Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

AND8076/D A 70 W Low Standby Power Supply With The NCP120x Series

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

AND8076/D

A 70 W Low Standby
Power Supply with
the NCP120x Series
Prepared by: Christophe Basso
ON Semiconductor

http://onsemi.com

APPLICATION NOTE
INTRODUCTION
The NCP1200 represents one of the cheapest solutions to
build efficient and cost-effective Switch-Mode Power
Supplies (SMPS). As this design example will show, the part
definition does not confine the component in low-power
applications only, but it can actually be used in Flyback and
Forward supplies for virtually any output power. The below
example depicts a universal mains 90-260 VAC power
supply delivering 16.5 V @ 4.5 A.
Beside its ease of implementation, the NCP1200 excels in
true low standby power designs. This application note
details how an amazing standby power of less than 100 mW
can be reached at high line with a nominal 70 W board.

this level. Thanks to the natural secondary / auxiliary


reflection, the primary auxiliary winding cannot
maintain a sufficient voltage on the control IC: Vcc
collapses and puts the controller in trouble, probably
entering an hiccup mode, similar to that of a startup
sequence. DSS being decoupled from Vout, you never
see that phenomenon.
As you can see, the DSS offers interesting features but, on
the other hand, it can sometimes compromise key design
parameters. Standby power and power dissipation are one of
these:
Standby power: the DSS standby power contribution
can easily be evaluated: VHV Iavg with Iavg, the current
consumption taken by the controller and VHV, the
high-voltage supply rail. If Iavg equals 1 mA, then we
have a standby power of 350 mW at a 350 VDC voltage
rail. Tricks exist to slightly reduce it, like the half-wave
diode, but you will only gain between 2030%.
Power dissipation: as stated above, all the current
consumed by the IC is seen through pin8. This is due to
the self-adaptive feature of the DSS. Should the IC
current move up or down, the DSS duty-cycle will
automatically adjust to deliver it. The controller current
depends on the internal IC consumption, but also on the
type of MOSFET connected to the output. It therefore
important to assess the total current drawn from the HV
rail and checks the right compatibility with the package
type. All details are given in the NCP1200 dedicated
data sheet and the application note AND8023/D.
As a result, the answer lies behind your design constraints.
If you would like to have a precise Over Current Protection
(OCP) trip point while driving a moderate size MOSFET,
DSS can be a good choice, provided low standby power is
not an absolute necessity. In our case, we want to drive a
large MOSFET for a better efficiency but we need to reach
the lowest possible standby power. We will thus adopt an
auxiliary winding configuration to permanently disable the
DSS. Solutions to various combinations of these constraints
are described in the application note Tips and Tricks for the
NCP1200, document number AND8069/D.

DSS or Not DSS?


The Dynamic Self Supply (DSS) lets you directly drive
MOSFETs from the high-voltage rail. This option brings
you several advantages, as stated below:
True overload detection: with UC384X-based systems,
the switching oscillations are stopped in case the Vcc
line drops below a given Undervoltage Lockout level
(UVLO). This principle considers a good coupling
between the primary auxiliary winding and the power
secondary winding. Unfortunately, leakage elements
often degrade this coupling and you only can detect true
short-circuit (when Vout is close to zero) and not
overload conditions. Thanks to the DSS, the NCP1200
does not need an auxiliary information to sense an
overload condition. By detecting a current setpoint
pushed to the maximum, the internal logic takes the
decision to enter into a safe burst operation,
auto-recovering when the default leaves. Precise
overload levels can thus be implemented.
Guaranteed operation at low output levels: the Vcc
delivered by an auxiliary winding moves with the
power output level because a coupling exists between
both windings. When the supply is used in battery
charging applications, Vout can move depending on the
charging state. That is to say, when the battery is nearly
empty, its voltage can be close to zero, forcing Vout at

Semiconductor Components Industries, LLC, 2003

April, 2003 - Rev. 2

Publication Order Number:


AND8076/D

AND8076/D
Self-Powering the Controller in Standby
An auxiliary winding does not usually cause any
self-supply problem with a continuous pulses flow. In
standby, whatever implemented frequency reduction
techniques (e.g. skip or frequency foldback), the recurrence
between pulses can become very low. By definition, the
feedback loop manages to keep the energy content in each
burst high enough to maintain the nominal output voltage.
However, on the auxiliary side, it can be difficult to keep the
Vcc above the controllers UVLO. Remember, to
permanently disable the DSS, you need to guarantee a level
above VccON max. which is 11 V for the NCP1200. Failure
to do this will re-activate the DSS in no-load conditions and
standby power will be degraded. Figure 1 offers a view of a
typical bunch of pulses captured in standby at a 127 VDC
input voltage.

auxiliary output gets clamped by a 15V Zener diode in


nominal operation. Figure 2 shows the option.
We measured a Vcc of 11.5 V @ 230 VAC and 12.2 V @
90 VAC. Rlimit on Figure 2 can easily be adjusted to move
these values up or down, depending on the final winding
ratios. Care must be taken to avoid over-dissipation of the
15 V Zener diode in nominal conditions.
Power Supply, Element-by-Element Design
Lets first detail the specs of our power supply:
Vin: 90265 VAC
Vout: 16.8 V @ 4.2 A (Pout = 70 W)
Short-circuit protection
Over-voltage protection
Efficiency > 80%
Pin = 70 / 0.8 = 87.5
The below sequence details step-by-step the calculation
procedure for every component of the power supply.
DC High-Voltage Rail

33 ms

From these above numbers, we can deduce the level of the


high-voltage rail, neglecting the dual Vf drop:
VHV max  265  2  374 VDC
VHV min  90  2  127 VDC

Vripple
Vinpeak

200 s

VHVavg

Figure 1. A Bunch of Auxiliary Pulses Captured


While the Supply Operates at No-Load
(Vin = 127 VDC)

1
2

10 ms

NCP1200

Rlimit
3

CVcc

1N4148
1

15 V

Caux

Laux

Figure 3. A Typical Ripple Voltage Over the Bulk


Capacitor

Figure 2. The Auxiliary Is Clamped to Avoid


Exceeding the 16 V Maximum Rating

Bulk Capacitor

Figure 3 portrays the typical waveform captured across a


bulk capacitor delivering power to a given charge. To
simplify the calculation, we will neglect the charging period
and thus consider a total discharge time equal to 1/(2 Fline).
From the design characteristics, we can evaluate the
equivalent current (Iload) drawn by the charge at the lowest
input line condition. Lets us adopt a 40% ripple level, or a
50 V drop from the corresponding Vinpeak. To evaluate the
equivalent load current (which discharges Cbulk between

As we previously stated, we want to deliver 70 W with a


16.5 V output level. The maximum rating for the NCP1200
states a level less than 16 V. As a result, the auxiliary Vcc
shall be less than 16V but also above VccON in any
conditions to ensure full DSS de-activation. A solution
consists in artificially raising the ratio between the power
winding and the auxiliary one to ensure adequate supply at
no-load. We successfully tested a 0.9 ratio, where the

http://onsemi.com
2

AND8076/D
the peaks), we divide the input power by the average
rectified voltage:
Iload 

1
tc 

4  Fline

(1)

Pout
Pin

Vripple
Vrectavg

  Vpeak 

(3)

360  Fline

During these 3 ms, Vbulk is the seat of a rising voltage equal


to Vripple or 50 Vpp. This corresponds to a brought charge
Q of:

 860 mA DC @ 90 VAC input voltage

Thanks to Figure 3 information, we can evaluate the


capacitor value which allows the drop from Vpeak down to
Vavg - (Vripple/2) to stay within our 50 V target, dV C =
iload dt:

Qbulk  Vripple  Cbulk  9 mC

Vripple
2

(4)

From Figure 4, we can calculate the amount of charge Q


drawn from the input by integrating the input current over
the diode conduction time:

(2)

Pout

2    Fline  Vripple  Vpeak 

Vmin

VACin 2

 3 ms @ Vin  90 VAC

Cbulk 

sin - 1

 idiode(t).dt
tc

Qin 

 171 F or 180 F for a normalized value

(5)

(Fline = 50 Hz worse case).


The expression of idiode(t) is:
Diode Bridge Selection

Ipeak  tc  t
tc

To select the right rectifiers, it is necessary to know the


RMS current flowing through its internal diodes. Prior to
reach this final result, we need to evaluate the diode
conduction time. From Figure 4, we can see that the diode
starts to conduct when VACin reaches Vmin and stops when
reaching Vinpeak:

After proper integration, it comes:


Qin  1  Ipeak  tc
2

If we now equate Qbulk and Qin and solve for Ipeak, it comes:
Ipeak 

tc
Vinpeak

Vmin

(6)

Qbulk  2
tc

(7)

or 6 A peak. We can now evaluate the RMS current flowing


through the diodes:

Vbulk

Irms 

Fline 

 Ipeak 

 (idiode(t))2  dt
tc

(8)

tc3  2  Fline

 1.9 A @ VAC  90
300 mA

Idiode

We selected a KBU4J diode bridge (600 V/4 A) for the


rectifying function. A small resistor, or best an NTC, can
however be put in series to limit the surge current (when you
plug the SMPS in the AC outlet) to less than the diode
maximum peak current (Ifsm) or what the standard imposes
you.
Thanks to these numbers, we compute the apparent power
at low line: 1.9 A 90 V = 170 VA which compared to our
87.5 Watts of active power (neglecting the input diode
bridge and Cbulk losses) gives a power factor of:

VAC in

X-2 ms/div

Figure 4. When VACin Reaches Vpeak, the Diode


Stops Conducting

From a mathematical point of view, we can calculate the


time VACin takes to reach Vmin, with Vmin = Vpeak Vripple:
VACin  sin(  t)  Vmin

PF  W  0.51
V.A

Since Vpeak is reached at the input sinusoid top (or one


fourth of the input period), then the diode conducting time
tc is simply:

(9)

conform to what we could expect from this kind of offline


power supply.

http://onsemi.com
3

AND8076/D
Transformer Calculation
Transformer calculation can be done in several manners:
a) you evaluate ALL the transformer parameters, electrical
but also physical ones, including wire type, bobbin stack-up
etc. b) you only evaluate the electrical data and leave the rest
of the process to a transformer manufacturer. We will adhere
to the latest option by providing you with a list of potential
transformer manufacturers you can use for prototyping and
manufacturing. However, as you will discover, designing a
transformer for SMPS is an iterative process: once you
freeze some numbers, it is likely that they finally appear
either over or under estimated. As a result, you re-start with
new values and see if they finally fit your needs. To help you
speed-up the transformer design, a design-aid spreadsheet
is available from the ON Semiconductor web site,
www.onsemi.com/pub/NCP1200. Lets start the process
with the turn ratio calculation.

conduction power is evaluating using the following


formula:

Rather than manually calculating these numbers, we will see


later on how a Spice simulator can do the job for us.
Primary Inductance and Peak Current

For AC/DC adapters delivering this amount of power in


a small place, it is of common practice to make the power
supply enter CCM in the middle of the total AC range
(around 180 VAC in our case). When the input AC voltage
diminishes, the on-time increases and the primary /
secondary RMS current go up. This implies a greater
heatsink for the MOSFET but also larger aluminum cans for
the secondary filters. For this reason, a transition from
Discontinuous Conduction Mode (DCM) to CCM will be
envisaged here. Figure 5 depicts these different modes.
Different methods exist to find the point transition takes
place (also called the critical or borderline point). The idea
consists in finding the critical inductance Lc that will make
the supply enter CCM at 180 VAC. From Figure 5, we can
write:

Turn Ratio and Output Diode Selection

The primary/secondary turn ratio affects several


parameters:
The drain plateau voltage during the OFF time: the
lowest plateau gives room for the leakage inductance
spike before reaching the MOSFETs BVdss:
(10)

Np
Vplateau 
 (Vout  Vf)  VinDC max
Ns

The secondary Peak Inverse Voltage (PIV) is linked to

ton 

Lp  Ip
VinDC

(14)

toff 

Lp  Ip
N  (Vout  Vf)

(15)

the turn ratio and the regulated output voltage by:


PIV  Ns  VinDC max  Vout
Np

(11)
CCM

If you lower the plateau voltage, you will increase the


reverse voltage the secondary diode must sustain.
With these numbers in mind, you can tweak the turn ratio
according to the MOSFET BVdss and the diode maximum
reverse voltage. A Schottky diode represents a good choice,
especially with a power supply that can possibly enter
Continuous Conduction Mode (CCM). The lack of reverse
recovery loss and a low forward drop play in favor of this
component. However, because of the metal-silicon
junction, moderate breakdown voltages are available for a
moderate cost. The MBR20100 represents an interesting
choice since it welcomes two 100 V Schottky in a TO-220
package. Being in thermal contact, a parallel wiring is
possible. The 100 V VRRM lets us calculate the minimum
turn ratio we can go down to, keeping an acceptable safety
margin:
N  PIV  Vout
VinDC max

(13)

Pdiodeavg  Vf  Idavg  Rd  Idrms2

L > Lc
IL

IP

Not 0 at
turn ON
L = Lc

ON

L > Lc

DCM

OFF

IL(avg)
0 before
turn ON
BCM
D/Fs

Dead-time
time

(12)

Figure 5. Depending on the Primary Current at Turn- On,


the Supply Crosses Various Operating Modes

Np:Ns 1:0.221. A final ratio of 1:0.166 offers an


adequate safety margin (Vreverse = 80 V max). The diodes

http://onsemi.com
4

AND8076/D
4.50
3.50
2.50
1.50
500M

DRV

Vramp

Radd1
CS

1
C

14.0
10.0

Rsense

6.00
2.00

Radd2

-2.00

150

10.0U

Figure 6. A Very Simple Way to Generate a


Ramp from a Square Wave Signal

 2Lp Pout
 FSW

(16)

90.0U

Ramp Compensation

With a supply entering CCM together with a duty-cycle


greater than 50%, we need to inject ramp compensation into
the controller to prevent subharmonic oscillations. An easy
way to generate a ramp, is to take the driving signal available
from pin5 and integrate it through a RC network. Figure 6
shows how to wire the components and Figure 7 shows the
signal obtain with a 18 k / 1 nF RC time constant.
To calculate the necessary amount of ramp m, several
methods exist. We will stick to the standard one which
consists in injecting between 50 and 75% of the off-time
downslope. The calculation is as follow:
Primary off-slope:

(Vout2  2  Vout  Vf  Vf2)  (eff  N2  Vin2) (17)


[Pout  [(N  Vout  N  Vf  Vin)2  FSW]]  2

The numerical application gives a 484 H inductance with


a peak current of 2.36 A. The NCP1200 incorporates a
skip-cycle feature that forces the controller to slice the
switching pattern when the power supply drives light loads.
Depending on the system time constants, the recurrence of
the burst can enter the audible frequency range. Since the
default skip-cycle takes place at one third of maximum peak
current, it is better to avoid working at high peak current in
normal operation. Should noise still appear in skip mode,
pin1 lets you select a different lower skip level
(unfortunately to the detriment of the standby power)
generating less mechanical noise. As a result, we slightly
increased the primary inductance to 700 H to further limit
the noise in standby operation.

N  (Vout  Vf)
 153 mAs
Lp

D

(20)
Vout
 45% @ Vin  120 VDC
N  Vin  Vout

From Figure 6 network, the maximum voltage is given by


R and Radd1 + Radd2. With a 11 V driving voltage delivered
by the NCP1200, we recommend a 18 k for R and 1 nF for
C. These values offer an acceptable tradeoff in terms of
power consumption but also in terms of noise immunity. The

The MOSFET drain voltage sees, in normal operation, a


maximum voltage of:
Lleak
Clump

(19)

Once reflected over Rsense, it becomes: 50.5 mV / s (S)


Duty-cycle in CCM:

MOSFET Selection

VinDC max  (Vout  Vf)  N  Ip 

70.0U

The first term represents the maximum rectified DC voltage


and goes up to 375 V. The reflected voltage pushes further
up by 101 V. Summing up these levels gives a total
steady-state drain voltage of 476 V. The last term in equation
18 depicts the leakage inductance action which further
stresses the MOSFET at the opening. If we select a 600 V
device, it leaves more than 100 V for this leakage action. A
clamping network will stop its rise anyway. A 2SK2843
from Toshiba can be a good choice. This is a TO-220 600 V
10 A component which features a 1.2 RDS(ON) @ Tj =
100C.

Ip = primary peak current


N = Np / Ns = 1/0.166 = 6
Pout = output power
= efficiency
Lp = primary inductance
Fsw = switching frequency
Vf = secondary diode forward drop
VinDC Vac 2 (neglecting ripple)
Combining equations 14, 15 and 16 we obtain an Lp value
to be in BCM at 180 VAC input voltage:
Lp 

50.0U

Figure 7. Simulations Show a Capacitor Voltage Ramping


Up from a Few Hundred of mV Up to Nearly 5 V

From the Flyback formula, we obtain:


Ip 

30.0U

(18)

http://onsemi.com
5

AND8076/D
Component Constraints
In this section, we will see how Spice can help us to
precisely determine all the component constraints and thus
calculate the necessary amount of heatsink they need. The
simulation schematic we adopted is given on Figure 8 and
shows the NCP1200 wired as recommended in the data
sheet. Please note that the simulation fixture has been
simplified to allow faster simulation time. For instance, the
TL431 and its network usually hamper the simulation time
to find out the right operating point. A Zener diode and a
resistor help finding it in a much quicker way.

Vcc  Vdrv  1  e 
 5 V

(21)

with t = 0.45 1/61 k. This provides an available ramp level


of 677 mV/s (S). By setting Radd2 to 1 k, Radd1 can be
computed using the following formula:
Radd1  1 k 

S  22.5 k
S  m

for m = 60%. Final tweak gives an 18k resistor for Radd1.

X1
XFMR
RATIO = -0.166

hv

R3
200 m
7

Iout

Vsec
X6
L1
MBR20100
100 H
15

R10
10 m
31

corresponding time constant of 18s gives a ramp maximum


peak voltage of:

Vout
Vout

17

Iprim

Vclamp

Istartup

Isec

C3
100 nF

Iripple1

12

Rload
3.7

Iripple2
32

C1
3.3 mF
IC = 16.5

Lleak
12 H

D4
MUR160

Resr2
300 m

24

+
Vinput
120

R2
4.7 k

Resr1
11 m

Lp
700 H

C2
220 F
IC = 16.5

Vdrain
Iclp
1

Vadj

2
21

Idrain

NCP1200

10

NCP1200
Fs = 65 k

Rg
Vcc drv 10

R15
1.0 k

19

X5
SFH610A

20

Drv
14

R8
1.0 k

R5
100
m

R10
18 k

D1
1N4148
18

D3
1N963

Vsense
R9
18 k

16

Cvcc
47 F
IC = 14

23

C6
1.0 nF

Rsense
0.33

VFB

Figure 8. The Simplified Simulation Schematic Helps to Determine All the Component Key Parameters

http://onsemi.com
6

AND8076/D

1 Idrain

2 Iprim

3 Iripple1

4 VCLAMP
RMS = 1.10 amps between
2.05 M and 2.06 M secs

3.00
2.00
1.00

0
-1.00
RMS = 1.53 amps between
2.05 M and 2.06 M secs

2.40

2.00
1.60
1.20
800 M
RMS = 4.65 amps between
2.04 M and 2.05 M secs

10.00
6.00

2.00
-2.00
-6.00
y (mean) = 152 volts
x (first) = 2.00 M secs

156
154

152
150
148
2.01 M

2.03 M

2.05 M

2.07 M

2.09 M

Figure 9. Complete Simulation Results of the 70 W Converter Operated at 120 VDC Input Voltage

Important results appear in Figure 9. Please note that the


maximum RMS current occur at the lowest line where the
duty-cycle is pushed to the limit.
As you can see, the ramp compensation works fine and no
subharmonic oscillations can be noted. Once everything is
extracted, below are summarized the most important design
constraints:

Lower Rheatsink- air resistances can of course be selected to


run the device cooler.
Diode

The MBR20100 welcomes two diodes that share nearly


equal current thanks to their equal forward drops. The total
forward drop dissipation will remain the same but the RMS
losses sensitive to the dynamic resistance will divide by two:
IRMS total = 6.8 A
IAVG total = Iout = 4.2 A
Rd @ 3.4 Arms = 27 m
Vf @ 2.2 Aavg = 0.7 V
Pcond for one diode = 3.42 0.027 + 2.2 0.7 = 1.85 W
or 3.7 W for the whole TO-220 package. Simulations gives
a bit less to 3.4 W. Heat calculations (Tj < 100C and 50C
ambient) recommend a heatsink of 8C/W for the
MBR20100. As stated before, lower Rheatsink- air
resistances can of course be selected to run the device cooler.

MOSFET

Rdson @ 100C = 1.2 ohms


RthetaJC = 2.8C/W
Pcond = 1.2 * 1.12 = 1.5 W
The conduction losses are the strongest at low line. The total
simulated losses, including switching events are evaluated to
be around 2.6 W. Further breadboard measurements confirmed
this number. If we want to keep the junction temperature
around 100C at an ambient of 50C, then we shall add a
proper heatsink according to the following calculation:
Rheatsink- air 

(Tj max  Tamb max)


P

Capacitors

Icapacitor RMS = 5 A
The paralleling of capacitors will help achieve the right
ripple current shared between all the devices. We selected
three 2.2 mF capacitors capable of handling 1.7 Arms each.

 RJunction- Case  RCase- Heatsink


 15CW

http://onsemi.com
7

AND8076/D
Transformer

network analyzer and confirms the validity of our approach


(Vin = 240 VAC).
Stability has been checked at various line/loads
combinations and gave good results. Final transient step did
not reveal any overshoot or unwanted oscillations.

Below are the key parameters you will pass to your


transformer manufacturer to help him select the right
winding size and tailor the internal gap:
Maximum peak primary current, including 160 ns
propagation delay: 1 / 0.33 + 374 160 n / 700 = 3.2 A
Maximum primary RMS current at low line: 1.6 A
Maximum secondary RMS current: 6.9 A
Primary inductance: 700 H
Turn-ratio, power section: Np:Ns = 1:0.166
Turn-ratio, auxiliary section: Np:Naux = 1:0.15

The Adapter Schematic


The final schematic implements a current-mode Flyback
architecture, driving a 600 V MOSFET. The 2SK2545
features a 10 A capability but a 6.0 A/600 V can also be
mounted, such as the FQP6N60 from Fairchild but to the
expense of increased conduction losses. Figure 13 offers a
complete view of the electrical sketch. The board can
actually be used with either auxiliary or without auxiliary
winding. By removing the resistance R4, you reactivate the
DSS on a NCP120X controller featuring this ability. The
board can therefore accept the following controllers:
NCP1200, featuring DSS.
NCP1200A, featuring DSS.
NCP1203, auxiliary winding only.
NCP1216, featuring DSS and internal ramp compensation.
Improved EMI jittering with DSS.
NCP1217, auxiliary winding only, internal ramp
compensation.
On NCP1216 and 1217, the internal ramp compensation
avoids using the external circuitry made of R1-R7-D4 and
C10. If one of these two parts are plugged in the demoboard,
you must disable this network by simply disconnecting R1
and growing R6 up to 2.7 k (for a 65 kHz operation).
The NCP120X takes place with two other bipolars that
implement a discrete SCR, activated in presence of an OVP,
e.g. an optocoupler failure. D5 senses the overvoltage
condition and can easily be adjusted to fit any other levels.
Thanks to R10, the OVP permanently latches-off the supply
and the user shall cycle VCC off and on again to restart the
supply. Shutdown is obtained by pulling the feedback pin
down through D6. The clamp resistor is split in two different
components to avoid an excessive heat burden on one single
device. Both main MOSFET and secondary diode are
mounted on an adequate heatsink to evacuate the heat.
To ease the designer task, or simply help evaluating the
board performance faster, we have experimented different
transformers, available through Appendix B manufacturers.
Please note that some include the auxiliary winding for DSS
deactivation whereas other only offer a dual winding
arrangement where the DSS no longer activates and offers
the best standby power. All details are given in appendix B.
The final demoboard will not accommodate with all these
transformers simply because multiple footprints was not
possible. They however have all been tested okay.
Measurements were taken with the Coilcraft transformer.
As a final note, the actual demoboard delivers
19 VDC/70 W versus the original design-based 16.5 V. As
a result, figure 1b circuit has been replaced by L3-R13 and
C12 to improve the short-circuit protection when using an
auxiliary winding. Output voltage can be adjusted by
changing the feedback network made of R12/R20/R21.

Clamping Network

The clamping network can be calculated using the


following formulae:
(22)
Rclamp
2  Vclamp  (Vclamp  (Vout  Vf sec)  N)

Lleak  Ip2  FSW

Cclamp 

Vclamp
Vripple  FSW  Rclamp

(23)

The power dissipated by Rclamp can also be expressed by:


PRclamp  1  Lleak  Ip2  FSW 
2

Vclamp
(VoutVf sec)N
Vclamp

(VoutVf sec)N

(24)

with:
Vclamp: the desired clamping level;
Ip: the maximum peak current (e.g. during overload);
Vout + Vf: the regulated output voltage level + the
secondary diode voltage drop;
Lleak: the primary leakage inductance;
N: the Ns:Np conversion ratio;
Fsw: the switching frequency;
Vripple: the clamping ripple, could be around 20 V.
With a measured leakage inductance of 12 H and a final
clamping level of 150 V, Rclamp is found to be 4.7 k/6 W
and Cclamp 100 nF. The RMS current flowing through
Cclamp is 220 mA. RC networks are economical clamping
devices and care must be taken to not exceed the MOSFET
BVdss in the most stringent conditions, e.g. a cold startup
sequence at high line. Worse case arises when Ip is
maximum and Vout reaches the target.
Stability Analysis
The stability analysis can be investigated using different
approaches. Spice has proven to be rather accurate for
feedback loop analysis with SMPS. We will use the
NCP1200 average model which is available to download
from our web site (www.onsemi.com/pub/ncp1200). Figure
10 shows the simulation template where the feedback
network on the TL431 has been simplified to a simple 100nF
capacitor.
Thanks to average modeling, the simulation time is kept
short and results are delivered in a snap-shot, as testified by
Figure 11. Figure 12 unveils the results obtained using a

http://onsemi.com
8

AND8076/D

CTRL

2.93
340

Vin

+
Vin
350

OUT

11

17.5
Resr1
30 m

Rload
4.2

16.8
9

C1
5.8 mF

C2
100 F

out2

R15
1.0 k

X3
SFH610A
2.93

16.8

out1

Vout

R17
300 m

16.9

X1
NCP1200_Av
FS = 61 k
L = 700
MC = 39605
RI = 0.33

0 14

out2
16.8

16.9

12

LoL
1.0 kH

CoL
1.0 kF

VStim
AC = 1

FB GND

Rs
10 m

NCP1200
averaged
IN

Iout

D1
X1x
MBR20100CT out1 L1
XFMR
10 H
16.9
105 RATIO = 0.166

16.1

15

R5
4.7 k
15.4

Rupp
39 k

C1
100 nF

10

C5
1.0 nF

2.50

X5
TL431

13

Rlow
6.8 k

Figure 10. The Simulation Schematic for Our 70 W Current-Mode Power Supply

Phase (deg)

Mag (dB)
48.0

160
Phase

24.0

80.0

48.0

160.0

36.0

120.0

24.0
12.0

Gain

80.0

Phase

40.0
0

-40.0

-12.0
-24.0

-80.0

-48.0

-160
0

10

1k

10 k

Mag

-24.0

-80.0

-36.0

-120.0

-48.0

-160.0
10

100 k

Figure 11. Simulated Bode Plot of the


Current Mode Flyback

100

1k

10 k

100 k

Figure 12. Measured Open-Loop Gain


with a Network Analyzer

http://onsemi.com
9

http://onsemi.com

10

Universal
Input

2.2
R2
5 W, fuse type

R3
1M
C2
X2
470 nF

2 27 mH
L2

B1
2KBP08M

C3
220 F
+

2N2222
Q2
R8
10 k

2N2907
Q1

R5
10 k

OVP
(optional)

C11
10 nF

C23
1.0 nF*

D6
BAT54

D5
27 V

Vcc 6
Drv 5

3 CS
4 GND

R1
18 k
Ramp Comp
if needed

C24
47 F

2 FB

R6
1.0 k *

HV 8

IC3

R10
12 k

1 Adj

* = Close to the IC

R1A
39 k

C10
1.0 nF

R1B
39 k

L3
47 H

D1
MBR20100

C5 C6 C7

C25
2.2 nF
Y1 type

IC1
SFH6156-2

M1
FQP6N60

IC2
TL431

C8
open

L1
10 H

R11
1.0 k

R18
1.0 k

2200 F
Lp = 700 H
T1
1:0.15 for Np:Ns_aux
1:0.166 for Np:Ns_power

3 1  SMD //

R7A, B, C
0.33, total 1 W

D4
1N4148

R9
47

D2

C12, 10 nF

R13
1.5 k

D3
1N4148

D8
MUR160

D9
15 V
R7
18 k

C4
100 F

R4
220

C1
100 nF

2 39 k 3 W in //

R20
10 k

R12
27 k

R21
5.6 k

C9
100 nF

GND

C220
220 F

R19
open

19 V
@ 3.6 A

AND8076/D

Figure 13. The Simulation Schematic for Our 70 W Current-Mode Power Supply

16.740

16.765

16.735

16.760
16.755

16.730

Output Voltage (V)

Output Voltage (VDC)

AND8076/D

16.725
16.720
16.715
16.710

240 VAC

16.745
110 VAC

16.740
16.735
16.730
16.725

16.705
16.700
100

16.750

16.720
120

140

160
180
200
Input Voltage (VAC)

220

16.715

240

Figure 14. Line Regulation Is Excellent Thanks to


Current Mode and a Good Open-Loop DC Gain

20

40
Output Power (W)

60

80

Figure 15. Load Regulation at Two Different


Input Voltages

Efficiency
We have designed two boards, one using the auxiliary
winding for best standby performance, and another one with
the Dynamic Self-Supply (DSS) left normally working.
Because of the auxiliary winding, it has been necessary to
further clamp the drain voltage in order to improve the
primary overload detection. It is not necessary with the DSS
and therefore the RCD drain clamp network can be less
aggressive, thus slightly improving the efficiency. Board 2
also features a 6 A MOSFET compared to a 3 A MOSFET
on board 1.
Board 1, aux. winding: Vin = 110 VAC, = 79%
Vin = 240 VAC, = 83.5%
Board 2, DSS:
Vin = 110 VAC, = 83.4%
Vin = 240 VAC, = 84.8%

Board Final Results


Standby Power

Measured on an Infratek watt-meter operated in


Watt-hour accumulation mode for best accuracy (run length
= 30 minutes).
Vin = 120 VAC, Vout = 16.76 V, Iout = 0 Pin = 78 mW
Vin = 240 VAC, Vout = 16.76 V, Iout = 0 Pin = 84 mW
Line Regulation

The array in Figure 14 shows the performance when the


input voltage is moving between both range ends. As one
can see, current mode control with good open- loop gain
ensures a Vout less than 1 mV for a 212 VDC input
variation (-106 dB DC audio susceptibility).
Load Regulation

By varying the load current between 11 W and 70 W, it is


possible to plot the load regulation of the board as shown in
Figure 15.

http://onsemi.com
11

AND8076/D
Appendix A, Bill of Material
All resistors are 5% 1/4 W SMD 1206 unless otherwise noted.
All SMD capacitors are 1206 SMD 16 V types unless otherwise noted.
All through-hole electrolytic capacitors are radial types unless otherwise noted.
Manufacturer references are given for specific components only.
R1

18 k

C9

100 nF

R2

2.2 , 5.0 W fuse resistor or


3.15 A/250 V T fuse

C10

1.0 nF

C11

10 nF

R1A, B

39 k, 3.0 W, PRO3, thru


holes

C12

10 nF

R3

1.0 M (not wired on demo)

C22

470 F/35 V/radial

R4

220

C23

1.0 nF

R5

10 k

C24

47 F/25 V/radial

R6

1.0 k

C25

2.2 nF-Y1 security device

R7

18 k

B1

600 V-4.0 A diode bridge


KBU4J

General Semi

R7A, B, C

1.0 1.0 W SMD

D1

MBR20100

R8

10 k

47 , thru holes

D2

1N4148

R9

D3

1N4148

R10

12 k

D4

1N4148

R11

1.0 k

D5

27 V/400 mW

R12

27 k, thru holes

D6

BAT54

R13

1.5 k

D8

MUR160

R18

1.0 k, thru holes

D9

15 V/400 mW

R19

Not wired, open for


feedback options

IC1

SFH6156-2

R20

10 k

IC2

TL431 TO-92

ON Semiconductor

R21

5.6 k

IC3

NCP1200P60

ON Semiconductor

C1

100 nF/400 V

Q1

2N2907

ON Semiconductor

C2

470 nF/X2 security device

Q2

2N2222

ON Semiconductor

C3

220 F/400 V snap-in

Philips
2222-157-46221

L1

PCV-2-103-05

Coilcraft

L2

B82724-A2142-N1

EPCOS

L3

47 H

ON Semiconductor

ON Semiconductor
ON Semiconductor
Infineon

C4

100 F/35 V

C5

2200 F/25 V/radial

Philips
2222-136-50222

M1

2SK2543 (Toshiba) or
FQP6N60 (Fairchild)

C6

2200 F/25 V/radial

Philips
2222-136-50222

T1

Z9260-A or Z9007-B

Coilcraft

2200 F/25 V/radial

Heatsink 1

KL194/38,1 SW (diode)

Seifert

C7

Philips
2222-136-50222

Heatsink 2

KL195/38,1 SW (MOSFET)

Seifert

Not wired, open for


feedback options

C8

http://onsemi.com
12

47 H

AND8076/D
Appendix B, Transformer Manufacturers
Thomson Multimedia - Orega
Route de Noiron
B.P. 24
70101 GRAY Cedex - France
Tel : 33 (0)3 84 64 54 26
Fax: 33 (0)3 84 65 18 45
www.thomsonmultimedia.com
Email: Bouillotj@thmulti.com
Ref. : G7086-01, no aux. winding, P = 70 W

Eldor Corporation Headquarter


Via Plinio 10,
22030 Orsenigo
(Como) Italia
Tel. : +39-031-636 111
Fax : +39-031 636 280
eldor@eldor.it
www.eldor.it
ref. : 2074.5059A, no aux. winding, P = 70 W
Pulse Engineering
Site dOrgelet
Zone industrielle
39270 - ORGELET
Tel. : 33 (0)3 84 35 04 04
Fax: 33 (0)3 84 25 46 41
http://www.pulseeng.com/
Email: vpelletier@pulseeng.com
ref. : PF0082, with auxiliary winding, P = 50 W
ref. : PF0091, without auxiliary winding, P = 50 W
Coilcraft
1102 Silver Lake Road
Cary, Illinois 60013 USA
Tel: (847) 6396400
Fax: (847) 6391469
Email: info@coilcraft.com
http://www.coilcraft.com
ref. : Z9260-A, with auxiliary winding, P = 70 W
ref. : Z9007-B, without auxiliary winding, P = 70 W

For Lower Volumes:

Atelier Special de Bobinage


125 cours Jean Jaurs
38130 ECHIROLLES - France
Tel. : 33 (0)4 76 23 02 24
Fax: 33 (0)4 76 22 64 89
Email: asb@wanadoo.fr
Ref. : NCP120035 WUM, no aux. winding,
RM10 P = 35 W

http://onsemi.com
13

AND8076/D

ON Semiconductor and
are registered trademarks of Semiconductor Components Industries, LLC (SCILLC). SCILLC reserves the right to make
changes without further notice to any products herein. SCILLC makes no warranty, representation or guarantee regarding the suitability of its products for any
particular purpose, nor does SCILLC assume any liability arising out of the application or use of any product or circuit, and specifically disclaims any and all
liability, including without limitation special, consequential or incidental damages. Typical parameters which may be provided in SCILLC data sheets and/or
specifications can and do vary in different applications and actual performance may vary over time. All operating parameters, including Typicals must be
validated for each customer application by customers technical experts. SCILLC does not convey any license under its patent rights nor the rights of others.
SCILLC products are not designed, intended, or authorized for use as components in systems intended for surgical implant into the body, or other applications
intended to support or sustain life, or for any other application in which the failure of the SCILLC product could create a situation where personal injury or death
may occur. Should Buyer purchase or use SCILLC products for any such unintended or unauthorized application, Buyer shall indemnify and hold SCILLC
and its officers, employees, subsidiaries, affiliates, and distributors harmless against all claims, costs, damages, and expenses, and reasonable attorney fees
arising out of, directly or indirectly, any claim of personal injury or death associated with such unintended or unauthorized use, even if such claim alleges that
SCILLC was negligent regarding the design or manufacture of the part. SCILLC is an Equal Opportunity/Affirmative Action Employer.

PUBLICATION ORDERING INFORMATION


Literature Fulfillment:
Literature Distribution Center for ON Semiconductor
P.O. Box 5163, Denver, Colorado 80217 USA
Phone: 303-675-2175 or 800-344-3860 Toll Free USA/Canada
Fax: 303-675-2176 or 800-344-3867 Toll Free USA/Canada
Email: ONlit@hibbertco.com

JAPAN: ON Semiconductor, Japan Customer Focus Center


2-9-1 Kamimeguro, Meguro-ku, Tokyo, Japan 153-0051
Phone: 81-3-5773-3850
ON Semiconductor Website: http://onsemi.com
For additional information, please contact your local
Sales Representative.

N. American Technical Support: 800-282-9855 Toll Free USA/Canada

http://onsemi.com
14

AND8076/D

You might also like