Phil and Bruce,
Since you asked for a core dump on tips for building a resonant L section
in a power supply filter, here it is. My own experience with resonant choke
comes from 1982. I also spent considerable time working with Carl Sievers
(now retired) of SNC transformer in Oshkosh, WI, along with another old
timer there who had designed one for a Heathkit amplifier (was it the
Warrior?). They shared their experience with me. At the time, I may have
built the highest power resonant choke input attempted. Maybe not, speak up
if anyone did it.
The reason that we wanted one was to reduce the cost, and size of the iron
choke for a L-input power supply filter in a single phase transmitter. 3
phase rigs are much easier to filter, and the critical inductance needed to
prevent voltage soaring is < 10 H. If you calculate the load range, there
is some point in which you remove enough bleeder load that the current in
the L becomes discontinuous and the power supply degenerates into a Cap
input topology. At this point, the voltage rises to the peak value. The
solution is to make L huge, unfortunately on the order of 30 - 60 H would
be giangantic. Lowering the bleeder value also works, but it means a lot of
power wasted all the time, hundreds to thousands of watts. All this is
calculated from the critical inductance/bleeder calculation in any
textbook. The other problem is building enough filtering in the shunt
capacitor to cut the 120 Hz line ripple. (or worse yet, 100 Hz for Europe).
The stored energy in the C was horribly high, and required a bigger series
R to the plate connection of the cavity amplifier. A resonant L section
filter would reduce stored energy in the power supply. I liked that idea.
I studied the following systems which used this configuration:
Harris HFL1000 SSB amp, used 7 H and 0.25 uF, 5 kV, followed by 6 uF shunt
cap, and 100 K bleeder for 2700 volts DC output.
Galaxy 300 SSB rig used 4 H and 0.1 uF, 2.5 kV, followed by about 65 uF
shunt cap, and 45 K bleeder for 800 VDC.
Other rigs also, I forget which ones at the time.
Now I quote from "SSB circuits and systems", by Pappenfus et al (1964):
"The shunt capacitor value is selected to tune the choke at the ripple freq
and with the load consisting of the highest value of Reff which will be
encountered. Filter choke inductance normally decreases to some extent as
the direct current through it increases. The effective L at low load
currents may be twice as much as the full-load rating. This is partially
due to the inductance increase caused by the large AC voltage across it.
When the load is increased and Reff is reduced, the choke may change
inductance enough to detune the circuit, but the critical inductance needed
is then lower, so regulation does not suffer. Proper tuning can permit a
several-times reduction in filter choke size."
..."Tuning the input choke will reduce the fundamental ripple component on
the order of 10 dB. The atten. of harmonics of the ripple freq. is less
with a tuned input choke however. The reduction is limited by the voltage
division of the tuning capacitance and the filter capacitance. With 0.1 uf
and 8 uf, this amounts to 38 dB. THe peak voltage of the second harmonic of
the ripple frequency is down 17.5 dB from the dc voltage. The ripple second
harmonic component, therefore will be down approx. 55 dB."
Another source to read about them is in Terman's Radio Engineering Handbook.
Also the red Radiotron Designers Handbook by Langford-Smith.
Bruce sez:
>I have a 10H choke rated at 1.3 A and hipot tested to 13KVA. The closest
>cap I could find is a .2uf 10kva oil >cap. I measured it at .21uf and ind
>at 9.1H or so, but I dont know how accurate my meter is.
You need to actually measure the choke with an impedance meter at 120 Hz,
since the ripple frequency will be 120 for a FWB rectifier. And you need to
DC bias the thing at the operating current, while measuring. An old General
Radio impedance bridge is handy, with a big blocking cap to prevent the DC
from going into the bridge.
You might could also sweep it with an audio oscillator, and somehow
calculate the L, but you need to put the bias on it. Also, figure that the
high peak AC across the choke will also push the BH curve up into the tips;
maybe give an incremental inductance change as mentioned above.
I used an audio oscillator, tube type like an HP 200 series, and a VTVM
acorss a load resistor. Fed the resonant choke section in series with the
load. I measured impedance, Q, from this, looking at the voltage across the
load R. Also used a GR 1650 bridge at times.
With a 1.9 H choke and 2 uF cap
Lequiv = 1.45 H, Q = 24 at 120 Hz, Rdc = 21 Ohms
With a 6.8 H, 0.27 uF
Lequiv = 9.31 H (?), Q = 22, Rdc = 46 Ohms
With another brand of 6.8 H, 0.27 uF
Lequiv = 6.33 H, Q = 3, Rdc = 35 Ohms
With a 6 H, 0.27 uF
Lequiv = 3.9 H, Q = 32, Rdc = 50 Ohms
Eventually I ran an optimization calculation on a hand calculator (this was
in 1982!) that gave me various L and C combos, and the resultant output
ripple versus freq.
One combo was 2.533 H and 1 uF cap, with Rdc = 30 Ohms. It gave 40.6 dB of
120 Hz rejection when the second shunt C was 2 uF. When 4, the ripple was 6
dB lower. Rbleeder was 200K. Full load was about 5 K Ohms, a class C
amplifier at max output. To get an equivalent filter without a resonant
section, would have required 10 H input L, and 20 uF shunt C, and a 10 K
bleeder to prevent soaring!
Later I power tested a couple of chokes and caps. One thing I noted was
that the inrush current into the power supply was identical for no load or
full load. Also noted that the power supply would ring during turn on, and
overvoltage. At no load, the ripple across the filter was essentially a
sinewave.
I tapped the choke for 120 Hz, and it was wound by SNC for 100 Hz operation
as well, for overseas customers. The caps I chose were 0.97 uF, 2.5 kVAC
polypropylene film, 5% tolerence spec'd (measured by buying about 50 caps
and doing a lot of measurements). I then series connected two caps to get
roughly twice the AC voltage rating, leading to about 0.48 uF as my
resonating cap. L was rated at 5.06 H, +/- 5%, 21 Ohms, measured at 1.2
ADC. Also a tap at 3.5 H for USA transmitter operation. I am reading this
from an old parts list. Then there was a 4 uF shunt C. In the final design
I followed the first section with a second series L, about 3.5 H (small and
inexpensive) and another shunt 4 uF. This gave excellent ripple rejection
for even 720 Hz, and also was very well regulated due to the resonant
choke. It was modelled to prevent any resonances at power line multiples,
or at 50 or 60 Hz, in case a rectifier failed open, and allowed half wave
operation.
Bruce,
a Pyranol cap is not good for this. it will probably blow eventually, then
a real mess. You really need a smaller choke since the L will dominate in
your scheme, and then it will be sensitive to DC current, the thing will
'swing' a bit. Special chokes can be gapped to try and reduce swing, but
yours probably isn't. To make that effect smaller, reduce L, and let C get
bigger. You can purchase the 1 uF HV AC capacitors used in doublers in home
microwave ovens for about $10-20. They have a polypropylene dielectric (or
some polymer) instead of oil/kraft paper. They have lower loss for 120 Hz
and the higher harmonics of the ripple. Since the cap is not really
sensitive to the DC voltage or current, it will be more stable this way.
And yes, it should be resonant, not off one side. The voltage will SOAR if
you shift the resonance, and possibly blow the cap or the following ciruit.
The voltage soars up to the peak of the ripple, if the thing becomes non
resonant, a pure L input choke, when the load is removed. ( such as SSB
operation with no speech). With a good resonant choke, you can load and
unload your HV power supply, with only small change in voltage. You also
have to watch for audio resonances in the circuit, getting excited from
speech amplification. Television transmission has even worse problem, the
bouncing video waveform. Many times they stick to simple C input filters
for this reason, very stiff, minimum bounce and resonances.
>How critical is being exactly resonate at 60hz?
>Will my oil cap work without blowing up, as it is pyranol .. a PCB?
>I just go the cap and have not tried at low voltage yet, but guess the
>real test is at high voltage. plan is for a very stiff 5500VDC amp supply.
>What is the best test method that you can recommend before I buy big xfmr?
Testing these things is risky, You need to look at the voltage across the
cap, maybe using two voltage divider probes on a 2 channel scope,
differential measurement (ch 1 - ch 2). If the peak voltage exceeds the cap
rating, it will eventually eat it. The peak is large in a resonant choke
circuit. Thats why i recommend the HV AC caps from microwave oven HV
doublers. At least they are rated for high peak voltages.
You also need to be aware that a resonant choke might run hot with the
circulating 120 Hz current. Check that, and check the cap (with power off,
discharged!).
Also, the resulting ripple after your resonant choke input section will be
predominantly 240, 360, 480 Hz... so be aware that this stuff might be
audible and must be filtered with a small smooting LC section. A lot of
people disregard that, and it needs to be measured to make sure you are not
making noise sidebands on your output. It is essentially AM from your plate
supply, although if the tube remains linear and not saturated, it has some
immunity to incindental AM on the power supply.
Good luck, let us know if you have success (or smoke!).
73
John
--
FAQ on WWW: http://www.contesting.com/ampfaq.html
Submissions: amps@contesting.com
Administrative requests: amps-REQUEST@contesting.com
Problems: owner-amps@contesting.com
Search: http://www.contesting.com/km9p/search.htm
|