In any case the back EMF is only about 0.7V with the diode so it won't radiate much energy. An unsupressed spike from the wire inductance would likely radiate more.

Well, I triple-dog-disagree with you. The lower the clamping voltage the larger the current spike. This is the reason for putting th diode as close to the source (inductor) as possible, the decrease the loop inductance/radiating antenna size.

MOV clamping action is slower and softer than tranzorbs or zeners, which is bad for protection but good for EMI.

IMHO

ak

 

Should be easy enough to run an experiment and resolve the question?

 

Well, I triple-dog-disagree with you. The lower the clamping voltage the larger the current spike. This is the reason for putting th diode as close to the source (inductor) as possible, the decrease the loop inductance/radiating antenna size.

MOV clamping action is slower and softer than tranzorbs or zeners, which is bad for protection but good for EMI.

IMHO

ak

It should be as close to it's source as possible - this is true with almost anything. Inductance from a length of wire is negligible (in this case) - unless you have under gauged it, but then you have other issues to worry about.
I have a diode at the FET - and it's not working. However it is working with one at the Coil. But I can't always put one at the coil, therfor the need of the additional electronics as mentioned.

http://en.wikipedia.org/wiki/Flyback_diode
Induction at the opening of a contact

According to Lenz's law, if the current through an inductance changes, this inductance induces a voltage so the current will go on flowing as long as there is energy in the magnetic field. If the current can only flow through the air, the voltage is therefore so high that the air conducts. That is why in mechanically-switched circuits, the near-instantaneous dissipation which occurs without a flyback diode is often observed as an arc across the opening mechanical contacts. Energy is dissipated in this arc primarily as intense heat which causes undesirable premature erosion of the contacts. Another way to dissipate energy is through electromagnetic radiation.
Similarly, for non-mechanical solid state switching (i.e., a transistor), large voltage drops across an unactivated solid state switch can destroy the component in question (either instantaneously or through accelerated wear and tear).
Some energy is also lost from the system as a whole and from the arc as a broad spectrum of electromagnetic radiation, in the form of radio waves and light. These radio waves can cause undesirable clicks and pops on nearby radio receivers.
To minimize the antenna-like radiation of this electromagnetic energy from wires connected to the inductor, the flyback diode should be connected as physically close to the inductor as practicable. This approach also minimizes those parts of the circuit that are subject to an unwanted high-voltage — a good engineering practice.

 

I have a diode at the FET - and it's not working. However it is working with one at the Coil. But I can't always put one at the coil, therfor the need of the additional electronics as mentioned.

so the coil is say 10ft from the fet, and when the diode is at the fet this fet dies, but if you move that same diode 10ft away onto the coil then the fet does not die? is all the wiring the same in both cases?

i would think the current looping would be dampened by the fact that 20ft of wire has some ohms in it. two parallel wires wont have much inductance compared to the coil itself.

interesting.

but i will lean back to zener on fet drain.

 

so the coil is say 10ft from the fet, and when the diode is at the fet this fet dies, but if you move that same diode 10ft away onto the coil then the fet does not die? is all the wiring the same in both cases?

i would think the current looping would be dampened by the fact that 20ft of wire has some ohms in it. two parallel wires wont have much inductance compared to the coil itself.

interesting.

but i will lean back to zener on fet drain.

Basically yes. But I think the diode at the FET isn't doing me any good because it is just connected to the +12v supply and not to the + side of the coil like it should be. Adding one at the coil takes out the spike. But I still like your idea of the zener as added protection.

I think the added resistance of the wire length "softens" the blow by the time it reaches the FET and that is why it does not pop instanly, sometimes it takes a few.

Thanks!

 

Basically yes. But I think the diode at the FET isn't doing me any good because it is just connected to the +12v supply and not to the + side of the coil like it should be. Adding one at the coil takes out the spike. But I still like your idea of the zener as added protection.

I think the added resistance of the wire length "softens" the blow by the time it reaches the FET and that is why it does not pop instanly, sometimes it takes a few.

Thanks!

ah, i suspect the +12v that you are tying the diode to on fet side has some other silicon or resistance in the path back to the +12v terminal of the coil making it difficult for that loop current to make its way back ?? the diode does need to bridge across the coil terminals (physically where doesnt matter for your application). the fet was burning up because the diode could not do its job correctly.

resistance in the wire only slows the current of the inductive emf (when diode is on fet side), should not mean anything to the fet. the voltage on the wire will be about the same on both ends of the wire. granted, Drain capacitance takes time to charge like a RC circuit, but that charge and resistance is very small. with diode properly working (i would still add zener), i bet ya if you scoped the single/same wire from coil end to fet drain, you will see a small voltage diff.

i have done heavy load switching through big inductor and all i used was zener+resistor to ground, as shown in the scope pics i posted. so if putting a diode on the coil is a difficult or non-reliable method, then zener on your board is the way to go.

 

It's just the inductance in the big loop from the power supply to both the solenoid and the board. Since you don't control what happens at the solenoid you have to protect the FET with either a big zener or a TVS.

 

Like I said in post #9, put a cap on the Gate-Source of the FET, then it will turn on/off slowly (ramping over 50m or 100mS etc).

There will be no back EMF "spike". Problems all solved.

It's a trick I learned from high-end audio equipment relay switching.

 

That or just a big resistor in series with the gate drive signal.

 

Well, I triple-dog-disagree with you. The lower the clamping voltage the larger the current spike. This is the reason for putting th diode as close to the source (inductor) as possible, the decrease the loop inductance/radiating antenna size.
................

Triple-dog-disagree, eh? That's pretty serious. But that doesn't make your statement correct. The current "spike" is not a spike. It remains the same no matter how it is suppressed, as it is equal to the inductor current at the instance the switch is open. The rate of decay of that current does depend upon what stops the current. A diode has a very low voltage so the current is slowed mainly by the resistance in the circuit (including the coil resistance). This relatively slow reduction in current radiates very little no matter how large the loop. However, an inductive current that is not suppressed by the diode, such as the inductance of the wire between the solenoid and the switch, and between the solenoid and the power rail, will create a voltage spike and a likely decaying oscillation. That is what generate EMI radiation. And that's why you want the diode to suppress the whole loop, not just the solenoid coil.

 

That or just a big resistor in series with the gate drive signal.

if you run the fet in the non-linear region the fet will run much hotter. turning the fet on/off fast is the better way for the fet. handling the flyback is as easy as adding a zener + resistor from drain to gnd.

 

Sure, especially since there are a lot of unknowns - Length of wire, inductance of the solenoid, and my favorite, no 12 volts to the board at all only to the solenoid. Why a resistor and zener, why not just a zener or TVS.

 

Sure, especially since there are a lot of unknowns - Length of wire, inductance of the solenoid, and my favorite, no 12 volts to the board at all only to the solenoid. Why a resistor and zener, why not just a zener or TVS.

a 10 or 20 ohm to "restrict" the current which protects the zener. it also dissipates some of the energy.

 

if you run the fet in the non-linear region the fet will run much hotter. turning the fet on/off fast is the better way for the fet. handling the flyback is as easy as adding a zener + resistor from drain to gnd.

He specified driving a relay coil at long distance. It's not PWM.

Adding a ramped turn on/off to the FET will not cause any significant heat difference in an infrequently switched 800mA relay coil. But it will massively reduce any issues of EMI or circuit glitching.

When I worked in heavy industry much of the control gear was deliberately designed for slower changing signals, it solves so many problems. Things like RC networks on remote switches etc.

It's popular fashion to try to do everything in nS these days, but it's not always better.

 

Well, I triple-dog-disagree with you. The lower the clamping voltage the larger the current spike. This is the reason for putting th diode as close to the source (inductor) as possible, the decrease the loop inductance/radiating antenna size.

Triple-dog-disagree, eh? That's pretty serious. But that doesn't make your statement correct. The current "spike" is not a spike. It remains the same no matter how it is suppressed, as it is equal to the inductor current at the instance the switch is open. The rate of decay of that current does depend upon what stops the current.

Gotta call this for crutschow here. First we all should know the current in an inductor cannot change instantaneously, so the current before the switch opens is the same as the current after the switch opens. Next, since the energy os proportional the the square of current... current is decreasing as energy drains out, and it must be draining out since you ain't puttin no mo into it!

The rate of discharge is controlled by the voltage across the coil. Since:

V = L dI/dT or dI/dT = V/L

the larger the voltage the faster the current decays. Give it a steady voltage and the current decays in a straight line from the max (just before opening) current down to zero.

(You'll see funny 2nd & 3rd order effects on your scope if you watch an actual coil but basically this is the predominant way the current decays.)

 

I have the feeling that the issue is placement of the Diode as a couple of you mentioned. As I stated before, the Diode is on board with the FET, Anode to the Drain and Cathode to +12v however as it is installed the +12v for the coil may be coming from a long distance away - or possibly a different source, that helps confirm my suspicions.

This should be the only way it can blow. Even if the line inductance is a 100 or 200uH (few hundred feet of wire, single turn) and the FET goes into avalanche break down until the diode clamps it, it is only a few uj of energy- well below the FET spec of 300+ mj. If the 12 volts is missing to the board or from another source and the inductance of the solenoid is fairly high (say 800mh) it will be toast.
If this is the case the turn off would need to be very very slow and the turn off power could still be close to being outside the safe operating area of the FET (75 watts).
To me the zener or TVS is the best solution. I would just use a 5 watt zener without the resistor and let the coil resistance limit the current. Noisy but a cool FET.

 

He specified driving a relay coil at long distance. It's not PWM.

Adding a ramped turn on/off to the FET will not cause any significant heat difference in an infrequently switched 800mA relay coil. But it will massively reduce any issues of EMI or circuit glitching.

When I worked in heavy industry much of the control gear was deliberately designed for slower changing signals, it solves so many problems. Things like RC networks on remote switches etc.

It's popular fashion to try to do everything in nS these days, but it's not always better.

slowing the coil in a relay will cause the contacts to move slower, and this will allow for the deterioration of the contacts at a faster pace. sames goes for just using a diode across the relay coil, the looping current causes the relay to work more slowly. a major concern for the application? dunno, no requirements have been posted.... but i will assume the OP wants the parts to last as long as possible, etc.

To me the zener or TVS is the best solution. I would just use a 5 watt zener without the resistor and let the coil resistance limit the current. Noisy but a cool FET.

+1, but adding a 10-20ohm is safe.

 

slowing the coil in a relay will cause the contacts to move slower, and this will allow for the deterioration of the contacts at a faster pace. sames goes for just using a diode across the relay coil, the looping current causes the relay to work more slowly. a major concern for the application? dunno, no requirements have been posted.... but i will assume the OP wants the parts to last as long as possible, etc.
...

That's a valid point, but doesn't apply that much. Even if you can instantly drop the coil current to zero the core and armature can take 50-60mS to de-saturate and throw. It's a slow process because the core and armature are not laminated but plain steel.

So ramping the coil current to zero over 50mS or so makes little difference on the throw out time of the armature. The mechanical times are much slower than the coil current change.

You are right of course if the application causes for very fast contact break then a diode across the coil is not good as it slows the current decay. But then it probably calls for a special fast break relay anyway, and zener coil snubber etc, which is heading into a specialist application.