SCR gate drive problem

[Noway2]

All,

I have been dragged into a problem at work that is a bit over my head and I am hoping that some of you could help shed some light on it for me. The problem relates to our motor soft start controllers. The production test department has learned that there is an apparent problem with getting SOME OF the SCRs to fire. I have been dragged into the situation to see if I can figure it out.

Over the past couple of days, I have been trying to come up to speed on the gate drive circuit and how it relates to the typical methods of triggering the SCRs. The firing circuit, originally designed circa 1980 generates periodic pulses to the SCR gate coupled via a pulse transformer. The circuit is an analog design that relies on the B-H curve of the transformer for timing and a bunch of other analog tricks.

So far, from measurements I have made I am seeing that the circuit is generating a firing pulse about every 24uS. These firing pulses are about 4uS wide. I am NOT seeing any 'back porch' portion to the pulse as at 4uS it is gone. The voltage amplitude applied to the gate is about 5V with a ~24V open circuit voltage. Readings with a current probe are indicating that the peak gate current is about 340mA. The anode - cathode voltage is 460Vac with a resistive load, which should be relatively easy to trigger due to the higher voltage and resistive load providing a fast latching current.

From app notes I have seen and the SCR data sheet, the voltage and current figures look like they are correct for a static (DC) scr firing, but I suspect that there isn't enough a$$ in these pulse to trigger dynamically like this. Specifically I believe that the 4uS pulse is too narrow and it should be more like 10uS. I believe that for dynamic triggering the gate current should be a couple of amps rather than a lousy 350mA.... Though I may be way off base here.

The data sheet says that the gate trigger voltage is a max of 3V with 200mA trigger current. The instantaneous gate power rating is 150W, 10W RMS - which I am nowhere approaching. Again, if I were looking at this from a DC perspective, ie load line equation I think I would be in the trigger region - but I don't think this applies to my case.

Can anyone shed some additional light on this situation and perhaps confirm or dismiss my analysis?

 

[Skogsgurra]

Hello!

Been there lots of times. I assume that you have an inductive load (motor) and that means that your current rate-of-rise is limited by the di/dt = U/L relation. If your thyristor's holding current is, say, 100 mA, and you do not get there before your 4 us trigger pulse is over, then you will not reach holding current.

The fact that you do not see a back porch (which is actual thyristor current voltage drop reflected back on the gate voltage) indicates that this is your problem.

I, too, would say that 4 us is way too narrow. What are the trigger transformers designed for? Why so narrow? Has anything been changed lately? New program in the controlling micro (if there is one) or any new component introduced? What do the older devices do?

A very effective way to test if narrow trigger pulses are your problem is to connect incandescent lamps (watch for right voltage, series connect if necessary) and see if you get correct triggering then.

Of course, soft starters have very special needs with regards to keeping current paths open in different parts of the thyristor bridge. But that is included in the basic design. Not something that you would expect to change just like that.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[sreid]

I certain you are on the right track. A google search found

http://books.google.com/books?id=rZ...A3Zg&sig=DsYnyeNbnmGga4wolSZR8JPiwBo#PPA56,M1

Which shows a gate trigger delay curve.

Two things may and probably have changed; SCR manufacturer and the core manufacturer.

 

[Skogsgurra]

Yes. It does! Indeed.

And lots of other details, too. Do not let that plethora of details lead you astray. Keep it simple. Check with incandescent lamps first. That will tell you what is going on.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[itsmoked]

Gunnar how does adding incandescents stretch the pulse or increase it to the point the SCRs start firing correctly?

Or am I missing something?

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[Skogsgurra]

Time constant.

L/R in a motor winding is hundreds of milliseconds.

L/R in a lamp is microseconds.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[itsmoked]

Oh oh! You mean as a load to the SCR! I thought these bulbs were going into the gate drive circuit.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[Skogsgurra]

Yes!

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[sreid]

The original post said, "The anode - cathode voltage is 460Vac with a resistive load."

In my first post, I was pointing out that there is a turn on delay called Gate Spreading Time in which gate current is flowing but the charge is only doing an electron-hole cancellation. So the 4 us pulses may not be transfering enough energy to start the gate turn on process (as the poster suspects).

 

[Noway2]

Hey Guys,

Thank you all for your responses.

It looks like everyone is in agreement here as to the fact that the firing pulse is too narrow. According to some engineers at the vendor that we buy the modules from, the devices are specified as taking a 5uS pulse. Hence, I would expect some devices to work at 4uS and some to not, just as we are seeing.

We have started to look at ways to modify the firing circuit to see if we can get a wider pulse out of it. As I understand it, the circuit uses a transformer in flyback configuration to generate the pulses, ie the pulse is generated at turn off. I am going to attempt to see if I can get more energy into the pulse by reducing the capacitance in the snubber (across the pulse transformer not the SCR) and see if I can switch the darlington BJT switching transformer to a FET to get a faster shutoff and hence bigger impulse.

Of course this is one of those situations where the correct solution would be a circuit redesign, but management wants a band aid fix to the present board

I will give these things a try and if I am still not having any luck, I will see if I can post a link to a schematic and some wave form pictures for everyone to reference.

Sried, that looks like a pretty interesting book. I was familiar with the concept of that once the SCR is triggered that it takes time for it to conduct over the entire region. If I recall correctly, the conduction region grows at about 50uM/uS or something along those lines, but yes, this circuit is on the edge of the curve.

 

[Skogsgurra]

Yes, I see that now. I was thinking about the soft start application. Yes, with a resistive load, it is definitely a question of too narrow pulses.

I do not understand the flyback operation. All trigger transformers I have been near are working in forward conduction mode. The flyback is usually only for fast demagnetization. Are you positive that the triggering is on the flyback? Is there a recent error introduced in the winding polarity?

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[sreid]

Reducing the snubber capacitor may help, it could be eating up some of the energy available. Remember that for a flyback drive, the energy available at the secondary is the energy stored in the primary inductance (I^2R/2).

 

[Noway2]

Skogsgurra,

No I am not positive that it is a flyback and the more I look at it, I think you are correct that it is a forward converter. I was told that it was a flyback by someone else who looked at the circuit.

I will have to see if I can get a link to a picture posted, but for the time being, here is a description.

The primary of the transformer is connected on one side to +15Vdc. The other side of the primary has about 20R of resistance to the collector of a darlington pair transistor that does the switching. On the darlington emitter is a 120R connected to -15Vdc. Also across the transformer primary is a parallel RC + diode snubber.

There is an analog control circuit that does line synchronization and turn on / turn off via comparison to a ramp signal to provide phase angle firing control.

I am starting to wonder about the operation of the transformer versus the saturation curve. As I understand it, when the primary coil is energized, current will begin to flow with the voltage drop primarily across the transformer primary. When the transformer core saturates, the effective impedance will drop sharply and the voltage drop will appear across the primary resistances instead.

If I understand the circuit correctly, it is this time, before the primary saturates, where the pulse will be generated on the secondary. If this is the case, then either the transformer is saturating too fast or the turn on signal to the Darlington is too narrow.

If the transformer core is saturating too quickly, could I slow this down by increasing the primary resistance?

I am not sure what I can do if it is a case where the control signal is too short....yet...

 

[LionelHutz]

I'm a little late but still.

I find it odd that there is a diode snubber across the transformer. I would have expected a diode and zener in series. If you're running 50% duty cycle you've got let the core reset at the same voltage as the turn-on.

You are correct that the gate current should peak at a few amps, not mA. It is not recommended to actually use the minimum gate turn-on current in real world applications. Big devices need something like a gate current that turns on at 4A/uS and 2A peak. The time of this initial pulse should be long enough to allow the anode current to reach at least 3-4 times it latching current value. Then, a "back porch" of >100mA should exist for at least 3-5 times the initial pulse width. You should also be able to supply a decent voltage with this pulse(maybe 20V @ 2A). Smaller devices can get away with maybe 0.5-1A at 10V.

Westcode has some good app notes on SCR gating.

 

[Noway2]

Thank you for the reply.

The duty cycle is not 50%. In actuality the duty cycle is very small. The way the circuit is supposed to work is that the voltage is applied across the pulse transformer (in forward mode) and the transformer is allowed to saturate. The saturation action is supposed to kick off some other circuit that allows the transformer to reset through the snubber across the transformer. From what I can tell this reset current is supposed to provide the "back porch" in that it is still primary current flowing that will appear on the secondary of the transformer. The transformer will see a voltage polarity shift typical of flyback configuration and hence should reset.

It turns out that last night I think we finally discovered what is going on. The circuit, which is one of the craziest that I have ever seen, us the Darlington transistor to apply the voltage to the pulse transformer. The Darlington has a resistor in the emitter circuit between the emitter and the -15V supply. What was (and still is) happening is that as the primary current builds, the emitter voltage rises. When the emitter voltage rises above the level of the base, the Darlington shuts off and the reset process begins and then the cycle repeats. There is also a capacitor at the emitter - resistor node that I believe is supposed to prevent the voltage rise, but I don't think it is doing its job.

Last night we found a potential error in the circuit and changed a component value. This changed adjusted the quiescent base voltage upwards, giving more room before the Darlington shuts off. Coupled with another change to get some more primary current flowing, we were able to get the pulse up to about 750mA at 4uS-5uS when driving into actual SCRs, though it is a lot wider when driving into a 10 ohm resistive gate simulator (lower voltage that is safer to play with).

In any advent, these efforts got the soft start working with our small motor and it was decided to ship the unit to the customer.

I still have some ideas I want to try to see if I can improve the circuit, but I am not certain that I will be able to get much more out of it.

 

[exrca]

An scr requires the gate pulse to be held at a certain minimum current for a minimum time to ensure maintenance of conduction, with positive anode voltage. The gate circuit has a minimum gate drive voltage and current as entry to the negative resistance switch on region, when the gate gets "joined" to the anode. Calling this vaue, Rin, I prefer to drive the gate through a resistance of several times this value and at a higher current, to absorb circuit tolerances. Normally if the gate is driven hard, within its limits, the scr will switch faster.

 

[Noway2]

Thank your for the reply and your explanation. I agree with your reasoning about using a resistance to ensure known quantities.

Unfortunately, this is one of those cases where "they" decided that the solution must be limited to component modifications to the existing circuit.

Personally, I believe that the design, even with its modifications is woefully inadequate. Unfortunately, "they" are playing the part of the proverbial ostrich and latching on to the concept that if it appears to work in their limited test environment that it means that the product is fully functional and will work under all conditions.