3
\$\begingroup\$

I am making a six-stage coilgun. I know the optocouplers are switched in the diagram, but the thyristors keep blowing up. They are blowing up randomly. That's my fourth thyristor I have replaced.

After each replacement the circuit works fine for some time (circa 40-50 shots) and then one of thyristors blows up (always while charging or when the capacitors are charged and it is waiting to be triggered by ESP32).

circuit of coilgun

pcb schematic

blown thyristor

I can't find a reason. I am using 450 V, 1000 µF capacitors with 70TPS12/70TPS16 thyristors.

The coils' resistances are about 0.49 Ω (241 µH), so the current and voltage are within specification of what the 70TPS12 and 70TPS16 can handle (using 4.5 V for gate voltage). The thyristor is blowing up while the capacitors are charging. Peak current is around 770 A. Capacitors are charged by this module hooked to 12 V 2 A (current limited) PSU.

What can be a problem? Ugh another thyristor just blew up... But now it is different. It didn't violently blew up; it just shorted. Also, it happened in the first couple of seconds of charging so it was at low voltage (between 30-50 V). Now I was testing ONLY first stage and after couple of launches the first thyristor was shorted. So I think its not connected to accidental trigger.


This is the thyristor discharge:

enter image description here

multile coils triggerd

Probe and ground on each side of the coil. I used an isolated oscilloscope.

\$\endgroup\$
20
  • 2
    \$\begingroup\$ The thyristor is blowing up while the capacitors are charging. What happens if that one thyristor is removed? Will another one blow up? The circuit diagram is very hard to read - it has very low resolution, and is drawn in a way that's hard to follow. Please edit the question and use the built-in circuit editor (press Ctrl-M in the editor) to draw the circuit. Include the charging circuit and the load (coil). You don't need to show the terminal blocks on the diagram, we would like to see how the whole system is designed. There may be a problem there. \$\endgroup\$ Commented Feb 22 at 17:04
  • 1
    \$\begingroup\$ @lukakopa Have you measured the current going through the thyristors? what source are you charging the caps from? \$\endgroup\$
    – Voltage Spike
    Commented Feb 22 at 22:22
  • 2
    \$\begingroup\$ I don't know if this is related, but your 'on' gate current is really low compared to the 100mA minimum required to trigger the thyristors for sure at room temperature (150mA in the cold). More like 15 or 20mA. \$\endgroup\$ Commented Feb 22 at 23:36
  • 3
    \$\begingroup\$ 1. Check that the SCR's are turning off after a discharge! If there's still charging going on, they might not commutate off. The SCR's could not dissipate the heat from the power supply trying to charge, while still conducting. 2. Also, this might be voltage related, not current related. Yes, they're rated for 1.6 kV, and there should only be 430 VDC across them, but check the actual voltage across them, anode-to-cathode and gate-to-cathode. \$\endgroup\$ Commented Feb 23 at 3:42
  • 1
    \$\begingroup\$ @Antonio51 cooling doesn't matter for ~ms pulses; heat has barely left the die and entered the backplate at that point. (Rule-of-thumb time constants: junction/channel ~µs, die ~100s µs, backplate ~ms, heatsink/PCB ~s.) The pulse duration and (apparent) peak current are within I_FSM (8ms) so it should be within ratings as claimed. \$\endgroup\$ Commented Feb 23 at 14:45

3 Answers 3

4
\$\begingroup\$

While this isn't a conclusive answer (and, it doesn't look like there is a conclusive answer given current information), here are some things to consider:

  1. As raised by Spehro in a comment, trigger current is quite low.

The datasheet suggests 150mA maximum trigger current. Mind the meaning of "maximum" here: this is the maximum value of least current required to trigger. That is: maybe it will trigger with less current, but maybe it will fail to sometimes: perhaps a given device behaves normally for the most part, but fails to operate on a cold day; or the design works for the most part, but sometimes you receive marginal devices that take more current to trigger, and there is production fallout during testing which is attributable to this as the cause. (Not that you're doing production, but engineers using this datasheet are.)

Often, parameters are specified in min-max and max-min ways, so that we can apply more than enough current to perform a function, or design a function to ignore some voltage, etc. In this case, we want to apply more than 100mA (at 25°C, I assume the case here) to ensure triggering.

Moreover, we can read a typical trigger current from the tgt parameter: they tested it at 300mA. Finally, anything up to 2.5A is acceptable, at least if we don't leave it on too long (there is a power limit as well). For pulsed application, more doesn't hurt; I might choose ~500mA here.

How much trigger current are you delivering?

PC817B is a phototransistor type optoisolator, having a CTR (current gain) of about 1-2.

You're applying, I believe 3.3V from the ESP32, through 220Ω, to a 1.2V typ. LED. The pin output-high resistance is typically 70Ω (ballpark CMOS figure), so we have If = (3.3V - 1.2V) / (290Ω) = 7.2mA. We expect up to about 15mA output current.

With a 4.5V supply, even if the phototransistor saturates hard 'on', and at a typical ~1.5V gate 'on' voltage, the 220Ω gate resistor limits current to 13.6mA. Somewhere between these two, will be your actual gate current: it can be lower, but not much lower (maybe 10mA, given the minimum CTR of the 'B' binned parts).

Note that phototransistors also switch slowly; it will take a few microseconds for the transistor to turn on. Compared to tgt, that's relatively long, and the SCR may be turning on before gate current reaches its peak. Maybe this doesn't matter much, but I would like to see it a bit faster, anyway.

A simple way to increase gate current is to use an emitter follower, such as 2N4401, after the opto. Add a pull-down resistor (B to GND) to sink opto leakage current.

  1. Respect the ratings.

Given the description, measurements, and comments, it seems ratings are generally obeyed:

  • Voltage: 450V is well below the min. 1200V rating.
  • Surge current: 770A is below the 930A ITSM rating. It is also shorter, at about 1ms, versus the 10ms half-sine pulse specified. You could visually extrapolate Fig.5 or 6 to shorter time scales, and figure the surge rating is well in excess of what's needed here -- not that this extrapolation is justified, and there are several reasons not to, but suffice it to say if it can handle more current at 10ms, less current at 1ms is not a problem.
  • dI/dt: if the coil is 241µH as claimed (i.e., in situ, linear, and as fired), we expect a current rate of (450V) / (241µH) = 1.87 A/µs, well below the 150 rating.
  • dV/dt: this is normally a concern when breaking an inductive load, or in active rectifier service; evidently the system is reasonably well damped, leaving no energy by the zero crossing (or, really no zero crossing at all), so we expect neither reversal nor rise of voltage. And indeed, the waveform doesn't show any obvious ringing after the discharge.

We note the system is well damped, as C = 1mF, L = 241µH, and a ballpark ESR figure for the capacitor is 100mΩ. Added to the coil's 500mΩ and we get 600mΩ total. In comparison, the resonant impedance \$Z_0 = \sqrt{\frac{L}{C}}\$ is 491mΩ, just a little lower, or a Q factor of 0.82. (A Q factor around 1 is considered well damped for practical purposes.*)

*Technically speaking, Q = 0.5 is critically damped, lower being "overdamped", and higher being "underdamped". For practical purposes, the difference is small, i.e. the oscillation or overshoot is minimal here; I use '"well" damped' in the "good enough" sense here.)

  1. Speculative causes.

While you seem to be triggering them successfully despite the meager gate current, it may be that triggering isn't complete, and better reliability would be had with full drive current. This does seem unlikely, given that dI/dt is quite low compared to ratings; the device should have plenty of time to allow current to spread out across the device as it turns on (and on and on, as current rises relatively gradually).

Turn-on might also not matter, or only hardly. Starting from 450V, quite a bit of charge is stored in the junction (they don't give a figure unfortunately, but I would expect something this size to have some 100s pF, maybe a few nF), and this provides an initial kick, a source of charges to make the whole device suddenly conductive. Particularly with certain types of SCRs, which use their own charge as a booster to enhance turn-on.

You can read about some types of SCRs, as well as the ratings, and how to operate them, here:
Philips Thyristors 1974

While the exact type in question won't be present in this or contemporary books (and this part likely uses improvements that weren't known back then), I think older databooks are a valuable source of information about fundamental device types. Particularly those before such devices were common knowledge, and before simulation and modeling tools took over -- notice all calculations are offered either in simple algebraic formulas, or as graphs or nomograms where values can be read out directly. All the hard work and calculation has been done for you.

You mention failures often on charging, rather than operation. It's not clear whether this is a red herring (a device might fail shorted during a shot, and, well, how would you know? :) ), but it sounds suspicious enough to raise a final point:

  1. Be careful of accidental turn-on.

This can be reduced with a G-K pull-down resistor, which sinks leakage current from the anode, through the gate, to the cathode. Otherwise, current flows into the gate, where it gets multiplied: this will be low enough not to cause accidental trigger up to rated voltage and temperature, but it will make it more sensitive, especially to dV/dt (current flowing through junction capacitance into the gate).

Considering the array of switches in your charging circuit, it may be that, either due to residual charge on one or more capacitors in the bank, or due to the charger itself (you aren't closing the switches when it's already powered, right..?), you are generating a sharp dV/dt that is causing accidental turn-on; and, when all capacitors are arrayed into one poor SCR that's accidentally fired, well, that's probably not going to be pretty. (The larger capacitance gives a somewhat longer pulse, though we have plenty of time to spare; the current, however, goes up proportionally, and we don't have nearly as much headroom there.)

Mechanical contacts are not to be underestimated: in the last most minute distance until the surfaces finally touch each other -- mere nanometers -- the resistance decreases from difficult-to-measure values (GΩ?), to mere milliohms. If there's enough voltage across the gap to ionize air (30V or so), a spark jumps before physical contact is made, and this has a similarly fast risetime (fractional nanoseconds). It will take much more time for the current to rise up to capacitor short-circuit levels (fractional to several kA), but the electromagnetic pulse emitted from contact closure can induce voltages in nearby wiring -- importantly, where you thought your SCRs were neatly in parallel, now some voltage is induced along the ground, and between charging paths, and accidental turn-on can occur potentially quite easily.

It's not clear how much residual voltage you are seeing typically, or what it gets up to by the time failure occurs, but the parallel connection can potentially mean not much voltage is required, to exceed the surge current rating.

One more comment, to expand on failure mode a bit: it could also be that a device is partially damaged somehow. This is quite unlikely, I think, for SCRs, at least outside of overvoltage (where avalanche can occur at a point within the die), but suppose for sake of argument, enough heating occurred that a hotspot formed, and the device got partially melted there. Typically the breakdown voltage will be lower as a result, and in a local area where it's now even more susceptible to failure. It could be that a damaged device fires suddenly at below the rated voltage, and takes itself out during the parallel connection.

The simplest solution for the charging circuit would seem to be, adding resistors in series with each switch: a few hundred ohms perhaps, enough that the voltage drop during charging is small, but large enough to greatly reduce inrush current on closing the switches.

\$\endgroup\$
7
  • \$\begingroup\$ Could a induced current (picture of oscilloscope I added) that is induced while multiple coils are triggered sequentially be somehow damaging thyristors? \$\endgroup\$
    – lukakopa
    Commented Feb 24 at 0:01
  • \$\begingroup\$ Perhaps. An RC snubber across the SCR might be added to reduce dV/dt, but if you aren't seeing them all fire at once, it's probably not that. \$\endgroup\$ Commented Feb 24 at 0:15
  • \$\begingroup\$ Oh ok. And thank you for deep analize nd help. Anything else that I could check, what can be shorting thyristors? Could it be just bad quality thyristors? \$\endgroup\$
    – lukakopa
    Commented Feb 24 at 0:28
  • \$\begingroup\$ @lukakopa Current isn’t induced. Voltage is. \$\endgroup\$
    – winny
    Commented Feb 24 at 9:45
  • \$\begingroup\$ @winny Ops sorry. I menat voltage. \$\endgroup\$
    – lukakopa
    Commented Feb 24 at 9:51
2
\$\begingroup\$

Here is what you should have as waveforms with T50RIA120 SCR.
Don't have the 70tps12 or 16 in my database components.

The wiring of a diode (D1 to top L1, curves in RED) does not "change" the oscillograms (excepted for the "remaining" voltage on capacitor ...).
The current pulse (i(L1)) is "longer".

enter image description here

enter image description here

Note that SCR peak power is as high as ~ 400 W ...

\$\endgroup\$
5
  • \$\begingroup\$ Why you didn't include flyback diode? \$\endgroup\$
    – lukakopa
    Commented Feb 23 at 10:39
  • \$\begingroup\$ Sorry, I forget it (?), I will insert it and update waveforms. \$\endgroup\$
    – Antonio51
    Commented Feb 23 at 12:51
  • \$\begingroup\$ Waveforms on oscilioscope are very similar to your simulation just there is small bump before because of induced current \$\endgroup\$
    – lukakopa
    Commented Feb 23 at 19:36
  • \$\begingroup\$ Ok. The Serie resistor (my R1) should be a "bit" lower, something as 0.1 Ohm. The current should then be a sinusoid arch. \$\endgroup\$
    – Antonio51
    Commented Feb 23 at 20:55
  • \$\begingroup\$ As a precaution, a resistor between G-K should also be used ... \$\endgroup\$
    – Antonio51
    Commented Feb 24 at 10:05
2
\$\begingroup\$

If one of the SCRs were to fire somehow while those charging switches were closed, the coil and SCR would effectively discharge the capacitance of ALL the capacitors. 6x more than normal. That could easily blow something up.

You could try adding diodes to the charger circuit and see if that prevents the problem. You should really have those anyway. 1V of drop hardly matters at 450V.

Other than that? Maybe overvoltage. Have you checked the charge voltage with more than 1 tool?

Because it fails during charging or idle, I think it must be overvoltage or a false trigger.

I'd also consider trying a different battery for each stage. The combination of all those isolated gate drives looks suspicious to me. I suspect that parastics would cause all sorts of transient voltages to occur in that circuit once one of the stages fires.

I've built a lot of coilguns, and one thing I've noticed is that the very steep DI/DT curves make parasitic inductance very important.

You should definitely simulate this in LT spice. You will learn a lot. Add an appropriate series inductor in the model wherever a wire longer than 1cm exists, to account for the parasitic inductance.

\$\endgroup\$
3
  • \$\begingroup\$ I think overvoltage (more than 1,2 kV) would blow up capacitors before thyristors would fail. Currently I don't have diodes with rating that would suit carging circuit so I will try the idea that @TimWilliams suggested in one of the answers to put pull down resistor(1Mohm) bettween G and K of thyristors because now I am almost certain that is accidental trigger what cause thyristors to blow up. Thank you for all help. \$\endgroup\$
    – lukakopa
    Commented Feb 24 at 9:48
  • \$\begingroup\$ Didn't solve the problem. I was testing ONLY first stage and after couple of launches the first thyristor was shorted. So I think its not connected to accidental trigger. \$\endgroup\$
    – lukakopa
    Commented Feb 24 at 10:27
  • \$\begingroup\$ I was thinking a pull-down more around 100 ohm, in combination with boosted gate drive. 1M won't do much, but even 10k might be enough to help rule it out. \$\endgroup\$ Commented Feb 24 at 13:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.