Isolation and protection for SCR gate circuits

[Paul E. Schoen]

I have the prototype mostly done for the "smart" SCR trigger circuit
described in a previous post, and it works well enough at low power. One of
my next steps will be to install it in a test set under actual operating
conditions, which are 480 to 600 VAC fused at about 200 amperes, with
voltage to ground of 277 to 450 VAC.

Previous designs have used dual bobbin transformers rated at about 4000
VDC, and optoisolators (4N35) rated at 3500 VDC. They have been in service
for about 20 years in various forms, and there have been no reported
failures due to isolation breakdown.

In order to reduce the board size (and allow DC power), my new design uses
DC-DC converters to provide gate power. My prototype uses C&D Technologies
NMV1205SA which is rated at 3 kVDC isolation. 5 VDC at 200 mA (current
regulated) seems to be enough to fire the gates on a small 90A SCR package.
I think for the large SCRs I will need something like 12 VDC at 250 MA, but
most 3 watt packages seem to be rated at only about 1000 or 1500 VDC. I
found one from http://www.tracopower.com/ with reinforced insulation that
is rated at 4000 VDC, but its working voltage is only 300 VAC, and it costs
about $40 each (I need two).

I will probably make my own isolated voltage supply using ferrite toroid
transformers and a simple switching circuit, but that is for another post.
I would like to discuss the requirements for isolation and safety at these
high line voltages, and ways to provide protection or minimize damage from
failure.

There is an IEC standard (IEC/EN 60950, UL 60950-1) that seems to state
that normal insulation requires a rating of 1000 VDC plus twice the peak
voltage, so this would be about 1400 VDC for 120 VAC, 1800 VDC for 240 VAC,
and 2000 VDC for 300 VAC. There is also a "double insulation" standard that
is about twice these values, and I think many European parts are required
to have a 4000 volt rating. However, it did not seem to be very specific
for 480 VAC or 600 VAC.

I am mostly interested in avoiding any sort of catastrophic failure that
could cause injury to a test set operator, or a chain reaction sort of
failure that could cause carbonization, arcing, and ionization that could
lead to a severe high current fault. The SCR and the controller PC board
are located in a steel enclosure that should contain low energy faults. The
SCR gate leads are typically twisted pairs of about #22 AWG Teflon wire,
which should ultimately limit the magnitude of a fault to a couple hundred
amperes, for a few milliseconds, without extreme damage.

I thought about adding 1/2 amp fuses on these leads, but I would need four
fuses rated at 600 V (such as KTK), which cost about $12 each. This is
probably overkill. The tracks on the PC board are only about 20 mils, so I
think they would burn open and limit the damage to repairing or replacing
the board. However, I'm not sure if this is reliable at such high voltages.
If arcing and carbonization occur, a more serious fault may result.

The worst case scenario would be if a fault current entered the control
circuitry, which is in turn connected to logic circuitry. Actually, there
is a second small solid state relay in the controller which provides
another level of isolation, but at that point the fault voltage will have
entered a wiring harness which has wires rated at 300 V, and there are
connectors and terminals which might not safely insulate this higher
voltage.

I know there are a lot of SCR controllers in use, and most of them use
pulse transformers with a good safety record. There are also motor
controllers using IGBTs on 720 VDC busses derived from 480 to 600 VAC
mains, and they commonly use optoisolators and DC-DC converters without
major horror stories that I know of. So, perhaps I am being overly
cautious, but I would like to draw on your experience and knowledge to make
sure I make properly informed decisions.

Many thanks,

Paul E. Schoen
www.pstech-inc.com

 

[legg]

On Sat, 29 Jul 2006 03:49:51 -0400, "Paul E. Schoen"

I am mostly interested in avoiding any sort of catastrophic failure that
could cause injury to a test set operator, or a chain reaction sort of
failure that could cause carbonization, arcing, and ionization that could
lead to a severe high current fault. The SCR and the controller PC board
are located in a steel enclosure that should contain low energy faults. The
SCR gate leads are typically twisted pairs of about #22 AWG Teflon wire,
which should ultimately limit the magnitude of a fault to a couple hundred
amperes, for a few milliseconds, without extreme damage.

I thought about adding 1/2 amp fuses on these leads, but I would need four
fuses rated at 600 V (such as KTK), which cost about $12 each. This is
probably overkill. The tracks on the PC board are only about 20 mils, so I
think they would burn open and limit the damage to repairing or replacing
the board. However, I'm not sure if this is reliable at such high voltages.
If arcing and carbonization occur, a more serious fault may result.

The worst case scenario would be if a fault current entered the control
circuitry, which is in turn connected to logic circuitry. Actually, there
is a second small solid state relay in the controller which provides
another level of isolation, but at that point the fault voltage will have
entered a wiring harness which has wires rated at 300 V, and there are
connectors and terminals which might not safely insulate this higher
voltage.

You are basically referring to the behavior of your hardware under
single-fault abnormal conditions.These are only considered possible
across basic insulation barriers or components without reinforced
insulation.

Your test equipment will be expected to provide reinforced insulation
between the hazardous potentials and the operator, so that single
faults do not result in a hazard to the operator.

Single faults are basically any single short circuit or open circuit
condition that may occur. For any specific single fault, a
satisfactory safe result is expected - no hazardous conditions
presented to user accessible terminals - no fire or sustained source
of combustion - no ejected or loose material or connections that might
produce a second fault - no opening of wiring that is not specifically
designed to open safely (forget about it - no open printed traces or
harness wires) - no loss of hipot test integrity.

Some components can reliably perform protective functions without
producing an unsafe result - a fuse or other limiter that is correctly
sized and placed can prevent many downstream devices from unsafe
behavior. Series or parallel components in some situations prevent any
abnormal conditions from occurring.

Single faults applied to components on the operator-accessible
portions of the circuitry also have to result in the same behavior.

Draw up a table of components and board spacings that don't provide
reinforced insulation levels or redundant connections. It is an
interesting exercise predict the result of te open or short condition.
These should be followed up with practical testing, using the exact
components intended in the end use, if you don't want your
assumptions overturned at an inconvenient later date.

RL

 

[Paul E. Schoen]

On Sat, 29 Jul 2006 03:49:51 -0400, "Paul E. Schoen"



You are basically referring to the behavior of your hardware under
single-fault abnormal conditions.These are only considered possible
across basic insulation barriers or components without reinforced
insulation.

Your test equipment will be expected to provide reinforced insulation
between the hazardous potentials and the operator, so that single
faults do not result in a hazard to the operator.

Single faults are basically any single short circuit or open circuit
condition that may occur. For any specific single fault, a
satisfactory safe result is expected - no hazardous conditions
presented to user accessible terminals - no fire or sustained source
of combustion - no ejected or loose material or connections that might
produce a second fault - no opening of wiring that is not specifically
designed to open safely (forget about it - no open printed traces or
harness wires) - no loss of hipot test integrity.

Some components can reliably perform protective functions without
producing an unsafe result - a fuse or other limiter that is correctly
sized and placed can prevent many downstream devices from unsafe
behavior. Series or parallel components in some situations prevent any
abnormal conditions from occurring.

Single faults applied to components on the operator-accessible
portions of the circuitry also have to result in the same behavior.

Draw up a table of components and board spacings that don't provide
reinforced insulation levels or redundant connections. It is an
interesting exercise predict the result of te open or short condition.
These should be followed up with practical testing, using the exact
components intended in the end use, if you don't want your
assumptions overturned at an inconvenient later date.

RL

Thank you for your detailed response. It appears that the safest approach
is to use 600 VRMS fuses for the four gate leads. The main disadvantage is
increased cost and complexity of installation. If I could find fuses of
smaller dimension and lower cost, preferably PCB mounted or in an in-line
holder, I would probably go with this option. Any recommendations on
alternatives to the KTK and similar fuses?

I also considered fusible resistors, but they seem to be rated at 300 volts
and less, and are essentially thermal time delay elements which would not
provide the desired fast trip and current limiting of a fast acting fuse.

I have looked at many SCR and IGBT trigger boards, and have not seen any
special means of protection from a failure of isolation. It may be that the
isolation devices (optoisolators and transformers) have special reinforced
insulation and are considered intrinsically safe. It is always possible
that any component could fail, and it is impossible to design for total
safety under all circumstances.

The DC-DC converters with reinforced insulation, rated at 4000 volts, are
probably quite safe under these operating conditions. If I use specially
made transformers, also rated and tested at 4000 volts, I think these will
also be safe. The 4N35 optoisolators (per Vishay) are specified with a 5300
VRMS rating, with double molding isolation, so I think they will be safe.

As a precaution, I may build a test jig on which I will mount one or more
of the optoisolators, DC-DC converters, and transformers. I can measure the
insulation resistance at 2500 VDC (or even 5000 VDC), and make sure it is
the same at all voltages. Then I can apply 2500 VAC and leave it on for an
extended time, like 24 hours, and then remeasure the insulation resistance.
I think this will be a fair test for the components themselves. Beyond
that, the most likely source of failure will be the PC board itself,
environmental contamination, and external wiring problems. We normally
hipot the completed SCR boards at 2500 VDC, although we do not specify how
long the voltage is applied. It might be good to specify one minute, and
also perform a before and after insulation resistance test.

Finally, please elaborate on the requirements for "reinforced insulation".
Is it in IEC950? Thanks.

Paul

 

[legg]

Thank you for your detailed response. It appears that the safest approach
is to use 600 VRMS fuses for the four gate leads. The main disadvantage is
increased cost and complexity of installation. If I could find fuses of
smaller dimension and lower cost, preferably PCB mounted or in an in-line
holder, I would probably go with this option. Any recommendations on
alternatives to the KTK and similar fuses?

I also considered fusible resistors, but they seem to be rated at 300 volts
and less, and are essentially thermal time delay elements which would not
provide the desired fast trip and current limiting of a fast acting fuse.

I have looked at many SCR and IGBT trigger boards, and have not seen any
special means of protection from a failure of isolation. It may be that the
isolation devices (optoisolators and transformers) have special reinforced
insulation and are considered intrinsically safe. It is always possible
that any component could fail, and it is impossible to design for total
safety under all circumstances.

Under what circumstances do you anticipate large SCR gate fault
currents? The paths from the anode and cathode are usually pretty
well defined - gate drive signal loops being independant; the gate
structure is fairly low-impedance to the drive circuit loop with an
internal shorting failure mode.

The impedance that normally limits gate current can also be used to
limit fault current, and semiconductor return paths can be integrated
into the drive section to shunt this predictably - a shorting
condition of the semiconductors or an opening condition of the
limiters can produce a satisfactory result.

IGBT and fet gates aren't so predictable due to the failure modes of
the gate oxide, but again, limiting impedances and shunting
semiconductors here can serve the same function.

Single fault fault behaviour doesn't usually depend on the normal
ratings of components - these ratings simply cover normal operation.
The aim is no ignition, no flying parts.

The DC-DC converters with reinforced insulation, rated at 4000 volts, are
probably quite safe under these operating conditions. If I use specially
made transformers, also rated and tested at 4000 volts, I think these will
also be safe. The 4N35 optoisolators (per Vishay) are specified with a 5300
VRMS rating, with double molding isolation, so I think they will be safe.

As a precaution, I may build a test jig on which I will mount one or more
of the optoisolators, DC-DC converters, and transformers. I can measure the
insulation resistance at 2500 VDC (or even 5000 VDC), and make sure it is
the same at all voltages. Then I can apply 2500 VAC and leave it on for an
extended time, like 24 hours, and then remeasure the insulation resistance.
I think this will be a fair test for the components themselves. Beyond
that, the most likely source of failure will be the PC board itself,
environmental contamination, and external wiring problems. We normally
hipot the completed SCR boards at 2500 VDC, although we do not specify how
long the voltage is applied. It might be good to specify one minute, and
also perform a before and after insulation resistance test.

Application of test stess voltages for longer periods than are
required to perform the test is abnormal, and unless you know what
you're trying to prove, its also pointless. Continuous stress outside
of the intended application can create corona, surface tracking and
non-recoverable component damage.

Finally, please elaborate on the requirements for "reinforced insulation".
Is it in IEC950?

It will be specified in the standard that you're intending to address,
though the insulation thicknesses, layer count, and creepage distances
may be device-specific. Lab equipment is EN61010, but transformers
could be covered by EN61558... other components may refer back to VDE
or BSI publications.

RL

 

[Robert Baer]

Thank you for your detailed response. It appears that the safest approach
is to use 600 VRMS fuses for the four gate leads. The main disadvantage is
increased cost and complexity of installation. If I could find fuses of
smaller dimension and lower cost, preferably PCB mounted or in an in-line
holder, I would probably go with this option. Any recommendations on
alternatives to the KTK and similar fuses?

I also considered fusible resistors, but they seem to be rated at 300 volts
and less, and are essentially thermal time delay elements which would not
provide the desired fast trip and current limiting of a fast acting fuse.

I have looked at many SCR and IGBT trigger boards, and have not seen any
special means of protection from a failure of isolation. It may be that the
isolation devices (optoisolators and transformers) have special reinforced
insulation and are considered intrinsically safe. It is always possible
that any component could fail, and it is impossible to design for total
safety under all circumstances.

The DC-DC converters with reinforced insulation, rated at 4000 volts, are
probably quite safe under these operating conditions. If I use specially
made transformers, also rated and tested at 4000 volts, I think these will
also be safe. The 4N35 optoisolators (per Vishay) are specified with a 5300
VRMS rating, with double molding isolation, so I think they will be safe.

As a precaution, I may build a test jig on which I will mount one or more
of the optoisolators, DC-DC converters, and transformers. I can measure the
insulation resistance at 2500 VDC (or even 5000 VDC), and make sure it is
the same at all voltages. Then I can apply 2500 VAC and leave it on for an
extended time, like 24 hours, and then remeasure the insulation resistance.
I think this will be a fair test for the components themselves. Beyond
that, the most likely source of failure will be the PC board itself,
environmental contamination, and external wiring problems. We normally
hipot the completed SCR boards at 2500 VDC, although we do not specify how
long the voltage is applied. It might be good to specify one minute, and
also perform a before and after insulation resistance test.

Finally, please elaborate on the requirements for "reinforced insulation".
Is it in IEC950? Thanks.

Paul

As far as a cheep fuze goes that can have a fair HV standoff for PCB
use, consider using a #30 bare wire that starts at a PCB pad, goes "up"
away from the PCB at least 1/2 inch, and then back "down" to a second
pad a few inches away; the "run" being reasonably parallel to the PCB.
If the current capability is not too high, then any arc would
nominally extinguish fairly fast, as the amount of material (the wire)
to vaporize and provide a continuing arc path is limited.
Having the wire a good distance away from the PCB prevents the hot
plasma from heating the PCB enough to start carbonizing or vaporizing it.
If the power is excessive, the fuze material needs to be in either a
hi-voltage liquid (?freons or transformer oil?) or special fire
retardant powder like that used in power cartridge fuses.
Failing that, use a few feet of space for the interrupter with a high
pressure air blast or triggered shotgun to disturb/kill the arc.


There is a tradeoff; the amount of fuzeable conductive material
needed for the fuse itself should be as little as possible so that when
vaporized, does not lend to a nice highly conductive path about as good
as the original conductor.
And that *includes* the ends if the current and energy is sufficent
to heat up and then vaporize them.

One time, in an Army barricks, someone continually ran thsir radio
and someone else objected and got nowhere - so gave them a "grid leak
detector" to "improve reception".
That was a plug where the wires were twisted together; the idea was
to trip the circuit breaker and shut off power.
Well, it was a poor connection, and the result was a plume arc
shooting about a foot out from the outlet; they tried kicking it to put
the fire out to no avail--it grew longer.
I discovered it and walked over to the breaker box and snapped all of
them off.
The estimated current was 10 amps on a 15 amp circuit; would have
continued "forever".
All of the "grid leak detector" was vaporized and most of that part
of the outlet was vaporized and the wiring was in the process of
providing ions for the arc.

 

[Paul]

Under what circumstances do you anticipate large SCR gate fault
currents? The paths from the anode and cathode are usually pretty
well defined - gate drive signal loops being independant; the gate
structure is fairly low-impedance to the drive circuit loop with an
internal shorting failure mode.

The impedance that normally limits gate current can also be used to
limit fault current, and semiconductor return paths can be integrated
into the drive section to shunt this predictably - a shorting
condition of the semiconductors or an opening condition of the
limiters can produce a satisfactory result.

IGBT and fet gates aren't so predictable due to the failure modes of
the gate oxide, but again, limiting impedances and shunting
semiconductors here can serve the same function.

Single fault fault behaviour doesn't usually depend on the normal
ratings of components - these ratings simply cover normal operation.
The aim is no ignition, no flying parts.

The only condition which concerns me would involve an actual breakdown
of the isolation barrier of the DC-DC converter or the optoisolator.
Hopefully that would be no more likely than a failure in the insulation
of the gate wires, although they only touch the heat sink of the SCRs.
In one case this will be the same potential as the gate, but the
opposite heat sink may differ by 480 to 600 VAC when the switch is off.
It would be impossible to fuse the gate wires for this, but possibly
standoffs could be used.

However, there is one retrofit application in which the SCR assembly is
located remotely from the trigger board, and the gate wires are brought
through an extensive wiring harness on two twisted pairs inside a
shielded cable. It is rated at 600 VAC but it makes me a bit nervous,
especially where it goes through a multi-pin connector with many other
signal and control wires. It would be nearly impossible to rewire the
entire test set, so we take a calculated risk that is (probably) the
responsibility of the original designer.

Application of test stess voltages for longer periods than are
required to perform the test is abnormal, and unless you know what
you're trying to prove, its also pointless. Continuous stress outside
of the intended application can create corona, surface tracking and
non-recoverable component damage.

I would only perform such tests as a sort of accelerated lifetime
failure test, or essentially a destructive test. I would certainly not
subject production units to any more than a one minute hipot at 2500
volts.

It will be specified in the standard that you're intending to address,
though the insulation thicknesses, layer count, and creepage distances
may be device-specific. Lab equipment is EN61010, but transformers
could be covered by EN61558... other components may refer back to VDE
or BSI publications.

We have not generally used any standards for these test sets, other
than what I have considered reasonable and safe. It would probably be a
good idea to research various standards to see what may apply. I have
some NEMA standards that I have referred to, and we try to use
components that are adequately rated. These circuit breaker test sets
are specialty devices, and we have not sought UL or any other
approvals. Some older test sets that we repair and retrofit have had
what I consider to be serious safety issues, and we attempt to correct
them when we work on them.

I'll check the standards you mentioned. Any others I should
investigate? Thanks!

Paul E. Schoen
www.pstech-inc.com
(via Google because Coretel is flaky again)

 

[legg]

The only condition which concerns me would involve an actual breakdown
of the isolation barrier of the DC-DC converter or the optoisolator.
Hopefully that would be no more likely than a failure in the insulation
of the gate wires, although they only touch the heat sink of the SCRs.
In one case this will be the same potential as the gate, but the
opposite heat sink may differ by 480 to 600 VAC when the switch is off.
It would be impossible to fuse the gate wires for this, but possibly
standoffs could be used.

Choosing a suitable DC-DC converter or optical isolator will involve
testing the most likely candidates for suitability.

Obviously the short of the gate lead to anode potential will fire the
SCR, protecting the harness wire connections but requiring a fusible
link to protect the load and possibly the SCR. If the drive limiting
impedance is present in the return path, then any voltage developed on
it will also fire the SCR - possibly an undesirable condition if the
section is subject to high dV/dT in normal operation.

However, there is one retrofit application in which the SCR assembly is
located remotely from the trigger board, and the gate wires are brought
through an extensive wiring harness on two twisted pairs inside a
shielded cable. It is rated at 600 VAC but it makes me a bit nervous,
especially where it goes through a multi-pin connector with many other
signal and control wires. It would be nearly impossible to rewire the
entire test set, so we take a calculated risk that is (probably) the
responsibility of the original designer.

I suggest that you limit your concerns to the equipment that has
developed a need for recognized safety approvals, once the actual
requirement is determined.

Drivers are generally located at the drive point for a number of very
good reasons, only one of which is the reduction in hazardous real
estate.

Safety certification will tend to reduce the practicality of a later
retrofit, for better or for worse.

I would only perform such tests as a sort of accelerated lifetime
failure test, or essentially a destructive test. I would certainly not
subject production units to any more than a one minute hipot at 2500
volts.

Accelerated life testing does not involve applying factors that are
outside the design limits intended. It is more likely to examine
operation under stated temperature extremes and rates of change,
vibration and design/functional limits, including start-up and
shutdown.

We have not generally used any standards for these test sets, other
than what I have considered reasonable and safe. It would probably be a
good idea to research various standards to see what may apply. I have
some NEMA standards that I have referred to, and we try to use
components that are adequately rated. These circuit breaker test sets
are specialty devices, and we have not sought UL or any other
approvals. Some older test sets that we repair and retrofit have had
what I consider to be serious safety issues, and we attempt to correct
them when we work on them.

I'll check the standards you mentioned. Any others I should
investigate? Thanks!

Your end-use market will determine and justify those standards that
are most sensible.

RL