Diode Protection for Mosfet from Back EMF

Thread Starter

Sitara

Joined May 2, 2014
57
Hi,
I'm investigating a circuit based on diodes, for protecting a mosfet switch from back emf due to transformer leakage inductance (in lieu of a snubber).
The basic idea is to (a) block off the mosfet's body diode, because this is too slow even though it is correctly oriented to conduct the back emf (b) to provide an additional, reverse biased, diode, parallel to the mosfet, to conduct the back emf. I'm using the 16A, 600V 45nS MUR1660CT diode for both (a) & (b). I have uploaded my LTSpice schematic for this circuit as well as spice file for the diode. Below is a screendump of the LTSpice simulation of this circuit:

upload_2017-3-11_21-9-37.gif

My questions are: (1) Why am I seeing a drain-source voltage spike of ~600V in the opposite direction to what the back-emf polarity should be ?
(2) why are the external diodes not shielding the mosfet from the back emf spike? Many thanks in advance.
 

Attachments

ian field

Joined Oct 27, 2012
6,536
Hi,
I'm investigating a circuit based on diodes, for protecting a mosfet switch from back emf due to transformer leakage inductance (in lieu of a snubber).
The basic idea is to (a) block off the mosfet's body diode, because this is too slow even though it is correctly oriented to conduct the back emf (b) to provide an additional, reverse biased, diode, parallel to the mosfet, to conduct the back emf. I'm using the 16A, 600V 45nS MUR1660CT diode for both (a) & (b). I have uploaded my LTSpice schematic for this circuit as well as spice file for the diode. Below is a screendump of the LTSpice simulation of this circuit:

View attachment 122295

My questions are: (1) Why am I seeing a drain-source voltage spike of ~600V in the opposite direction to what the back-emf polarity should be ?
(2) why are the external diodes not shielding the mosfet from the back emf spike? Many thanks in advance.
Traditionally; the back emf diode shunts the inductor - current flows when the diode clamps, so it builds up lines of flux round the core again. With a relay; that slows down unlatching. Mains operated flyback converters usually have a snubber circuit - the snubber diode dumps the back emf into a capacitor. the capacitor in turn has a parallel resistor to bleed the charges away each time they arrive.
 

AlbertHall

Joined Jun 4, 2014
12,338
The back emf at the drain will take the drain positive of the positive supply not negative so your diode (and the diode in the MOSFET) will do nothing to stop the back emf spike. The diode, or the snubber, should be across the load, cathode to positive supply.
 

crutschow

Joined Mar 14, 2008
34,201
.........
My questions are: (1) Why am I seeing a drain-source voltage spike of ~600V in the opposite direction to what the back-emf polarity should be ?
(2) why are the external diodes not shielding the mosfet from the back emf spike? Many thanks in advance.
The answer to both questions is because you have not correctly determined what the "back-emf" polarity will be.

When the inductor is conducting the top is positive with respect to the bottom.
When you turn the inductor off it tries to keep the current flowing and the bottom of the inductor becomes positive with respect to the top (back-emf), giving a large positive spike at the MOSFET drain.

Thus for protection, the diode has to go from the transistor drain (anode) to the inductor power supply (cathode).
 

Thread Starter

Sitara

Joined May 2, 2014
57
The back emf at the drain will take the drain positive of the positive supply not negative so your diode (and the diode in the MOSFET) will do nothing to stop the back emf spike. The diode, or the snubber, should be across the load, cathode to positive supply.
Bingo! and thank you Albert! Here is the proof of your assertion:

upload_2017-3-11_22-8-15.gif
 

Thread Starter

Sitara

Joined May 2, 2014
57
The answer to both questions is because you have not correctly determined what the "back-emf" polarity will be.

When the inductor is conducting the top is positive with respect to the bottom.
When you turn the inductor off it tries to keep the current flowing and the bottom of the inductor becomes positive with respect to the top (back-emf), giving a large positive spike at the MOSFET drain.

Thus for protection, the diode has to go from the transistor drain (anode) to the inductor power supply (cathode).
With hind-sight (and the proof pasted below), its easy to see the logic in what you say. I was misled by my (perhaps incorrect) concept of how a mosfet body diode conducts away back-emf in say a Full Bridge switcher. So I thought to use a faster external diode as an ersatz body diode for the same purpose.
 

crutschow

Joined Mar 14, 2008
34,201
.................
I was misled by my (perhaps incorrect) concept of how a mosfet body diode conducts away back-emf in say a Full Bridge switcher. So I thought to use a faster external diode as an ersatz body diode for the same purpose.
If you look at the current path for the inductive load current in a bridge switch it's through the MOSFET body diodes of the MOSFETs opposite from those being turned off.
 

MrAl

Joined Jun 17, 2014
11,342
Hi,
I'm investigating a circuit based on diodes, for protecting a mosfet switch from back emf due to transformer leakage inductance (in lieu of a snubber).
The basic idea is to (a) block off the mosfet's body diode, because this is too slow even though it is correctly oriented to conduct the back emf (b) to provide an additional, reverse biased, diode, parallel to the mosfet, to conduct the back emf. I'm using the 16A, 600V 45nS MUR1660CT diode for both (a) & (b). I have uploaded my LTSpice schematic for this circuit as well as spice file for the diode. Below is a screendump of the LTSpice simulation of this circuit:

View attachment 122295

My questions are: (1) Why am I seeing a drain-source voltage spike of ~600V in the opposite direction to what the back-emf polarity should be ?
(2) why are the external diodes not shielding the mosfet from the back emf spike? Many thanks in advance.
Hi,

That circuit is different than a converter circuit with at least a half bridge as output rather than just one MOSFET. With one mosfet only there are other considerations for a lone inductor that are not the same as with a converter with two sets of upper and lower mosfets (H bridge).

In the H bridge mosfet circuit, when the lower of one side turns off the upper of the other side probably turns off (there are other ways but this is the most reasonable). When that happens, the reverse diodes act as a full wave bridge rectifier that rectifies the inductor current and feeds in back into the input filter caps. Thus only some of the back emf inductor energy is lost and most is recycled.

In the lone inductor circuit, to achieve the same result as with the dual mosfet circuit you'd have to find a way to get the back emf energy back into the power source. One way is to use a second winding with a bridge rectifier that pumps the current back into the main supply caps. When the back emf occurs, the energy is transferred to the second winding and that allows that floating current source to be rectified and pumps the current back into the input caps. There is also the possibility of using a center tapped inductor with some rectifiers.

I have a feeling you are not that interested in the lone inductor circuit though. Be aware though that if you are, the back emf energy will be lost if you shunt the inductor with anything that is there ti eat up that back emf energy.

So really the lone inductor circuit is not a good example of what happens in a full H bridge converter and it sounds like you were looking into the H bridge circuit.

The main reason for using an external diode with a mosfet it the heat in the diode is contained inside the mosfet package, which adds to the thermal problems. With an external diode, some of that heat is moved outside the case and so the transistor itself stays cooler.

The SNUBBER is a different animal altogether. That goes right across the transistor because it is there mainly to protect the transistor. The leads should be as short as possible. In a circuit like an H bridge with four transistors, one snubber for each transistor is required.
 
Last edited:

ian field

Joined Oct 27, 2012
6,536
Hi,

That circuit is different than a converter circuit with at least a half bridge as output rather than just one MOSFET. With one mosfet only there are other considerations for a lone inductor that are not the same as with a converter with two sets of upper and lower mosfets (H bridge).

In the H bridge mosfet circuit, when the lower of one side turns off the upper of the other side probably turns off (there are other ways but this is the most reasonable). When that happens, the reverse diodes act as a full wave bridge rectifier that rectifies the inductor current and feeds in back into the input filter caps. Thus only some of the back emf inductor energy is lost and most is recycled.

In the lone inductor circuit, to achieve the same result as with the dual mosfet circuit you'd have to find a way to get the back emf energy back into the power source. One way is to use a second winding with a bridge rectifier that pumps the current back into the main supply caps. When the back emf occurs, the energy is transferred to the second winding and that allows that floating current source to be rectified and pumps the current back into the input caps. There is also the possibility of using a center tapped inductor with some rectifiers.

I have a feeling you are not that interested in the lone inductor circuit though. Be aware though that if you are, the back emf energy will be lost if you shunt the inductor with anything that is there ti eat up that back emf energy.

So really the lone inductor circuit is not a good example of what happens in a full H bridge converter and it sounds like you were looking into the H bridge circuit.

The main reason for using an external diode with a mosfet it the heat in the diode is contained inside the mosfet package, which adds to the thermal problems. With an external diode, some of that heat is moved outside the case and so the transistor itself stays cooler.

The SNUBBER is a different animal altogether. That goes right across the transistor because it is there mainly to protect the transistor. The leads should be as short as possible. In a circuit like an H bridge with four transistors, one snubber for each transistor is required.
In a SMPSU - the snubber consisting of a diode, capacitor and resistor invariably goes across the primary. There may also be a simple C/R snubber across the transistor - but not always.
 

Thread Starter

Sitara

Joined May 2, 2014
57
In a SMPSU - the snubber consisting of a diode, capacitor and resistor invariably goes across the primary. There may also be a simple C/R snubber across the transistor - but not always.
I am presuming that, by different animal, Mr Al meant the simple RC snubber. It is different in the sense that it is 'tuned' - the capacitive reactance must cancel the (leakage) inductive reactance at the 'ringing' frequency as described, for example in this useful app note:
https://www.maximintegrated.com/en/app-notes/index.mvp/id/3835

But however nice, neat & tidy the theory looks, my practical experiments with using these snubbers with practical push-pull smps, frequently ended with a bright blue flash accompanied by a loud bang and phenolic Magic Smoke, reminiscent of a burnt Bakelite saucepan handle (when saucepans had such handles). The cause of this was, more likely than not, a lightly loaded secondary and not so much leakage inductance. With hand-wound bobbins for ferrite core transformers, increasing primary inductance to lower the magnetising current to safe levels (where the mosfets don't blow up), is impractical. There simply isn't enough space for the windings plus the interlayer insulation on the bobbin.

And this is with over-current protection. The ideal form of protection is to continuously sense the secondary load and to shut down the gate drives when that load falls below a given threshold.
 

MrAl

Joined Jun 17, 2014
11,342
In a SMPSU - the snubber consisting of a diode, capacitor and resistor invariably goes across the primary. There may also be a simple C/R snubber across the transistor - but not always.
Hi,

Ok then you have to explain why there are thousands of converters out there with snubbers right across the transistor, and that is the only type in there and it has a diode, cap, resistor. and short leads.
While your at it, explain why the boards for the drivers that are mounted right on the heat sink also have the snubber right on the same board and have short traces, and four transistors means four snubbers, and six transistors (three phase) means six snubbers.
These are units that were shipped to different military governments as well as Sandia Labs and other companies. Some were regular converters and some were made for line tied applications.

Or perhaps i can explain it for you :)
The reason is that there is inductance in the leads going to the output transformer not just in the transformer primary itself, so it makes sense to install a snubber on each transistor but get it as close to the transistor as possible.
This wasnt my idea per se, but was the idea of engineers who where much older than me at the time and where in the game long before i was.

It would still be interesting however to hear where you got your idea from, the idea of just putting a snubber across the primary of the output transformer. Where and why that idea originated.
 

MrAl

Joined Jun 17, 2014
11,342
I am presuming that, by different animal, Mr Al meant the simple RC snubber. It is different in the sense that it is 'tuned' - the capacitive reactance must cancel the (leakage) inductive reactance at the 'ringing' frequency as described, for example in this useful app note:
https://www.maximintegrated.com/en/app-notes/index.mvp/id/3835

But however nice, neat & tidy the theory looks, my practical experiments with using these snubbers with practical push-pull smps, frequently ended with a bright blue flash accompanied by a loud bang and phenolic Magic Smoke, reminiscent of a burnt Bakelite saucepan handle (when saucepans had such handles). The cause of this was, more likely than not, a lightly loaded secondary and not so much leakage inductance. With hand-wound bobbins for ferrite core transformers, increasing primary inductance to lower the magnetising current to safe levels (where the mosfets don't blow up), is impractical. There simply isn't enough space for the windings plus the interlayer insulation on the bobbin.

And this is with over-current protection. The ideal form of protection is to continuously sense the secondary load and to shut down the gate drives when that load falls below a given threshold.
Hi,

Actually the snubbers i was talking about each had a diode, resistor, and capacitor. They were mounted right on the same board that mounted to the heat sink of each transistor, and each transistor had it's own heat sink in many cases, or two shared one heat sink so there were two drivers and two snubbers on the same PC board that bolted to the heat sink.

The reason for this is in my previous post, but the simple explanation is that the inductance that can cause a high voltage spike is not all in the output transformer primary. Any lead has some inductance and a good way to make sure you've covered all of it is to place the snubber right across the transistor, not a mile away from it :)
We just have to keep in mind that the reason for the snubber in the first place is to protect the transistor, not to protect the transformer primary.

In some cheap designs (low power) you'll see just one snubber but in high quality converters you'll see four, if they know what they are doing that is :)
 

ian field

Joined Oct 27, 2012
6,536
Hi,

Ok then you have to explain why there are thousands of converters out there with snubbers right across the transistor, and that is the only type in there and it has a diode, cap, resistor. and short leads.
.
Some of them do - I've even seen one or two with D/C/R snubbers across the transistor, but simple C/R snubbers are more likely if there's any at all.

I can't actually ever remember seeing a flyback switcher with no snubber across the primary - once or twice a simple C/R, but the D/C/R snubber is far more usual.

You can explain all you like - I'm telling you what I see out there "on this planet".
 

MrAl

Joined Jun 17, 2014
11,342
Some of them do - I've even seen one or two with D/C/R snubbers across the transistor, but simple C/R snubbers are more likely if there's any at all.

I can't actually ever remember seeing a flyback switcher with no snubber across the primary - once or twice a simple C/R, but the D/C/R snubber is far more usual.

You can explain all you like - I'm telling you what I see out there "on this planet".
Hi,

Well of course you are free to reason it out on your own in your own way, and it is not as if i dont believe you, but my reasoning covers all cases not just some. I reasoned it as the main component that will fail if there is a spike, from any location, included but not limited to the primary. So by that reasoning, a snubber right across the transistor will protect it no matter where that spike energy comes from so it must work if any snubber is going to work in the first place.
To the contrary but also within the same frame of reasoning, if the spike does not come from the primary then the energy could get to the transistor and cause damage.

Be aware that i also report what i have seen and had been taught by some of the best experts in the industry and even some in cutting edge science. But also be aware that most of these converters where high power...most over 1000 watts. The lower power devices may not be as bad because the kickback is not as severe with lower current.

It's not that i am saying that you have never seen a converter without snubber across the transistors, i am just saying that is it better practice to put them across the transistors. The electrical theory behind where we find inductances in a converter actually would dictate this practice, although i have two converters of my own that i would bet do not have individual snubbers for cost reasons and also because they are lower power.

The theory behind the reasoning is that there is not only inductance in the primary, but each wire leading to each primary and each wire in series with each transistor contains inductance, any of which can cause a kickback. If we were to draw a circuit we would see a somewhat larger primary inductor in series with at least two smaller inductors and those two smaller inductors are what connects the transistors to the primary, because they represent the real wire not a fictitious short from one primary lead to one transistor set and another short from the other primary lead to some other transistor set.
There's also inductance in other leads to the transistor also, and sometimes there is even a small physical inductor placed in series with an upper transistor to help mitigate transistor shoot through conduction.
Now if the leads from primary to transistor sets is short and current low, then there is less risk of significant back emf from an actual wire rather than the primary. So there could be cases where we dont see too much of this damaging action. In smaller converters this might be the case, if they can survive for hours of operation at a time. Some of the stuff we did when i worked in the industry also had to run 24/7 as computer power supplies and various other full time AC power sources.

There is also the power point view, where we could see more wasted power with a snubber across the primary. It depends on the design though.
 

crutschow

Joined Mar 14, 2008
34,201
Of course a flyback transformer can't have just a diode across the primary since it will clamp the flyback output voltage to only 0.7V times the transformer turns ratio.
Thus you want a snubber circuit that keeps the flyback voltage safely below the transistors breakdown voltage but not so low that it reduces the output voltage.
That could be some form of RC snubber or a zener connected back-to-back with a diode.
 

ian field

Joined Oct 27, 2012
6,536
Hi,

Well of course you are free to reason it out on your own in your own way, and it is not as if i dont believe you, but my reasoning covers all cases not just some. I reasoned it as the main component that will fail if there is a spike, from any location, included but not limited to the primary. So by that reasoning, a snubber right across the transistor will protect it no matter where that spike energy comes from so it must work if any
.
Make your mind up - you mentioned tuning in an earlier post. A tuning capacitor wouldn't have a "Q spoiler" resistor anyway.

You seem to be considering push pull circuits when the TS clearly stated right from the start that it was a flyback control issue.

It can be entertaining when self righteous college boys try to tell the old hands how it is.............
 
Top