Varistor across two SCRs

[Paul E. Schoen]

We are using an SCR switch to apply a voltage to a high current transformer
for circuit breaker testing. The two SCRs are connected antiparallel, and
the gates are turned on at about 70 degrees phase angle, and then left on
until the desired length AC pulse has occurred, or the breaker on the
output trips and stops current flow.

The applied voltage is selected from transformer taps from 0 to 600 VAC.
Sometimes when switching higher taps, the SCRs self-initiate, causing a
very high output current that trips the breaker instantly, but
unexpectedly. We were able to reduce this effect with a 30 uF cap and 1 ohm
snubber across the tap switch, but the problem still occurred. The
manufacturer had mounted a 575 VRMS MOV varistor across the SCRs, which are
rated 1800 V. When we disconnected the MOV, the problem disappeared.

My question is to determine if there could be a danger to the SCRs. I have
found an MOV rated 1800 V, so that might provide some protection. The
maximum peak voltage that should be imposed on the SCRs is about 850 volts,
and the MOV supplied breaks down at 850 to 900 volts. The snubber across
the SCRs is 0.033 uF and 50 ohms. The capacitance of the MOV is probably
about 0.1 uF.

I think the initial surge through the MOV is magnetizing the transformer,
and then there may be inductive spikes during the contact bouncing of the
switch. I think these spikes caused the MOV to conduct, resulting in a
partial cycle of current. Without the MOV, the SCRs probably do not see a
spike high enough to cause them to trigger on. My understanding of SCRs is
that a voltage above their breakdown voltage will cause spurious
triggering, and possible degradation of blocking voltage.

My other concern is about voltage spikes that may occur when the circuit
breaker on the output opens up. I would expect an inductive kick, which
will initially cause arcing on the breaker contacts, dissipating most of
the stored energy. Since the SCRs are still turned on, I think any other
inductive energy will be dissipated in the snubber on the tap switch, and
the wiring of the primary circuit.

We are still having some problems when the firing circuit board is
connected to the SCRs, however. I will be looking into that Monday. I think
the current in the tap switch snubber is being picked up by the board, and
somehow turning on the gates. The board is mounted on the SCR heat sink, so
maybe it needs to be moved away or put into a shielded enclosure.

Thanks for any ideas or thoughts.

Paul

 

[Phil Allison]

"Paul E. Schoen"



** Why not post this on " alt.engineering.electrical" ??

Thousands of volts and amps at the same time scares hell out of most
electronics types.





........ Phil

 

[Winfield]

[ snip ]

I think the initial surge through the MOV is magnetizing the
transformer, and then there may be inductive spikes during the
contact bouncing of the switch. I think these spikes caused the
MOV to conduct, resulting in a partial cycle of current. Without
the MOV, the SCRs probably do not see a spike high enough to
cause them to trigger on. My understanding of SCRs is that
a voltage above their breakdown voltage will cause spurious
triggering, and possible degradation of blocking voltage.

It's excessive dV/dt that's more likely to cause unintended
SCR turn-on, rather than high voltages. I suggest you get out
a scope with a 1500-volt probe to see what's really happening.

 

[[email protected]]

We are using an SCR switch to apply a voltage to a high current transformer
for circuit breaker testing. The two SCRs are connected antiparallel, and
the gates are turned on at about 70 degrees phase angle, and then left on
until the desired length AC pulse has occurred, or the breaker on the
output trips and stops current flow.

The applied voltage is selected from transformer taps from 0 to 600 VAC.
Sometimes when switching higher taps, the SCRs self-initiate, causing a
very high output current that trips the breaker instantly, but
unexpectedly. We were able to reduce this effect with a 30 uF cap and 1 ohm
snubber across the tap switch, but the problem still occurred.

There should be no current flowing when you change taps, so your
snubber shouldnt be doing anything usefull. You can remove it once you
find the real problem.


The

manufacturer had mounted a 575 VRMS MOV varistor across the SCRs, which are
rated 1800 V.

575V mov accross 600V supply, should be obvious. Are you sure the scr
is conducting? If that were the case why would removing the mov cure
it?

When we disconnected the MOV, the problem disappeared.

My question is to determine if there could be a danger to the SCRs.

Very likely, consider what happens to the secondary current when your
breaker trips. You need to provide an alternative path for it or you
will get a high voltage pulse reflected back through the transformer
to your scrs.


I have

found an MOV rated 1800 V, so that might provide some protection. The
maximum peak voltage that should be imposed on the SCRs is about 850 volts,
and the MOV supplied breaks down at 850 to 900 volts. The snubber across
the SCRs is 0.033 uF and 50 ohms. The capacitance of the MOV is probably
about 0.1 uF.

I would suggest you increase the snubber to 0.47uF, Your mov is
unlikely to have any capacitance worth speaking of.

I think the initial surge through the MOV is magnetizing the transformer,
and then there may be inductive spikes during the contact bouncing of the
switch. I think these spikes caused the MOV to conduct, resulting in a
partial cycle of current.

There shouldnt be an initial surge through the mov. The rest of this
statement doesn't make any sence either.

Without the MOV, the SCRs probably do not see a

spike high enough to cause them to trigger on. My understanding of SCRs is
that a voltage above their breakdown voltage will cause spurious
triggering, and possible degradation of blocking voltage.

My other concern is about voltage spikes that may occur when the circuit
breaker on the output opens up. I would expect an inductive kick, which
will initially cause arcing on the breaker contacts, dissipating most of
the stored energy. Since the SCRs are still turned on,

You hope your scrs are still turned on, it depends how your fireing
them and how lucky you are.


I think any other

inductive energy will be dissipated in the snubber on the tap switch,

I thought you said that snubber was accross the switch, ie shorted out
by the contacts.


and

the wiring of the primary circuit.

We are still having some problems when the firing circuit board is
connected to the SCRs,

I expect this is where most of your problems lie maybe you could post
the circuit.

however. I will be looking into that Monday. I think

the current in the tap switch snubber is being picked up by the board, and
somehow turning on the gates. The board is mounted on the SCR heat sink, so
maybe it needs to be moved away or put into a shielded enclosure.

Thanks for any ideas or thoughts.

What your doing is fairly straight forward, there are no real
"gottchas" apart fron not saturating your output tramsformer. Phase
control on the scrs is a much better way to control the output
current, a whole lot cheaper than a tapped transformer to.

 

[John Popelish]

We are using an SCR switch to apply a voltage to a high current transformer
for circuit breaker testing. The two SCRs are connected antiparallel, and
the gates are turned on at about 70 degrees phase angle, and then left on
until the desired length AC pulse has occurred, or the breaker on the
output trips and stops current flow.

The applied voltage is selected from transformer taps from 0 to 600 VAC.
Sometimes when switching higher taps, the SCRs self-initiate, causing a
very high output current that trips the breaker instantly, but
unexpectedly. We were able to reduce this effect with a 30 uF cap and 1 ohm
snubber across the tap switch, but the problem still occurred. The
manufacturer had mounted a 575 VRMS MOV varistor across the SCRs, which are
rated 1800 V. When we disconnected the MOV, the problem disappeared.

My question is to determine if there could be a danger to the SCRs. I have
found an MOV rated 1800 V, so that might provide some protection. The
maximum peak voltage that should be imposed on the SCRs is about 850 volts,
and the MOV supplied breaks down at 850 to 900 volts. The snubber across
the SCRs is 0.033 uF and 50 ohms. The capacitance of the MOV is probably
about 0.1 uF.

I think the initial surge through the MOV is magnetizing the transformer,
and then there may be inductive spikes during the contact bouncing of the
switch. I think these spikes caused the MOV to conduct, resulting in a
partial cycle of current. Without the MOV, the SCRs probably do not see a
spike high enough to cause them to trigger on. My understanding of SCRs is
that a voltage above their breakdown voltage will cause spurious
triggering, and possible degradation of blocking voltage.

SCRs will trigger on excess voltage and some value of dv/dt
(via capacitive coupling to the gate). As the applied
voltage rises, the tolerated dv/dt goes down.

I can't speculate on the effect the MOVs are having unless I
see the whole circuit. Are both the snubbers and MOVs
connected cathode to anode? Is there any other snubber
across the tap switch or transformer windings?

My other concern is about voltage spikes that may occur when the circuit
breaker on the output opens up. I would expect an inductive kick, which
will initially cause arcing on the breaker contacts, dissipating most of
the stored energy. Since the SCRs are still turned on, I think any other
inductive energy will be dissipated in the snubber on the tap switch, and
the wiring of the primary circuit.

I agree that most of the stored inductive energy will be
dumped into the arc, not the SCRs, but, again, the whole
schematic would be helpful.

We are still having some problems when the firing circuit board is
connected to the SCRs, however. I will be looking into that Monday. I think
the current in the tap switch snubber is being picked up by the board, and
somehow turning on the gates.

Gate loading also has a big effect on the dv/dt rating of
many SCRs. So, exactly what the gate driver connects to the
gates and how the wiring intercepts stray fields can get
involved.

The board is mounted on the SCR heat sink, so
maybe it needs to be moved away or put into a shielded enclosure.

If the driver is very sensitive, perhaps. But adding length
to the gate lead is also problematic, and is probably the
reason it is so close.

 

[john jardine]

[...]
Paul

Sounds like classic dv/dt single cycle switch through.
I'd suggest putting a string of 10k resistors across all the transformer
tapping points. The idea is that whatever the tapping switch is wont to do,
the thyristor is always 'pre warned'of the source sine wave shape, hence
never sees excessive edges.

 

[john jardine]

[...]

Belay the previous. I can't see your switching arrangements allowing it.

 

[Fred Bloggs]

We are using an SCR switch to apply a voltage to a high current transformer
for circuit breaker testing. The two SCRs are connected antiparallel, and
the gates are turned on at about 70 degrees phase angle, and then left on
until the desired length AC pulse has occurred, or the breaker on the
output trips and stops current flow.

The applied voltage is selected from transformer taps from 0 to 600 VAC.
Sometimes when switching higher taps, the SCRs self-initiate, causing a
very high output current that trips the breaker instantly, but
unexpectedly. We were able to reduce this effect with a 30 uF cap and 1 ohm
snubber across the tap switch, but the problem still occurred. The
manufacturer had mounted a 575 VRMS MOV varistor across the SCRs, which are
rated 1800 V. When we disconnected the MOV, the problem disappeared.

My question is to determine if there could be a danger to the SCRs. I have
found an MOV rated 1800 V, so that might provide some protection. The
maximum peak voltage that should be imposed on the SCRs is about 850 volts,
and the MOV supplied breaks down at 850 to 900 volts. The snubber across
the SCRs is 0.033 uF and 50 ohms. The capacitance of the MOV is probably
about 0.1 uF.

I think the initial surge through the MOV is magnetizing the transformer,
and then there may be inductive spikes during the contact bouncing of the
switch. I think these spikes caused the MOV to conduct, resulting in a
partial cycle of current. Without the MOV, the SCRs probably do not see a
spike high enough to cause them to trigger on. My understanding of SCRs is
that a voltage above their breakdown voltage will cause spurious
triggering, and possible degradation of blocking voltage.

My other concern is about voltage spikes that may occur when the circuit
breaker on the output opens up. I would expect an inductive kick, which
will initially cause arcing on the breaker contacts, dissipating most of
the stored energy. Since the SCRs are still turned on, I think any other
inductive energy will be dissipated in the snubber on the tap switch, and
the wiring of the primary circuit.

We are still having some problems when the firing circuit board is
connected to the SCRs, however. I will be looking into that Monday. I think
the current in the tap switch snubber is being picked up by the board, and
somehow turning on the gates. The board is mounted on the SCR heat sink, so
maybe it needs to be moved away or put into a shielded enclosure.

Thanks for any ideas or thoughts.

Paul

Bouncing a HV with inductive source impedance through a mechanical
switch onto dv/dt sensitive components, what do you expect. Unless you
have some kind of overarching throughput requirements, it seems that the
tap switching should be sequenced, power down the primary, make the
switch, then synchronously reapply primary power.

 

[Paul E. Schoen]

Bouncing a HV with inductive source impedance through a mechanical switch
onto dv/dt sensitive components, what do you expect. Unless you have some
kind of overarching throughput requirements, it seems that the tap
switching should be sequenced, power down the primary, make the switch,
then synchronously reapply primary power.

(This reply also relates to other posts. Thanks to all for responding)

Please see schematic:
http://www.smart.net/~pstech/dwgs/F-R116 BTS-300 Retrofit Schematic.pdf

This is a retrofit of a test set that was originally designed and built
using a contactor. We replace the contactor with an SCR to provide better
control of initial phase angle to reduce DC offset which can greatly affect
instantaneous trip time.

There is no way to power down the tapped autotransformer because it also
supplies control power, as well as allowing a range of input voltages
208/240/480/575. These units are often quite old, dating to the 1970s, and
the tap switches may be in questionable condition, so they may have quite a
bit of contact bounce.

One idea I had was to keep the contactor in series with the SCR, and
connect the coil to the NC push-to-turn interlock switch on the tap switch,
so the SCR would only have voltage applied to it after the tap change. But
it is still possible for the voltage to be applied at the peak, and the
contactor might have contact bounce as well.

Another idea was to put a contactor on the variable transformer T4A&B, so
that the inductance of the boost transformer T2 would limit the dV/dt to
the SCR when the tap is changed. But that involves a lot of extra wiring.

This is an unusual situation. Most SCRs have power applied to them only
when the system is first powered up, and then all control is done by means
of gating. Also, the load is usually something like a motor or heaters or
other devices which do not react to a brief 1/2 cycle conduction due to
dV/dt. In this case, a half cycle of conduction at maximum output setting
applies a current of 20,000 to 50,000 amperes to a circuit breaker, and
will probably cause it to trip if it is a "small" breaker of 1600 amps or
less.

It is not possible to use phase-modulated control for output current
adjustment because the circuit breaker must be tested with a sine wave
current similar to a fault current in normal use. The precise control of
initial phase angle helps to eliminate DC offset that can cause the first
1/2 cycle to be as much as twice the usual peak voltage (this happens if
zero-crossing initiation is used on an inductive load).

We have determined that removing the MOV takes care of the problem when the
gates are removed from the firing board. There is still a small snubber
across the SCRs. It needs to be smaller than usual because the leakage can
cause a very large output current before the SCRs are turned on. If the
variable transformer is adjusted to 5% of maximum, the overall ratio of the
primary to the secondary circuit is almost 1000/1, so a snubber current of
100 mA will produce an output of 100 amperes. The 0.033 uF snubber allows
only 7 mA, which can be over 7 amps output. A more typical snubber of 0.47
uF allows 100 amperes or more. The varistor capacitance is probably no more
than 1000 pF, so it should not contribute much to the output current.

There are some test sets by other manufacturers, that have the gate wires
running in a long shielded cable to the firing circuit. I think this is a
dangerous condition, as the only protection is provided by the mains
breaker of 350 amps at 480 VAC, and the gate wires are at least 18 AWG. In
our retrofits, as well as our new production units, we mount the SCR board
on the heat sink of the dual hockey puk SCRs. The heat sink is electrically
connected to the main power, which may be fixed 480 VAC, or a variable
source as in the retrofits where we have problems.

There is obviously something more happening if having the firing PCB
connected still causes problems. It might be through the control wiring, or
possibly capacitively coupled from the changing potential of the heat sink
to sensitive circuitry on the PCB. I'll learn more Monday.

Thanks for all your thoughtful replies. I hope the schematic and this
further explanation might provide what is needed for better analysis.

Paul

 

[Terry Given]

[ snip ]

I think the initial surge through the MOV is magnetizing the
transformer, and then there may be inductive spikes during the
contact bouncing of the switch. I think these spikes caused the
MOV to conduct, resulting in a partial cycle of current. Without
the MOV, the SCRs probably do not see a spike high enough to
cause them to trigger on. My understanding of SCRs is that
a voltage above their breakdown voltage will cause spurious
triggering, and possible degradation of blocking voltage.

It's excessive dV/dt that's more likely to cause unintended
SCR turn-on, rather than high voltages. I suggest you get out
a scope with a 1500-volt probe to see what's really happening.

driving the gates from a stiffer source and placing a large Gate-Cathode
cap helps too.

actually snubbing the SCRs would probably also help; according to the
schematic:

http://www.smart.net/~pstech/dwgs/F-R116 BTS-300 Retrofit Schematic.pdf

the R-C snubber connects across the series combination of the SCRs *and*
the primary winding of T3. it should be across the SCRs.

Cheers
Terry

 

[Michael A. Terrell]

"Paul E. Schoen"

** Why not post this on " alt.engineering.electrical" ??

Thousands of volts and amps at the same time scares hell out of most
electronics types.

....... Phil

You only speak for yourself, 'Toaster Boy'


--
Service to my country? Been there, Done that, and I've got my DD214 to
prove it.
Member of DAV #85.

Michael A. Terrell
Central Florida

 

[Paul E. Schoen]

[ snip ]

I think the initial surge through the MOV is magnetizing the
transformer, and then there may be inductive spikes during the
contact bouncing of the switch. I think these spikes caused the
MOV to conduct, resulting in a partial cycle of current. Without
the MOV, the SCRs probably do not see a spike high enough to
cause them to trigger on. My understanding of SCRs is that
a voltage above their breakdown voltage will cause spurious
triggering, and possible degradation of blocking voltage.

It's excessive dV/dt that's more likely to cause unintended
SCR turn-on, rather than high voltages. I suggest you get out
a scope with a 1500-volt probe to see what's really happening.

driving the gates from a stiffer source and placing a large Gate-Cathode
cap helps too.

actually snubbing the SCRs would probably also help; according to the
schematic:

http://www.smart.net/~pstech/dwgs/F-R116 BTS-300 Retrofit Schematic.pdf

the R-C snubber connects across the series combination of the SCRs *and*
the primary winding of T3. it should be across the SCRs.

Not shown on the schematic is a snubber on the SCR trigger PCB. It consists
of a 0.033 uF 2000 V capacitor and a 50 ohm power resistor.

The gate drive is a current regulated source fixed at about 250 mADC. There
are 0.005 uF capacitors across each gate to cathode. The SCRs have built-in
resistors in parallel with the G-K junction, so they read anywhere from 15
to 35 ohms. I was surprised to find such wide variation, even on new SCR
packages, but some of them are more consistent.

I'll post more about my findings after I do tests Monday.

Thanks,

Paul

 

[John Popelish]

Not shown on the schematic is a snubber on the SCR trigger PCB. It consists
of a 0.033 uF 2000 V capacitor and a 50 ohm power resistor.

(snip)

How are these snubbers connected to the SCRs?

 

[Paul E. Schoen]

(snip)

How are these snubbers connected to the SCRs?

They are in series, with one end to one cathode and the other end to the
other cathode. On the old style trigger board, separate wires were used. On
the new board, the cathode connections to the gate wires are used. The new
board seems to work better than the old one.

It is possible that the current surge through the snubber could cause a
small voltage on the gate by virtue of voltage drop through the wire, but I
don't think it would be enough to cause triggering with gate wires of no
more than 12", and about 18 AWG. The snubber resistor of 50 ohms limits the
surge current to about 20 amps, so the wire would produce only about 0.12 V
at 0.006 ohms/ft. This might be a factor if the board is moved further from
the SCRs, in which case we would use a separately wired snubber. We may do
that anyway.

Thanks for the idea.

Paul

 

[Terry Given]

They are in series, with one end to one cathode and the other end to the
other cathode. On the old style trigger board, separate wires were used. On
the new board, the cathode connections to the gate wires are used. The new
board seems to work better than the old one.

that sounds a bit scary.

It is possible that the current surge through the snubber could cause a
small voltage on the gate by virtue of voltage drop through the wire, but I
don't think it would be enough to cause triggering with gate wires of no
more than 12", and about 18 AWG.

work out the inductance, then see whether or not you think it can turn
the SCRs on. I would have a snubber per SCR, mounted with as low an
inductance method as possible. And I would ensure that the snubber R's
were low inductance (eg carbon composition)


The snubber resistor of 50 ohms limits the

surge current to about 20 amps, so the wire would produce only about 0.12 V
at 0.006 ohms/ft.

and if its 20nH/inch x 12" ~ 250nH, and the 20A snubber current ramps up
in say 100ns, then V = LdI/dt = 250nH*20A/100ns = 50V. these numbers are
of course all imaginary, but you ought to check. sharing gate and
snubber current on a single wire is a bad idea.


This might be a factor if the board is moved further from

the SCRs, in which case we would use a separately wired snubber. We may do
that anyway.

Thanks for the idea.

Paul

Cheers
Terry

 

[Terry Given]

[ snip ]

I think the initial surge through the MOV is magnetizing the
transformer, and then there may be inductive spikes during the
contact bouncing of the switch. I think these spikes caused the
MOV to conduct, resulting in a partial cycle of current. Without
the MOV, the SCRs probably do not see a spike high enough to
cause them to trigger on. My understanding of SCRs is that
a voltage above their breakdown voltage will cause spurious
triggering, and possible degradation of blocking voltage.


It's excessive dV/dt that's more likely to cause unintended
SCR turn-on, rather than high voltages. I suggest you get out
a scope with a 1500-volt probe to see what's really happening.

driving the gates from a stiffer source and placing a large Gate-Cathode
cap helps too.

actually snubbing the SCRs would probably also help; according to the
schematic:

http://www.smart.net/~pstech/dwgs/F-R116 BTS-300 Retrofit Schematic.pdf

the R-C snubber connects across the series combination of the SCRs *and*
the primary winding of T3. it should be across the SCRs.

Not shown on the schematic is a snubber on the SCR trigger PCB. It consists
of a 0.033 uF 2000 V capacitor and a 50 ohm power resistor.

The gate drive is a current regulated source fixed at about 250 mADC. There
are 0.005 uF capacitors across each gate to cathode. The SCRs have built-in
resistors in parallel with the G-K junction, so they read anywhere from 15
to 35 ohms. I was surprised to find such wide variation, even on new SCR
packages, but some of them are more consistent.

Ive been using some little 40TPS12 SCRs, which are about 40R G-K, and I
have a 100nF shunt cap as close to the gate as I can get it (and a gate
driver that can still turn it on PDQ)

I'll post more about my findings after I do tests Monday.

Thanks,

Paul

Cheers
Terry

 

[Tim Williams]

these numbers are of course all imaginary

Of course they are, reactance is all fake numbers! ;-)

Tim

 

[John Popelish]

They are in series, with one end to one cathode and the other end to the
other cathode. On the old style trigger board, separate wires were used. On
the new board, the cathode connections to the gate wires are used. The new
board seems to work better than the old one.

It is possible that the current surge through the snubber could cause a
small voltage on the gate by virtue of voltage drop through the wire, but I
don't think it would be enough to cause triggering with gate wires of no
more than 12", and about 18 AWG. The snubber resistor of 50 ohms limits the
surge current to about 20 amps, so the wire would produce only about 0.12 V
at 0.006 ohms/ft. This might be a factor if the board is moved further from
the SCRs, in which case we would use a separately wired snubber. We may do
that anyway.

Normally, one connects snubbers across the SCRs, because
those are the components you are trying to protect from the
dv/dt at zero current turn off, that does not produce zero
voltage after turn off, because of an inductive load.

But in this case, as you say, any snubber current through
such a snubber produces a multiplied current at the
secondary of the output transformer.

You might model this system as 3 blocks. An input block
that consists of a switched source that includes inductive
energy storage, an SCR switch, and an output block that also
includes inductive energy storage.

The input block is already snubbed by a large RC pair (1 ohm
in series with 30 uF), the SCRs are snubbed by the pair on
the driver board (which I think is a very iffy way to do
that) in addition to having a capacitive MOV across it, but
the output transformer is not snubbed at all.

I suspect that the capacitance of the MOV is ringing with
the leakage inductance of the output transformer, jacking up
the voltage steps caused by input voltage switching.

I think I would experiment with opening the snubber on the
driver board (to lower the off state output current), and
adding a more significant snubber (say, 3 times the
capacitance) across the output transformer.

The RC values on the input side may not be optimal, either.
I suspect the resistance was lowered in an attempt to try
to solve this false firing problem, and that it is lower
than optimum for damping the voltage ringing when the
voltage is stepped, with no other load. The capacitor may
also be larger than optimum. Since the turn off occurs when
the load breaker trips, there is really no need to do much
of anything to protect the SCRs at that time. The MOV is
more important to protect the driver board from excess
voltage than to protect the SCRs.

 

[Paul E. Schoen]

(This reply also relates to other posts. Thanks to all for responding)

Please see schematic:
http://www.smart.net/~pstech/dwgs/F-R116 BTS-300 Retrofit Schematic.pdf

This is a retrofit of a test set that was originally designed and built
using a contactor. We replace the contactor with an SCR to provide better
control of initial phase angle to reduce DC offset which can greatly
affect instantaneous trip time.

There is no way to power down the tapped autotransformer because it also
supplies control power, as well as allowing a range of input voltages
208/240/480/575. These units are often quite old, dating to the 1970s,
and the tap switches may be in questionable condition, so they may have
quite a bit of contact bounce.

One idea I had was to keep the contactor in series with the SCR, and
connect the coil to the NC push-to-turn interlock switch on the tap
switch, so the SCR would only have voltage applied to it after the tap
change. But it is still possible for the voltage to be applied at the
peak, and the contactor might have contact bounce as well.

Another idea was to put a contactor on the variable transformer T4A&B, so
that the inductance of the boost transformer T2 would limit the dV/dt to
the SCR when the tap is changed. But that involves a lot of extra wiring.

This is an unusual situation. Most SCRs have power applied to them only
when the system is first powered up, and then all control is done by
means of gating. Also, the load is usually something like a motor or
heaters or other devices which do not react to a brief 1/2 cycle
conduction due to dV/dt. In this case, a half cycle of conduction at
maximum output setting applies a current of 20,000 to 50,000 amperes to a
circuit breaker, and will probably cause it to trip if it is a "small"
breaker of 1600 amps or less.

It is not possible to use phase-modulated control for output current
adjustment because the circuit breaker must be tested with a sine wave
current similar to a fault current in normal use. The precise control of
initial phase angle helps to eliminate DC offset that can cause the first
1/2 cycle to be as much as twice the usual peak voltage (this happens if
zero-crossing initiation is used on an inductive load).

We have determined that removing the MOV takes care of the problem when
the gates are removed from the firing board. There is still a small
snubber across the SCRs. It needs to be smaller than usual because the
leakage can cause a very large output current before the SCRs are turned
on. If the variable transformer is adjusted to 5% of maximum, the overall
ratio of the primary to the secondary circuit is almost 1000/1, so a
snubber current of 100 mA will produce an output of 100 amperes. The
0.033 uF snubber allows only 7 mA, which can be over 7 amps output. A
more typical snubber of 0.47 uF allows 100 amperes or more. The varistor
capacitance is probably no more than 1000 pF, so it should not contribute
much to the output current.

There are some test sets by other manufacturers, that have the gate wires
running in a long shielded cable to the firing circuit. I think this is a
dangerous condition, as the only protection is provided by the mains
breaker of 350 amps at 480 VAC, and the gate wires are at least 18 AWG.
In our retrofits, as well as our new production units, we mount the SCR
board on the heat sink of the dual hockey puk SCRs. The heat sink is
electrically connected to the main power, which may be fixed 480 VAC, or
a variable source as in the retrofits where we have problems.

There is obviously something more happening if having the firing PCB
connected still causes problems. It might be through the control wiring,
or possibly capacitively coupled from the changing potential of the heat
sink to sensitive circuitry on the PCB. I'll learn more Monday.

Thanks for all your thoughtful replies. I hope the schematic and this
further explanation might provide what is needed for better analysis.

Paul

We did extensive testing today and here are the results (so far):

1. The capacitor of the snubber across the switch sometimes charged up to a
peak value of 800 volts when the tap switch opened, and when it closed on
the next tap voltage, if it was at a peak of opposite phase, there could be
as much as 1600 volts imposed on the SCR. The problem seemed less when the
snubber was disconnected, but we might use one with a smaller capacitor
(about 1 uF), and a bleeder resistor to reduce the voltage to near zero
between taps.

2. The snubber on the board (0.033 uF + 50 ohms) did not seem to cause much
problem with output leakage, so we can use a more effective snubber of
about 0.47 uF and 20 ohms. The leakage current problem is only evident in a
different test set design.

3. The raw 12 VDC power supplies for the gates were bypassed only with 1000
uF capacitors, and we saw some high voltage spikes when there was an
applied voltage transient. Adding a 1 uF ceramic bypass capacitor greatly
reduced the spike, although it actually seemed to make the circuit more
prone to false triggering. Of course, now the circuit could provide a much
more solid high speed pulse.

4. There were occasional gate trigger spikes of several uSec, which
apparently were enough to trigger the gate, The PNP series pass transistors
which turn on the gate current were driven at the base through an
optoisolator. There was no bypassing, and no resistor from B-E. A 0.1 uF
capacitor from B-C greatly reduced the output gate spike. I will also add
an appropriate B-E resistor which should reduce the turn-on threshold.

These circuit changes have reduced the problem to a very rare occurrance.
The MOV was the main cuplprit originally, and the high voltage caused by
the snubber explained why it sometimes conducted so strongly (and, in one
case previously, exploded). The proposed 1800 V MOV will probably be an
inexpensive precaution.

Many thanks for the ideas. I think we have a good handle on the situation
now. By the way, we used a TEK THS730A digital storage scope, which has two
isolated input channels, and some very powerful features. The measurements
we made, on 480 VAC mains power, would have been impossible with an
ordinary DSO. These scopes list for about $4000, can be found on eBay for
about $1500, and we got ours for about half that because of cosmetic
blemishes. I have used Fluke Scopemeters before, and did not like them
much, but this scope is great.

Paul