Bidirectional current switching with single type transistor

[artis]

So here is my dilemma. I have a test setup with multiple coils all lined up in parallel in such a way that one end of each coil is connected together while the other is connected to a switch (mosfet) the switches then connect each coil to ground but that ground can become + based on which coil is switched ON because one type of coils have current in one direction while the other have current in the opposite direction.
The coils generate the current within them from a primary source.

The problem is this , I have basically a row of parallel switches. A mosfet can easily conduct current in both direction but it also has the body diode into it.
This means that if my voltage drop across the mosfet is less than the threshold of the body diode it will stay OFF when other fet's are ON but if that voltage is above that it will conduct, if it also has the right polarity.
This means that those coils that have the polarity in the "right" way for the body diode to conduct will conduct without gate charge every time the PD across the switch increases above the body diode threshold.

Now having just one switch is the best solution but it doesn't work in this case.
What other solutions are there besides this ?
Back to back mosfet aka a two series mosfet bidirectional switch?
This solves the current conduction in reverse direction problem but now I have to drive two mosfet instead of one for every switch I need and I need a gate drive IC etc,
What about IGBT's ? They too have body diodes , most of them anyways

 

[Baluncore]

Some bipolar switches use SCRs or TRIACs, others use sampling bridges.
I would start by suggesting relays, or CMOS analog gates.

But first we need to know the magnitude of the problem.

Are the voltages across the individual coils bipolar ?
What is the maximum voltage present being switched ?

Does current flow both ways through an individual coil ?
What is the maximum current being switched ?

What is the maximum frequency of the coil current ?
How much time is available for switching ?

If this was the windings of a multi-phase PM rotor generator, then you could use diodes, allowing it to inherently switch itself. Use a MOSFET backwards, the body diode then conducts, the gate allows reverse conduction.

Some problems cannot be solved in an ideal way.

 

[artis]

Some bipolar switches use SCRs or TRIACs, others use sampling bridges.
I would start by suggesting relays, or CMOS analog gates.

But first we need to know the magnitude of the problem.

Are the voltages across the individual coils bipolar ?
What is the maximum voltage present being switched ?

Does current flow both ways through an individual coil ?
What is the maximum current being switched ?

What is the maximum frequency of the coil current ?
How much time is available for switching ?

If this was the windings of a multi-phase PM rotor generator, then you could use diodes, allowing it to inherently switch itself. Use a MOSFET backwards, the body diode then conducts, the gate allows reverse conduction.

Some problems cannot be solved in an ideal way.

Let me make this easier by making a drawing.
My voltages are low, from couple of volts to double digits not more.
Assume I have two batteries with opposing polarities, I want to switch them through a load to create an AC current through it, say I have just two of them so two switches, the problem as already described.
Current could get high, so low loss/resistance would be beneficial in terms of switches.

My wish is so that I could create a simple driving circuit as I wish to use semiconductors for the switches because they are fast , relays are not an option.
So a fast switch , not necessarily high voltage , but one that can block reverse current when met with reverse polarity, so a single mosfet doesn't work.

I wonder what are the easiest most common alternatives?

 

[Rive]

Assume I have two batteries with opposing polarities, I want to switch them through a load to create an AC current through it

Are you sure you would not be better off with a simple H-bridge instead?

 

[Baluncore]

What you have there is half an H-bridge.

 

[TonyStewart]

Why re-invent the wheel?, rather learn how things work first.

also learn this ... https://www.physicsforums.com/posts/6895630/

then learn everything about this half-bridge design

 

[Baluncore]

then learn everything about this half-bridge design

I certainly had not realised that artis was designing a self-resonant push-pull induction heater.

I thought something like this would be closer to the topological requirements.

 

[TonyStewart]

The load may be changed to resistive but then needs a dual clock with dead-time to prevent shoot-thru (short circuit on DC.) This (fault) will happen to your circuits when "control" Vgs > Vgs(th).

All CMOS logic is designed this way but unipolar, such that the RdsOn is limited by the Vdd and controlled for each family , but with a 50% tolerance for Vol/Io=R i.e. "controlled shootthru" such as 50 ohms for 5.5V logic and 25 ohms for 3.6V logic. Whereas old CD4000 was much higher RdsOn and made for high voltage (<18V)

 

[Baluncore]

@TonyStewart
You should identify who you are addressing with your comments.
This is not a social media free-for-all.

I believe you are making this discussion more complex than it needs to be at this stage, by posting circuits that are quite inapplicable to the situation, and so derailing this thread.

Please get back in your box, and give the OP time to work out what they want to do.

 

[TonyStewart]

With respect;
I do understand the problem of commutating Nch FETs all the same type ( vs complementary) and how to clamp one side off when driving the other using an inductive load.

known: user has a row of "like transistors (FETs)" and multiple coil load and a commutation problem
unknown: purpose of this design, engineering specs, logic diagram
understood: there are many basic types of commutation, inductive feedback with diode clamps (aka also known as ZVS oscillator) and logic-level forward-commutation with dead-time control using PWM on low side to generate Vboost voltage for high side Nch. It all depends on type of load and purpose of the design. There are also mono and bipolar supply designs.

@TonyStewart
You should identify who you are addressing with your comments.

I thought my comments were directed at everyone, including your simplified half-bridge models.

missing: a reference to my schematic. [my bad]

https://christianmalherbe.com/Induction Heater.html

 

[artis]

@TonyStewart I appreciate your effort but indeed @Baluncore is right that we need to slow down a bit here. You took my MS Paint made analogy which is a simple analogy to my actual problem and thought that I'm trying to make a half bridge switching network, this is not the case.
Although I can be partly blamed for not making a better case explanation in the first place, for that I apologize.

What I actually have and what I tried to make clear in the simple analogy is a single wire (literally a wire) that passes through a changing B field periodically , so a current is generated due to Lorentz Force , this creates a voltage across the wire.
In the end of the wire/conductor there is my switch, you can just imagine a single mosfet.
The problem is this , with a single FET when current/voltage is in one direction the mosfet can shut OFF aka control the current , when the voltage reverses, the MOSFET is continually ON due to the body diode.

So there came my problem how to create a switch with transistors that can be applied to a single wire in which current/voltage changes direction.

It is essentially a switch for an AC wire, the closest I know is the so called "solid state relay" where two N mosfets or P mosfets (usually N due to simple driving and lower RdsON) are connected back to back but such a two FET switch is more complex to control and requires twice the parts.

So my curiosity was can one make something simpler , is it even possible

 

[DaveE]

Putting the switch inside a bridge is simple, in some ways. Extra diode voltage drops and difficult FET drive are issues though.

 

[artis]

@DaveE I think that this configuration doesn't differ from a two switch series bidirectional switch, because either way one gets two diodes in place of the other transistor which together is either equal or higher of a series resistance isn't it?

 

[TonyStewart]

I think you must define all inputs and out waveforms and relevant spectral parameters better before choosing any design with a logical diagram (schematic).

Remember that a transistor is only bipolar if biased on some DC current otherwise unipolar and flyback effects reverse in voltage not DC current.

@DaveE 's suggestion will rectify bipolar source current from a unipolar switch since it interfaces with a full wave rectifier. Is that all you need? That is just a simple concept.

We need a much clearer problem definition.
- you initially said " multiple coils all lined up in parallel "

 

[DaveE]

@DaveE I think that this configuration doesn't differ from a two switch series bidirectional switch, because either way one gets two diodes in place of the other transistor which together is either equal or higher of a series resistance isn't it?

Yes. I think you're right if you only have one transistor. Most topologies will use two transistors (usually in series) for this reason.

But no (if I understand your comment), transistors, properly selected, will have a lower voltage drop than diodes. Synchronous rectifiers are a good example of this. If you need low voltage drop, you may want to replace your diodes with FETs. Of course, the MOSFET Rds(on) becomes an issue at very high currents. But then you can switch to BJTs, SCRs and such.

 

[DaveE]

Some bidirectional switch topologies shown at the beginning of this paper:

https://www.researchgate.net/public...g_SiC-MOSFET_for_Power_Converter_Applications

 

[DaveE]

With respect;
I do understand the problem of commutating Nch FETs all the same type ( vs complementary) and how to clamp one side off when driving the other using an inductive load.

known: user has a row of "like transistors (FETs)" and multiple coil load and a commutation problem
unknown: purpose of this design, engineering specs, logic diagram
understood: there are many basic types of commutation, inductive feedback with diode clamps (aka also known as ZVS oscillator) and logic-level forward-commutation with dead-time control using PWM on low side to generate Vboost voltage for high side Nch. It all depends on type of load and purpose of the design. There are also mono and bipolar supply designs.

I thought my comments were directed at everyone, including your simplified half-bridge models.

missing: a reference to my schematic. [my bad]

https://christianmalherbe.com/Induction Heater.html

There is often great value in focusing on the fundamental question/concepts in answering questions. We know you know a lot about complex circuits, you don't have to prove that at the expense of clear communication about the OPs question. Your half-bridge resonant converter schematic looked familiar to me, but my immediate reaction was "I don't care. It's too complicated for this thread." He asked about a switch topology, that is all.

You are better at filtering through the extraneous information in your examples than any of us are. There is also an advantage that you can do that once for many people, instead of expecting every reader to figure out what your fundamental point is.

Good educators start simple and query the questioner for what to say next.

 

[TonyStewart]

Did you follow how alternating current follows from DC using Full-Bridges using single supplies and Half-Bridges from dual supplies? There is also a need to have a dead time during transitions that depends on load reactance and current to avoid shoot-through failures.

Did you follow my link from my ZCS Oscillator (zero-crossing switch) example in #4?

When coils or chokes or transformers are used, several options exist but most popular are single driver with centre tap high , also used in stepper motors, or bipolar H-bridge drivers. Bipolar drivers can be both active or using a flyback diode for inductive loads. These topologies are common to SMPS as well.

There are many books on this subject. and short links https://www.wikiwand.com/en/H-Bridge

 

[DaveE]

I would probably just put two N-Mosfets in series. Either with the drains connected or the sources connected together, depending on your gate drive circuit choices. You can either use the body diodes or you can turn them both on together to get lower voltage drop.

Much of this design will depend on how you want to design the gate drive circuits. It's probably best to have the G-S nodes in a part of the circuit that doesn't generate a big dV/dt across the isolation barrier during switching, I think. Although that isn't necessarily possible or even optimal, depending on the application. There are lots of options here.

Note that many comments related to half bridges aren't really bidirectional switches, topology wise; which is fine if that works for you too.

 

[TonyStewart]

Your array of wires is like a Y or star-connected motor which can be as many phases as your want, typically 3 or more. Thus is defined as a "half-bridge" for each phase or wire with dead-time commutation.
If your power is low, it is pretty uncomplicated to simulate if you have interface specs.

 

[DaveE]

OK, changed my mind. Since you have lots of coils each being switched to a common node, I guess I would opt for the easy gate drive option. This would be an N-FET and a P-FET in parallel each with a series blocking diode to prevent reverse current. Then all of the sources can be at a common node which will make the control circuit much simpler.

 

[TonyStewart]

OK, changed my mind. Since you have lots of coils each being switched to a common node, I guess I would opt for the easy gate drive option. This would be an N-FET and a P-FET in parallel each with a series blocking diode to prevent reverse current. Then all of the sources can be at a common node which will make the control circuit much simpler.

I don't think those diodes will prevent shoot-through @DaveE .

 

[Baluncore]

I don't think those diodes will prevent shoot-through

Spreading fear, uncertainty and doubt about other member's suggestions is not of value here at PF. You seem to delight in immediately undermining suggestions made by other members. We do not need that destructive behaviour here.

You need to come up with positive suggestions directly related to the thread.
If you must be critical, then practice self-criticism, before you post.

 

[DaveE]

I don't think those diodes will prevent shoot-through @DaveE .

Sorry, I guess I should have drawn a schematic for you. When it's on everything "shoots through". When it's off, nothing "shoots through". It's a bidirectional switch, that's what it's supposed to do. Where does the current "shoot" to when you turn the switch on? IDK, that's not our problem, yet.

The OP hasn't really described the rest of his system in any detail. Perhaps we could be more helpful if we didn't assume we know the details of his application. He can ask for more if he needs it, you know, unless he's fed up at this point.

 

[artis]

@TonyStewart can you please just pause first? Once in a while there is a PF member who just throws the whole cyclopedia on a topic , even if you have the actual experience of building and working with every schematic and topology you have mentioned , it still is somewhat overkill to just throw it all out there on a topic that has no association with that.
I already said, please excuse me if I wasn't clear enough from the start , the reason for that is that I wanted to make the problem as simple and reduce it to the main part as much as possible, that caused some confusion.

I am also familiar with switching topologies to a certain level but my problem doesn't deal with any switching topology.
Shoot through is only a problem for switching topologies because of the stored DC energy in the caps, I have no caps, no DC, no transformer , no flyback voltage no nothing. Just a tiny bit of inductance and some capacitance , no more than a typical transmission cable would have.
I as @DaveE correctly said ,I actually need the "shoot through" to happen.

The idea is basically this, I have many parallel wires, each wire has a switch in it, the wires carry bidirectional current. That's it. If it makes it simpler for you @TonyStewart imagine many mains phase wires all in parallel and all of the same phase, current goes forth current goes back, put a switch across such wire and you need the capability to block current in both directions which with a single mosfet isn't possible due to the body diode. That is it, this is all the problem.So what I was originally looking at, and still am is this
How to switch bidirectional current with
1) as little of switch resistance as possible
2) as simply as possible (low parts count)
3) the switch driving to be as simple as possible

At the same time I can paraphrase Einstein, I need the switches and driving to be as simple as possible but not simpler.... in other words, to just do the job.

The simple two FET back to back series configuration is known to me, I have implemented that already in a solid state relay in an amplifier output.
The reason I was asking is because that kind of configuration needs a more complicated gate drive circuit, no problem for a once on , once off amplifier output relay, but becomes trickier for many parallel switches that also need fast switching.Maybe I can sacrifice a little bit of that RdsON switch resistance by omitting a switch and putting instead a fast recovery diode in it's place or something like that.

 

[artis]

OK, changed my mind. Since you have lots of coils each being switched to a common node, I guess I would opt for the easy gate drive option. This would be an N-FET and a P-FET in parallel each with a series blocking diode to prevent reverse current. Then all of the sources can be at a common node which will make the control circuit much simpler.

I thought about this, but I think there are two problems, possibly, first is that there actually isn't a common node on either part of the switches, because there are additional devices "down the line" so whenever current reverses there would a PD of some couple of tens of volts that also reveres across any closed switch.
Secondly , isn't it the case that all P fet's have higher RdsON than their N channel counterparts due to the inherent structure of the device itself?
Doesn't this affect voltage drop for low voltages ?

 

[DaveE]

So what I was originally looking at, and still am is this
How to switch bidirectional current with
1) as little of switch resistance as possible
2) as simply as possible (low parts count)
3) the switch driving to be as simple as possible

I doubt you'll get all of that. It'll be a trade off.

Maybe I can sacrifice a little bit of that RdsON switch resistance by omitting a switch and putting instead a fast recovery diode in it's place or something like that.

Schottky diode then? Lower Vf and I think you said it's a low voltage application.

 

[artis]

Schottky diode then? Lower Vf and I think you said it's a low voltage application.

low voltage yes, but high current, will have to look into it, don't know from the top of my head but IIRC schotky's were usually in low power applications?

 

[DaveE]

Secondly , isn't it the case that all P fet's have higher RdsON than their N channel counterparts due to the inherent structure of the device itself?

Yes, they suck, mobility wise. So you pay more and buy a bigger die, or you use NMOS with a more complex circuit.

schotky's were usually in low power applications?

I don't know the V, I, P you are working with, but there are definitely big Schottkies, current wise.

https://www.vishay.com/docs/94171/vs-243nq100pbf.pdf

 

[Baluncore]

Power MOSFETs often have a Schottky diode in parallel with the body diode.

If you use the body diode of a MOSFET as a forward conducting diode, you can turn on the MOSFET while the diode is conducting to reduce the diode voltage drop. That is the principle of an ideal diode, or as used in a synchronous rectifier.

Since this is high-current low-voltage, if you connect two N-channel MOSFETs source to source, and gate to gate, do you not have a bidirectional switch between the drains, controlled by ±15 volts on the common gate terminal?

 

[artis]

Some bidirectional switch topologies shown at the beginning of this paper:

https://www.researchgate.net/public...g_SiC-MOSFET_for_Power_Converter_Applications

I had seen this paper before, some interesting options, not sure how well BJT's take prolonged operation near their SOA current wise, doesn't it cause heating of the die and secondary breakdown or what was the thing happening to them when the die temp increased beyond a limit?

Since this is high-current low-voltage, if you connect two N-channel MOSFETs source to source, and gate to gate, do you not have a bidirectional switch between the drains, controlled by ±15 volts on the common gate terminal?

In theory yes, by the way isn't it the case that in common source mode for two series N fets the voltage drop for either current direction forms only across one of the two switches given the other is bypassed through the body diode therefore before applying voltage to gate to turn ON the PD is across one switch in either current direction?

 

[Baluncore]

In theory yes, by the way isn't it the case that in common source mode for two series N fets the voltage drop for either current direction forms only across one of the two switches given the other is bypassed through the body diode therefore before applying voltage to gate to turn ON the PD is across one switch in either current direction?

When both MOSFETs are turned on, neither body diode conducts.
When both MOSFETs are turned off, one body diode conducts, the other blocks current.

The source floats when off, but quickly resolves the situation if it is negative, since a MOSFET then turns on sufficiently, to turn both off. The source cannot float high, because then a body diode will conduct.

 

[Rive]

need the capability to block current in both directions which with a single mosfet isn't possible due to the body diode.

I'm still struggling with understanding the circuit you intend to build, so please just take it as noise if does not fit: once we had a high current switch circuit with two serial MOSFETs where OFF-current was terminated by keeping the middle point voltage level close to the output point (by an extreme input resistance opamp circuit).
Keeping the body diode closed just did the trick nicely.

 

[artis]

When both MOSFETs are turned on, neither body diode conducts.
When both MOSFETs are turned off, one body diode conducts, the other blocks current.

The source floats when off, but quickly resolves the situation if it is negative, since a MOSFET then turns on sufficiently, to turn both off. The source cannot float high, because then a body diode will conduct.

But I can also ground the source with respect to the gates at all times within the gate drive circuit, can I not?
It shouldn't affect the voltage/current of the main path through the FET's nor the gate drive circuit itself because the two circuits - the power and gate drive one don't interact.

 

[Baluncore]

 

[Baluncore]

But I can also ground the source with respect to the gates at all times within the gate drive circuit, can I not?
It shouldn't affect the voltage/current of the main path through the FET's nor the gate drive circuit itself because the two circuits - the power and gate drive one don't interact.

I don't think it is necessary to ground the sources if the current path is low-voltage and the gate voltage is high enough to turn the switches on, while being small enough to not turn on the Zener diodes used for gate voltage protection.

You can ground the sources if you do not ground any other part of the switched current circuit. There may be problems with grounding the sources if the inductors share a common node.

You need to consider protection of the MOSFETs from the flyback of the inductor voltage.

 

[artis]

Ok, so I've thought about it and here is a partial solution I've come up with.
The coils are all parallel shown by the numbers "1 and 2" , so each parallel coil has the switching part in the middle of the of the coil (because there are two coils in series) So as @Baluncore already suggested , I'm using a plain "old" N mosfet bidirectional switch made from two back to back common source N fets.
Then I am using a zener in series with a diode arrangement. Making two of these with opposing polarities since inductor current is bidirectional.

So whenever the mosfets open up into OFF position, the flyback voltage is clamped by the zener + diode series clamp, but the zener addition I think clamps the inductor current at higher voltage which might be beneficial for faster inductor energy discharge as compared to clamping at regular diode ~0.7 volts .

The zener also makes it possible to use the bidirectional mosfet switch at all since while switching the regular current through the inductors the zener will allow not to short circuit and circumvent the actual N fet switch, otherwise simple diodes would conduct all the time , everytime the voltage is above 0.7

Additionally I have two diodes parallel to each mosfet body diode as they are fast recovery and can handle larger currents.

Besides I think that by having two such diodes parallel to the mosfet ones, I can actually just commutate one FET at a time by leaving the other open or closed depending on the need.

Anyway what do you think?
Remember the current through the inductors/coils is bidirectional so I can switch both while it passes one way as well as it passing the other way so flyback is both directions

 

[Baluncore]

Anyway what do you think?

Zeners are a reverse breakdown diode with normal conduction forwards. Therefore, two Zeners, reversed in series, will clamp bi-directional transients. The breakdown voltage will not be perfectly symmetrical since two different Zeners are being used.

Zener diodes are more expensive than power diodes, so a single Zener in a rectifier bridge, is the lowest cost solution, and has voltage symmetry.

Additionally I have two diodes parallel to each mosfet body diode as they are fast recovery and can handle larger currents.

I don't think the FRDs are necessary.

The decision whether to use common source or common drain MOSFET pairs will have implications to how the gates of the bidirectional switches are biased and driven.

 

[DaveE]

Therefore, two Zeners, reversed in series, will clamp bi-directional transients.

Indeed. You can buy them that way as one part.

https://www.littelfuse.com/~/media/.../littelfuse_tvs_diode_1_5ke_datasheet.pdf.pdf

 

[DaveE]

I would also consider leaving out the FRD diodes and turning both fets on together so that you get lower on voltage drop as the current flows through both channels and not the body diodes (synchronous rectifier).

Anyway, I don't think reverse recovery is an issue since the turnoff time is only dependent on turning off the forward biased fet. It doesn't matter how long the other forward biased diode takes to switch, does it?

 

[artis]

Zeners are a reverse breakdown diode with normal conduction forwards. Therefore, two Zeners, reversed in series, will clamp bi-directional transients. The breakdown voltage will not be perfectly symmetrical since two different Zeners are being used.

Right I forgot about that simple fact that I can use only zeners since they too function also as a regular diode below the reverse breakdown voltage by which it starts to conduct in the opposite direction.

Zener diodes are more expensive than power diodes, so a single Zener in a rectifier bridge, is the lowest cost solution, and has voltage symmetry.

I'm not sure I imagined correctly how this can be made possibly in a N fet bidirectional switch?

The decision whether to use common source or common drain MOSFET pairs will have implications to how the gates of the bidirectional switches are biased and driven.

I have the impression that using them with sources tied together is more forgiving in terms of driving them , especially if the overall switched voltage is low.
Given those switches would come parallel to one another I could tie the common sources all together but in order to not create a ground loop I could tie them together using a resistors, what do you think? Or is that a bad idea? It would probably decrease the available upper switching frequency

Anyway, I don't think reverse recovery is an issue since the turnoff time is only dependent on turning off the forward biased fet. It doesn't matter how long the other forward biased diode takes to switch, does it?

That is true.

Anyway so here's my updated schematic based on the suggestions/corrections

 

[Baluncore]

I'm not sure I imagined correctly how this can be made possibly in a N fet bidirectional switch?

You can use B2B Zeners, or a Zener in a bridge. But Zeners and mains frequency rectifiers are slow, so you may need snubbers to reduce the slew rate.

An alternative would be to use two FRDs on each inductor, connected to the supply rails. That recovers the magnetic energy and stores it where it can be recycled, just make sure it cannot pump the supply voltage up beyond what is safe.
Here is the bridge circuit, and my preferred FRDs to the supply rails.

Given those switches would come parallel to one another I could tie the common sources all together but in order to not create a ground loop I could tie them together using a resistor, what do you think?

Without a greater circuit diagram, I cannot comment.

I would try to avoid placing the bidirectional switch between two inductors, because that requires twice as many flyback protection diodes, and moves the source further from the ground or common.

 

[DaveE]

Be mindful of the power/energy dissipation in those zener diodes. IDK your situation, but it's easy to blow up a normal tiny zener in PS clamping circuits. This is an advantage in clamping to the PSs with normal diodes as @Baluncore suggested. That's a much more common approach (except when you can't, LOL).