bill ramsay wrote:
dear all
I am playing with some 813's at the moment. I am trying to get the
heater supply sorted out.
the 813 needs 10v at 5 amps, the do glow v. bright when lit up.
pretty neat.
I have a breadboard amp going at the moment which someone else has
given me.
The current heater supply uses a 20 odd volt ac transformer, bridge
rectiifer, feeding a 2n3055 based regulator, thing is putting out
about 50watts in heat. as it is stereo there is 100w in heat being
generated.
whenever i model a 5 amp 10V heater supply in PSUD i get some really
whacky readings, ie.
28V transformer to feed 100,000u cap, with a 0.5ohm resistor, then
1000u cap into the tube. ie. 2 ohm load,
or am i missing something here, 28V ac seems a long way away from 10v
dc, or is this an artifact of the high current pulsing away and
charging the caps?
I wanted to build a dc supply that does not generate so much heat in
itself.
any thoughts, ideas etc. especially the printable ones would be
greatly appreciated.
kind regards
bill ramsay.
To get enough heater power for 4 x 813 for a stereo PP amp,
you would need 4 separate 10v x 5A power supplies.
I guess many folks would use centre tapped winding for each tube,
to allow for normal AC operation, but let's assume you insist on DC.
A separate PT rated fo about 350 VA would be suitable, and a toroidal
rated at 500VA would be nice.
You could try gettting a supplier to give you a toroidal tranny
with no secondaries wound on, and then do your own secs,
because the turns required for each 10v may only be between 15 and 20
turns
of say 1.5mm thick wire.
To reduce transformwer noise to a minimum, I would use a choke input
supply
for each tube, so the tranny AV would have to be about 12vrms to make
+10DV.
10v at 5A is an RL = 2 ohms, so the choke value needs to be at least
2 / 940 = 2.1 mH, which would be easy to make using about 2.0 mm dia
wound around a stack of iron with a core area of about 25mm x 20 mm,
with a very generous gap.
In fact, if you made a choke with 20 mH, it would ber a lot lot better.
The choke size quickly goes up with the inductance needed,
and the DCR, which needs to be minimized because Pd = I x I x R.
The capacitance would preferably need to be as high as you can afford,
but no need for ultra high ripple current ability, because the choke
input
avoids the high charge currents to any caps,
so I would use multiple paralleled 10 x 4,700 uF caps normally used for
SS amp rails supplies,
say to make to make 47,000 uF.
The AV input to the choke after the diodes is in the vicinity of 8vrms at
twice the
mains F, ( just guessing ) and the LC filter with 20 mH and 47,000 uF has
an impedance ratio
of 12.5 ohms to 0.034 ohms at 100 Hz, ( and better with 60 Hz mains ),
and so the 8v of hum is reduced to 8 x 0.034 / 12.5 = 0.021 vrms, which
imho will not modulate the cathode signal or be heard.
It would even be possible to make the chokes using a normal transformer
bobbin
with a centre hole say 1"H x 1.5"T and with a winding area of 2" x 0.75"
and use 1.6mm dia wire, and use some cut up old transformer
cores for a solenoid type of core, which won't saturate, but which would
give you about 4 times the inductance for the same turns using an air
core.
Such chokes are prone to hum, and should be well varnished and potted.
The choke in my class A SS amp which has a CLC type of 3 amp supply
has a 1.25" stack of 1" tongue lams, and its about 15 mH at 3 amps,
and I use 9,400uF off the high current bridge rectifiers, and 45,000 uF
after the choke,
and the ripple is much lower than a choke supply, just a few mV.
No need to waste power using SS regulator pass elements.
With a choke in a choke input supply, you will never get the theoretical
maximum voltage of 0.63 x the peak value of the AV input voltage from the
transformer
because of the DCR of the choke, and only 0.5 ohms means a V drop of
2.5 volts with 5 Amps.
And it means 12.5 watts of heat liberated in the copper winding.
So it is imperative to get the DCR of the choke to a low value,
preferably
less than 0.25 ohms.
Get the supply done for one tube, and then make all the rest the same.
The turns required for 12vrms on the power tranny can be trialled using
some thin wire to establish the voltage, so start with 20 turns, and
measure the
voltage, and remove turns until you drop to a few volts above where you
need
to be, say at 15v. Then use some thick wire, wound on tight,
taped down well onto the core.
Removing more turns is easy.
If you get 11v at 5 amps for a single filament supply,
then with 4 such supplies expect the voltage to drop to 10v each.
For a regulated supply, you would need a couple of low voltage high
current
BJT in parallel as a pass element, with a drop across the BJT
of say 4 volts, and so the supply from a cap input filter has to be 14v.
The AVrms would have to be about 12vrms, about the same as for a choke
input filter.
12v gives a theoretical max of
(1.414 x 12v) = 17DV, but there is a drop across two silicon diodes of
the bridge,
and then the loss due to the ripple characteristic, and one is lucky
to get 14DV from 12vrms of AV.
But using 9,400 uF caps for C1 will give a ripple voltage of
1.2 vrms at 5 amps DC.
To limit the effects and noise of the quite high peak charging currents
instead of resorting to
a choke input, the use of series R between the diodes and C1 is
recommended,
and a value where R = 5 x ZC is recommended, so at 100 Hz, Z9,400uF =
0.17 ohms,
so the R could be 0.82 ohms, and rated at 20 watts.
This would however raise the voltage needed by the cap input filter to
give the same
DV output.
The ripple voltage will not change, but the charge period will increase,
thus reducing the peak charge current occuring when no series R is used.
A CRC filter is also possible, and if C1 is 9,400 uF, then R should be at
least
10 times the reactance of C1, and the C2's reactance at least 1/10 of R,
so if ZC = 0.17 ohms, then R could be 1.7 ohms, rated at 100 watts,
then another 9,400 uF, which would reduce the ripple from 1.2vrms to 0.12
vrms,
which is a poor result compared to a choke, and the V drop across 1.7
ohms
would be 8.5 v, which liberates 42.5 watts in the R!
And this is a real lousy result.
R cannot be much less than 10 times C1's reactance value, because
otherwise the
diodes begin to see the C2 as their load as well.
If we said to hell with peak charge currents, and used
100,000 uF caps straight off the rectifiers, then the ripple would be
0.11 vrms.
( I have such a supply in an SS amp with two mighty 100,000 x 75 volt
Sprague computer grade
caps each 220mm long, and 75mm dia, with about a 20 amp ripple current
rating.
No chokes, and the noise induced into the amp is quite negligible because
of careful
layout of wiring, and the use of 50 dB of global NFB. the DCR of the
tranny windings and mains
wiring in the house walls provides the charge limiting DCR. )
Since you only want 10v, the caps will be higher C value for the size,
and 100,000 uF at 16v at high ripple won't be too big.
If a CRC suply is needed, then R needs to be about 0.22 ohms to
get a 10 times , -20 dB drop in ripple at C1, if C2 is also 100,000 uF,
from 0.11vrms
to 0.011 vrms.
It dissipates only 5.5 watts, and the V drop is only 1 v.
But a choke of 1mH would have a Z of 0.628 ohms at 100 Hz,
and C2 of 100,000 uF has 0.016 ohms, so the attenuation factor
in the LC filter is 0.025, 25 times better than the R.
Somewhere there is an efficient compromise between C used and choke
size, and what you are prepared to pay for.
The choke input filter or CLC would be more reliable than 4 x solid state
regulated
power supplies, imho.
Patrick Turner.
|