Driving high inductance Solenoid using H-Bridge

[koniho]

I've been building an IGBT based H-Bridge using irs2108 drivers. I've varied the model of IGBT around in hopes of over coming a problem that I've been having of IGBTs exploding. With much frustration I think I've narrowed the issue down to the Voltage build-up when switching the bridge into the free-wheeling "off-state". I'm planning on running the system at 300V but so far, I've been running into issues around 180V. Power supplies all look clean and the driving logic has been verified. The HS and ~LS signals of the irs2108 are driven by the same signal so that the shoot-through condition should not occur unless driver is damaged.

What I'm seeing is that the IGBT's go into a latch up mode or similar and hit a short circuit fault until they burn up in a puff of magic smoke.

My next inclination was to place a snubber network in parallel with the load but as far as I can see on the scope, there's little improvement in the turn-off voltage transients. I've read many application notes now on snubber networks across the IGBTs but they seem to be mainly concerned with reducing the switching losses. In my application, I'm only switching the solenoid 'on' with a short 5-50ms pulse and not doing any PWM.

The last thing I've tried was a Varistor in parallel with the load and I did see improvements in suppressing voltage transients but once I upped the voltage to 180V, I again experienced an explosive bridge.

If anyone has experience with driving a highly inductive load with an h-bridge using IGBT's I'd love to hear your input on my problem. Thanks!

Attached is the H-Bridge schematic:

https://lh3.googleusercontent.com/_...kN024/s800/10seg_2dir_actuator_oneChannel.png

 

[crutschow]

Deleted incorrect post.

 

[BrownOut]

The fact that the varistor improved things may point to the problem. If the varistor helps protect the transistors, then the problem would seem to be the inductive kickback from the motors while switching off, a common problem. You might find transistors that are specially made for this application ( haven't looked up yours ) Also, snubber networks can do more than inprove switching looses. They can also help suppress the kind of inductive spikes you might be experiencing. You can also try transorbe in place of the varistors, as I think they might act faster. If you safely can run the system below 180V, then you can use a scope to probe the motors and see if the inductive spikes are there, how fast the rise, how long they last, etc. I don't actually have alot of experience driving motors, but I believe suppressing inductive spikes will be your top priority.

 

[crutschow]

If you can tolerate slowing down the release time of the solenoid slightly, you could add diodes across each transistor from collector to emitter. That will keep all transients across the transistors equal to no more than one diode-drop above the power supply voltage.

The diodes should be rated at 500V or more. The current rating can be less than the solenoid current, depending upon the operate frequency of the solenoid. For NPN transistors connect the diode's cathode to the collector and anode to emitter.

 

[MrAl]

I've been building an IGBT based H-Bridge using irs2108 drivers. I've varied the model of IGBT around in hopes of over coming a problem that I've been having of IGBTs exploding. With much frustration I think I've narrowed the issue down to the Voltage build-up when switching the bridge into the free-wheeling "off-state". I'm planning on running the system at 300V but so far, I've been running into issues around 180V. Power supplies all look clean and the driving logic has been verified. The HS and ~LS signals of the irs2108 are driven by the same signal so that the shoot-through condition should not occur unless driver is damaged.

What I'm seeing is that the IGBT's go into a latch up mode or similar and hit a short circuit fault until they burn up in a puff of magic smoke.

My next inclination was to place a snubber network in parallel with the load but as far as I can see on the scope, there's little improvement in the turn-off voltage transients. I've read many application notes now on snubber networks across the IGBTs but they seem to be mainly concerned with reducing the switching losses. In my application, I'm only switching the solenoid 'on' with a short 5-50ms pulse and not doing any PWM.

The last thing I've tried was a Varistor in parallel with the load and I did see improvements in suppressing voltage transients but once I upped the voltage to 180V, I again experienced an explosive bridge.

If anyone has experience with driving a highly inductive load with an h-bridge using IGBT's I'd love to hear your input on my problem. Thanks!

Attached is the H-Bridge schematic:

https://lh3.googleusercontent.com/_...kN024/s800/10seg_2dir_actuator_oneChannel.png

Hi there,


When one of the transistors turns off (such as one of the uppers) the inductive kick makes the voltage across the inductance rise quickly. In a perfect world where we have no wire inductance the diodes across the transistors act like a full wave bridge rectifier and push the inductive energy back into the power supply. This sudden burst of energy could cause the power supply itself to rise suddenly to levels that were not supposed to happen so the key here is to make sure the power supply has adequate bypassing to keep the spike low. This is what would happen if we were *lucky*, but really it may only happen to a lesser extent which means not only do we have to make sure the power supply source impedance is low, we also have to incorporate another means to make up for the wire inductance.
This is where the snubber comes in. The snubber is built with short wires so that it can connect across the transistor as close as possible. It is there to literally eat up that extra energy from the inductive kickback, before the diodes get to conduct some of that energy back to the power supply. Of course this usually means one snubber per transistor which means one high speed diode, one capacitor, and one power resistor per transistor.
Another way to deal with this is to use the transistor itself as a snubber through a controlled turn off. This is a bit more difficult to accomplish however and something that is usually used with MOSFETs so im not sure if it works with IGBT's, but i thought i would mention it in case you want to look for more information.

What is it that is being driven there, is it just a solenoid?

 

[ronv]

Just because you are driving the oposite sides of the bridge with the same signal doesn't mean you don't have a shoot thru problem due to the difference in turn on and turn off times of the IBGTs. I would scope the current during reversals. If you don't have reversals then it almost has to be the inductive kick. A little more detail on how the system works might help. (IGBT type swithing times etc.)

 

[koniho]

If you can tolerate slowing down the release time of the solenoid slightly, you could add diodes across each transistor from collector to emitter. That will keep all transients across the transistors equal to no more than one diode-drop above the power supply voltage.

The diodes should be rated at 500V or more. The current rating can be less than the solenoid current, depending upon the operate frequency of the solenoid. For NPN transistors connect the diode's cathode to the collector and anode to emitter.

The IGBT's that I'm using have an integrated free-wheeling diode. I've been concerned that it may not be enough though and maybe failing in some conditions. Are failing free-wheeling diodes a common failure mode? I'm guessing not due to the relatively small amounts of energy that's due to the inductive kick, but at this point, I can't be sure of anything.

 

[koniho]

Hi there,


When one of the transistors turns off (such as one of the uppers) the inductive kick makes the voltage across the inductance rise quickly. In a perfect world where we have no wire inductance the diodes across the transistors act like a full wave bridge rectifier and push the inductive energy back into the power supply. This sudden burst of energy could cause the power supply itself to rise suddenly to levels that were not supposed to happen so the key here is to make sure the power supply has adequate bypassing to keep the spike low. This is what would happen if we were *lucky*, but really it may only happen to a lesser extent which means not only do we have to make sure the power supply source impedance is low, we also have to incorporate another means to make up for the wire inductance.
This is where the snubber comes in. The snubber is built with short wires so that it can connect across the transistor as close as possible. It is there to literally eat up that extra energy from the inductive kickback, before the diodes get to conduct some of that energy back to the power supply. Of course this usually means one snubber per transistor which means one high speed diode, one capacitor, and one power resistor per transistor.
Another way to deal with this is to use the transistor itself as a snubber through a controlled turn off. This is a bit more difficult to accomplish however and something that is usually used with MOSFETs so im not sure if it works with IGBT's, but i thought i would mention it in case you want to look for more information.

What is it that is being driven there, is it just a solenoid?

I am just driving a solenoid. Actually 10 of them to be exact. At this stage, I'm just driving them with a 10 ms pulse, in one direction or the other and freewheeling on the low side during their nominal state. The duty cycle is low, the fastest I'd be firing them would be once every 30 seconds or so. In the actual application, they might fire once every hour or so.

I'm currently putting together a number of snubbers to connect to each of the IGBT's on the bridge and will report back results.

Thanks all for all the input. Can't wait to solve this!

 

[koniho]

Just because you are driving the oposite sides of the bridge with the same signal doesn't mean you don't have a shoot thru problem due to the difference in turn on and turn off times of the IBGTs. I would scope the current during reversals. If you don't have reversals then it almost has to be the inductive kick. A little more detail on how the system works might help. (IGBT type swithing times etc.)

In this application, there are no instant reversals. The time between reversals is essentially infinite wrt to the 'on' times.

A typical IGBT that I've tried is:
https://www.electro-tech-online.com/custompdfs/2011/03/irgb4062dpbf.pdf

I've tried a number, all with similar specifications to the one above.

 

[MrAl]

I am just driving a solenoid. Actually 10 of them to be exact. At this stage, I'm just driving them with a 10 ms pulse, in one direction or the other and freewheeling on the low side during their nominal state. The duty cycle is low, the fastest I'd be firing them would be once every 30 seconds or so. In the actual application, they might fire once every hour or so.

I'm currently putting together a number of snubbers to connect to each of the IGBT's on the bridge and will report back results.

Thanks all for all the input. Can't wait to solve this!

Hello again,

Oh ok sounds good. If the snubbers dont appear to work properly it could be the power resistor value is too large or the connecting wires are too long. Short wires and an appropriate size resistor should do it. Good luck with it.

 

[koniho]

I've been scoping around and I've noticed that upon de-energizing the solenoid, I'm getting a current reversal, in the solenoid itself, and definitely on the power supply line as well.

The current reversal on the power line is worrisome. If my thinking is correct, the inductive kick is enough to push the V- above V+. Am I correct in thinking this is the only way the current reversal is possible? If the current was being measured inline with the solenoid, I'd expect to see current reversal, but definitely not from the power supply.

Attached are plots of the Voltages at the solenoid terminals (wrt to V-) and the power supply current (disregard the offset). These were recorded with a large High J 270V Varistor across the terminals of the Solenoid. Would it be better to connect the Varistors from the solenoid terminals to V-? I've got some transorbs now and will see if they make a difference.

It definitely looks like an inductive kickback issue.

 

[MrAl]

Hi,

The current reversal in the power supply is typical of the energy of the solenoid moving back into the power supply as per my previous explanation of how the diodes redirect the current after a transistor turns off. That's going to be normal. The thing to watch out for is that the impedance near the top of the bridge is low enough to handle that surge without having the voltage rise too much. The snubbers are there to protect the transistor mostly during the time it takes for the line inductance to settle, so they may not eat up all that surge but only part of it.

If there is significant current reversal in the solenoid it would seem that the opposite pair of transistors are turning on. Is there any way that would normally happen with the drive circuit?

 

[crutschow]

I've been scoping around and I've noticed that upon de-energizing the solenoid, I'm getting a current reversal, in the solenoid itself, and definitely on the power supply line as well.

The current reversal on the power line is worrisome. If my thinking is correct, the inductive kick is enough to push the V- above V+. Am I correct in thinking this is the only way the current reversal is possible? If the current was being measured inline with the solenoid, I'd expect to see current reversal, but definitely not from the power supply.

There is no current reversal in the solenoid. The inductance just attempts to keep the current moving in the same direction. But this has the effect of reversing the voltage across the solenoid since it now goes from being a current sink (load) to being a current source.

Thus, if you have the solenoid energized by it being connected between the plus and minus supplies and then turn off the transistors, the solenoid will keep the current flowing for a short period by pulling current from the minus supply and pushing it into the positive supply through the transistor's free-wheeling diodes. The power supplies should have enough capacitance to absorb these current spikes. If not then you will get large positive voltage spikes on the positive power supply and negative spikes on the negative supply.

 

[tcmtech]

For the more obvious questions how many amps and at what actual working voltage will these be running normally?

Also what size of capacitor do you have on your rails of the power supply?

 

[koniho]

For the more obvious questions how many amps and at what actual working voltage will these be running normally?

Also what size of capacitor do you have on your rails of the power supply?

I'm aiming to drive the solenoids, which have a nominal resistance of 20 ohms, and about 20 mH inductance with 300V. At 300V, each solenoid would be driven with 15A for a maximum of 10 ms. Currently there is 36,000 uF across the rails.

 

[koniho]

I'm currently free-wheeling by keeping the low-side of the bridge on after the solenoid is energized. Would it be better to free-wheel on the high-side? Are there any advantages/disadvantages in this application?

I've never really been clear on this decision.

 

[tcmtech]

I think that there may be a possibility that if you have 36,000 uf of capacitors with working voltage ratings up over the 300 volt level your ESR of the set may be too high to be able to handle the kickback from the solenoids.
In effect they are producing a very short voltage spike well past the 600 volt working range of your IGBT's and that causes them to pop.

If it was me I would be tempted to try putting a load dump circuit of sorts on the power supply rails that triggers when ever the rail voltage goes over 350 volts and then shuts of when it drops below again. The simplest method would be to turn another IGBT like you are using now that dumps excess power directly through a 20 ohm resistive load.

Load dumping and other power diversion circuits are normal on VFD's and similar devices that deal with large inductive loads that are prone to creating substantial power spikes and feedback. At 15 amps on a 300 volt system I would count that as being more than justifiable to need it.

 

[koniho]

I think that there may be a possibility that if you have 36,000 uf of capacitors with working voltage ratings up over the 300 volt level your ESR of the set may be too high to be able to handle the kickback from the solenoids.
In effect they are producing a very short voltage spike well past the 600 volt working range of your IGBT's and that causes them to pop.

If it was me I would be tempted to try putting a load dump circuit of sorts on the power supply rails that triggers when ever the rail voltage goes over 350 volts and then shuts of when it drops below again. The simplest method would be to turn another IGBT like you are using now that dumps excess power directly through a 20 ohm resistive load.

Load dumping and other power diversion circuits are normal on VFD's and similar devices that deal with large inductive loads that are prone to creating substantial power spikes and feedback. At 15 amps on a 300 volt system I would count that as being more than justifiable to need it.

Previously, I was using a bank of 9+ caps and the system had worked a lot more reliably. At some point, it was decided to get caps with higher capacitance and reduce the number (for space savings). Assuming that both caps had similar ESR, the bank of 9 caps would have a much lower ESR than the 3 12,000 uF caps that are currently being used, so this may be a clue.

I'll do some reading on Load dumping circuits. Would a transorb (varistor, or any other TVS) be sufficient or does the load dumping need to be active?

 

[tcmtech]

If you changed capacitors and the problem started I would have a lot of suspicions its related to the ESR of the new capacitors which means that you may have a very good chance of filtering that problem out just by adding a number of much smaller capacitors that have very low ESR ratings.
I would try putting several .1 and 1 uf and one or two 5 - 10 uf poly capacitors on the main bank and see what that does for feedback spike clamping.

Regarding the load dump circuit for what power level you are using I would go with an active system so that it can handle the high power levels and frequent cycling. Perhaps start with a big wire wound resistor or common electric heater element.

 

[BrownOut]

I'll do some reading on Load dumping circuits. Would a transorb (varistor, or any other TVS) be sufficient or does the load dumping need to be active?

These are overvoltage suppression devices, and as I said earlier, they can help protect your driving transistors. Snubbers can also help to dissipate the extra energy. You should give both a try, but don't just connect them and start running, make sure you monitor the trun off transent while you ramp up the voltage. That will give you the best clue to what is working. You can use the suppression devices across the inductive load, the drivers and the rails, if needed.