1
\$\begingroup\$

I would like to use 2 SCRs in antiparallel configuration to control the power to a all resistive load (20A) using 240Vac (line to Line). Pretty typical configuration and part of a lot of SCRs datasheets. The application has a ON/OFF switch that controls in part if the SCRs are ON (using zero crossing firing)

The problem to overcome is the failure on an SCR. If an SCR fails short you lose control to the load. So I have added a second set of SCRs for redundant control. One set controls the "Line side" of the load and the "Return-side", such that if one SCR fails the other set of SCRs would allow control of the power.

Yet the problem with that, is the fact that, if an SCR fails short, the load will have power for half a cycle on one end of it, while the other end is effectively disconnected. That causes a potential corner case safety failure if a person "turned OFF" the switch but the load has a potential on one of the legs.

My question is how can you create a FAIL safe SCR. A fuse will not work as the current will be within the limits. My device needs to work at -40C.

\$\endgroup\$

3 Answers 3

1
\$\begingroup\$

You can't make something totally fail-safe. However, you can make failure less likely, and you can make that failure less problematic.

First, make it less problematic. If you have two switches, one on each side of the load, then a failure of one could leave the load connected to the 'wrong' mains pole. That can be mitigated by putting the two SCRs in series in your 'higher reliability' switch, and putting those in the pole you want to switch.

If the failure of any given switch is independent of that of the others (and this may be true for some mechanisms, like a comsic ray strike, and not true for others, like an overvoltage spike, overtemperature, overcurrent or old age), then putting multiple units in series will make failure less likely. If one switch has a failure in 1 in 1000 operations, then 2 switches will be 1 in 1 million, and 3 switches will be i in 10^9.

As you see from my list of what failures would be random, and what related, it's probably better to protect the SCRs from fault conditions, than to simply series up a number of them.

\$\endgroup\$
2
  • \$\begingroup\$ I am protecting against over current, ESD, filtering any voltage spikes and lightning, the load is pretty consistent as well. Nevertheless in failure analysis a short SCR is also possible for some reason maybe aging, cosmic rays. Perhaps the best way to have a fail safe SCR controlled circuit would be to add a crow bar to blow a fuse in case of SCR deteced failure \$\endgroup\$
    – DIODEX
    Feb 12, 2017 at 17:57
  • \$\begingroup\$ BY the way, by Fail-Safe I mean that it fails in a safe manner (always open). Not that it just won't fail. \$\endgroup\$
    – DIODEX
    Feb 12, 2017 at 17:59
1
\$\begingroup\$

Protecting against single point failure in the SCR's you have to separate them.

Perhaps something such as this:

schematic

simulate this circuit – Schematic created using CircuitLab

Note: This only protects against (provides redundancy) an SCR going short circuit, assuming that this is the failure mode you see most often. I'd assume this is true if you have an intial current surge in your resistive load (as it warms up).

\$\endgroup\$
6
  • \$\begingroup\$ The diodes do not provide the isolation to a short. Also notice I updated the description, as instead of being "Live" to "Neutral" the AC connection is "Line to Line" allowing one end of the load to be connected to a "Live" wire in case of a short, unless I use 4 pairs of back to back SCRs. \$\endgroup\$
    – DIODEX
    Feb 12, 2017 at 17:53
  • \$\begingroup\$ The diodes allow you to use two SCR's instead of four and are required from that perspective. If either SCR goes short the diodes still conduct of course. Line to line makes no difference in the answer I posted. The Diodes can't protect against a short, they simply allow the problem to be separated into two parts using single SCR's in a redundant configuration instead of antiparallel pairs. This configuration allows a single SCR short failure to be handled. If an SCR fails open it's handled by default. \$\endgroup\$ Feb 12, 2017 at 19:38
  • \$\begingroup\$ Interesting design, I had seen this configuration with IGBTs. The biggest draw back would be that each diode will have about 1.2V with 4 of them it is 4.8V@20A seems as I would be wasting over 96W of power (plus SCRs). My current design uses 2 pairs of anti-parallel SCRs with the load in the middle, in an attempt to remove power from both ends of the load for service. If you picture the load being remote, the bottom leg on your figure would always be at line potential, unsafe for someone to touch. \$\endgroup\$
    – DIODEX
    Feb 13, 2017 at 4:16
  • \$\begingroup\$ Yup, the power dissipation may be a problem, though there are power diodes with Vf down at 1.05 V ...such as BR-E35/50 media.digikey.com/pdf/Data%20Sheets/Comchip%20PDFs/… ....still 80 W is a bunch of heat. Is there a reason you can't use 2 serial Triacs? \$\endgroup\$ Feb 13, 2017 at 6:42
  • \$\begingroup\$ it goes back to the original problem that if a TRIAC fails short then one side of the load would be left connected to a "hot" line since the load is connected line to line \$\endgroup\$
    – DIODEX
    Feb 14, 2017 at 15:36
1
\$\begingroup\$

I would suggest a series mechanical relay (or, more appropriately at high currents, a contactor) as a backup to the SCRs.

You will not get the voltage drop and associated power dissipation you would get from an SCR, and properly applied contactors very rarely fail closed.

Keep in mind that a momentary short applied to series SCRs will subject both series SCRs to the same current surge and is likely to cause similar damage (for example, causing them to fail short) to both. They will also likely be subjected to the same temperature, another likely cause of failure. You should also detect when one fails and take appropriate action, for reasons that should be obvious.

A frequently applied principle in safety systems is to use different types of devices to get a higher level of safety. For example, a press may have an electrical interlock, a hydraulic and a mechanical bar that stops motion. All three would have to fail (or be deliberately bypassed) in order to cause injury to the operator.

\$\endgroup\$

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.