Jim:
Thanks for the response.
Wow, with cap banks like you describe, there's some significant energy
there... What are you working on, a pulsed cyclotron?
Let's see, given your 10 banks with 200 caps each, that's 2000 caps and only
1 failure every three years, your failure rate is nearly insignificant. With
a failure rate that low and a cycle rate you described, I'm supposing you are
charging them with nearly pure DC, but without knowing the application, it's
hard to guess... Of course, 200 caps in parallel mean leakage of about
800ma, so you've got to have a pretty good supply charging those caps.
A bank of 90 caps at 16000uf each looks like 1.4 farads. That's an
incredible amount of storage at first look. The leakage alone on a bank like
that would be about 400ma. To charge them slowly, I'd be interested in
knowing the charge time, and what is being used to limit the charge current.
Is it an active supply with a ramped current, or as simple as really high
wattage limiting resistors?
In the linear amplifier, there are really a couple of things that tend to
kill the caps. First, the ripple current tends to heat the caps, which are
in a relatively warm location to begin with, sitting right next to a fairly
large transformer, and a full wave bridge just above them. The transmitting
tube is on the "other side of the wall" so to speak, but I'm sure it does
contribute to some heat in the area. I haven't had this particular amplifier
up and running yet, so can't instrument it to find out what the temperatures
are, but I wouldn't be surprised to find it at 50C or higher, as the cooling
air is all directed into the tube chimney and no particular attention has
been made to cooling the capacitor bank. Given the ripple current, I wouldn't
be surprised to see 80C, which is dangerously near the rated temperature
rating (at least for me). Next, we have the typical cycle service an
amateur amplifier puts these through, with a nearly uncontrolled charge
(limited somewhat by a series resistor in the primary supply for the first
second or so on an initial power up), and the demands as the amplifier is
used. Now, given that the capacitor bank is never fully discharged, and the
voltage across the bank is 3600 volts, the transformer rated at only 0.8
amperes, and the amount of current required during typical transmission, the
heating from ripple current is probably the biggest concern for the initial
design.
However, given that I live in an area with less than perfect regulation of
the incoming AC line, I tend to be a bit gun-shy about what I see personally
as a "marginal" design. At my last QTH, the line voltage per our local
Hawaiian Electric Light Company (HELCO) should have been between 230 and 240
volts. I was the last house on the end of a 1 mile run of utility poles, and
the transformer, which was over 300 feet from the building, served only my
house and workshop at the time. In my workshop, I ran a single phase to 3
phase solid state inverter for a 3 horsepower wood lathe. During a period of
two years, even with rather sporadic use after the first year, I had four
inverters fail, and the failure was always the input diode bridge. After the
first failure, I instrumented the AC line and kept graphs on it. The line
voltage ranged from 235 to 243 volts over the typical 24 hour period, and
tended to be around 239 volts much of the time, which I see as the nominal
voltage for that location. Unfortunately, there were times when the voltage
would rise to as much as 255 volts for brief periods of time (usually only a
few seconds, but one prolonged instance went over 15 seconds). I didn't
actually catch any high line spikes during operation of my lathe, but there
were times I forgot to turn off the inverter and would come back and find it
dead in the water a couple of days later. I finally eliminated the problem
by putting a buck/boost transformer in the circuit, reducing the nominal
voltage to 227 volts. Now there is a definite possibility the particular
inverter input bridge wasn't properly rated, or wasn't up to spec, but all
problems went away once I got the line voltage down (and the observed spikes
at or below 243 volts).
Now thinking about the amplifier, which is supposed to produce 3600 volts
with a nominal 240 volt input. Given some likelihood of seeing spikes to 255
volts, I could easily imagine the capacitor bank seeing 3825 volts for as
much as 15 seconds. OK, not a long time, considering the really big picture,
but looking inside the amplifier and seeing that three of the bridging
resistors (50K ohm, 7 watts) and 4 of the capacitors were replaced at some
time in its history, I am tending towards getting more margin into the
design.
My tolerance for failure is low. I would prefer to get this amplifier up to
snuff, put it into service, and never have to open the case again. I know
that's a fairly high goal, but the AL-1200 uses one of the longest lived
tubes in this type of service, and I'm unlikely to overdrive it so long as I
load it properly, as it can withstand 130 watts and both of my transceivers
only put out 100 watts..
I can purchase the 450 volt capacitors for about $10 each, so the 10pc bank
will cost me about a hundred dollars. Add another $25 for new resistors and
$20 for the supplies to layout the PCB for 10 caps, I'm looking at $145 to
have a 4500 volt capacitor bank. If I can get the 500 volt capacitors, they
will likely cost $25 to $30 each (lower volume, fewer dealers, etc.) and I
would have a 5000Volt capacitor bank for $295 to $345 out of pocket. If I
just replaced the caps and resistors on the original PCB, I'd be out $125,
but have a 3600 volt bank. With 500 volt capacitors I'd be at 4000 volts.
Given the circumstances of apparent previous failure(s) and my low tolerance
for failure 'on my watch', I would feel better having the extra headroom,
even with a price premium well over double. (Funny how it comes around to
the warm fuzzy feelings isn't it?) Of course, f I don't find a supplier for
the 500 volt capacitors fairly soon, impatience will take over and I'll end
up buying the 450 volt ones anyway!
Of course, in the back of my mind (and not previously mentioned) is the
possibility of running the amplifier from the 220 volt tap on the
transformer, which would give me 3925 volts nominal on the bank at 240 volts,
and 4010 volts on the bank at 245 volts, and 4175 if I see the same sort of
255 volt spike at the new QTH. I've done dumber things in my life and could
see a possibility of it happening in a moment of weakness, so I'm also buying
some insurance by going to the higher voltage capacitors and the new PCB.
The biggest reason to run a higher voltage would be to allow getting full
legal output with a lower drive from the transceiver. I would prefer to run
with less than 100 watts out from the transceiver, but thinking about it, it
would likely only be about 10% or so less drive, so perhaps it isn't worth
all the work for that reason alone.
Thanks
--Rick
Jim Adney wrote:
On Fri, 14 Oct 2005 07:08:21 GMT Rick Frazier
wrote:
I was in high tech engineering (computers and peripherals) for 20 years,
and we never ran electrolytics at anywhere near their rated voltage.
Permissible margins for low voltage DC circuits were 50% or more
margin. Typical rule of thumb was 100% margin (a 10 volt cap running on
nominal 5 volt line, etc.). Even in an industry where every penny of
component cost was significant, submitting a design with electrolytics
running with as little 25% margin typically got a less than stellar
response during design reviews, to say the least.
Your margin would be determined by your tolerance for failure. If
you're making 10,000 units per year, each with 10 caps in them and
want to keep your cap related failures to less than, say, 10 per year,
then you need to be a lot more conservative than if you have a single
device with 10 caps in it that you want to keep your odds for failure
under 1% per year.
For most of the items I fix around here, I like to approach the
problem in a way that I can think of the solution as permanent, but
when you have a sample of just one, a 1% annual failure rate is about
as close to perfect as you'll ever get,
I'd say that for 450V caps, you're being much too conservative. We
have banks of 450V caps at work, each bank with about 200 caps in it.
They are used for energy storage and charged slowly and discharged
rapidly about every 5 minutes all day long. We have about 10 such
banks. We have occasional failures, maybe 1 cap every 3-4 years. We
don't think that's too bad.
I'm currently working on some upgrade banks which will have about 90
16,000 uF, 450V caps per bank. These will be run at 450V, and we'll
have 16 such banks in the second phase of the project. United
Chemi-Con doesn't seem to have any problem with this. We expect
occasional failures, but we also realize that anything else is just
being unrealistic.
C-D lists computer grade caps rated at 500 and 550 V, each with surge
ratings higher than that. United C-C does, too, but they admit that
their etched alum foil for those voltages is not as advanced as at
450V, so the energy density is not as high. This does not sound like
it would be a problem for you.
Another thing we noted with interest was that the C-D catalog said
that grading resistors were not necessary when installing caps in
series for higher voltages, especially if all the caps were from the
same batch.
If I were you, I would just buy the caps you need/want at 450V, and
then buy a couple of spares from the same batch. You would want to
reform these if you ever needed them, but they would give you the
necessary assurance that you'd be set for life.
-
-----------------------------------------------
Jim Adney
Madison, WI 53711 USA
-----------------------------------------------