SPD3303X Catastrophic Fault (Design Flaw?) - (Edit: Nope! User Error!)

[christopher.robot]

Hey folks - looking for something of a sanity check here...

I recently had to teardown & repair my SPD3303X because of what appears to be an absurd design oversight.

I was charging a large lithium ion battery (yes, through a BMS & active balancer) when we had a brownout in my building. 

(Before someone shouts at me - it wasn't just banal charging, I was performing charging tests which utilized some features of the PSU).

When the PSU re-powered it went through an internal relay cycling which immediately flooded the room with the stench of conflagrated FR4.

Upon inspection, the supply had shorted the inputs on power-up (which seems an intended part of the process) and turned two PCB traces into fusible links and took out a triac (UB19, a BTA08-600C).

I've never had any other bench PSU do something like this.  That includes 4 Agilents, 2 Tenmas, and a host of other cheapie uprights (I'm not saying it couldn't happen, I'm just saying it hasn't).

What on earth is the point of this shorting?  Some kind of self test for the relays?  Why would it not check for voltage on the output first?  Why wouldn't there be fuses somewhere?  (Crowbar circuits typically have fuses, no?)

Am I crazy?  Is this a normal implementation and I should always externally fuse or Schottky any high-current capable sources connected to the output?  Would it be silly to modify the circuit to include some chunky SMT fuses (higher than the output rating, but below "blowing traces off the board" current)?  I get that adding components to a calibrated output is usually a no-no, but some beefy fuses right at the output lugs aren't likely to change the specs, I'd wager...

This seems like madness to me, and cost me a day of repairs - and I'm frankly afraid to use this thing for anything with batteries or big capacitors connected to the output.

Forgive the somewhat scrubby repairs, I went back and cleaned them up after I verified the supply was again functional, but I didn't take any pics.

[jjoonathan]

Yeah, the "social contract" around power supplies can have a lot of caveats. One way or another, we're all familiar with "current limit doesn't account for output cap," leading to first-time EE lab students exploding LEDs everywhere  . "If anything happens to the input power, all bets are off" is another common for cheap supplies, though I'm surprised to see it from Siglent. My GPP-4323 doesn't do this, fwiw -- there's a reverse protection diode and a 500 ohm resistor, but no momentary short on boot.

My first power supply was a KKMoon that dumped its 48V intermediate voltage caps into the output if you removed power before disabling output first, so it *can* get even sillier.

[AVGresponding]

One should always use a diode when using a psu to charge batteries, to prevent this exact situation. Many psus have the capability to sink a certain amount of current from the DUT (two or four quadrant operation, basically operating as a load) and might well default to 0V after a mains failure, and with a large battery this might easily exceed the unit ratings.

[christopher.robot]

One should always use a diode when using a psu to charge batteries, to prevent this exact situation. Many psus have the capability to sink a certain amount of current from the DUT (two or four quadrant operation, basically operating as a load) and might well default to 0V after a mains failure, and with a large battery this might easily exceed the unit ratings.

Yes yes, sure sure, I could have been "more careful" - but the point is that this seems like it was easily preventable in the design.  It seems like when implementing the shorting behavior, they didn't consider or didn't care about potential failure modes.  A simple chunky SMT fuse (which I guess the PCB trace sorta is, albeit unpredictable) would suffice - though it's a PSU with output sensing - it seems a simple firmware solution to have it check for voltage on the output before shorting during power up...

[nctnico]

One should always use a diode when using a psu to charge batteries, to prevent this exact situation. Many psus have the capability to sink a certain amount of current from the DUT

I agree 100% ! Some PSUs even have a crowbar (thyristor) which shorts the output if the output voltage is higher compared to the set voltage to prevent a failure inside the PSU to cause the DUT to be destroyed. I've seen the damage that can happen if you charge a battery using such a PSU when the PSU decides it is overheating and shuts down...

Bottom line: never ever charge a battery from a lab PSU without a series diode.

[Martin72]

Hi,
Exactly that, the output must be decoupled - the power supply unit does not expect any external voltage at its outputs.
Therefore, this is a :

Catastrophic Fault

From the OP and no

(Design Flaw?)

.

The outputs are presumably short-circuited so that voltage peaks are not passed on to any loads when they are switched on.
A certain cheap power supply unit at work did not have this feature a few years ago.
The result was that when it was switched on, the maximum voltage was briefly present before it went back to the set value.
And a circuit board full of ICs that were expecting 15V was only a complete failure.

[zrq]

If I recall right, my HPAK 66332A/6632B have the same crowbar mechanism, and it makes them (although 2 quadrants capable) not really suitable as electronic loads.

[RoGeorge]

I was charging a large lithium ion battery (yes, through a BMS & active balancer) when we had a brownout in my building. 

That's a thing you should never do with a power supply.
You must add a series diode when charging batteries.

For example, in the Rigol DP832 supply manual it is explicitly told NOT to charge batteries without a series diode.  The series diode must be added to prevent the situations where the battery backfeed into the power supply output.

Many power supply also have an internal analog only "crowbar protection" for when somehow the output voltage becomes higher then the programmed voltage (to protect the load circuit against overvoltage, for example if the internal regulator fails, and start to deliver full voltage instead of programmed voltage).  The crowbar will short-circuit the outputs if it sees more voltage than expected.  Often it's triac meant in parallel with the output.

[christopher.robot]

Still seems entirely preventable from a design standpoint, with this particular implementation.

That said, the general consensus is this was a setup failure - so I suppose I must concede. 

Still, for the testing I was doing, adding a diode would have affected the logging, and an external fuse might not have prevented internal damage (though better than naught, I suppose). 

I can add a diode and a downstream DAQ, but the latter seems much more convoluted and prone to error than anything introduced by a Schottky.

Maybe I'll build some kind mains-sensing interruptor between the mains and the PSU, so it stays off if the mains get sketchy (another thing that could probably be implemented in the original design). 

Probably a lot of work for a one-off, but then again, so was the repair...

Thanks for the input everyone, (and my apologies to Siglent for the slander - [Fixed, edited the title]). 

At least I learned something!

[RoGeorge]

The good part when charging through a diode is that most often the charging current is constant, which means the voltage drop on the diode is also constant, so it can be only a subtract of constant 0.6V or so when logging, instead of an additional DAQ.  That's what I usually do while logging a charging.

In case it's not constant current charging, then you can trace first the drop voltage of the diode against current, memorize the diode's I-V characteristic, then subtract the curve of the diode during charging.

For 3A or so power supply, the diodes must be quite beefy, and mounted on a radiator.  I've uses a former CPU radiator, and two double diodes in parallel, such that I can connect 1 battery to 2 of the power supply channels at the same time, to double the charging current.  The double diodes were recovered from a broken ATX power supply for PC.



Another thing you should have is a fusible right near the battery, because a big battery can easily heat a short circuit wire to red hot, which might start a fire.  The short circuit might happen in ways you can not control or predict.  For example, once a pigeon entered the room and managed to fly-bump erratically into almost everything around.