FWIW department:
Some observations noted over the last 40+ years designing, building, and
testing power supplies:
Determining the required ratings for power supply rectifiers, no matter
if they are vacuum tubes, mercury vapor, selenium, or silicon, can be
easily done using the information presented in this paper: O.H. Schade,
"Analysis of Rectifier Operation", Proc. IRE, Vol 31, No. 7, July 1943.
This information is also reprinted in various forms in ARRL publications
and manufacturers app notes. The most comprehensive source of design
data I have come across is the "Motorola Silicon Rectifier Data Manual"
for 1980, Series A. Everything a guy needs to know about silicon
rectifiers is contained in this one book. Other manufacturers may have
similar guides but I have not run across them. Also, Motorola
discontinued this series of design data books in the 90's sometime.
Much has been said over the last few days about diode PIV and equalizing
components but the information has been, in my opinion, misleading and
sometimes incorrect.
As important or probably even more important than the PIV rating of the
diode is the "Ir" or maximum reverse current rating. In a series string
of diodes, the reverse current characteristics of the individual diodes
will be the dominant factor in determining the PIV seen across each diode.
"Ir" is not as consistent as "PIV" and is highly temperature dependent.
The reverse current is typically measured at the rated reverse bias of
the part and is indicative of the effective resistance of the part when
reversed biased at that particular voltage. If you run the part on a
curve tracer, the the reverse current (and therefore the resistance)
will change in a more or less linear fashion until the avalanche voltage
is reached. Unfortunately there is no guarantee that the reverse current
(and the resistance) at any given reverse bias will be the same for
every diode.
In fact, the spec for a 1N4007 runs from 0.05 microamp (typical) to 10
microamp (max) at 25° C junction temperature and from 1.0 microamp
(typical) to 50 microamp (max) at 100° C junction temperature.
Lets look at a string of two diodes with 1000 volts peak reverse voltage
across them. If one diode has say 1.0 microamp reverse current at 500
volts reverse bias and the other has 3.0 microamp reverse current at the
same reverse voltage the effective series resistances are 500 megohms
and 166 megohms. Assuming a linear reverse current curve, one diode will
have a voltage about 750 volts across it and the other will have about
250 volts across it. No problem if both diodes have a PIV rating of 1000
volts or more but you can see what happens if the PIV across the diode
string is actually 1500 volts instead of 1000 or if there is a wider
variation in "Ir" characteristics..
Recommendations for eliminating the compensation components are based on
the presumption that the individual diodes have virtually identical
reverse current characteristics *AND* are operated with identical
junction temperatures *AND* there are far more diodes in the string than
you might actually need in order to cover up any inconsistencies.
If you can buy "lots" of diodes and they are all the same date code and
you insure they are all operating at the same junction temperature (NOT
ambient but junction) and put 3x what you really need, you can be pretty
sure that it will be reliable.
Now hams being as cheap as they are will never go to a major
semiconductor manufacturer and buy a batch of diodes with identical date
codes. They will buy them at a fraction of that cost and wind up with
many date codes from a distributer or even many manufacturers from a
"surplus" source. In this case the reverse resistance characteristics
will be all over the map and the only way to control the PIV each
individual diode sees is to swamp out the "Ir" by using a parallel
resistor across each diode with a resistance much lower than the reverse
resistance of the diode.
Historically, over the last 40 years I have never seen a diode that came
off a major manufacturers production line that didn't conform to the
above characteristics.
In the distant past, silicon diodes were expensive and manufacturers did
not want to use more than the bare minimum required. Resistors (and
capacitors) were 100 times cheaper than diodes. Made sense to use them
to reduce the number of diodes required.
The junction capacitance of a typical rectifier diode is small enough
that the impedance differences between individual diodes at 50/60 Hz is
insignificant. However, the old Motorola, GE, RCA, etc, tube type
commercial mobile radios used multivibrator DC-DC converters running at
around 2 kHz. The risetimes and transients involved in those systems
were fast enough to warrant compensating individual diode junction
capacitance with external parallel capacitors. I suspect this is where
the idea of putting capacitors across every diode in the world came
from. I have never seen a commercial/military power supply running from
50/60 Hz mains have compensating capacitors.
In those instances where high frequency transients are expected, usually
a simple capacitor across the transformers winding is enough to fix the
problem. I will frequently incorporate this cap no matter if I use
compensating resistors or not.
As far as deciding how much PIV you need, once you have "enough" to
match the design maximum PIV expected, anything more is at the
discretion of the designer as to how much he wants to trade off
reliability vs. cost. Personally, if I have the space available, I will
use about 3x the expected PIV (diodes are cheap these days!).
What is the expected PIV? It depends......
Assume a typical single phase full wave bridge rectifier running into a
capacitor filter:
Each leg of the bridge will see a reverse voltage equal to the maximum
DC output voltage. This does not include any possible transients or
surges nor any safety margin. I usually make each leg 3X this number.
Each leg of the bridge will see an average current equal to 1/2 of the
DC output current. This means a bridge made from 1 amp diodes will
handle up to 2 amps of output current assuming the peak repetitive
current rating is not exceeded and the temperature is controlled.
The transformer secondary voltage will be 0.707 times the output DC
voltage (roughly depending on many component characteristics) so
therefore the transformer secondary current rating will need to be 1.414
times the output DC current. Nobody ever gets this right!
The peak current through each diode leg will be *EIGHT* times the output
DC current! Pay attention to the peak repetitive current rating of the
diodes. They derate very fast with temperature and this IS a critical
parameter.
Junction temperature is a big factor in reliability. The heat generated
in a diode junction is due primarily to the current through the junction
and the forward voltage drop. The heat is mostly dissipated through the
LEADS and not through the epoxy body. Every major manufacturer gives
derating curves for various mounting schemes. I have never once seen
encapsulation recommended. Potting the diodes in epoxy will decrease the
allowable forward current and the maximum PIV by a significant amount
due to the inability of the diode leads to transfer heat efficiently
when embedded in an insulating material. Commercial manufacturers of
encapsulated diodes take this into account when designing the stack. If
you don't do likewise when encapsulating your own, be advised that you
will probably be disappointed in your efforts.
Although I sometime stack diodes without compensating resistors due to
space considerations, I normally use the resistors. I'm careful to use
resistors that are appropriate and that the diode/resistor strings are
mounted in a fashion which allows proper heat dissipation. Using this
approach I have had many many years of proven reliability using "bargain
basement" diodes in HV applications both in Choke input and Capacitor
input supplies.
YMMV
73, Larry
Larry - W7IUV
DN07dg
http://w7iuv.com
_______________________________________________
Amps mailing list
Amps@contesting.com
http://lists.contesting.com/mailman/listinfo/amps
|