Snubber circuit design

[Ignoramus13229]

I am trying to make a Dc to AC inverter using IGBT.

There is a big inductor before the bridge. I do not yet know its
unductance, but I will measure it.

I want to safely handle switching the circuit off (both possibly
during active switching, as well as if the bridge is simply turned
off).

I can try doing two things.

1. Place an appropriately sized capacitor and resistor between the DC
power rails coming in.

2. Also place a big ass varistor between power rails. It could be
rated for appropriate amperage. I can see that there are varistors
rated for quite large currents.

I read a little about varistors here:

http://www.radioradar.net/docs/varistor.php

It appears, from my current thinking as well as previously made
suggestions, that doing both things together makes sense. For short
interruptions done during routine switching (if they occur at all),
the capacitor and resistor would be used, and for anything worse than
that, the varistor would save me.

If that is the case, I would like to see some guidelines on picking a
varistor, given certain inductance (to be found out) and interrupting
voltage. One thing that I am not so certain about is the breakdown
voltage. I figure that for a 85 OCV welder, and 1,200V IGBT, it would
be sensible to select breakdown voltage of, say, 300 volt or
thereabouts.

I will try to measure inductance tonight.

i

 

[Terry Given]

I am trying to make a Dc to AC inverter using IGBT.

There is a big inductor before the bridge. I do not yet know its
unductance, but I will measure it.

I want to safely handle switching the circuit off (both possibly
during active switching, as well as if the bridge is simply turned
off).

that means you have to give the inductor current somewhere to commutate
to, whilst the bridge is off.

I can try doing two things.

1. Place an appropriately sized capacitor and resistor between the DC
power rails coming in.

thats not gonna work so well. that R carries the full load current, so
will need an impressive peak pulse power rating.


an RCD snubber, with a monstrous D, is probably what you want.

2. Also place a big ass varistor between power rails. It could be
rated for appropriate amperage. I can see that there are varistors
rated for quite large currents.

I read a little about varistors here:

http://www.radioradar.net/docs/varistor.php

you can read more languages than I can.

they may not mention the most important part - the wear-out
characteristics of MOVs. basically they can clamp a finite amount of
Joules, then they go short circuit. Then (often within a few hundred
microseconds) they go *BANG*

applications using MOVs to clamp repetitive surges are generally best
suited for breaking MOVs.

find a manufacturer (eg Seimens, or whatever they are called now) that
actually makes the things, and read their app notes.

It appears, from my current thinking as well as previously made
suggestions, that doing both things together makes sense. For short
interruptions done during routine switching (if they occur at all),
the capacitor and resistor would be used, and for anything worse than
that, the varistor would save me.

If that is the case, I would like to see some guidelines on picking a
varistor, given certain inductance (to be found out) and interrupting
voltage. One thing that I am not so certain about is the breakdown
voltage. I figure that for a 85 OCV welder, and 1,200V IGBT, it would
be sensible to select breakdown voltage of, say, 300 volt or
thereabouts.

I will try to measure inductance tonight.

i

Cheers
Terry

 

[Chris Jones]

that means you have to give the inductor current somewhere to commutate
to, whilst the bridge is off.


thats not gonna work so well. that R carries the full load current, so
will need an impressive peak pulse power rating.


an RCD snubber, with a monstrous D, is probably what you want.

you can read more languages than I can.

they may not mention the most important part - the wear-out
characteristics of MOVs. basically they can clamp a finite amount of
Joules, then they go short circuit. Then (often within a few hundred
microseconds) they go *BANG*

applications using MOVs to clamp repetitive surges are generally best
suited for breaking MOVs.

find a manufacturer (eg Seimens, or whatever they are called now) that
actually makes the things, and read their app notes.


Cheers
Terry

I just had a look at the Epcos application notes. They specifically
recommend using their varistors for clamping spikes due to switching off an
inductive load. Take a look: (sorry for the long URL, I'll break it across
a few lines, since I don't know if I'll mess things up otherwise.)

http://www.epcos.com/web/generator/Web/Sections/ProductCatalog/
NonlinearResistors/Varistors/PDF/
PDF__Applications,property=Data__en.pdf;/Applications.pdf

Certainly, if a varistor is used to divert a lightning strike then at best
only one operation can be expected before the varistor becomes useless, but
I do not see any mention of a slower wear-out mechanism in the Epcos
application notes, except the warning that the varistor must not be allowed
to reach an excessive temperature.

I believe that the OP intends to put a large motor-run type capacitor in
parallel with the varistor so that it will only be in unusual circumstances
that the varistor will carry the full welder current.

Chris

 

[Ignoramus3498]

that means you have to give the inductor current somewhere to commutate
to, whilst the bridge is off.

I thought that it could simply flow into the capacitor through the
resistor.

There are two kinds of events, really, a transient switch off during
bridge switching, and a complete shutdown (turn off) of the
bridge. Note that most complete shutdowns would happen while NOT under
power, for obvious reasons. I cannot be switching AC switch off while
welding.

I suppose that shutdowns due to arc extinguishing are rare and softer.

For transient switch offs...

Suppose that, for worst case example, 300 A is flowing into the
circuit. Then a switching even occurs and the circuit opens for 2
microseconds (a very long time for IR22141SS that can likely be
reduced severalfold).

That means that if the current continues to flow without any
reduction, into the capacitor, without change, the charge of the
capacitor would be

300 * 2E-6 = 0.0006 coulomb.

A 10 uF capacitor charged to 0.0006 coulomb would have a voltage
across its leads of 0.0006/10e-6 = 60 volts. That's quite survivable,
especially in an RCD snubber.

thats not gonna work so well. that R carries the full load current, so

will need an impressive peak pulse power rating.


an RCD snubber, with a monstrous D, is probably what you want.

That makes perfect sense. I have the snubber diagrams right in front
of me, RCD has a diode in parallel with resistor. Right?

What kind of diode should serve as that D. It should survive currents
up to, say, 300-400 A and voltages of a few hundreds of volts.

you can read more languages than I can.

they may not mention the most important part - the wear-out
characteristics of MOVs. basically they can clamp a finite amount of
Joules, then they go short circuit. Then (often within a few hundred
microseconds) they go *BANG*

applications using MOVs to clamp repetitive surges are generally best
suited for breaking MOVs.

find a manufacturer (eg Seimens, or whatever they are called now) that
actually makes the things, and read their app notes.

Also, if I read you correctly, a MOV is not necessary, is that right?

I also measured inductance of my reactor. It seems to be about 1 mH.

I used this technique

http://et.nmsu.edu/~etti/fall96/electronics/induct/induct.html

and used both high and low Ohms on my decade resistor. High ohms were
useless, I did it more for curiosity.

i

 

[Ignoramus3498]

I just had a look at the Epcos application notes. They specifically
recommend using their varistors for clamping spikes due to switching off an
inductive load. Take a look: (sorry for the long URL, I'll break it across
a few lines, since I don't know if I'll mess things up otherwise.)

http://www.epcos.com/web/generator/Web/Sections/ProductCatalog/
NonlinearResistors/Varistors/PDF/
PDF__Applications,property=Data__en.pdf;/Applications.pdf

Certainly, if a varistor is used to divert a lightning strike then at best
only one operation can be expected before the varistor becomes useless, but
I do not see any mention of a slower wear-out mechanism in the Epcos
application notes, except the warning that the varistor must not be allowed
to reach an excessive temperature.

I believe that the OP intends to put a large motor-run type capacitor in
parallel with the varistor so that it will only be in unusual circumstances
that the varistor will carry the full welder current.

You are 100% right. I will use a relatively large capacitor, which
should be no problem with an RCD type snubber (if I am not
mistaken). It would not make starting conditions severe and would not
make starting welding arcs make a sound of a .22. Then the varistor
would be for catching emergency situations, etc, and not strictly
necessary.

i

 

[Terry Given]

I thought that it could simply flow into the capacitor through the
resistor.

There are two kinds of events, really, a transient switch off during
bridge switching, and a complete shutdown (turn off) of the
bridge. Note that most complete shutdowns would happen while NOT under
power, for obvious reasons. I cannot be switching AC switch off while
welding.

I suppose that shutdowns due to arc extinguishing are rare and softer.

For transient switch offs...

Suppose that, for worst case example, 300 A is flowing into the
circuit. Then a switching even occurs and the circuit opens for 2
microseconds (a very long time for IR22141SS that can likely be
reduced severalfold).

That means that if the current continues to flow without any
reduction, into the capacitor, without change, the charge of the
capacitor would be

300 * 2E-6 = 0.0006 coulomb.

A 10 uF capacitor charged to 0.0006 coulomb would have a voltage
across its leads of 0.0006/10e-6 = 60 volts. That's quite survivable,
especially in an RCD snubber.





That makes perfect sense. I have the snubber diagrams right in front
of me, RCD has a diode in parallel with resistor. Right?

What kind of diode should serve as that D. It should survive currents
up to, say, 300-400 A and voltages of a few hundreds of volts.

yep^2



Also, if I read you correctly, a MOV is not necessary, is that right?

not strictly speaking, no.

I also measured inductance of my reactor. It seems to be about 1 mH.

I used this technique

http://et.nmsu.edu/~etti/fall96/electronics/induct/induct.html

and used both high and low Ohms on my decade resistor. High ohms were
useless, I did it more for curiosity.

not necessarily much use for your inductor, as it only pumps very low
current thru the inductor. If the core material permeability varies
strongly with H, then this will give an incorrect reading - invariably
an over-estimate. google groups "splat test" for a trick that doesnt
have that problem.

i

Cheers
Terry

 

[Ignoramus24693]

yep^2

Got it. Are any particular ones more suitable? I looked at digikey and
found too many dazzling choices.

not strictly speaking, no.

I see. It's more for protection when "shit happens", like electrical
ground. It does not normally see any use.


CORRECTION, it is about 2-5 mH, see my another post Measured
INDUCTANCE of my reactor. I used some wrong data when I made my
previous post.

not necessarily much use for your inductor, as it only pumps very low
current thru the inductor. If the core material permeability varies
strongly with H, then this will give an incorrect reading - invariably
an over-estimate. google groups "splat test" for a trick that doesnt
have that problem.

Thanks. I looked up the splat test. I am not quite sure how would I
measure current with my oscilloscopes. Tek 2465 and 475. I kind of
like this idea though, that's definitely the way to go.

I did some calculations of capacitance of snubbers for given
parameters. The I used formula from Fuji

Cesp = L*Io^2 / (Vcep - Ed)^2

L inductance
Io normal current
Vcep snubber peak voltage
Ed DC supply voltage

The results are mind boggling. My calculations suggest that I need
about 50,000 uF (!) for 250V caps. Either my math is wrong somewhere,
or else I am missing something very big.

The above calculation applies to complete turnoff of the H
bridge. Which should almost never happen under load. I assume that
conditions of arc extinguishing are much milder.

Complete turnoff at high current, I think, should happen rarely. Maybe
never. In those instances, a snubber may provide appropriate
protection.

In regular operation, turn off time would be very short, perhaps 0.5
uS or so. To absorb 0.5 uS's worth of high current, I need a lot less
capacitance. (microfarads)

So, I can say that I am quite confused.



i

 

[Terry Given]

Got it. Are any particular ones more suitable? I looked at digikey and
found too many dazzling choices.

pass. its gotta carry the right amount of current, and have an
appropriate voltage rating - at least peak snubber cap voltage.

I see. It's more for protection when "shit happens", like electrical
ground. It does not normally see any use.





CORRECTION, it is about 2-5 mH, see my another post Measured
INDUCTANCE of my reactor. I used some wrong data when I made my
previous post.




Thanks. I looked up the splat test. I am not quite sure how would I
measure current with my oscilloscopes. Tek 2465 and 475. I kind of
like this idea though, that's definitely the way to go.

use a pulse generator (555 timer) and a decent FET, a current shunt and
a low-voltage supply with shitloads of capacitance - ensure Ecap >>
Einductor at measured current.

I did some calculations of capacitance of snubbers for given
parameters. The I used formula from Fuji

Cesp = L*Io^2 / (Vcep - Ed)^2

L inductance
Io normal current
Vcep snubber peak voltage
Ed DC supply voltage

The results are mind boggling. My calculations suggest that I need
about 50,000 uF (!) for 250V caps. Either my math is wrong somewhere,
or else I am missing something very big.

the latter.

inductor energy Eind = 0.5*L*I^2
initial cap energy Ecap_init = 0.5*C*Vinit^2
final cap energy Ecap_final = 0.5*C*Vfinal^2

Efinal = Einit + Eind

0.5*C*Vfinal^2 = 0.5*C*Vinit^2 + 0.5*L*I^2

C*Vfinal^2 = C*Vinit^2 + L*I^2

C*[Vfinal^2 - Vinit^2] = L*I^2

C = L*I^2/[Vfinal^2 - Vinit^2]

so yeah, there's a bit of a mistake in your numbers, because the
equation is wrong.

The above calculation applies to complete turnoff of the H
bridge. Which should almost never happen under load. I assume that
conditions of arc extinguishing are much milder.

Complete turnoff at high current, I think, should happen rarely. Maybe
never. In those instances, a snubber may provide appropriate
protection.

In regular operation, turn off time would be very short, perhaps 0.5
uS or so. To absorb 0.5 uS's worth of high current, I need a lot less
capacitance. (microfarads)

yeah Q = CV = It, can assume I is constant. So 200A for 0.5us = 100uC =
100V across 1uF. You've also got stacks of volts to play with....

So, I can say that I am quite confused.

the typo in the fuji app note doesnt really help, does it

 

[Ignoramus5275]

pass. its gotta carry the right amount of current, and have an
appropriate voltage rating - at least peak snubber cap voltage.

Got it.

I think that I can buy 15 Panasonic ERZV20D271 varistors, they would
provide max voltage of 225 VDC, clamping voltage of 455 VDC, and 135
joules each. I would parallel them to get appropriate joule rating.

The hope is that these varistors would never actually get a chance to
conduct current. The H bridge should not be stopped at full load
unless someone moves the current switch handle during welding, which
is unlikely. I hope that interruptions of the welding arc would happen
more slowly and thus the highest peak voltages/currents would not
occur. I may be mistaken.

For a diode, I could use a 300U60AMA diode, 300 A, 600 V rated. It's
kind of pricey, maybe I can parallel several cheaper diodes? Such as
3-4 Fairchild diodes RURU10060?

use a pulse generator (555 timer) and a decent FET, a current shunt and
a low-voltage supply with shitloads of capacitance - ensure Ecap >>
Einductor at measured current.

I see. I may do that. Good idea.

I did some calculations of capacitance of snubbers for given
parameters. The I used formula from Fuji

Cesp = L*Io^2 / (Vcep - Ed)^2

L inductance
Io normal current
Vcep snubber peak voltage
Ed DC supply voltage

The results are mind boggling. My calculations suggest that I need
about 50,000 uF (!) for 250V caps. Either my math is wrong somewhere,
or else I am missing something very big.

the latter.

inductor energy Eind = 0.5*L*I^2
initial cap energy Ecap_init = 0.5*C*Vinit^2
final cap energy Ecap_final = 0.5*C*Vfinal^2

Efinal = Einit + Eind

0.5*C*Vfinal^2 = 0.5*C*Vinit^2 + 0.5*L*I^2

C*Vfinal^2 = C*Vinit^2 + L*I^2

C*[Vfinal^2 - Vinit^2] = L*I^2

C = L*I^2/[Vfinal^2 - Vinit^2]

so yeah, there's a bit of a mistake in your numbers, because the
equation is wrong.

Thanks, you are right. Still, the required capacitance is tremendous.

yeah Q = CV = It, can assume I is constant. So 200A for 0.5us = 100uC =
100V across 1uF. You've also got stacks of volts to play with....

Very nice to know, makes life easier.

Thanks Terry. Your thoughts about component selection for the snubber
(see my text in the middle of the message)will be appreciated...

i

 

[Ignoramus5275]

I was kinda shocked by the price of diodes for 300+ amps and thought
about something.

In my spare parts pile, I have a big heavy TT170N silicon controlled rectifier:

http://igor.chudov.com/projects/Homemade-TIG-DC-to-AC-Inverter/Components/TT170N.pdf

It is rated for 1,200V, 350 amps. Can I just use it as a diode in an
RCD snubber if I supply proper voltage to its gate continuously to
turn it on?

i