@#$%^&*!!!! rebote!!!!

[R O Tiree]

I might be way off base here, but I've found some interesting info on failure modes of 78XX regulators.

One of the subtle problems I read about has to do with the caps either side of the chip, connected from V_IN to GND and the other(s) from V_OUT to GND.  If the one connected to the OUT pin is bigger than the one connected to the IN pin then, when you disconnect the power, the V_OUT cap takes longer to discharge which effectively reverses the voltage across the regulator and kills it.  The schem you posted to has the IN cap at 47µF and a 100µF cap connected from pin1 of the 2399 (5VDC) to GND.  The Echo Base schem has those cap values the other way around - 100 at V_IN and 47 at V_OUT.  Now, if your V_IN cap has failed, you could be getting huge current spikes the wrong way through the 7805 when you power down.  That will kill the reg very quickly, according to the web page I just read.

Another site pointed (possibly) towards a dodgy ground connection to the reg causing problems... V_OUT too high and reg getting hotter than it should.  It was a bit difficult to follow that thread, because it seemed that most of the contributors didn't have English as their first language... maybe I got the wrong end of the stick, but it seemed that the guy's problem was solved re-flowing the GND connection.

There's always the old chestnut about getting the pinouts wrong - the TO220 package pinout is "reversed" compared with the diddy little TO92 one.

Finally, the thermal protection system in this IC is less than stellar.  It seems that, once it starts to shut down, the voltage at the output pin drops.  This actually increases the voltage across the chip and, in extreme circumstances, can actually cause the chip to continue to get even hotter.  Once the chip gets heat-damaged internally, it's not unresonable to find the voltage climbing again... to destruction (and taking out your 2399 along the way).

If it were mine, I'd think about bread-boarding a brand-new 7805 with a dummy load on the V_OUT pin (say 10k), with a V_IN cap of 100µF and a V_OUT cap of just 0.1µF and see that I actually got a stable 5V out of it and minimal heating.  I'd test continuity on every trace with no ICs installed, just to check no shorts or dodgy connections increasing the current draw.  I'd replace all the electro caps on spec, making the V_IN cap 100µF and the V_OUT cap 22µF or so.  Next, I'd pop the 7805 in there and check that the V_OUT was a stable 5V with minimal heating - do a soak test for a good half-hour or so, to be absolutely certain.  Try it with the PSU as well.  Once I was absolutely positive that the supply voltage was to spec and stable, I'd think about popping the 2399 in and seeing what happens.

[Jdansti]

I might be way off base here, but I've found some interesting info on failure modes of 78XX regulators.

One of the subtle problems I read about has to do with the caps either side of the chip, connected from VIN to GND and the other(s) from VOUT to GND.  If the one connected to the OUT pin is bigger than the one connected to the IN pin then, when you disconnect the power, the VOUT cap takes longer to discharge which effectively reverses the voltage across the regulator and kills it.  The schem you posted to has the IN cap at 47µF and a 100µF cap connected from pin1 of the 2399 (5VDC) to GND.  The Echo Base schem has those cap values the other way around - 100 at VIN and 47 at VOUT.  Now, if your VIN cap has failed, you could be getting huge current spikes the wrong way through the 7805 when you power down.  That will kill the reg very quickly, according to the web page I just read.

Right. The cap can discharge back to the regulator even if the cap is OK. Here's the recommended way to protect the regulator.

[pinkjimiphoton]

I might be way off base here, but I've found some interesting info on failure modes of 78XX regulators.

One of the subtle problems I read about has to do with the caps either side of the chip, connected from V_IN to GND and the other(s) from V_OUT to GND.  If the one connected to the OUT pin is bigger than the one connected to the IN pin then, when you disconnect the power, the V_OUT cap takes longer to discharge which effectively reverses the voltage across the regulator and kills it.  The schem you posted to has the IN cap at 47µF and a 100µF cap connected from pin1 of the 2399 (5VDC) to GND.  The Echo Base schem has those cap values the other way around - 100 at V_IN and 47 at V_OUT.  Now, if your V_IN cap has failed, you could be getting huge current spikes the wrong way through the 7805 when you power down.  That will kill the reg very quickly, according to the web page I just read.

Another site pointed (possibly) towards a dodgy ground connection to the reg causing problems... V_OUT too high and reg getting hotter than it should.  It was a bit difficult to follow that thread, because it seemed that most of the contributors didn't have English as their first language... maybe I got the wrong end of the stick, but it seemed that the guy's problem was solved re-flowing the GND connection.

There's always the old chestnut about getting the pinouts wrong - the TO220 package pinout is "reversed" compared with the diddy little TO92 one.

Finally, the thermal protection system in this IC is less than stellar.  It seems that, once it starts to shut down, the voltage at the output pin drops.  This actually increases the voltage across the chip and, in extreme circumstances, can actually cause the chip to continue to get even hotter.  Once the chip gets heat-damaged internally, it's not unresonable to find the voltage climbing again... to destruction (and taking out your 2399 along the way).

If it were mine, I'd think about bread-boarding a brand-new 7805 with a dummy load on the V_OUT pin (say 10k), with a V_IN cap of 100µF and a V_OUT cap of just 0.1µF and see that I actually got a stable 5V out of it and minimal heating.  I'd test continuity on every trace with no ICs installed, just to check no shorts or dodgy connections increasing the current draw.  I'd replace all the electro caps on spec, making the V_IN cap 100µF and the V_OUT cap 22µF or so.  Next, I'd pop the 7805 in there and check that the V_OUT was a stable 5V with minimal heating - do a soak test for a good half-hour or so, to be absolutely certain.  Try it with the PSU as well.  Once I was absolutely positive that the supply voltage was to spec and stable, I'd think about popping the 2399 in and seeing what happens.

dude, i believe you nailed it...some of the stuff you're describing is exactly what seemed to happen last nite, right down to the negative voltage across the regulator!!! i thought i was seeing things.. i gotta digest this post a little more when i have a little less beer in my system and no family visiting.

THANK YOU!! I THINK YOU CRACKED THE PROBLEM!!!!! 

[oldschoolanalog]

Good call guys.
I have always put the larger cap before the reg. The thought process being it is best to have the unregulated DC as smooth as possible before it reaches the reg. Now I realize how many designs do not do this.
Case in point: http://www.generalguitargadgets.com/pdf/ggg_pt80_sc.pdf
A very popular and much built design (also, the MAX 1044 doesn't supply enough current for this circuit; but that's another story....).
The "insurance" diode is good form and for the few cents it costs is well worth installing.
@ jimi: Yay beer!

[Jdansti]

I'd still put a couple of diodes in as extra insurance. 

[pinkjimiphoton]

can't wait til tomorrow to try and implement the "fix"...thanks!!

man... beer is good. here's to you and you and you.... 

[R O Tiree]

I'd strap a diode across it as well, as jdansti and oldschoolanalog advise - I read about that on several forums, but they were less than clear about how to do it, so I didn't mention it.  Props to jdansti who found a picture that painted a thousand of my words

I think, perhaps, that the reason you see so many designs for the PT2399 with huge caps from pin 1 to GND is that the (appalling) datasheet shows these high values in that position.  This is possibly because not all 5V supplies are sufficiently "stiff", stable and ripple-free and Princeton Tech are hedging their bets.  It might be sufficient to just go with a 10µF tant and a 0.1µF metal film in parallel there, with something like a 47µF at the V_IN terminal of the reg... and strap a protection diode in there as well.

[Ronan]

Just like to add my 2c to to all the brilliant info above.
I find it interesting that, at work, the items I most commonly repair/service are the ones that I least read the documentation for. It's like "familiarity breeds contempt". Same with "common" 5V regs, they are used on such a huge scale, for maybe 20 years, and few people bother to check the documentation anymore. Here is part of a Fairchild doc from Aug 2012, very recent. It is resonably clear that the power supply filter cap (the big one) should be placed before the reg, and just a 0.1uF after the reg to "improve stability and transient response". And, if the main filter cap is a fair distance away, as in the case of using a wall-wart, then place a 0.33uF cap immediately before the reg.

[Jdansti]

Good to know. Sometimes I just look at the pretty pictures. 

[Ronan]

John, the picture you posted is fine, and shows an extra diode protection measure. Often the datasheets will specify the minimum cap values to make the product work reliably, since bigger caps generally cost more money, and the regulator manufacturers want to make their products look economical for consumer electronics manufacturers. The regulator manufacturer assumes you know what size filter cap is required BEFORE the regulator.  R O Tiree's point about the PT2399 datasheet calling for a large cap on the 5V supply is a good one, maybe they were thinking a zener-regulated circuit would be used.

[pinkjimiphoton]

guys, dumb question, but in john's picture, the band (cathode) of the diode on the left is facing away from the power... is that right?
i thought you had to have the cathode go to +?

just wanna check, before i screw this up MORE...

i'm thinking 220uf across the power jack with the diode, and changing the first cap on the board to 100uf, and the one off pin 1 down to 10uf or so...

should i leave the other values the same when i replace them? do they all have to be bigger than the one at pin 1 to ground?

[Ronan]

You could leave that diode out jimi, put the 220uF across the power jack, and leave the pcb alone except for removing the 100uF cap altogether. Leave the stock 47uF in circuit. That's my take on it, with the proviso that the 0.1uF cap connected to pin 1 of the PT2399 is physically close to pin 3 of the 5V regulator, i.e. within say an inch of pcb track, otherwise I would fit another 0.1uF right at pin 3 of the reg to ground. 

[oldschoolanalog]

That diode is not for polarity protection. You already have that @ your DC in jack. It's there to protect the reg in case the output cap discharges. The current will flow through the diode instead of into the output of the reg.

> i'm thinking 220uf across the power jack with the diode, and changing the first cap on the board to 100uf, and the one off pin 1 down to 10uf or so...<
As long as the total capacatance before the reg exceeds the total @ the regs output. A .01uF ceramic disc in parallel w/the output cap is also a good idea. Helps filter high frequency hash & trash.
Happy solder sniffin!  

[Ronan]

Just to be clear, there's 2 diodes in that pic. I believe the diode on the left is not strictly needed, its for input polarity protection, that's the one I said to leave out. The diode on the right is to protect the reg, and it would be a good idea to fit that one.

[pinkjimiphoton]

thanks guys, but i think the part i need to know is if the orientation of the diodes, as shown, is right... as IIRC, the cathode (band) needs to face the + on the input side? i'm just trying to be cautious, cuz the two things that always seem to be backwards are the cathode end of diodes, and the pins on pots in layouts and diagrams.

[oldschoolanalog]

This is 100% correct. 

[pinkjimiphoton]

thanks dave, what threw me was (for some reason) that the diode was in series with b+, not between b+ and earth.

brian dead. i need to smoke a couple more...beers. 

[Ronan]

jimi, this might answer your question, about the cathode end of the diode being connected to power. There's 2 basic protection methods, illustrated below.

[pinkjimiphoton]

gotcha...thanks ian, i understand it now. i usually do it the first way, tho the second way may work better for the rebote problem.