[dorkbotpdx-blabber] ARRGGHH - Please check this circuit

[Paul Stoffregen]

I would definitely use that 7805 to power the microcontroller from the 
12 volt line.

Most those power supplies have only 1 transformer, and usually 1 
feedback loop, so both the 5 and 12 volt lines will sag when you draw 
massive current.  Hopefully the 12 volts will stay above 7 volts so the 
7805 keeps working.

If it sags under the 7805's dropout voltage momentarily, you can use a 
schottky diode and big capacitor to hold the voltage up (don't forget 
the schottky diode loses about 0.4 volts).  Because the AVR doesn't draw 
much current, a reasonably sized capacitor should work.  The equation is 
C = (I * dT) / dV, where I is the processor+regulator current in amps 
(probably about 0.03), dT is the amount of time the capacitor has to 
provide the current, and dV is the amount of drop in voltage during that 
time.  If you want to hold up for a long time with little drop, plan on 
a giant capacitor!

It's also quite possible the power supply is keeping the 5 volts up, at 
its own circuit board, but you're dropping voltage due to the resistance 
in the wires (and any connectors).  If you're not using every single red 
and black wire, I'd trying hooking them all up to get as low a 
resistance as possible.

Many of the ATX supplies have one special 3.3 volt line that senses the 
3.3 volts at the motherboard, and all the other 3.3 volt lines actually 
deliver the current.  Of course, the same transformer is delivering the 
5 volts.... so if you draw heavily on the 5 volt line and not the 3.3 
volt line, the power supply might not be increasing its PWM to the 
transformer as much as it would if that sense line were connected.  This 
gets into a lot of guesswork about what's actually inside that power 
supply, but if you can figure out which line is the 3.3 volt sense (if 
any), you might really improve things by connecting it to a resistor 
divider from the 5 volt line.  For example, maybe 110 ohms from 5 volts 
to the sense, and 220 ohms from sense to ground.

Also worth considering might be loading the unused voltages.  Most ATX 
supplies have ordinary diodes feeding the output capacitors, not 
synchronous rectification.  So you can get an extra 0.7 drop on one line 
but not the others, and the feedback network takes the average (making 
the heavily loaded line 0.35 too low and the unused line 0.35 too 
high).  So if you can draw enough current on the "unused" lines to get a 
diode drop about 0.7 volts too (even 50 mA will help a lot), the 
feedback network should respond trying to keep them all, on average, 
around the right voltage.  Then again, some power supplies have 
resistors inside that do this already to some degree.

I would not recommend opening the power supply up and directly fiddling 
with its feedback loop.  Do as I say, not as I do...  :-)


-Paul




Hans Lindauer wrote:
> Yes, R3 is really .56 ohms.  I'll try checking the rail voltage as I 
> increase the PWM - that's a good idea.  I have a friend who's going to 
> let me borrow a LabView DAQ so I can try to figure out better what's 
> going on, but the DVM should give me some idea.
>
> One of the reasons I wanted to explore using 12V, was so that I could 
> place a 7805 regulator between 12V and the logic portion of the board, 
> with the idea that it would help to isolate it further while providing 
> lots more overhead.
>
> It seems like the power sags more when I strobe the LEDs to full power 
> than when I leave them on, so maybe the PSU has too slow of a 
> response?  I can definitely see the Dorkboards' power LEDs go dim when 
> the LEDs flash.  To address Mykle's concern, I am pretty much right at 
> the upper limit of current draw for this PSU.