![](http://img135.imageshack.us/img135/4205/ps3hu.jpg)
I simply use the parts I have and PSU Designer II (Duncan Amplification). I have 370+370V/0,3A transformer, and L-input PS is not possible, I need B+ about 410-430V. Then, I choose C-input with small first cap, two independent PS on the same HV winding, each with its own tube rectifier (5V4G) and CLCLCLC configuration... Let`s see:
-I1 current tap (4mA) is actually resistor load, about 100k
-I2 current tap is 300B (80mA)
-I3 current tap is driver (18mA)
Actual parts I used:
C1 = two 4,7µF/630V Wima MKP4 in parallel
C2 = 200µF/500V Mundorf M-Lytic HV electrolytic
C3 & C4 = 47µF/600V M-Tubecap polypropylene
L1 = 10H/250mA/98 Ohms, AE-Europe
L2 = 10H/150mA/140 Ohms, Schmidbauer
L3 = 10H/~50mA/285 Ohms, salvaged/old stock ![](http://www.audioroundtable.com/emoticons/smiley.gif)
Current ratings of the chokes is overdimensioned. All parts (especially caps) must be for 500V working voltage minimum.
On PSUD diagram we can see 30ms "segment", 1s after "turn on".
-The green line is ripple voltage, Vc1~80Vpp, from 400-480V. The DC voltage is Vc1 = Vc1p - Vripple/2 = 480 - 80/2 = 440V.
-Vc2 is yellow line, after first Pi-filter (L1/C2).
Vc2 = Vc1-RwL1*IL1 = 440 - 98*(0,018+0,08+0,004) = 430V. We have little ripple, Pi filter lower it S1=Vrc1/Vrc2 = (2Pi*f)^2*L*C =
= (2*3,14159*100)^2*10*200*10^-6 = 789,57 times, and Vrc2 =
= Vrc1/S 0 80/789,57 = 0,1Vpp.
-Vc3 is blue line, our B+ for output stage, after L2/C3 filter.
Vc3 = Vc2 - RwL2*IL2 = 430-140(0,08+0,018) = 416,5V. Our second Pi-filter lowers the ripple voltage 185,5 times and we theoretically have just 0,54 mVpp ripple here.
-Vc4 is a white line, our B1+ for the driver, after L3/C4 filter.
Vc4 = Vc3 - RwL3*IL3 = 416,5 - 285*0,08 = 411V. Third Pi-filter lowers the ripple voltage to just 0,003mVpp theoretically.
-Red "spikes" are ripple current waveforms on C1. Although we "draw" about 158mArms from our transformer, C-input means sharp peaks of 465mAp! That`s why we prefer L-input, if we can...
-Actual measured voltages on breadbord shows only a few volts difference in comparison with simulation.