Weird Power Generator?

[Silesius]

Hi everybody,
I've just finished a bipolar power supply using Ray's design. The one with the wallwart adaptor. Well, this one:



I've built it on a stripboard, and it works well... sometimes
When I plug it, it gives me +-12v, but not always. Sometimes I plug it and the modules don't work. Then I take the multimeter and I find -12v and only 0.65v on the positive side. But then I plug it again and it works. I can try three times and I can't get it to work; but then, suddenly, it works again: two, three, four times... Until it stops working.
I can't understand what's going on here. I've checked everything. I've used the exact parts. The schematic it's very simple and I think it's unlikely I've made a mistake. I've looked at it several times. But I don't know the theory behind this, so maybe there's something I'm overlooking.
I'm using an AC adaptor taken from a JoeMeek compressor: it gives 12vac and 1000mA, so I think this is not the problem.
I am really lost here... any ideas??

[Silesius]

Well, I've tried another AC adaptor, the one from my Doepfer eurorack. This one is 16vac and 2.5A. And I've blown one of these [C9]:


I must confess I've doubted a little with these caps. It was the first time I use it. I assumed the side with the + is the positive. Am I wrong? Is it possible I have all the 1uF tantalums reversed? Would the circuit work that way?

[JingleJoe]

In my experience polarized capacitors backwards don't conduct, untill you reach their reverse breakdown voltage because they have the effect of a diode along with the standard capacitor effect. The voltage of 0.65 volts is the diode voltage so I'd say you were seeing the voltage accross that cap or there's something up with one of your diodes, could be in backwards or something.

I allways find Mr Wilson's diagrams really really confusing, for example, this one has two ground rails in it and I find that maddeningly unhelpful because you can't easily comprehend the common point of the circuit. And whats going on with all those diodes? Are they supposed to be a bridge rectifier or what?

Capacitors in parallel just add up so he's got about 10mF (milli farads) of capacitance there. Thats overkill if you ask me, but it all depends how much current you are drawing.
Why not buy two 10mF (ten milli farad) capacitors instead of all those 3300uF ones?

[Silesius]

Hey, thaks for your answer JingleJoe.
That's what it is said in the MFOS website:

"Diodes CR3, CR4 and CR5, CR6 are protection diodes recommended in National Semiconductor Application Note AN-182 for protecting three terminal regulators from unwanted current paths that the output capacitors can inadvertently cause under certain conditions. Although the ones for the negative regulator are not considered as critical, better safe than sorry I always say. The 1uF tantalums close to the regulator pins are also recommended by National Semiconductor for operational stability".

Anyway, how is it possible one of the caps exploded if they are not reversed?
Also, I've checked everything, and my +12 output is directly connected to the 7812 output pin. There's nothing in between. Caps, diodes and regulators are in place, nothing reversed as far as I can see [and there's not much to see...]. The stripboard is clean, with plenty of space, and I can't find any shorts, so, what else can I do? thanks!

[Silesius]

Well, I found another weird thing: using the multimeter, I find the 7812 input "sometimes" is shorted with ground. I mean when I touch it with the multimeter, the multimeter beeps, because it finds a short to ground. But if I keep it there, the beep stops, and there's no short anymore. Then I try again and the meter say it's OK. But if I wait some seconds and I try again, the meter beeps. Sorry for the childish explanation, but that's the best I can do with my knowledge of electronics [and english language ]

[alieneYe]

if your multi beeps at all when testing if the regulator input and ground are crossed, then they probably are even with the intermittent "beeping"

so it sounds like you've got live voltage and ground crossed somewhere. i would check your traces on your board very closely with a magnifying glass.

this has happened to me in the past, and has lead to blowing up plenty of caps.

that schematic has only one ground plane. even tho there are two ground symbols on it, they still represent one ground plane, and that is how it will translate to the board

the 3300uf caps are "smoothing" caps, in theory, the more you use, the cleaner you power supply will be... if i am recalling correctly, early here.

good luck with your troubleshooting

[jordroid]

It looks like you have it almost working, so it's probably a quick fix away.

When you probe a node with a capacitor to ground, like the input to your regulators, it is normal for it to show a short for a brief period, and then a quickly rising impedance, this is caused by your multimeter actually charging up the cap. If the short lasts for about a second or so and then goes away you are probably ok.


Lots of people have had tantalum caps blow up on them, they seem to be more susceptible than normal electrolytics. If it was me i would first quadruple check that all the diodes are in the right way and not shorted, and then test it without that cap there, while monitoring it very closely for a while. If everything seems ok you could instal a new capacitor at the output, and then watch it like a hawk during the first power up.

The capacitors don't need to be tantalum for it to work, you can use electrolytics or whatever you can find easily. They don't need to be exactly 1uF either, anything around there will probably work, i see anything from 0.1uF to 10uF in that position in similar power supplies.

Let us know how it goes,

regards,

jordan

:edit: regarding jingle joe's question about many parallel caps instead of fewer large ones, one reason might be to get a lower ESR (equivalent series resistance), with several caps in parallel the capacitance will add up while the series resistance will go down, due to the nature of resistance in parallel. Another reason might be that smaller caps are smaller, cheaper and easier to find than huge single caps. i dunno, the power supply has worked well for many people though. cheers

[Silesius]

Thanks a lot for your help. I've been testing it without two of the tantalums [the one I blowed and its counterpart]. It does the same. I get always -12v, but the positive side sometimes it's OK, sometimes just gives me 0.65v.
I've been trying with the multimeter and I've noticed another thing when I unplug the adaptor from mains.
When it's working and you unplug it you see the +12v slowly going down. I think that's normal. But when I get only 0.65v and I unplug it the voltage rises quickly to +12, stays there for a couple of seconds, and then goes down again. I think that's not normal Could it be the regulator maybe?

[Skrog Productions]

Hi there , silly question , are the regulators bolted to the same bit of metal / heatsink , that could cause problems without mica insulators.
good luck.
Dave.

[Silesius]

No, no. I made two aluminum heatsinks, but they are not near each other. Besides, I'm testing it again without them and it does the same weird things. So I have a board with only 16 components, and still I can't see what's going on. I'm using 18 traces in the stripboard, and there's no cuts. So it's easy to see if there are shorts using the multimeter. I guess it has to be something wrong with some component. I think I'm going to change the regulators and see what happens. But I will have to order them
Or maybe I build it again from scratch. I took me less than an hour, and I've spent a lot more with the troubleshooting, so maybe I start again.
Its very strange... I can't understand why, when you have 0.65v and unplug it, it rises to 12v. So, it gives you what you want, but only when you turn it off .

[Osal]

Do you have connected the regulators correctly? Positive and negative regulators have different pin configuration. If it was this the regulator is burnt.
positive==== input-ground-output
negative====ground-input-output

By the way, you don't need C4 and C9 if the filter capacitors are close to the regulators. If tantalum C5, C10 give you problems, you can replace them by 22uF electrolytic. [/i]

[Osal]

[prgdeltablues]

I have the same problem (occasional failure of the positive power rail to work properly) with my test power supply, based on the same circuit - but using 78L12/79L12 100mA regulators, no heatsinks, much less smoothing capacitance. Unplugging and reinserting the wall-wart connector seems to solve it, but annoying.

Looks like we are not alone: http://www.groupdiy.com/forums.html?topic=440.0
and also see:
http://wwwd.national.com/national/PowerMB.nsf/88bc465ea04c97a58525620b0040cc4a/FB19F9D1B2E1D9B888256DE2003DB534?OpenDocument

found a reference in a book Mechatronics by Singh and Joshi (p 73) - seems like this latch-up with dual (+ and - lines) regulator setups is a known problem, caused by the negative regulator being turned on first, which somehow latches the positive rail to one diode drop. Supposed to be solved by the two diodes on the output side, but clearly isn't. some suggestions that different makes of regulator may be better than others.




Peter

[Silesius]

Well, it seems we have the answer !!!
I've been reading the links you posted, Prgdeltablues. Thanks a lot for the info. There's people in there describing my exact problem. But they explain it much better than me... [and I was saying I had 0.65v while actually I was having -0.65v].
They call it a "latch up" problem, so it even has a name Apparently, it can be solved with another brand of 7812, so I'm going to order some of them.
Apart from that, it seems I can use it without burning anything in the meantime, and that's great.
Thanks everybody; lots of nice people here!

[JingleJoe]

There's why your tantalum blew too; the -0.65 of a volt was probably just a little too much for it.

If it's due to the negative regulator activating before the positive, then I'm thinking; give it a delay by using a monostable and a relay.

[Silesius]

Just to let you know, and give a happy ending to the thread, I can confirm it works now with another 7812. No more latch up problems. On, off, on, off, always fine

[tazer]

Could you tell us the brand of 7812 you used to solve the problem please? as i also have this problem.

[elmegil]

I have an amusing story from a friend of mine about tantalums:

For my Sr. project, we built a generic I/O board (serial, parallel and timer chips—both Intel and Motorola—on an external board, interfaced via ISA bus to a PC). Since power came via ribbon cable, we used 10µF tantalum caps at the connector to do bulk power supply filtering, and 0.1µF film caps at each chip.

Our tantalum caps were marked oddly. Neither our prof nor our lab director knew quite what to make of the marking:

http://spatula-city.org/~im14u2c/images/tantalum_cap.png

Notice that the + sign is in the exact center, and both leads are equal length. My lab partner and I, along with my prof and the lab director all scratched our collective heads at this. My prof guessed that maybe the line indicated which lead was plus—that it was leading from the plus sign to the plus lead. So, we went with that.

We had around 8 or 10 of these tantalum caps on our board, since we needed to filter out both the +5v and the +12v supply rails. (Yep, we had some +12v stuff on there too. We planned to locally re-regulate the +12v down to some lower voltage to drive some ADC/DAC parts. We never got there.)

So, anyway, I wirewrapped all this stuff up, with the tantalum caps soldered to wire-wrap posts. I had my board hooked up to the 286 PC we were using for testing, and I decided to run some tests on whatever was the latest chip we had wired up. (I think it was the Intel 8255 parallel I/O chip.) I fired it up, did some tests, and shut it down.

After that brief test, my lab partner came to me and told me he'd run some of the spare tantalum caps through the capacitance tester. The cap tester was polarized, naturally, and he said he saw a major difference in capacitance reading depending on which way he put in the cap. He got the correct reading (10µF) if he put them in the opposite way from how we had ours wired.

"No biggie," we think. We decided to unsolder all the tantalum caps on the board, and just solder them back in the correct way. No worries, right?

We resoldered the caps in the right orientation by the weekend. That same weekend, I decided to test the board some more by running stress tests on the parallel I/O chip and the other chips. I had the board component-side down, so I could hook the oscilloscope up to some signals-of-interest. Once I had my stress test code cooking away on the 286, I went to get a soda.

I walked down the hall, bought myself a soda from the IEEE fridge, and walked back. Just as I walked back into the room, I hear

*pop* *crack* *shooooooooooooooosh*

as a bright red-purple flame erupted under my board, and deep purple smoke churned out. It was roughly the same color as the cap itself (drawn above). And boy did it smell awful.

I ran up, shut off the power supply and opened the window. Thankfully, our lab station was right next to the window, and so I was able to air the place out reasonably quickly.

Given that we had 8 or 10 of these caps on board, I thought to myself, "I must've missed one!" Uh huh.

So, I cut the failed cap out of the circuit while admiring how the top side of the board was now very black. There was a matching burn mark on the wooden lab bench. Nice.

I flipped the board back over, hooked my probes back up, and fired everything back up again. A few minutes later...

*pop* *crack* *shoooooooooooosh* *pop* *crack* *shooooooooooooooosh*

as two more caps give up the ghost in rapid succession.

Apparently, that short power-up I had given them while they were all wired backwards had basically killed them all, and even though they were now wired correctly, they were all ticking time bombs. Very colorful, acrid time bombs.

----------------

So that reverse voltage has probably made any/all tantalums on the board suspect....

[richardc64]

[tazer]

the 7812's i'm using have TS7812 on the body with a big S on it not sure what brand that is.

If anyone know's of a more reliable brand it would be great to know.

thanks

ian

[Silesius]

The same here: the working 7812 I use now has L7812CV written on it. I found it in an old bag after searching for a while, so I don't know the brand.
And I can't seem to find the glitchy one, sorry . My working area is a complete mess. Well, at least we know the cause of the problem; that's something...