REV.
ADN8830
–15–
OUT B
C1
R1
R
L
L1R2
PVDD
P1
Q1
N1
Q2
OUT A
V
X
DENOTES
PGND
Figure 14. Equivalent Circuit for PWM Amplifier and Filter
In this circuit, R
L
is the TEC resistance, R2 is the parasitic
resistance of the inductor combined with the equivalent r
DS, ON
of Q1 and Q2, and R1 is the ESR of C1. The voltage, V
X
, is the
pulse-width modulated waveform that switches between PVDD
and ground. This is a second-order low-pass filter with an exact
cutoff frequency of
f
RR
RRCL
C
L
L
=
+
+
()
1
2
2
111
π
(25)
Practically speaking, R1 and R2 are several tens of milliohms and
are much smaller than the TEC resistance, which can be a few
ohms. The cutoff frequency can be roughly approximated as
f
CL
C
=
1
2
1
11π
(26)
This cutoff frequency should be much lower than the clock
frequency to achieve adequate filtering of the switched output
waveform. Also of importance is the damping factor, , of the
L-C filter. Too low a damping factor will result in a longer
settling time and could potentially cause stability problems for
the temperature control loop. Neglecting R1 and R2 again, the
damping factor is simply
ζ
=
1
2
1
1R
L
C
L
(27)
Using the recommended values of L1 = 4.7 μH and C1 = 22 μF
results in a cutoff frequency of 15.7 kHz. With a TEC resistance
of 2 Ω, the damping factor is 0.12. The cutoff frequency can be
decreased to lower the output voltage ripple with slower clock
frequencies by increasing L1 or C1. Increasing C1 may appear
to be a simpler approach as it would not increase the physical
size of the inductor, but there is a potential stability danger in
lowering the damping factor too far. It is recommended that ζ
remain greater than 0.05 to provide a reasonable settling time
for the TEC. Increasing ζ also makes finding the proper PID
compensation easier as there is less ringing in the L-C output
filter. To allow adequate phase and gain margin for the PWM
amplifier, Table III should be used to find the lower limit of
cutoff frequency for a given damping factor.
Table III. Minimum L-C Filter Cutoff
Frequency vs. Damping Factor
f
C, MIN
(kHz)
0.05 8
0.1 4
0.2 2
0.3 1.9
0.5 1.6
> 0.707 1.5
Calculating PWM Output Ripple Voltage
Although it may seem that f
C
can be arbitrarily lowered to reduce
output ripple, the ripple voltage is also dependent on the ESR of
C1, shown as R1 in Figure 14. This resistance creates a zero
that turns the second-order filter into a first-order filter at high
frequencies. The location of this zero is
Z
RC
1
1
211
=
π
(28)
With a clock frequency greater than Z1, and presumably greater
than f
C
, the output voltage ripple is
ΔΔOUT A I R
L
1
(29)
ΔOUT A
VD DR
Lf
for f Z
DD
CLK
CLK
=
()
>
()
11
1
1
(30)
The worst-case voltage ripple occurs when the duty cycle of the
PWM output is exactly 50%, or when OUT A = 0.5 V
DD
. As
shown in Equation 31
OUT A
VR
fL
for f Z
MAX
DD
CLK
CLK
≈>
()
1
41
1
(31)
Here it can be directly seen that increasing the inductor value or
clock frequency will reduce the ripple. Choosing a low ESR
capacitor will ensure R1 remains low. Operating from a lower
supply voltage will also help reduce the output ripple voltage
from the L-C filter. With a clock frequency equal to Z1 but
presumably greater than f
C
, the worst-case output voltage ripple is
ΔOUT A V
RC f
LC f
for f Z
MAX DD
CLK
CLK
CLK
=
+
()
=
()
16 1 1 1
32 1 1
1
22
2
(32)
Which, if f
CLK
< Z1, can be further simplified to
ΔOUT A
V
LC f
for f Z
MAX
DD
CLK
CLK
=<
()
32 1 1
1
2
(33)
A typical 100 μF surface-mount electrolytic capacitor can have
an ESR of over 100 mΩ, pulling this zero to below 16 kHz, and
resulting in an excess of ripple voltage across the TEC. Low ESR
capacitors, such as ceramic or polymer aluminum capacitors,
are recommended instead. Polymer aluminum capacitors can
provide more bulk capacitance per unit area over ceramic ones,
saving board space. Table IV shows a limited list of capacitors
with their equivalent series resistances.
This is by no means a complete list of all capacitor manufacturers
or capacitor types that can be used in the application. The 22 μF
capacitor recommended has a maximum ESR of 35 mΩ, which
puts Z1 at 207 kHz. Using a 3.3 V supply with the recommended
inductor and capacitor listed with a 1 MHz clock frequency will
yield a worst-case ripple voltage at OUT A of about 6 mV.
External FET Requirements
External FETs are required for both the PWM and linear amplifiers
that drive OUT A and OUT B from the ADN8830. Although it
is important to select FETs that can supply the maximum current
required to the TEC, they should also have a low enough resis-
tance (r
DS, ON
) to prevent excessive power dissipation and improve
efficiency. Other key requirements from these FET pairs are
slightly different for the PWM and linear outputs.
D
REV. –16–
ADN8830
The gate drive outputs for the PWM amplifier at P1 (Pin 21)
and N1 (Pin 22) have a typical nonoverlap delay of 65 ns.
This is done to ensure that one FET is completely off before
the other FET is turned on, preventing current from shooting
through both simultaneously.
The input capacitance (C
ISS
) of the FET should not exceed 5 nF.
The P1 and N1 outputs from the ADN8830 have a typical output
impedance of 6 Ω. This creates a time constant in combination
with C
ISS
of the external FETs equal to 6 Ω C
ISS
. To ensure
shoot-through does not occur through these FETs, this time
constant should remain less than 30 ns.
The linear output from the ADN8830 uses N2 (Pin 10) and
P2 (Pin 11) to drive the gates of the linear side FETs, shown as
Q3 and Q4 in Figure 1. Local compensation for the linear ampli-
fier is achieved through the gate-to-drain capacitances (C
GD
) of
Q3 and Q4. The value of C
GD
, which can be determined from
the data sheet, is usually referred to as C
RSS
, the reverse transfer
capacitance. The exact C
RSS
value should be determined from a
graph that shows capacitance versus drain-to-source voltage,
using the power supply voltage as the appropriate V
DS
.
To ensure stability of the linear amplifier, the total C
GD
of the
PMOS device, Q3, should be greater than 2.5 nF and the total
C
GD
of the NMOS should be greater than 150 pF. External
capacitance can be added around the FET to increase the effective
C
GD
of the transistor. This is the function of C6 in the typical
application schematic shown in Figure 1. If external capacitance
must be added, it will generally only be required around the
PMOS transistor.
In the event of zero output current through the TEC, there will
be no current flowing through Q3 and Q4. In this condition,
these FETs will not provide any small signal gain and thus no
negative feedback for the linear amplifier. This leaves only a
feedforward signal path through C
GD
, which could cause a
settling problem at OUT B. This is often seen as a small signal
oscillation at OUT B, but only when the TEC is at or very near
zero current.
The remedy for this potential minor instability is to add
capacitance from OUT B to ground. This may need to be deter-
mined empirically, but a good starting point is 1.5 times the
total C
GD
. This is the function of C12 in Figure 1. Note that
while adding more C
GD
around Q3 and Q4 will help to ensure
stability, it could potentially increase instability in the zero current
dead band region, requiring additional capacitance from
OUT B to ground.
Bear in mind that the addition of these capacitors is only
for local stabilization. The stability of the entire TEC appli-
cation may need adjustment, which should be done around the
compensation amplifier. This is covered in the Compensation
Loop section.
There is one additional consideration for selecting both the
linear output FETs; they must have a minimum threshold
voltage (V
T
) of 0.6 V. Lower threshold voltages could cause
shoot-through current in the linear output transistors.
Table V shows the recommended FETs that can be used for the
linear output in the ADN8830 application. Table V includes the
appropriate external gate-to-drain capacitance (external C
GD
)
and snubber capacitor value (C
SNUB
) connected from OUT B to
ground that should be added to ensure local stability. Table VI
shows the recommended PWM output FETs. Although other
transistors can be used, these combinations have been tested
and are proved stable and reliable for typical applications.
Data sheets for these devices can be found at their respective
websites:
Fairchild – www.fairchildsemi.com
Vishay Siliconix – www.vishay.com
International Rectifier – www.irf.com
Calculating Power Dissipation and Efficiency
The total efficiency of the ADN8830 application circuit is simply
the ratio of the output power to the TEC divided by the total
power delivered from the supply. The idea in minimizing power
dissipation is to avoid both drawing additional power and reduc-
ing heat generated from the circuit. The dominant sources
of power dissipation will include resistive losses, gate charge
loss, core loss from the inductor, and the current used by the
ADN8830 itself.
The on-channel resistance of both the linear and PWM output
FETs will affect efficiency primarily at high output currents.
Because the linear amplifier operates in a high gain configuration,
it will be at either ground or V
DD
when significant current is
flowing through the TEC. In this condition, the power dissipation
through the linear output FET will be
PrI
FET LIN DS ON TEC,,
2
(34)
using either the r
DS, ON
for the NMOS or the PMOS depending
on the direction of the current flow. In the typical application
setup in Figure 2, if the TEC is cooling the target object, the
PMOS is sourcing the current. If the TEC is heating the
object, the NMOS will be sinking current.
Table IV. Partial List of Capacitors and Key Specifications
Value (F) ESR (m) Voltage Rating (V) Part Number Manufacturer Website
10 60 6.3 NSP100M6.3D2TR NIC Components www.niccomp.com
22* 35 8 ESRD220M08B Cornell Dubilier www.cornell-dubilier.com
22 35 8 NSP220M8D5TR NIC Components www.niccomp.com
22 35 8 EEFFD0K220R Panasonic www.maco.panasonic.co.jp
47 25 6.3 NSP470M6.3D2TR NIC Components www.niccomp.com
68 18 8 ESRD680M08B Cornell Dubilier www.cornell-dubilier.com
100 95 10 594D107X_010C2T Vishay www.vishay.com
*Recommend capacitor in typical application circuit Figure 1.
D
REV.
ADN8830
–17–
Although the FETs that drive OUT A alternate between Q1 and
Q2 being on, they have an equivalent series resistance that is
equal to a weighted average of their r
DS, ON
values.
RDr Dr
EQIV DS P DS N
+
()
×
,,
11
1
(35)
The resistive power loss from the PWM transistors is then
PRI
FET PWM EQIV TEC,
2
(36)
There is also a power loss from the continuing charging and
discharging of the gate capacitances on Q1 and Q2. The power
dissipated due to gate charge loss (P
GCL
) is
PCVf
GCL ISS DD CLK
=
1
2
2
(37)
using the appropriate input capacitance (C
ISS
) for the NMOS
and PMOS. Both transistors are switching, so P
GCL
should be
calculated for each one and will be added to find the total power
dissipated from the circuit.
The series resistance of the inductor, R2 from Figure 14, will
also exhibit a power dissipation equal to
PRI
R TEC2
2
2
(38)
Core loss from the inductor arises as a result of nonidealities of
the inductor. Although this is difficult to calculate explicitly, it
can be estimated as 80% of P
RLS
at 1 MHz switching frequen-
cies and 50% of P
RL
at 100 kHz. Judging conservatively
PP
LOSS RL
08.
(39)
Finally, the power dissipated by the ADN8830 is equal to the
current used by the device multiplied by the supply voltage.
Again, this exact equation is difficult to determine as we have
already taken into account some of the current while finding the
gate charge loss. A reasonable estimate is to use 40 mA as the
total current used by the ADN8830. The power dissipated from
the device itself is
PVmA
ADN DD8830
10
(40)
There are certainly other minor mechanisms for power dissipa-
tion in the circuit. However, a rough estimate of the total power
dissipated can be found by summing the preceding power dissi-
pation equations. Efficiency is then found by comparing the
power dissipated with the required output power to the load.
Efficiency
P
PP
LOAD
LOAD DISS TOT
=
+
,
(41)
where
PIV
LOAD LOAD LOAD
The measured efficiency of the system will likely be less than the
calculated efficiency. Measuring the efficiency of the application
circuit is fairly simple but must be done in an exact manner to
ensure the correct numbers are being measured. Using two high
current, low impedance ammeters and two voltmeters, the cir-
cuit should be set up as shown in Figure 15.
A
V
DD
POWER SUPPLY
GND
ADN8830
V
V
A
TEC
LOAD
Figure 15. Measuring Efficiency of the ADN8830 Circuit
Table V. Recommended FETs for Linear Output Amplifier
Part Number Type C
GD
(nF) Ext. C
GD
(nF) C
SNUB
(nF) r
DS, ON
(m)I
MAX
(A) Manufacturer
FDW2520C* NMOS 0.17 18 6.0 Fairchild
PMOS 0.15 2.2 3.3 35 4.5 Fairchild
IRF7401 NMOS 0.5 22 8.7 International Rectifier
IRF7233 PMOS 2.2 1.0 3.3 20 9.5 International Rectifier
FDR6674A NMOS 0.23 9.5 11.5 Fairchild
FDR840P PMOS 0.6 1.0 3.3 12 10 Fairchild
*Recommend transistors in typical application circuit Figure 1.
Table VI. Recommended FETs for PWM Output Amplifier
Part Number Type C
ISS
(nF) r
DS,ON
(m) Continuous I
MAX
(A) Manufacturer
FDW2520C* NMOS 1.33 18 6.0 Fairchild
PMOS 1.33 35 4.5 Fairchild
Si7904DN NMOS 1.0 30 5.3 Vishay Siliconix
Si7401DN PMOS 3.5 17 7.3 Vishay Siliconix
IRF7401 NMOS 1.6 22 8.7 International Rectifier
IRF7404 PMOS 1.5 40 6.7 International Rectifier
*Recommend transistors in typical application circuit Figure 1.
D

ADN8830ACPZ-REEL7

Mfr. #:
Manufacturer:
Analog Devices Inc.
Description:
Laser Drivers HIGH PRECISION/EFFICIENCY TEC CONTROLLER
Lifecycle:
New from this manufacturer.
Delivery:
DHL FedEx Ups TNT EMS
Payment:
T/T Paypal Visa MoneyGram Western Union

Products related to this Datasheet