
specification. But when we tried to build the prototype the results were
disappointing. We found out later that the transformer primaries were wound
with many more turns than needed for the voltage we were applying, and we
could get much more output just by removing about 20% of the primary
windings. They really didn't understand our application, and I did not have
enough information on the tape wound cores to be able to calculate the
optimum dimensions and winding data.

We could have completed the design by removing the windings, and it would
have been a very good transformer, but the boss had become frustrated and
totally abandoned the project, opting instead to use a "proven" EI or C core
design. I still think the previous design would have been better and cheaper
in the long run, but the head honcho starts getting loud and red in the face
whenever I suggest that, so I've had to "bite the bullet" and just let it
be.

Paul