Thyristor burning up!

[AngraMelo]

Hello friend,
I just build the circuit bellow. My issue is with the crowbar circuit (made beautifuly in paint) I already burned up 2 thyristor and I have no clue what is happening.
The Thyristor parameters should be waaay within specs for this project. It burns up (either goes short, or open forever) whenever the supply is on for around 3 min or so.
By probing TestPoint, I get a constant varying voltage of 1mV up to 17mV, it just goes up and down.
Is my zener bad?
I have no clue what is going on

Thyristor Datasheet: http://www.ween-semi.com/documents/BT152_SERIES.pdf

[james_s]

What current is your PSU capable of putting out? With multiple 2N3055's that implies quite a lot, that SCR is only capable of 13A max continuous current and that's going to be with a very good heatsink. In order for a crowbar circuit to be effective the PSU needs to be fused such that the crowbar can blow the fuse.

[AngraMelo]

It is configured to give 23A max. The thyristor is burning in the off state. Or so I think. The zener does not start conducting until around 16.5V.
It is fused aswell, as it shows in the schematic

[AngraMelo]

I just checked and you are right, this is not the right thyristor for this application, but that still does not answer why it is going short or open, given that the output voltage cant get over 15V given the resistor configuration of the 723. Thats what is bugging me

[T3sl4co1l]

Output voltage can get well over 15V for the fraction of a microsecond required to turn on the SCR.

Have you used an oscilloscope to perform a transient (step load) test?

Tim

[AngraMelo]

How stupid of me! No I havent! I will do that,
if that is the issue, how do we solve it?

[T3sl4co1l]

I don't see any compensation in the circuit.  Maybe it doesn't need it, but probably there is a way to apply it.  Read the datasheet or appnotes, they should give some ideas.

Tim

[german77]

A capacitor should be used in parallel with the resistor. This will delay the time it takes to the SCR to turn on. And a fuse is needed to prevent the SCR from failing.

[AngraMelo]

I don't see any compensation in the circuit.  Maybe it doesn't need it, but probably there is a way to apply it.  Read the datasheet or appnotes, they should give some ideas.

Tim
[/quote
Isnt the compensation between pin 13 and 4 of the LM723? Or do you mean something else?

A capacitor should be used in parallel with the resistor. This will delay the time it takes to the SCR to turn on. And a fuse is needed to prevent the SCR from failing.

The fuse is the F2 (25A)

[Ian.M]

However a 13A Thyristor is very unlikely to blow a 25A fuse.   It probably needs at *LEAST* a 30A thyristor, and its I²t rating needs to be significantly greater than the fuse's.  Also it needs either foldback current limiting, or thermal protection for the output transistors to avoid them burning up if the thyristor trips for any reason other than a shorted output transistor, as the current limiting as diagrammed will vastly delay or even prevent the fuse blowing, but the output transistors will be operating far in excess of their normal dissipation.

[schmitt trigger]

Last but not least...... all crowbar circuits must have a sacrificial element, i.e a fuse.

If you replace the SCR with -let's just say- a 1 kiloamp device, then your pass transistors will now burn up unless there is a fuse.


But the best is to prevent false triggering. Definitively something to slow the pulse down.

EDIT: I have seen that you indeed have a fuse. But it is 25 amps.

[T3sl4co1l]

Finally, note that the transistors protect the fuse, not the other way around.

The crowbar fires because the transistors have already failed, or are about to, and the only thing the SCR does is blow the fuse faster while keeping output voltage lower (hopefully).

Why would you need that?

The crowbar is ONLY relevant, if you are powering a bunch of circuitry that absolutely cannot be subjected to excessive voltage, lest it cost tens of thousands of dollars in servicing to fix.

You saw crowbars often in classic and mainframe computing.  It's rarely used today.  With everything consumable and unserviceable* today, such a failure results in needing new equipment anyway, so why bother?

*That is, a single failure is basically BER (beyond economic repair).  You can if you want, but you'll pay almost as much for a replacement board as a new unit.  Or you can do component-level replacement, but now you need orders of magnitude more knowledge and capability, which isn't cheap either.  (If you can amortize that over many fairly-quick repairs, you get something like what Louis Rossmann does, but keep in mind also that most of his repairs are stop-gap, no-guarantee, "it runs again but you want to backup your files and move them onto something newer" sort of thing.  As-new repairs are possible, but that doesn't mean you want to do them.  Corroded traces and burned up PCB can be repaired, but it is truly a heroic process!

Tim

[AngraMelo]

However a 13A Thyristor is very unlikely to blow a 25A fuse.   It probably needs at *LEAST* a 30A thyristor, and its I²t rating needs to be significantly greater than the fuse's.  Also it needs either foldback current limiting, or thermal protection for the output transistors to avoid them burning up if the thyristor trips for any reason other than a shorted output transistor, as the current limiting as diagrammed will vastly delay or even prevent the fuse blowing, but the output transistors will be operating far in excess of their normal dissipation.

You are right my friend, the thyristor, as pointed by other friends is insuficient, although it still does not explain what is triggering it.
I already located a suitable and obtainable thyristor the BTA41-600 seems like a suitable candidate

datasheet https://www.st.com/resource/en/datasheet/cd00002263.pdf

Also, could please direct me to some reading about foldback current limiting? Im a beginner and Ive never heard of it

Finally, note that the transistors protect the fuse, not the other way around.

The crowbar fires because the transistors have already failed, or are about to, and the only thing the SCR does is blow the fuse faster while keeping output voltage lower (hopefully).

Why would you need that?

The crowbar is ONLY relevant, if you are powering a bunch of circuitry that absolutely cannot be subjected to excessive voltage, lest it cost tens of thousands of dollars in servicing to fix.

This PSU is to power my ham equipment, nothing insanely expensive but a couple of 3055's and a fried 723 would be a waaaay better case.
This whole thing is to prevent my ham rig going bust given that there will be several things connected to the PS.
The main issue is that I live in Brazil, servicing a power supply is faaar better than ham rig. Insanely expensive to import parts or buy another rig. Do you have another suggestion for how to protect the equipment? If I were to buy a good power supply with all necessary safety features instead of building my own, I looking at spending around 6x more than building this one.

[AngraMelo]

Ok, so I did the capture of the TestPoint on the schematic by hooking up this project to another PSU.
The first test was with no capacitor to slow down the thyristor.
With a load that draws 3.5A attached to the project I captured a ~200mV transient going to the TestPoint (see picture)
With no load I get around half of it.
Besides that, there is no relevante voltage at that point after the power up process (constant 5mV up to 20mV)

Now for the interesting part, I tried using a capacitor (47uF) to slow down the voltage at that point, the only thing it did was introduce noise that would surppass the 100mV trigger point that I need to capture the transient. So the scope would trigger even without me turning on the PS that was powering the project.

As I was limiting the current of the PS that powered the whole experiment, I believe that when the transformer is the one providing (plus the other around 15000uF of capacitance waiting for me already mounted at the enclosure) the transient would be even worse.

Now, how on earth do I stop that voltage peak from frying everything?

[german77]

Now for the interesting part, I tried using a capacitor (47uF) to slow down the voltage at that point, the only thing it did was introduce noise that would surppass the 100mV trigger point that I need to capture the transient. So the scope would trigger even without me turning on the PS that was powering the project.

The capacitor added noise? That's kind of unexpected. But there is another approach. Use a low pass filter.  My example will ignore those fast transients without adding noise to the test point. But the turn on voltage will be the zener + 0.4v from the transistor. Try to use the smallest capacitor posible and a lower zener value to get the same turn on voltage and a short delay.

[AngraMelo]

German,

Thank you! I will try that as soon as I get home! It looks promissing! When you talk about the smallest possible capacitor, are you saying something like 10pf? or 1uf?

thanks again

[David Hess]

The zener becomes low impedance when it conducts so the effect of the capacitor on the response time is compromised.  This configuration is only good for suppressing RF.  More commonly a resistor is placed in series with the gate which both limits the gate current and allows the gate bypass capacitor to have a more significant effect.

Once the SCR triggers, the continuous current and forward voltage drop of the SCR can still result in considerable power dissipation.  I would at least configure the 723 for fold-back current limiting to help with this.

[Kartika]

you can use crowbar activate before drive collector 3055 fuse  no emitter side...
 

[Benta]

Probably solved 4 years ago...