some comments on linear power supplies: was : RE: [sdiy] Power Supply Design Questions

[Czech Martin]

Some comments: (all from my silly mistakes in the past)

-All transformers I know spec the RMS voltage. 
Assuming sinusoidal shape, the peak voltage will be SQRT(2)
higher, i.e. factor 1.4142...

-230V installations in Germany (and also in the EU)
should not have more than 10% voltage variations (?),
see EN 50160

-110V installations (e.g. in the USA) are said to possibly
have much more, but I do not know this

-local power generators (open air) can have much more, too,
but I do not know this

-the diodes will steal some of that voltage, 0.6V up to 1.0V,
depending on diode and current. Depending on the type of rectifier
you have one diode drop (half wave or center tap) or two
(full bridge). Just look at the current path.

-the diodes must be able to take the peak voltage (2x transformer
peak, if they not
conduct), and the peak current, and even short circuit current

-one could think that a very large power supply capacitor
(infinitely large) would be best. This is not the case!
A very large cap will look like a short for the poor transformer.
So a very high current will flow.  The ugly thing is that a
real large cap will allow charging only during the very peak
of the rectified sine wave, so the large charge current
will only flow during a short moment, i.e. very high peak.
This can damage the cap and also may be the transformer due
to losses (heat).
I have made the experience that such high peak currents
can actually loosen the coil wires due to the mechanical
force on them, so a pretty much acoustical quiet trafo
can be audibly humming after such abuse.

-a too low capacitor value will of course give too low valleys
in the waveform, so you must choose a too high transformer
output voltage in order to not go too low in the valleys.
More losses.

-depending on the core size, wire gauge, torroid or not,
transformers will have a voltage drop under load conditions.
Only few catalogues spec this.

-there are special caps for power supplies, these have a high volume,
good contacts (screw on) and they can stand the substantial heat
that will be generated due to the hum current in the capacitors
parasitic resistance. And , errr, they have a relief vent flap,
so to speak, so if the thing gets too hot, it will not explode like
a grenade (this can really happen, if it gets into your face
you are likely to be blind after the chemicals and splinters
hit your eye)

-if you want your surplus electrolytic caps (why do we always end up
with a large bag of those?), or you do not want to pay for the
real stuff, you can use multiple caps in parallel.
I have the impression that this will pretty much cure the hum current/
heat problem, however, if one of them is faulty, it will draw
all the current and explode, too

-use caps with sufficient voltage rating. Perhaps 20% higher than
the maximum peak voltage (keeping line/load voltage variations in mind)
Otherwise you might get a life time reduction, especially if large
heat will be produced

-do not use caps in series. If this can not be avoided,
add a leakage path (voltage divider) to the point in between, to make
sure that this point will really have half of the voltage, not less
or more.

-old electrolytics can not be used at once. They must be prepared,
using a 100k -1Meg series resistor in order to rebuild the isolation
film on the aluminum. You will notice "kaputt" capacitors by the simple fact
that the voltage then will not come up. Dispose them environmentally correct.
Using old caps at once will explode in your face

-choosing the right transformer and the right filter cap is not
so easy, the problems get bigger if the output power gets higher.
A wall wart is certainly more "fool proof" in comparison to a
1000W system. I think in the old "Tietze-Schenk" editions there
is some chapter about this. Perhaps I look also in the "Arts
of electronics" and wrap it up.

-anyhow, keeping all the tolerances (line, load, actual winding,
core saturation) in mind, and also the fact that
the cap must not be too large, there will be a lot of hum voltage
on the "DC" side under full power conditions. Therefore the average
voltage must be quite high in order to prevent drop outs.
Hence a lot of power dissipation in linear supplies, i.e. unwanted heat

-the idea of a lot of distributed regulators is really good.
Given that they have enough capacitance on their output, 
this should really decouple all circuits of the system.
In cheaper designs this is done by RC low passes, which of course
will spoil DC accuracy and power rejection. 
Note that high capacitance after regulator
gives no hum current heat problem. On start up the regulator
over current sensor should work, on shut down you may need a beefy
diode in order to redirect the discharge current around the regulator.
Most app notes show this.  

-in order to not introduce too much heat into the system, it is perhaps
wise to run the regulators at the minimum allowed drop out voltage
plus some safety. Keeping the wide voltage variations in mind
this calls for a big pre regulator, i.e. +-17V for a +-15 V system.
So all the tolerance is eaten up in the pre regulator.
In that way a lot of the heat will be kept outside.

-in a +-15V system and with 2V dropout only 13% of the heat will come
from the local regulator, the rest will come from the circuit in this way.
Other packages than TO3 or TO220 (?) can be used, 100mA regulator
rating may be fine. But I didn't test the residual noise of these
little buggers yet.
If we don't have a pre regulator, we may need perhaps 5V or 6 V of
dropout, so ~ 33% of the power will come from the local regulator!
This ratio will be much more worse for a 5V supply, of course.

-a real beefy supply (perhaps with toroid trafo) will have a large
start up current. This will be due to completely discharged caps
and to the fact that no magnetization is in the core, so this 
will look like a short until it has come up. Surprisingly even expensive
gear will show this problem, i.e. will blow or trigger your local
installation fuse. The cure is a start up circuit
for current limiting. Note: the mechanical force during startup
can damage trafo coils and leave them humming.
http://www.geocities.com/SoHo/Museum/4459/circuits/powsup.html
may help. 

-this allows for a sharper selected (slow) fuse on the primary
side, giving some more safety. 
note: all fuses I know are built for AC! Do not use a fuse in DC circuits
unless you know exactly what you are doing.

-the switch should be always on the primary side, nearly all switches
can only be operated with AC for larger currents. Overdimension.
Use a snubber network to suppress transients

-overdimension all parts of the supply. Go only to 80% of the
allowed rating. Life time will be enhanced a lot in this way.

-last one: after you have spent so much time on your diy equipment,
you might not want it to be destroyed due to power transients.
A circuit like
http://www.geocities.com/SoHo/Museum/4459/circuits/powsup.html
gives some hope that this will never happen.
It will also prevent RF frequencies to enter or to leave your 
circuits. Today I would like to have spark gaps between N and PE
and L1 and PE also.

-toying arround with mains power can cost your life! You should exacly
know what you are doing, especially grounding regulations (code).


m.c.