[kriskalish]

I guess my question is summed up by the title pretty well.  I already have an LM317-based supply that I built as a kid. Long story short, it doesn't work that well and I'm going to rebuild it. It makes sense to just go for a bipolar supply since I pretty much exclusively work with audio.
 
I was thinking about building the dual rail regulated supply that's basically right out of any LM337 datasheet.  I think I found a more complete schematic on the forums here at some point in time. Here it is:

 
I didn't really know much about power supplies a few days ago until I started reading.  I've heard from multiple sources that linear regulators can't sink current.  However, it seems to me like it's one of those things that's a matter of perspective. Like, the LM317 can sink current for the LM337 and vice versa. However this only makes sense in my mind if current flows from rail to rail.
 
A more likely scenario in a headphone amplifier is that current will flow from rail to ground.  I can't image how the electrons would flow such that the complementary regulator sinks current.
 
In short, how much current can this ground sink and why?

 

[Avro_Arrow]

Seeing as the "ground" (more accurately 0 volt) path is all passive (just a wire or trace),
then the amount of current it can handle is determined by it's impedance.
In practical terms, we try and keep the ground path a short as possible and use
a star configuration if practical.

 

[fubar3]

The fact that the 317 is not a current sink is probably intentional since it allows some flexibity. For example, it can be configured as a current limiter to provide short circuit protection.
 
A bipolar supply for audio usually has a common return from the audio channels. Maintain low impedance to minimize cross-talk, as far as I know.

 

[kriskalish]

Thanks for the knowledge guys. 
 
Is it fair to say that I need to attach the signal ground in this power supply to the real earth ground then? I just don't see where excited electrons (current) would go after travelling through my headphones otherwise.  It seems to me that the ground in this schematic is a "dead end" and its potential would be modulated by any amount of current going into it.
 
Sorry to bother you guys with such basic questions, but Wikipedia and Google are kind of failing me right now.

 

[jcx]

current always flows in closed loops - the sums around the loop(s) always "add up" - otherwise you'd get charges "piling up" somewhere making a huge V
 
in a AC power supply the current loops are a little complex - to simplify just consider the ps caps as batteries that get topped up by the rectified xmfr secondary every 1/120 s
 
no more "gnd current" ever flows than is sourced from the Vregs, ps Caps - some people do have ontological problems with "DC" current flowing through caps - but it works for this simplified analysis
 
 
if you have separate secondary windings it is now common to use 2 bridges, and the same regulator circuit duplicated for the neg supply - 317 perform better than 337

 

[kriskalish]

 
Quote:

current always flows in closed loops - the sums around the loop(s) always "add up" - otherwise you'd get charges "piling up" somewhere making a huge V
 
in a AC power supply the current loops are a little complex - to simplify just consider the ps caps as batteries that get topped up by the rectified xmfr secondary every 1/120 s
 
no more "gnd current" ever flows than is sourced from the Vregs, ps Caps - some people do have ontological problems with "DC" current flowing through caps - but it works for this simplified analysis

 
Thanks! This actually makes a lot of sense; treating the caps as batteries did the trick for me.
 
 
 
Quote:

if you have separate secondary windings it is now common to use 2 bridges, and the same regulator circuit duplicated for the neg supply - 317 perform better than 337

 
Unfortunately I'm trying to build this supply on a shoestring budget so I'm using an old Bando transformer I pulled out of an amplifier in high school.  It's actually quite huge and has three secondary windings, but unfortunately only one of the secondaries is useful to me. It's a 66VAC winding with a center tap.  The remaining two secondaries are about 20 VAC which seems too low to be usable. It looks like with proper bypass caps and such that the 337 performs close enough to its complement. It's not really meant to be an ideal permanent supply for anything; just something that is convenient to use for prototyping but also doesn't wreck my sound quality.
 
Speaking of sound quality, is it really worth keeping R1 and R2 (both 2.7ohms) from the original design? It looks like they just form RC filters with C6 and C5 with a corner frequency of about 59hz.  Isn't this only like -9dB of attenuation at 120hz (the frequency of rectifier ripple)? That seems like a marginal payoff to me.
 
-Kris

 

[jcx]

if you only need up to ~ +/-24 V then 2x 22 VAC could be fine - peak is 1.414 * 22 ~=30  - 1.5 (2 diode drops) - 2 (reg dropout)
 
2 v ripple from 2000uF 120 Hz full bridge ~= 480 mA
 
the R between the 2 caps adds some Vdrop, doesn't cut ripple V usefully but greatly improves high frequency line rejection - the Vreg chips line rejection starts rolling off at only a few kHz