Professional Documents
Culture Documents
THEORY OF OPERATION
&
DESIGN ANALYSIS
OF
by
Isaac Ozkaynak
1. INTRODUCTION
1) Power: 500 W
15) The final Klystron power supply will consists of four of these modules
connected in parallel on the input and in series on the output so that they can
The approach taken for the implementation and its in depth discussion had already
been furnished in the previously submitted Design Concepts document and it will not be
Figure 1 shows the power supply system arrangement. Each sub unit generates
1750V dc output voltage. The subunits are connected in parallel at their input ports and in
series at their output ports. Each sub unit is envisioned as identical in every circuit detail.
+
+
Unit 1 1750Vdc +
−
+
Unit 2 1750Vdc
−
100Vdc 7000Vdc
+
Unit 3 1750Vdc
−
+
Unit 4 1750Vdc −
−
and external interface circuits for the external output voltage control and for
Figure 2 shows the topology of the main power processor, which is called Push-
Pull Forward Converter (PPF) . It has 14 isolated and center tapped secondaries, which
are connected in series. Each center tapped output, which we call cell, generates 125Vdc
and the total is 1750V. The turns ratio of secondary windings to primary windings is 2.
The choice of 125V for cellular outputs is based on the voltage rating of Radiation
Hardened Mosfet Switches. Currently, the only Megarad hardened MOSFETs, which are
available as off the shelf, and suitable for employment at the secondary side by their
Figure 3 shows the gate drive waveforms of the Push-Pull Forward converter in
Figure 2. The primary switches are driven by pulse widths dictated by the control system.
However, the secondary switches are driven at fixed 50% duty ratio. The advantage of
this scheme is to prevent the voltage spikes at the drain terminals of the secondary
switches, since they are not clamped like the primary switches, when they are turned off.
Thus, the secondary switches are turned off at the end of the commutation sub intervals,
Before we proceed further, we need to develop both the steady state and the
VGS1, VGS 2
TS
TS
2
TS
D⋅
2
0
t
VGS 3, VGS 4
0
t
Figure 3 – Gate Drive Waveforms of the PPF Converter.
For establishing the Large and Small Signal models, the PPF converter output
stages, Figure 2, need to be “folded in” to a single output. In multi output converters, only
one output can be subject to PWM control due to the singularity of output state variables
7
i.e., there can be only one output state variable, voltage or current. This holds true on the
double, since we are not applying any “Weighted Control” on any other outputs, either.
In this folding-in process, all the other outputs along with their circuits are referred to the
output at which the PWM Control is applied. In this application, this process is facilitated
The “referred to single output or folded-in” topology is shown in Figure 4. During the
Q2
NP1 D2
Vg =100V NS1
Lf
C1
Q1 D3 Cf R
NP2 NS2
D1
Q3 Q4
−125V
ON time sub interval of either one of the primary switches, the circuit connections are the
RS1 vo
Lσ 12 iL Lσ 34
RL n ⋅iL n
RP1 f n2 n2 f n ⋅io
Lf
n2 RC
Vg f
n2
N P1 vnp1 2 R
n2
Ron3 n2 ⋅ C f vC f
1 n2 n
Ron1
TS
Figure 5 – Switching State during Q1 on time subinterval, i.e., D ⋅ .
2
The State Space expressions for this switching state are derived and collected as
⎤ ⎡ ⎤
R R
⎡ diL f − e1 −
⎢
⎢ dt
⎥
⎥=⎢
⎢ Le1 (
n ⋅ Le1 ⋅ R + RC f ) ⎥ ⎡ 1 ⎤
⎥ ⎡ iL f ⎤ ⎢ ⎥
⎢ dvC f ⎥ ⎢ ⎥ ⋅ ⎢ v ⎥ + ⎢ Le1 ⎥ ⋅ vg (1)
⎥ ⎣⎢ C f ⎦⎥ ⎢⎣ 0 ⎥⎦
R 1
⎥ ⎢C ⋅ R + R −
⎢
⎣ dt ⎦ ⎢⎣ f Cf ( ) (
C f ⋅ R + RC f ) ⎥
⎦
where,
RS 1 RL Ron 3 R ⋅ RC f
Re1 = RP1 + n ⋅ Ron1 + + + + (2)
n n n n ⋅ R + RC f ( )
and
9
Lf
Le1 = Lσ 12 + n ⋅ Lσ 34 + (3)
n
During the sub intervals of commutation, both of the primary switches are off. This sub
TS
interval is denoted as D′ ⋅ and defined as
2
TS TS T T
D′ ⋅ = − D ⋅ S = (1 − D ) ⋅ S → D′ = 1 − D (4)
2 2 2 2
During this sub interval, the secondary side commutates and the primary side clamp
capacitor charges. Figure 6 shows the secondary side switching state circuit.
RS1
n2 ⋅ Lσ 34 RL Lf iL vo
f io
RC
f
iL
f
1 iL
f iC R
2 f
2
Ron3 Cf
n2 ⋅ Lσ 23
RS 2
2
iL
f iL
2 f
D4
2
Q3 continues to conduct and the external anti-parallel diode of Q4, which is D4 is turned
on by the commutation, and the center tapped secondary windings are practically shorted.
n 2 ⋅ Lσ 23 ≅ n 2 ⋅ Lσ 34
(5)
RS 1 ≅ RS 2
Therefore, there is no need for writing KVL equation for Loop 2. Furthermore, since
iL f
⋅ Ron 3 ≅ VD 4 (6)
2
in this particular application, the commutation currents flowing through the center tapped
secondary windings would approximately be equal. The State Space expressions for the
⎤ ⎡ ⎤
R R
⎡ diL f − e2 −
⎢
⎢ dt
⎥
⎥=⎢
⎢ Le 2 (
Le 2 ⋅ R + RC f ) ⎥
⎥ ⎡ iL f ⎤
⎢ dvC f ⎥ ⎢ ⎥ ⋅ ⎢v ⎥ (7)
⎥ ⎣⎢ C f ⎦⎥
R 1
⎥ ⎢C ⋅ R + R −
⎢
⎣ dt ⎦ ⎢⎣ f ( Cf ) (
C f ⋅ R + RC f ) ⎥
⎦
where
RS 1 + Ron 3 R ⋅ RC f
Re 2 = + RL + (8)
2 R + RC f
n2
Le 2 = ⋅ Lσ 34 + L f (9)
2
and
Le1 1
= (10)
Le 2 n
state solutions as
1
Vo = n ⋅ Vg ⋅ D ⋅ (11)
R R L ⎡ L ⎤1
n ⋅ D ⋅ e1 + n ⋅ e 2 ⋅ e1 ⋅ (1 − D ) +
⋅ ⎢ D + n ⋅ e1 ⋅ (1 − D ) ⎥
R ⎣
R Le 2 Le 2 ⎦RC f
1+
R
parasitics
and
n ⋅ Vg ⋅ D 1
I Lf = Io = ⋅ (12)
R R R L ⎡ L ⎤1
n ⋅ D ⋅ e1 + n ⋅ e 2 ⋅ e1 ⋅ (1 − D ) +
⋅ ⎢ D + n ⋅ e1 ⋅ (1 − D ) ⎥
R ⎣
R Le 2 Le 2 ⎦RC f
1+
R
parasitics
1 R
+ e2
RC f R
1+
D= R (13)
Vg Le 2 1 L ⎛ L ⎞ R R L
⋅ − ⋅ e 2 ⋅ ⎜ 1 − n ⋅ e1 ⎟ + e 2 − e1 ⋅ e 2
⎛ RC ⎞ L
n ⋅ ⎜ 1 + f ⎟ e1 ⎝
Vo Le1 Le 2 ⎠ R R Le1
⎝ R ⎠
The above expressions illustrate the effects of the Leakage Inductances in the
xˆ
= ( s ⋅ I − A) ⋅ b
−1
(14)
vˆg
12
where
⎡ iˆL f ⎤
xˆ = ⎢ ⎥
⎢⎣ vˆC f ⎥⎦
⎡ s 0⎤
s⋅I = ⎢ ⎥
⎣0 s ⎦
⎡ Re1 Re 2 R⋅D R ⋅ (1 − D ) ⎤
⎢ − L ⋅ D − L ⋅ (1 − D ) − − ⎥
⎢ e1
A= ⎢
e2 (
n ⋅ Le1 ⋅ R + RC f ) (
Le 2 ⋅ R + RC f ) ⎥
R 1 ⎥
⎢ − ⎥
⎢
⎣ (
C f ⋅ R + RC f ) (
C f ⋅ R + RC f ) ⎥
⎦
⎡D⎤
b = ⎢ Le1 ⎥
⎢ ⎥ (15)
⎣⎢ 0 ⎦⎥
and
vˆo
= Gvd = C T ⋅ ( s ⋅ I − A) ⋅ b
−1
(16)
vˆg
where
⎡ R ⋅ RC f R ⎤
CT = ⎢ ⎥ (17)
⎣⎢ R + RC f R + RC f ⎦⎥
The exact expressions, and especially the system eigenfunction is too long to be furnished
here. Interested readers will find the exact expressions in the accompanying MathCAD
these transfer functions here. Figure 7 shows the input voltage to inductor current transfer
13
function. Figure 8 shows the plots of input voltage to output voltage transfer function.
xˆ
= ( s ⋅ I − A) ⋅ ⎡⎣( A1 − A2 ) ⋅ X + ( b1 − b2 ) ⋅ Vg ⎤⎦
−1
dˆ
(18)
vˆo
= C T ⋅ ( s ⋅ I − A) ⋅ ⎡⎣( A1 − A2 ) ⋅ X + ( b1 − b2 ) ⋅ Vg ⎤⎦
−1
dˆ
A1 , b1 and A2 matrices are given by Eqn. (1) and Eqn. (7), respectively, and b2 = 0 .
Figure 9 shows the plots of duty ratio to inductor current transfer function. We
can see, that the Current Injection Control (CIC, i.e., Current Mode) is not feasible for
this application. −20dB / dec. slope starts around 10 KHz. For taking advantage of the
slope and −90o phase, we need to arrange the intersection of voltage loop and inner
Figure 10 shows the plots of duty ratio to output voltage transfer function. The
shape of the plots suggest that the voltage mode control is feasible.
Voltage Mode Control implements two tier control via variable reference scheme.
With a near constant reference, the first tier controls the top most cellular output voltage,
Figure 2, to near -125V. The second tier controls the total output voltage to -1750V via
changing the variable reference of the first tier. Thus, the cellular output voltages may
experience limited deviation from 125V, but the total would be constant at -1750V. The
tacit assumption is that the limited deviations are acceptably small due to the fact that the
Figure 11 shows the error amplifier for the first tier. It senses the top most cellular
output voltage, whose positive port is tied to the load ground. It uses a variable reference
16
vˆo′′
C p′′
Rz′′ R f ′′ C f ′′
R1′′
Cz′′
vˆc
R2′′
Ao
which is generated by the Buffer Amplifier, Figure 12. In Figure 11, the error amplifier
where
⎛ ⎞
vˆo ⎜⎜⎜ −1750V ⎟⎟⎟
⎝ ⎠
CP
CZ RF
R1
R2 RZ Ao′
Rb
CO R1′
Ra
R2′ Ao′
VR
vˆR ⎛⎜⎜ s⎞⎟⎟
Ra ⎝ ⎠
Rb
VR =5.1V
and
1 + s ⋅ C ′′f ⋅ R′′f
Z ′′f =
s ⋅ ( C ′′f + C ′′p ) ⋅ (1 + s ⋅ Ce′′ ⋅ R′′f )
(21)
C ′′f ⋅ C ′′p
Ce′′ =
C ′′f + C ′′p
Notice, that Eqn. (19) would revert to an inverting amplifier transfer function, if we had
vˆR
used a constant reference voltage, since = 0 , then. For a Non-Ideal opamp, the Eqn.
vˆo′′
(19) becomes
where Ao = open loop transfer function of the non-ideal opamp. In this particular
application, we use UC1825N for PWM IC. The built-in error amplifier in UC 1825N is
modeled as
27000
Ao ( s ) = (23)
⎛ s ⎞ ⎛ s ⎞
⎜1 + ⎟ ⋅ ⎜1 + 6 ⎟
⎝ 2 ⋅ π ⋅ 200 ⎠ ⎝ 2 ⋅ π ⋅ 5.5 ⋅ 10 ⎠
from the published data on the device by the manufacturer, Unitrode/Texas Instrument.
The Buffer amplifier in Figure 12 scales the output voltage of -1750V (HV) and
buffers it with a low gain. It uses 800 KRad hardened power opamp, LM6172, with
50mA sink or source capability. LM6172 is also a wide bandwidth and high slew rate
device. One slight disadvantage in LM6172 employment is, that it requires dual polarity
bias supply, particularly, ±15V . The open loop transfer function of LM6172 is modeled
as
s
172 1+
Ao′ ( s ) = 10 20
1
⋅ ⋅ 2 ⋅ π ⋅ 50 ⋅ 106 (24)
2
s ⎛ s ⎞
⎜ 1 + 6 ⎟
⎝ 2 ⋅ π ⋅ 80 ⋅ 10 ⎠
from the published data by the manufacturer. Figure 13 shows the plots of Eqn. (23) and
R1′′ = 121K Ω
R2′′ = 5.11K Ω
R′′f = 4.99 K Ω
Rz′′ = 499Ω (25)
C z′′ = 470 pF
C ′′f = .018µ F
C ′′p = 0
21
Figure 14 shows the plots of error amplifiers of ideal and non-ideal opamps.
The first amplifier in Figure 12 generates an output voltage across CO of 10V, when the
output voltage is 1750V. This is further scaled and applied to the following differential
amplifier, which generates a variable reference voltage at its output. This variable
reference voltage is centered at the median value of +5.1V, and changed according to the
deviation of HV from the targeted value of -1750V. In turn, the variable reference is used
The differential amplifier output voltage expression for an ideal opamp is given
by
Rb
vˆR = ⋅ ( vˆ A − vˆB ) + VR (26)
Ra
vˆR ( s ) R2′ R
A1 ( s ) = =− ⋅ b (27)
vˆo′ R1′ + R2′ Ra
for the ideal opamp. For the non-ideal opamp, the eqn. (26) becomes
Rb Ra
vˆR ( s ) = ⋅ ( vˆ A − vˆB ) + ⋅ VR (28)
Rb ⎛ 1 ⎞ Rb ⎛ 1 ⎞
+ Ra ⋅ ⎜ 1 + ⎟ + Ra ⋅ ⎜ 1 + ⎟
Ao′ ⎝ Ao′ ⎠ Ao′ ⎝ Ao′ ⎠
vˆR ( s ) R2′ Rb
A1′ ( s ) = =− ⋅ (29)
vˆo′ R1′ + R2′ Rb ⎛ 1 ⎞
+ Ra ⋅ ⎜ 1 + ⎟
Ao′ ⎝ Ao′ ⎠
vˆc ⎛ Z ′′ ⎞ vˆ Z ′′ vˆ′′
= ⎜1 + f ⎟ ⋅ R − f ⋅ o (30)
vˆo ⎝ Z i′′ ⎠ vˆo Z1′′ vˆo
where
24
and
vˆo′′ 1
= , m = 14 (32)
vˆo m
The Eqn. (32) is based on the assumption of 14 identical output stages were stacked. The
vˆo′ R2 R ⋅ R 1 + s ⋅ CZ ⋅ RZ 1
Avb ( s ) = = ⋅ F O⋅ ⋅ (33)
vˆo R1 + R2 RZ 1 + s ⋅ CP ⋅ RF 1 + s ⋅ CO ⋅ RO
for the ideal opamp. Using the assumption of RZ R2 , the same transfer function
becomes
R2 ⋅ RO ZF 1
′ ( s) =
Avb ⋅ ⋅ (34)
R1 + R2 Z F ⎛ 1 ⎞ 1 + s ⋅ CO ⋅ RO
+ Z I ⋅ ⎜1 + ⎟
Ao′ ⎝ Ao′ ⎠
where
RF
ZF =
1 + s ⋅ CP ⋅ RF
(35)
RZ
ZI =
1 + s ⋅ CZ ⋅ RZ
for the non-ideal opamp. Using Eqn.(27), (31), (32) and (33) in Eqn. (29), we obtain
vˆc ⎛ Z ′′ ⎞ Z ′′ 1
Fv ( s ) = = ⎜ 1 + f ⎟ ⋅ A1 ( s ) ⋅ Avb ( s ) − f ⋅ (36)
vˆo ⎝ Z1′′ ⎠ Z1′′ 14
for the ideal opamps. Using Eqn. (29), (31), (32) and (34) in Eqn (22), we obtain,
for the non-ideal opamps. Figure 15 shows the plots of Eqn. (36) and (37).
Tv ( s ) = FM ⋅ Gvd ( s ) ⋅ Fv ( s ) (38)
R1 = 1.74 M Ω
R2 = 5.11K Ω
RZ = 49.9 K Ω
CZ = 680 pF
RF = 100 K Ω
CP = 8.2 pF
(40)
RO = 49.9
CO = .022 µ F
R1′ = 4.99 K Ω
R2′ = 5.11K Ω
Ra = 49.9 K Ω
Rb = 10.0 K Ω
Figure 16 shows the plots of voltage loop expressions of Eqn. (38) and (39). We have
intentionally brought about an approximate cross over frequency of 6KHz with 87o
phase margin for cautionary reasons due to HV noise. In a fabricated physical unit, one
⎛ ⎞
(R+ s ⋅ L f ) ⋅ ⎜ Rc +
L ⎜
1
s ⋅Cf
⎟⎟
Z P = 14 ⋅ ⎝ ⎠ (41)
1
RL + s ⋅ L f + Rc +
s ⋅Cf
where
27
RL = .251Ω
L f = 300 µ H
(42)
Rc = 11.69mΩ
C f = 18µ F
The factor of 14 in Eqn. (41) stems from the fact, that all output stages are stacked up.
ZP
Zo ( s ) = (43)
1 + Tv
for the voltage loop control. Figure 17 shows the plots of Eqn. (43). The maximum closed
loop output impedance is predicted as Z o = 110Ω for all frequencies, and the step load
deviation is predicted as 1.82% at the output of -1750V for .287 A load current
excursion.
Gvg
G A ( s ) = 14 ⋅ (44)
1 + Tv
Figure 18 shows the plots of Eqn. (44). From Figure 18, we predict about 3.3% deviation
of the output voltage of -1750V for a ±10V step input voltage change.
29
5. FAULT PROTECTION
Throughout the load current range, i.e., from no load or open load to the nominal
maximum load, which is about .286A at -1750V, the PPF Converter would operate, as if
it were under CCM regime due to inductor current reversal facilitation at the secondary
through synchronous switch use. What should be done, when the load current exceeds
We can not employ dynamically control over the DC load current in this
application. Theoretically feasible though, but impractical due to very small duty ratios,
that it would impose. For example, under high quality short circuit, the output voltage at
.1
the cellular level may be .1V. With a 220V input, this would entail = .045% duty
220
ratio, and about 4.5ns On Time for the primary switches at 100KHz . This very reasoning
is also valid, if the CIC were applicable. Only the quadratic converters can cope with this
large input-output voltage difference under fault conditions, but they are inefficient for
The next best alternative is to shutdown the converter upon fault condition,
subject to automatic recovery with periodicity. In this way, when the DC output current
exceeds a certain threshold, the unit is shutdown. This shutdown removes the “cognition”
of the fault condition, and thereby restarts the converter, subject to soft start each time.
The fault condition is defined as externally imposed over load and/or short circuit
condition. It does not, however, include internal causes, i.e., internal to the PPF
Converter.
Figure 19 shows the fault current sense arrangement. Since the return of control
electronics and the load is common, we would be sensing a positive voltage drop across
32
the DC current sense resistor, Rcs . Although, we considered sensing a negative voltage
VCC = +15V
1KΩ
1KΩ
CF
VR =5.1V
R1 1KΩ
drop via placing the load ground at the other end of Rcs , but we chose the above approach
based on the Bode Plots of sense amplifier transfer functions. However, there is no
difference between the two alternatives with regard to first 2ms – 5ms of their unit step
response, examined by taking the inverse Laplace transforms of their transfer functions.
At the fault current level of .35A , the sensed voltage is amplified to a voltage
level of about 5.1V across CO , Figure 20. This voltage is then compared to 5.1V
reference and the Mosfet is driven to off state, if the current exceeds .35A. The sudden
rise of Mosfet drain voltage triggers a fast and Retriggerable monostable, Figure 21, and
brings about a forced low at the Soft Start, and forced high at the Shutdown
34
terminals of the PWM IC. By the design of the monostable IC, MC14538BCD, the Q
T = Rx [ Ω ] ⋅ C x [ F ] (45)
In this application, we set it at 124ms , but it can easily be extended to longer durations.
Q and Q outputs drive very low input capacitance Mosfets, which, in turn, bring about
forced low at the Soft Start, and forced high at the Shutdown terminals of the PWM IC.
Due to both the conducted EMI Filter at the main converter input and the clamp
capacitor of the PPF Converter, an in-rush current is experienced when the power is
applied to the 100V Bus. This inrush current is mitigated by use of a temporary resistor in
series with the power bus. The magnitude of the resistor is chosen to reduce the peak
35
inrush current to a magnitude to less than 10 amperes. Figure 22 shows the in-rush
current limit circuit. The two paralleled Mosfet switches are turned on after an about 2s
of time out period elapses. The intend here is to turn them on, when the EMI filter
capacitors are completely charged up the to Bus Voltage level. Figure 23 shows the
simulation plots. The top plot is the EMI Filter Capacitor Voltage. The middle plot
depicts the rail voltage build up for the HKPS PWM IC, and the bottom plot is the in-rush
current profile. The HKPS PWM IC rail voltage reaches the threshold for the HKPS to
start. It declines to the Under Voltage Lock Out (UVLO) level within about 30ms, which
Once the HKPS reaches steady state, the Mosfet switches of the in-rush limit
circuit, Figure 22, are not turned on for another one second. The reason is to allow the
EMI Capacitor Voltages to completely reach the bus voltage level for preventing any
likely excessive current in the P-Channel Mosfet switches. Then, in the further next one
The power dissipated in the in-rush current limiter during this approximately 1.5
1 ⎛ ⎞ ⎛ Vg −τt ⎞
T t
−
T ∫0 ⎝
P= ⎜ g
V ⋅ e τ
⎟⋅⎜ ⋅e ⎟ ⋅ dt ≅ 31.7W (46)
⎠ ⎝R ⎠
where
τ = .0942 s
R = 12Ω (47)
Vg = 110V
τ = exponential time constant, measured from Figure 23. We would be using a 100W
rated and TO-247 Case resistor. The power dissipation by the transient in-rush current
limit function does not pertain to the efficiency concern of the sub unit.
38
7. OPERATIONAL PERFORMANCE
Other than the discussed topics above, the most important next issue is the
operational performance. In the above, we furnished how the primary elemental blocks
worked. In this section, we examine “how does the sub unit work?” along with the
uncovered minor details in the above. In the proceeding, we will also cover the power
The operation starts with DC Power being applied to the main power bus.
Through the in-rush current limit circuit, the initial in-rush current peak is limited to less
than 8A. Figure 24 shows the simulation circuit. As we observe, the Input EMI Filter is a
Pi Filter. R3, C1, Z1, R4 and DIAC2 functionally simulate the Bootstrap start up circuit
of the PWM IC. When the voltage across C1 reaches 16V, the load of about 100mA is
activated via DIAC2. Figure 23 shows the related salient waveforms. During this input
capacitor charge up period, in the first 400ms, the HKPS starts. HKPS generated bias
voltages quickly power the In-Rush Control Time Delay circuit, Figure 25. Until about
one second delay introduced by this circuit elapses, nothing happens. EMI Filter
39
capacitors sit there, all charged up. When the In-rush Control Time delay elapses, The P-
Channel Mosfets in Figure 22 are turned on and locked in the On State. Thus, at this
moment,
1 – All Primary Side Capacitors are charged up to the Bus Voltage level,
2 – All HKPS outputs already settled at their steady state values, and
Meanwhile, the Start-up Delay Circuit, Figure 26, which was powered up at the same
with the In-Rush Control Delay Circuit by the HKPS, but has twice the time delay length,
is ramping up to the moment, at which it will release the Soft Start port of the PWM IC
for starting the main power processor. When 2s time delay elapses, Q60 in Figure 26 is
turned off , and thus, the soft start of the main power processor is initiated.
40
The simulation circuit of the main power processor is too large for embedded here.
Instead, we will discuss its performance through the simulation waveforms. The reader is
advised to view the circuit schematic of the main power processor in sheets 1 & 2 of the
inserts. PSIM Software used in these simulations has some important limitations,
2 – Simulations of time durations greater than 100ms takes too long and often
soft start of the Main Power Processor. Instead, we furnish here what appears to be an
Figure 27 – Startup of the main power processor, referred to the cellular voltage.
feedback control design. On a physical sub unit, the ramp up section in Figure 27 would
take about 1s. The reason for that is to mitigate the output filter inductor currents to a
level close to their steady state nominal maximum. Otherwise, we have to design those
inductors for several times larger DC and HF AC current levels, thus ending up with
heavier magnetics.
The above mentioned limitations of the PSIM Software do not prevent obtaining
the steady state switching waveforms of the converter circuit. Figure 28 shows the
primary switch current under full load of 500W. We measure 5.7 A as the rms value of
Since there are two primary switches, the total power consumption would be 7.14W. This
is at the low line voltage of 90V. Figure 29 shows the primary winding currents. Again
42
(49)
PPT = .366W ⋅ 2 = .732W
The clamp capacitor does not experience any appreciable power dissipation due to
extremely low ESR of the MLC capacitor used. Finally, on the primary side, the input
current to the converter and the input current to the EMI Filter is shown in Figure 30.
On the secondary side, we examine the cellular level waveforms. Figure 31 shows
the output filter inductor current waveform under full load. We measure .199A as the rms
PLcore = 15.2mW
Figure 32 shows the secondary synchronous switch current and voltage waveform. These
switches are driven at almost 50% duty ratio. During their off state time interval, their
anti-parallel diodes turn on for the commutation sub interval. From Figure 32, we
measure .142 A as the switch rms current. Thus, the secondary synchronous switch loss is
Figure 30 – Main converter input current (Top) and EMI filter input current (bottom).
The total secondary synchronous switch losses is 24 mW per cellular output The grand
total secondary switch losses for the main converter is then .336W. We should not feel
concerned about the glitch seen in Figure 32, bottom trace, since the device current is
Since the secondary winding currents are the same as the synchronous switch
Figure 31 – Cellular output filter inductor current waveform under full load.
Figure 33 – Magnified view of Figure 32 at the vicinity of the device voltage glitch.
From Figure 30, we measure the input current to the EMI Filter as 5.6893A. Thus, the
power loss in the P-Channel Mosfet switches for the In-Rush Control,
Using the HF transformer core losses from the previously submitted document, we can
now arrange Table 1 for an initial assessment on the efficiency of the Main Power
Converter. We will update Table 1 later, with the loss estimates from the HKPS.
47
TABLE 1
EFFICIENCY 97.57%
HKPS is based on the same topology with the Main Power Processor, i.e., PPF
Converter. This is because, it is one of the two most efficient topologies, and it is par with
the BIFPPF in most applications. Figure 34 shows the topological circuit of the HKPS,
which generates two sets of ±15V . One set is return referenced to the Main PPF
Converter Secondary, and it powers all control and sense electronics, which are also
return referenced the secondary. The other set is return referenced to the primary side of
the Main PPF Converter, and powers all the electronics at that side. The HKPS PWM IC
is return referenced to the Primary Side of the Main PPF Converter, therefore, the
feedback control is primary return referenced. Neither at the primary side, nor at the
48
secondary side, the peripheral support electronics do require a very precise regulation of
49
the bias or rail voltages of ±15V . Furthermore, the loads, which appear as averaged DC
currents, are pseudo constant, determined by the average current need of the constant
number of “entities” they represent. During the switching transition sub intervals, the
relatively large load current excursions, especially at the +15V rails, are absorbed or
provided by the relatively large value output filter capacitors or intentional large value
ceramic bypass capacitors at the bias ports of the power client ICs.
The HKPS open loop transfer functions are derived in as much the same way as
the Main Converter. Therefore, we would furnish the summary plots of some important
ones, here.. Like in modeling the main power processor, we “fold” the secondary circuits
and refer them to the secondary circuit, where the feedback control is applied. Figure 36
shows the duty ratio to inductor current transfer function. The same reasons of why we
can not apply CIC are valid here, also. Any CIC arrangement would entail a cross over
frequency in the range of 50 KHz − 100 KHz , thus it is impractical, unless we use larger
inductance values, and therefore, heavier magnetics for the output filter inductors.
Nevertheless, the voltage mode control is quite satisfactory, and it is employed here.
Figure 37 shows the duty ratio to output voltage transfer function plots. We
observe that the HKPS is near open loop stable with an approximate 1 KHz cross over
(with FM included) and some sufficient phase margin for better than conditional stability.
The satellite output of the HKPS, which does not have the feedback control, would be
The error amplifier for the HKPS is practically the same circuit as in Figure 11
with the exception that it uses a constant reference, which is generated by the PWM IC.
Figure 38 shows the voltage error amplifier of the HKPS. It is the same circuit in
50
Figure 11, with the exception of component values and the use of constant reference
voltage, which is taken from its PWM IC. For the component values of
51
vˆo
Cp
Rz Rf Cf
R1
Cz
vˆc
R2
Ao
VREF = 5.1V
R1 = 10.0 K Ω
R2 = 5.11K Ω
R f = 48.7 K Ω
Rz = 499Ω (55)
C f = .033µ F
C z = .0047 µ F
Cp = 0
Figure 39 shows the error amplifier transfer function plots for both the ideal and
53
non-ideal opamps. Figure 40 shows the plots of voltage closed loop expression. We have
placed the cross over frequency to a cautious 4 KHz for keeping the second order peak at
around 50 KHz, -20dB or less. We measure approximately over 87o phase margin.
Figure 41 shows the closed loop output impedance transfer function plots. We measure
TABLE 2
EFFICIENCY 97.25%
55
9. REMOTE TELEMETRY
The main converter output voltage is scaled down to about -5V and amplified to
+10V across CO in Figure 42. This 10V becomes the bus voltage to the mini PPF
converter. Using 1 1 turns ratio, and constant duty ratio of near 50%, we transfer the
⎛ ⎞
Vo ⎜⎜⎜ −1750V ⎟⎟⎟
⎝ ⎠
CZ RF
R1
RO
R2 RZ Ao′
Q2
CO
RZ
NP1 D2
NS1
Vo
scale
C1
D3
Cf R Q1
NS2 D4 NP2
D1
output voltage info to an external circuit via galvanic isolation. Employing .1% resistors
58
in the sense amplifier circuit, telemetry with 5% or less worst case accuracy if feasible.
Selecting,
CO ⋅ RO = CZ ⋅ RZ
(56)
RZ R2
R2 R
Vo′ = ⋅ F (57)
R1 + R2 RZ
The estimate of worst case accuracy of this approach to the telemetry is not within the
scope of this report. However, it is strongly believed that a 5% or less accuracy should
be feasible with trimming of R2 and/or RF The load of the mini PPF Converter at its
secondary output port is assumed to be several KΩ . We need to point out here, that the
bandwidth of the mini PPF converter is limited to several hundred Hertz, and therefore
The remote control of the main PPF converter is established via another mini PPF
whose bus voltage is envisioned as variable 10V and the secondary output voltage is 5V
through 2/1 turns ratio. This PPF converter is driven by the telemetry PPF converter gate
signals through gate drive transformers. Its circuit is similar to that shown in Figure 42
without the buffer amplifier circuit. Its output follows the input bus voltage and becomes
The main power processor has total 28 Mosfet Switches. IRHM7C50SE, at its
secondary. Their maximum total gate charge is 150nC. Thus, each device, switching at
as maximum Gate Power Consumption. The total gate power consumption of the
The main PPF converter has four Mosfet switches, IRHM8360, at its primary. Their
maximum total gate charge is 210nC. Thus, the total gate drive power consumption at the
primary,
The total gate drive power consumption of the main PPF converter is then, 7.56W .
secondary. Their total gate charge is 50nC, and each one may consume maximum 75mW.
′ = .075W ⋅ 8 = .6W
PGST (61)
Moreover, the two remote telemetry PPF Converters may experience a maximum of
TABLE 3
Thus, the total sub unit is highly likely to meet about 95% efficiency, when