I have a short ( i think) and don't know where to start

[Hextejas]

Progress ! The V-, left and right show -13V !

Hmmm, can I check the MUR860 while still in the circuit ? I checked the resistance of a new one before soldering and it showed a resistance of 0.

The leg with the new diodes and that yields -13v, the 4 diodes all read 0 resistance. The other leg, all the diodes show some resistance.
I think that I will replace all 4 of them.

[sokoloff]

Your circuit has two logical "halves".

(If you wanted to, you would) Connect the 9VAC wall wart across either U1 and U2 or U3 and U4.

However, given that you have a DC bench power supply with current limiting, just use that to find the short. It's not quite as useful for troubleshooting the bridge rectifier, but your problem is not in the AC rectification part of the circuit. (I asked if you had a wall wart because I bet many more people can put their hands on that than on a bench power supply. You have the bench supply. That's way more effective in short-finding.)

[Hextejas]

I think its fixed. I hooked up the full mains voltage and no smoke or fire. Success as the output voltages read correctly.
Sheesh, what a hassle but finally solved and  thanks to all the excellent help here.
Thank you again. I hope that all my projects will go smoother. 

[sokoloff]

What did the problem end up being? 

[nick_d]

To check the diodes use the diode setting on the meter. Ohms is meaningless for diodes, since current increases super-linearly with voltage, not linearly. The diode setting puts a specified current through the diode and tells you the voltage. In forward orientation it should show about 0.6V, larger for beefy power diodes (up to about 1V) or LEDs (up to about 3V). In reverse orientation it should show infinite. You should be able to do this test with them soldered into the board and no supply or load connected. If result is correct then it is a pretty good indication of diode health (though not absolutely definitive). Likely your previous issue was with damaged but not visually obviously damaged diodes.
cheers, Nick

[alanb]

There is nothing between the diodes and the caps to limit the current. The caps are very large and therefore at start up the current is going to be high. Could this be the cause of your troubles rather than a short?

[xavier60]

There is nothing between the diodes and the caps to limit the current. The caps are very large and therefore at start up the current is going to be high. Could this be the cause of your troubles rather than a short?

Also, fast recovery diodes usually don't tolerate current surges as well as standard diodes.

[Hextejas]

There is nothing between the diodes and the caps to limit the current. The caps are very large and therefore at start up the current is going to be high. Could this be the cause of your troubles rather than a short?

I still have a problem. When I wire up both sides, I blow the 3.15a fuse. I can power either side but not both.
I am wondering if I have too large a transformer for the fuse.
The spec called for this :

Avel Lindberg Y236651 250VA 18V+18V Toroidal Transformer

And I have this:
Avel Lindberg Y236803 500VA 35V+35V Toroidal Transformer

Could  I be overpowering the fuse ?
I wired a meter into the circuit to see what kind of current was being pulled and it blew a 10amp fuse in the meter.

So, how could I slow down the initial rush of current ?

[alanb]

A resistor in series would limit the current this will waste energy as heat although this may be acceptable. You could also use a thermistor this would also waste energy but using the correct value would waste less than a normal resistor as the resistance drops as it warms up.   

[james_s]

When you say "wire up" do you mean when you put a load on them? Or when you just hook something up to them? If connecting wires blows the fuse then it sounds like you're missing the insulators between one or more power transistors and the case.

[Hextejas]

When you say "wire up" do you mean when you put a load on them? Or when you just hook something up to them? If connecting wires blows the fuse then it sounds like you're missing the insulators between one or more power transistors and the case.

Yes, wired up meant to put a load to both sides.
Question. How do i go about sizing a thermistor and would i need one for each side (rail) ?

[SiliconWizard]

Could you try without C10? (Maybe you already did?)

[nick_d]

I feel that something obvious must be getting missed here.

I suggest to get a low value high wattage resistor and put it in series with the incoming AC. If all works totally fine with a moderate load on both outputs then yeah maybe I buy the theory that it's inrush current to the many large caps that causes the issue. In that case, vary the resistance to see what you need to stop it misbehaving. And test the hell out of it -- that the supplies CAN float independently of each other and that it can drive almost the designed load on both channels independently once this possible startup issue is worked around with the small resistor to limit the inrush. You might also try a slow-blow fuse if it's really the inrush issue.

I bet that won't be the case. There's gotta be another issue.

cheers, Nick

[fsr]

If connecting the load blows fuses, then maybe your load has a problem. What is your load, anyways?
Also, if you replaced a 18+18v transformer by a 35+35v, there can be problems. Over a resistive load, the current is the voltage divided by the resistance, and you doubled the voltage.
You can test the psu by connecting both grounds together. I would first connect the grounds at the transformer, with nothing else connected, just the transformer by itself, and chech that you have your 35+35v. Of course, before connecting the grounds, make sure that they're two independent secondaries, like you think. Measuring voltage over the wires of separate secondaries should give you no voltage at all. With the transformer unpowered you can also test resistance. It should be infinite. The transformer's datasheet should tell you this, anyways.
With the grounds connected at the transformer, you can connect the grounds at the output of your psu. Nothing should blow, or you still have a problem in your psu. Then you can test by using two power resistors as loads, from + to gnd and from gnd to -
Use resistors of enough power rating, and their resistance should be enough for the current to be below the limits of your psu.
I also think that using fast recovery diodes on a bridge rectifier is strange.

[AngraMelo]

This is strange. I am using a benchtop power supply ( ps ), a GPC3020,  that can put out 0-35v DC. I have it set to 20v and this is what happens.
It has 3 output connections. (---   GND   +++)
I connected the +++ to U1 and when i connected the --- to U2, the ps blinked and turned off the C.V. led and turned on the C.C. led.
I think that it means it switched from constant voltage to constant current. I don't know if that was good or bad and the implications. Anyhow, what measured 20Vdc before now measures a lot less.
I am thinking I need to do something to fix that.
Secondly, when I try the same thing with U4 +++ and U3 ----, the ps stays at CV (20v8 dc), but none of the output shows any voltage.
Let me see if I can find a 1N4148 or some such.

A typical bench power supply will supply a set voltage or a set current, whichever is less. If you set it to 20V and a max of 100mA, it will provide 20V and 0mA when not connected to anything (and will be in "constant voltage mode"). If dead shorted, it will supply 100mA and way less than 1 volt. (It will be limited the voltage to whatever can push 100mA through the short. This is "constant current mode".)

In your case, it's telling you that you have a very low resistance path (maybe a dead short) on the U1/U2 side.

I am thinking I need to do something to fix that.

Yes. Find the short. 

So it seems your board has a short, if you are having problems with 20V I doubt it is a clearance issue. It might just be a dead short.

[AngraMelo]

Wait a minute, the circuit says you should use a 18V transforme and you are using a 35V one?
Your caps are rated for more than 35V?
I have waaaaay larger capacitor banks here that never had that problem.
They do make a fun noise when turning on but they dont blow any fuses (normally a slow blow fuse is needed)

[Hextejas]

Hooray, I think that we got it !!!
I added a small ohm (2ohm) resistor, large Watt, I dont know what it is, to the schematic as shown and it seems to work ok. It's a very beefy resistor and it did not get hot after powering up. 
The power supply does not blow fuses and is putting out the proper DC voltages. 
My question now is:
1) Should I add a similar resistor to the other rail ? Does it need to be "balanced" with the other side ? Is there such a thing as balance between the 2 sides ? This electronics hobby is a strange beast.

Picture attached.

[metrologist]

I suspect that resistor is rated 2 watts, maybe 5W. Your transformer is rated 500W (input)?

[Hextejas]

I suspect that resistor is rated 2 watts, maybe 5W. Your transformer is rated 500W (input)?

Yes on the transformer.
And I spoke too soon. Alas.
I put another 2 ohm in front of AC2, plugged everything in, bolted it into the chassis, turned it on, and smoke and a gunshot.
 
Maybe adding the 2nd resistor caused it bu i am at a loss to explain it.
Soooooo , I will take out the resistor, replace the exploded diode and try again,

[sokoloff]

If you're not discharging all the caps between tests, it's possible that a previous test has left some charge in the caps and the subsequent power-on is not the same test as a power-on from a "cold" (all discharged) state.

Adding the second resistor in that location did not cause the failure.

[metrologist]

You said when you put your bench supply to U1 - U2, it went to CC mode. You need to identify what caused that. In that state, if it is not drawing too much current, because you set the current limit to a decent value, you might be able to detect something getting hot. A thermal camera is great for that, or maybe a finger...

[Hextejas]

You said when you put your bench supply to U1 - U2, it went to CC mode. You need to identify what caused that. In that state, if it is not drawing too much current, because you set the current limit to a decent value, you might be able to detect something getting hot. A thermal camera is great for that, or maybe a finger...

Excellent advice and i will give it a try.

[nick_d]

Yes you need both resistors (for the test at any rate, I am not sure I would recommend the resistors for production). Now let us think about this. A fully discharged capacitor charging from 35V through 2 ohms will take 17.5A initially, this dropping off as the capacitor charges. According to http://www.mouser.com/ds/2/149/RURP860-244175.pdf the peak non-repetitive surge current of MUR860 is 100A. So I believe you have proved that inrush current isn't the issue. You may want to check RC time constant to be sure that this counts as a "surge" but realistically I think that it would.

I think more likely you have an intermittent short or a PCB design rule issue or a reject PCB or all three. Can you make up another PCB the same and test it? And in case the halves are shorted together, maybe try running with the top half (+ve supply) from the first PCB and the bottom half (-ve supply) from the second PCB.

Otherwise, the original advice applies. Validate everything with the 9V battery and possibly small signal diodes instead of the MUR860s (but keep the MUR860s if they are already soldered in and pass multimeter tests on diode setting).

Then, if that's okay give it low voltage AC with much current limiting. What I do sometimes for this is grab a 12V AC plugpack and then use the 12VAC as if it was 240VAC. So basically you have an extra step down and then your normal transformer, result about 2V. And put some resistors in too. Then see if you actually have a reliable 2V supply over time.

cheers, Nick

[ebastler]

Picture attached.

But you did not actually install the resistor as shown in your schematic (in parallel with the wire), right? It needs to be in series, i.e. you must disconnect the original wire or trace. Apologies if this was already clear.

[nick_d]

Another thought: check your sums carefully. If you increased transformer voltage increase capacitor voltage in proportion. That's because 35VAC means 35V RMS so the peak is 1.414 times 35 in VDC, I often see 400V capacitors used in 240VAC supplies, 450V or 500V are also common. If you use say 35V rated capacitor with 35VAC input, the capacitor charges normally until the AC gets near the peak and then the capacitor begins to short as the AC peak exceeds 35VDC. It will short as long as more than the rated voltage is across it, which tends to limit the voltage stored in the capacitor to approximately its rated voltage. There may be some margin of safety but typically the margin is tiny.

Another possibility is that your capacitors could be damaged from previous experiments causing them to either short (easy to detect) or short above some voltage that is less than the rated voltage (much harder to detect). Although it would be a major pain a good test would be to try running the circuit with all capacitors removed. The low voltage test won't pick up this issue. The low current test MIGHT but the problem is fairly benign at low currents so may be missed.

cheers, Nick