Description on page 2

COMPAQ PROLIANT PS4060 Power Supply Primary board Secundary board

+382V

0.5 mH 220uF

350 VA
230V/ 50Hz 1.8 mH 2.5 kHz
1:1 115V/ 3A/ 400Hz
660uF
2 uF
PFC
2 uF
120k
45uH
+24V
47nF

220uF 15k
+ +

5R6
V~ A~
1mA / 150Vac 20mA / 3A ac
9V
100k
27R

15.61 V
1uF BYV26E
+382V
+ 2n2
CD40174 400Hz
12k CD4049 8k 5uF
10n
R 6k8 12k 27
150p IRFP460
56k
5V6 22k 3k3
33k NE5561
33n
CD4093 27k
33k 7.57 V 27k
22k
PWM
56k IRS2109
OP470
3.8V
FREQ 100k 11 kHz
5k Vcc 15k 5.76 V 15k
18k (0)
Mid voltage 27
15 nF 7.55 V +15
7k8 oscillator
+8Vref 1k
4800Hz 270k 1N4148 0.135R
220k
2n2
39k
100Hz ripple
feed forward

3k3
+15
AMPL 10k 10 uF
10 cycle ramp-up

350W 400Hz converter

16 april 2008
400Hz converter
8 August 2008 K.Bouwknegt
The 400Hz converter is based on a Compaq Proliant
Server power supply. This hot-swappable PSU type
PS4060 is rated for 750W at 220Vac input, or 500W
at 120Vac input. It should be possible to make at
least 300W output at 400Hz with this PSU.
From the PS4060, the chassis, fan, and primary
electronics board were used, plus some coils and flat
cable connectors from the secondary board.
Almost everything on the main board (on top of
photo) was used: the input rectifier, the PFC circuit
to get 382Vdc, the auxiliary converter to provide
+9V and +15V voltages for the fan and PWM
electronics, the heatsinks and 2 out of the 4 power
fets that once drove the converter transformers on the
secondary board.
Board 2 was added to the heatsink on the
primaries board (orange board on photo) to create the
400Hz PWM drive signals for two power FETs on
the main board. This board contains a 4800Hz
oscillator ( 12 x 400Hz), a divide-by-12 shift register
to make 6 square waves at 400Hz, each 30 degrees
phase shifted from the previous one; a buffer
amplifier to give these square waves a variable
height, a summing amplifier that adds 5 square
waves to a stepped wave at 400Hz that has no lower
harmonics than the 11th, and the PWM chip.
This 400Hz sine wave is compared to a 18kHz saw
tooth in a dedicated PWM circuit, resulting in one bit
information whether the lower or upper fet should be
"on".
A special "pole" driver circuit splits this signal into
drive signals for the upper and lower power fet.
Using this circuit, no drive transformers are needed.
Board 3 on the bottom of the case contains the
power output circuit.
On the left, two large capacitors form the midpoint
of the DC supply as a return for the output circuit. A
2mH/4A ferrite core choke in series with a ring core
choke and 4x 1uF capacitors filter the PWM
waveform into a sine wave
The cut-off frequency is 1.8 kHz, near the fourth
harmonic of 400Hz. This frequency is almost absent
in the PWM modulator, as well as in the expected
load current.
The filtering of the 18kHz carrier is enhanced by
giving half the capacitors a series choke, thus
forming a notch at the pwm frequency.
Finally, a C-core transformer with 1:1 turns ratio
provides isolation between the mains and the 400Hz
output. Behind the meters are some components that
were salvaged from the original PS4060 secondary
board to provide a 24Vdc/3A output, needed as well
by most avionics .
Metering
The voltmeter is straightforward. A 1mA moving
coil instrument measures the average value of the
rectified output voltage. This value is shown on a
Magnetics
PWM filter inductor 1.8mH/4Arms
The inductor is based on a ferrite core E55/28/21
from either Ferroxcube or EPCOS. The core halves
are gapless, so I added 1.6mm printed circuit board
as gap. The coil with 1.1mm dia wire has 88 turns in
4 layers, that is 14 metres, with 0.24Ω dc resistance.
Output transformer
The transformer was made on a C-core from
Vacuumschmelze type SU48b with an iron section
15 x 25mm. Both primary and secondary have 144
turns of the same 1.1mm dia enameled wire in 4
layers, 2 layers on each leg. In total 40 meters, with
0.34Ω dc resistance for either primary or secondary.
The copper loss is 6W at 300W output. The (only)
core I had was made for 50/60Hz, wound from
0.3mm strip. To keep the losses reasonable at 400Hz,
the peak induction is only 1.2T. With the fan used,
both iron and copper stabilize at 48°C at 300W
output.
The isolation between primary and secondary is 3
layers mylar, and withstands 2kV rms
volts rms scale, assuming an (almost) sinusoidal
wave. Full scale at 1mAdc is 150Vac.
The amp meter is a 20mA moving coil instrument,
connected to 200:2 current transformer via a
schottky-diodes bridge. The secondary of the CT
can produce 1Vrms above 300Hz without saturation.
This gives a perfect linear scale to 3A ac.
Lateron I added a small switch to show either the
115V ac voltage and current or the 24Vdc voltage
and current.