Bleeding it dry

Shocking!

Cor blimey I nearly blew my head off. Shorting capacitors, particularly large ones is a very scary experience. I had removed the rectification board from the mains board of the power supply project. Here’s the board:

The burn mark where I shorted a fully-charged cap

The burn mark where I shorted a fully-charged cap


Mains rectification and filtering board, note the humungoid caps!

Mains rectification and filtering board, note the humungoid caps!

I have noted rather absent mindedly in the past how I should remember to add some bleeder resistors to make the board safe when I power it off so I don’t kill myself when I try to make modifications to it. Well the short of it was “I was just about to”. Last time I used the power supply, I was only using the positive side of the symmetrical supply to test my pre-regulator circuit. When I was finished, I connected to a 30Ω power resistor to discharge the capacitors. Thinking it safe, I gingerly removed the power board just now and thinking “oh I’ll just be double safe” I shorted one of the caps with my screwdriver. BANG.

How on earth did this happen? Didn’t I discharge the cap last time I used it?

The answer is no. I did drain the caps from the positive side, but I never touched the negative side and both were still fully charged! My extra diligence and paranoia just saved my life. Had I shorted that with my finger I would probably be dead as a door nail now.

Best practices (or how not kill yourself by being stupid)

It always pays to be doubly, triply and quadruply sure. Those who own firearms will always tell you to always assume it is loaded and treat it gingerly and with appropriate respect for a deadly weapon. The same applies to electronics. If you know you are working with high voltages, high currents and/or have large capacitors present, always assume they are charged and treat them carefully.

Keep fingers and metal objects well away from exposed contacts, PCB traces, or anything vaguely conductive. I would go a step further and say do not touch it at all unless you are 100% sure it is not plugged in and all capacitors are completely discharged.

I was going on the assumption that these caps were indeed discharged, but just to be safe I would short the terminals using an insulated screwdriver just to be doubly sure. I was not expecting a big spark and a bang, but I’m alive, unharmed and very glad of my cautious instincts.

Let that be a lesson to us all.

Bleeding it dry

I did mention that I drained 2 of four capacitors last time I used it using a 30Ω power resistor. This is common and accepted practice for discharging capacitors safely. Shorting the terminals with a screwdriver is also commonly done but isn’t the safest thing in the world and anyone with any experience will tell you to use a resistor and measure the voltage across the terminals to be sure it’s drained. Using a screwdriver or other conductive object, since the resistance is very very low, will make a dramatic discharge, spark and bang which could scare you into touching something you shouldn’t, start a fire, damage the screwdriver (it will weld metal) and if the tool is not insulated or you are accidentally touching a conductive part – kill you anyway.

This is why there are warnings on the back of every piece of equipment you own telling you emphatically not to open it. All power supplies have capacitors in them, and even if measures are taken to discharge them on power-down, those could fail so it’s best to keep inexperienced hands out.

Going with the safer option, we can manually use a power resistor, one that can handle the power going through it, to safely and steadily discharge our capacitors. Well you may ask, why not just build the resistor into the design so that when it’s powered off the caps are automatically discharged? This is, in fact, done all the time and are known as “bleeder resistors”.

Though it would be great to whack in a 30Ω power resistor and quickly discharge them (about half a second) so that it’s already safe by the time you have the case open. The difficulty with this is it really dissipates a lot of power. If you take my power supply as an example, it’s output voltage is 42.6V each side (±42.6V) so each capacitor will be charged up to the rail voltage. Using Ohm’s law that means our 30Ω resistor will draw 820mA and thus burn 20.2W! That’s a bit hot and dramatic and we’d need a resistor rated 25W or higher so it could do this consistently without melting on us.

As you probably already guessed, since these would be permanently installed, that means we’re pissing away a combined 40.4W (for both sides) just doing nothing! Unacceptable that.

How fast, how long?

The trick is to find a balance. A high enough resistor value so it doesn’t draw a lot of current under normal operation, nor dissipate a lot of heat warranting a beefy power resistor, but low enough in value that it discharges the cap as quickly as possible.

One could figure this out off the top of your head. The discharge time of the capacitor can be roughly figured as T = R*C (charge/discharge time is roughly equivalent to the resistance times the capacitance in the RC circuit). In my case, each side of my power supply has two capacitors in parallel which add up to 11,500µF of capacitance (crazy I know, but the output is so smooth!). If I choose, say, a 1kΩ resistor, then 1000*0.0115=11.5 seconds. That’s pretty good! Count 12 steamboats and we’re safe!

Using the same math above, 42.6V over 1000Ω is 42.6mA of current, but however is 1.8W which still puts it in the the category of power resistor. I like having a 20% margin of error at least in anything I design, so even a 2W resistor is pushing it a bit. Though it would probably be fine on it’s own, it’s good to keep in mind that it will be inside a case and around other components that are hot or are heat sensitive. Fortunately, I have a pair of 1kΩ 5W power resistors in my bin which I will doubtless use.

Conclusions

This is pretty good actually. I can discharge my caps in 11.5 seconds, and only draw 85.2mA of current or 3.6W combined. Some of the more experienced designers reading this will laugh at me pissing away this much power, but it is after all a linear power supply and will be horribly wasteful anyway. I’m building this for functionality and safety over efficiency and sanity. In the end, the quicker I can discharge those caps, the better.

Also, here’s a lovely online calculator to assist hobbyists in calculating there bleeder resistor values. I noted when I plugged my values in that I’m discharging 10.43J of energy! wow.

Update

So okay, I soldered in the two 1kΩ resistors. There wasn’t any room on the top side of the board so I mounted them on the bottom. Figured since they were touching the metal plate that holds the whole thing together it would make a great heatsink for them and prevent them from heating up the caps on the top. See here:

Soldered the 1kΩ resistors on the underside

Soldered the 1kΩ resistors on the underside


Here's another angle, note i filed out the pads in between the rails for some quick isolation slots

Here’s another angle, note i filed out the pads in between the rails for some quick isolation slots

On the whole it works! Just one caveat: it doesn’t drain them in 11.5 seconds as above. This is something I always screw up with capacitor charging/discharging calculations. Apparently, the formula T=RC is used to figure out the time constant of the capacitor. It works out to about 68% (if memory serves) of the total charge of the capacitor. Also, the charge/discharge curves are not linear but exponential, so it’s kind of like imagining half lives of radionuclides. The full charge/dischange is actually 5 x R x C. So 12*5 is 60 seconds which according to empirical measurements is how long it takes to discharge my caps so that there is less than a volt potential on them (i.e. the point where it couldn’t hurt a fly). So I can assume my equipment is safe to work on after one minute therabouts.

It was good I tried the 1kΩ option, I had contemplated using 10kΩ for less power consumption, but that would make it 115s to be 68% discharged, and a whole 9.5 MINUTES for it to almost totally discharge. That’s a bit too long. I’m happy with my solution.

Comments are closed.