TPS2398, TPS2399
www.onsemi.com
10
APPLICATIONS INFORMATION
General
When a plug−in module or printed circuit card is inserted
into a live chassis slot, discharged supply bulk capacitance
on the board can draw huge transient currents from the
system supplies. Without some form of inrush limiting,
these currents can reach peak magnitudes ranging up to
several hundred amps, particularly in high−voltage systems.
Such large transients can damage connector pins, PCB etch,
and plug−in and supply components. In addition, current
spikes can cause voltage droops on the power distribution
bus, causing other boards in the system to reset.
The TPS2398 and TPS2399 are hot swap power managers
designed to limit these peaks to preset levels, as well as
control the slew rate (di/dt) at which charging current ramps
to the user−programmed limit. These devices use an external
N−channel pass FET and sense element to provide
closed−loop control of current sourced to the load. Input
supply under−voltage lockout (UVLO) protection allows
hot swap circuits to turn on automatically with the
application of power, or to be controlled with a system
command via the EN input. External capacitors control both
the current ramp rate, and the time−out period for load
voltage ramping. In addition, an internal overload
comparator provides circuit breaker protection against
shorts occurring during steady−state (post−turn−on)
operation of the card.
The TPS2398 and TPS2399 operate directly from the
input supply (nominal −48 VDC rail). The –VIN pin
connects to the negative voltage rail, and the RTN pin
connects to the supply return. Internal regulators convert
input power to the supply levels required by the device
circuitry. An input UVLO circuit holds the GATE output low
until the supply voltage reaches a nominal 30 V level. A
second comparator monitors the EN input; this pin must be
pulled above the 1.4 V enable threshold to turn on power to
the load.
Once enabled, and when the input supply is above the
UVLO threshold, the GATE pull−down is removed, the
linear control amplifier (LCA) is enabled, and a large
discharge device in the RAMP CONTROL block is turned
off. Subsequently, a small current source is now able to
charge an external capacitor connected to the IRAMP pin.
This results in a linear voltage ramp at IRAMP. The voltage
ramp on the capacitor actually has two discrete slopes. As
shown in Figure 2, charging current is supplied from either
of two sources. Initially at turn−on, the 600 nA source is
selected, to provide a slow turn−on rate. This slow turn−on
helps ensure that the LCA is pulled out of saturation, and is
slewing to the voltage at its non−inverting input before
normal rate load charging is allowed. This mechanism helps
reduce current steps at turn−on. Once the voltage at the
IRAMP pin reaches approximately 0.5 V, an internal
comparator de−asserts the SLOW signal, and the 10 mA
source is selected for the remainder of the ramp period.
The voltage at IRAMP is divided down by a factor of 100,
and applied to the non−inverting input of the LCA. Load
current magnitude information at the ISENS pin is applied
to the inverting input. This voltage is developed by
connecting the current sense resistor between ISENS and
−VIN. The LCA controls the gate of the external pass FET
to force the ISENS voltage to track the divided down
IRAMP voltage. Consequently, the load current slew rate
tracks the linear voltage ramp at the IRAMP pin, producing
a linear di/dt of the load current. The IRAMP capacitor is
charged to about 5 V; however, the LCA input is clamped at
40 mV. Therefore, the current sourced to the load during
turn−on is limited to a value given by I
MAX
40 mV / R
SNS
,
where R
SNS
is the value of the sense resistor.
The resultant load current, regulated by the controller,
charges the module’s input bulk capacitance in a safe
fashion. Under normal conditions, this capacitance
eventually charges up to the dc input potential. At this point,
the load demand drops off, and the voltage at ISENS
decreases. The LCA now drives the GATE output to its
supply rail.
The device detects this condition as the GATE voltage
rises through 7.5 V, latches this status and asserts the /PG
output. If the full sourced current limit is not yet available to
the load, as evidenced by the IRAMP voltage being less than
5 V, then the /PG assertion is delayed until that condition is
also met.
The peak, steady−state GATE pin output, typically 14 V,
ensures sufficient overdrive to fully enhance the external
FET, while not exceeding the typical 20 V V
GS
rating of
common N−channel power FETs.
Fault timing is accomplished by connecting a capacitor
between the FLTTIME and −VIN pins, allowing
user−programming of the timeout period. Whenever the hot
swap controller is in current control mode as described
above, the LCA asserts an overcurrent indication – OC
signal in the Figure 2. Overcurrent fault timing is inhibited
during the slow turn−on portion of the IRAMP waveform.
However, once the device transitions to the normal rate
current ramp (V
IRAMP
0.5 V), the external capacitor is
charged by a 50 mA source, generating a voltage ramp at the
FLTTIME pin. If the load voltage ramps successfully, the
fault capacitor is discharged, and load initialization can
begin. However, if the timing capacitor voltage attains the
4 V fault threshold, the LCA is disabled, the pass FET is
rapidly turned off, and the fault is latched. Fault capacitor
charging ceases, and the capacitor is then discharged. In
addition, latching of a fault condition causes rapid discharge
of the IRAMP capacitor. In this manner, the soft−start
function is then reset and ready for the next output enable,
if and when conditions permit.
Subsequent to a plug−in’s start−up, and during the
module’s steady−state operation, load faults that force
current limit operation also initiate fault timing cycles as
TPS2398, TPS2399
www.onsemi.com
11
described above. In this case, a fault timeout also clears the
previously latched power good status.
The TPS2398 latches off in response to faults; once a fault
timeout occurs, a large NMOS device is activated to rapidly
discharge the external capacitor, resetting the timer for any
subsequent device reset. The TPS2398 can only be reset by
cycling power to the device, or by cycling the EN input.
In response to a latched fault condition, the TPS2399
enters a fault retry mode, wherein it periodically retries the
load to test for continued existence of the fault. In this mode,
the FLTTIME capacitor is discharged slowly by a about a
0.4 mA constant−current sink. When the voltage at the
FLTTIME pin decays below 0.5 V, the LCA and RAMP
CONTROL circuits are re−enabled, and a normal turn−on
current ramp ensues. Again, during the load charging, the
OC signal causes charging of the FLTTIME capacitor until
the next delay period elapses. The sequential charging and
discharging of the FLTTIME capacitor results in a typical
1% retry duty cycle. If the fault subsides, the timing
capacitor is rapidly discharged, duty−cycle operation stops,
and the /PG output is asserted.
Note that because of the timing inhibit during the initial
slow ramp period, the duty cycle in practice is slightly
greater than the nominal 1% value. However, sourced
current during this period peaks at only about one−eighth the
maximum limit. The duty cycle of the normal ramp and
constant−current periods is approximately 1%.
The FAULT LOGIC within the TIMER BLOCK
automatically manages capacitor charge and discharge
actions, and the enabling of the GATE output.
Supply Transient Response
The TPS2398 and TPS2399 also feature a fast−acting
overload comparator which acts to clamp large transients
from catastrophic faults occurring once the pass FET is fully
enhanced, such as short circuits. This function provides a
back−up protection to the LCA by providing a hard gate
discharge action when the LCA is saturated. If sense voltage
excursions above 100 mV are detected, this comparator
rapidly pulls down the GATE output, bypassing the fault
timer, and terminating the short−circuit condition. Once the
spike has been brought down below the overload threshold,
the GATE output is released, allowing the circuit to turn on
again in either current−ramp or current−limit mode. A 4 ms
deglitch filter is applied to the OL signal to help reduce the
occurrence of nuisance trips.
In redundant−supply systems, the sudden switchover to a
supply of higher voltage potential is one more source of
large current spikes. Due to the low impedance of filter
capacitance under such high−frequency transients, these
spikes are generally indistinguishable from true
short−circuit faults to a hot swap controller. However, the
TPS2398 and TPS2399 transient response addresses this
issue by providing rapid circuit−breaker protection for load
faults along with minimal interruption of power flow during
supply switching events. The scope plots in Figure 26
illustrate how.
Figure 26 is a scope capture of the TPS2398/99 response
in a diode−OR configuration to such an input transient event.
(All waveforms are referenced to the −VIN pin.) In this
example, the module is initially operating from a nominal
−43 V supply (relative to the backplane supply return node).
At the first major time division, another power supply, with
an output of −48 V, is suddenly hot swapped into a
secondary, or INB, input. This sudden voltage step is
reflected in the −48V_RTN trace. On this board, the 5 V
potential difference caused an 8 A spike, as shown by the I
IN
trace (I
IN
trace has been measured after the diode−OR). The
GATE pin is rapidly pulled low, which quickly terminates
the overload spike. However, it is quickly released, and seen
to drive back to the pass FET ON−threshold, in this case,
about 5 V. The resultant current−limit operation of the
circuit is evidenced by the 2 A load on the B supply. Once
supply current is flowing again, the filter capacitance is
charged up to the new input supply level, seen here on the
V
DRAIN
trace. Once the capacitance is fully charged, the
load demand rolls off to the operating 1 A level. As an added
benefit, this event is transparent to the /PG signal, which
remains asserted throughout the disturbance.
Figure 26. Input Transient Response
V
RTN
(5 V/div, offset 43 V)
V
IRAMP
(2 V/div)
I
IN
(2 A/div)
V
GATE
(5 V/div)
V
DRAIN
(5 V/div)
V
/PG
(50 V/div)
C
IRAMP
= 3.9 nF
C
FLTTM
= 100 nF
C
OUT
= 100 mF
R
OUT
= 50 W
R
SNS
= 20 W
In order for downstream loads (bricks, etc.) to operate
through the distribution bus transient, it is important to
properly size the filtering capacitance to supply the needed
energy during the OFF−time of the pass FET. In this
example, once the RTN node jumps by 5 V higher than the
original potential, about 6 V develops across the FET,
TPS2398, TPS2399
www.onsemi.com
12
indicating approximately a 1 V droop across the brick input.
Therefore, due to the fast response of the TPS2398/99
devices, the 100 mF capacitor achieves excellent hold−up of
the brick input voltage. Actual requirements depend heavily
on the individual application. Whether the device turns back
on in either current−ramp or current−limit mode depends in
part on the size of the ramp capacitor (C
IRAMP
) and the input
capacitance of the pass FET. But in any case, the circuit turns
back on in a controlled−current manner after rapidly
clamping the potentially damaging spike.
Setting the Sense Resistor Value
Due to the current−limiting action of the internal LCA, the
maximum allowable load current for an implementation is
easily programmed by selecting the appropriate sense
resistor value. The LCA acts to limit the sense voltage
V
ISENS
to its internal reference. Once the voltage at the
IRAMP pin exceeds approximately 4 V, this limit is the
clamp voltage, V
REF_K
. Therefore, a maximum sense
resistor value can be determined from Equation 1.
R
SNS(MAX)
v
V
REF_K(MIN)
I
MAX
(eq. 1)
R
SNS(MAX)
v
33 mV
I
MAX
Where:
R
SNS
is the sense resistor value,
V
REF_K(MIN)
is the minimum ref. clamp voltage, and
I
MAX
is the desired current limit.
When setting the sense resistor value, it is important to
consider two factors, the minimum current that may be
imposed by the TPS2398 or TPS2399, and the maximum
load under normal operation of the module. For the first
factor, the specification minimum clamp value is used, as
seen in Equation 1. This method accounts for the tolerance
in the sourced current limit below the typical level expected
(40 mV / R
SNS
). (The clamp measurement includes LCA
input offset voltage; therefore, this offset does not have to be
factored into the current limit again.) Second, if the load
current varies over a range of values under normal operating
conditions, then the maximum load level must be allowed
for by the value of R
SNS
. One example of this is when the
load is a switching converter, or brick, which draws higher
input current, for a given power output, when the
distribution bus is at the low end of its operating range, with
decreasing draw at higher supply voltages. To avoid
current−limit operation under normal loading, some margin
should be designed in between this maximum anticipated
load and the minimum current limit level, or
I
MAX
> I
LOAD(MAX)
, for Equation 1.
For example, using a 20 mW sense resistor for a nominal
1 A load application provides a minimum of 650 mA of
overhead for load variance/margin. Typical bulk capacitor
charging current during turn−on is 2 A (40 mV / 20 mW).
Setting the Inrush Slew Rate
The TPS2398 and TPS2399 devices enable
user−programming of the maximum current slew rate
during load start−up events. A capacitor tied to the IRAMP
pin (C
2
in the typical application diagram) controls the di/dt
rate. Once the sense resistor value has been established, a
value for ramp capacitor C
IRAMP
, in microfarads, can be
determined from Equation 2.
C
IRAMP
+
11
100 @ R
SNS
@
ǒ
di
dt
Ǔ
MAX
(eq. 2)
Where:
R
SENSE
is in ohms, and
ǒ
di
dt
Ǔ
MAX
is the desired maximum slew rate, in
amperes/second.
For example, if the desired slew rate for the typical
application shown is 1500 mA/ms, the calculated value for
C
IRAMP
is about 3.7 nF. Selecting the next larger standard
value of 3.9 nF (as shown in the diagram) provides some
margin for capacitor and sense resistor tolerances.
As described earlier in this section, the TPS2398 and
TPS2399 initiate ramp capacitor charging, and
consequently, load current di/dt at a reduced rate. This
reduced rate applies until the voltage on the IRAMP pin is
about 0.5 V. The maximum di/dt rate, as set by Equation 2,
is effective once the device has switched to the 10 mA
charging source.
Setting the Fault Timing Capacitor
The fault timeout period is established by the value of the
capacitor connected to the FLTTIME pin, C
FLTTM
. The
timeout period permits riding out spurious current glitches
and surges that may occur during operation of the system,
and prevents indefinite sourcing into faulted loads swapped
into a live system. However, to ensure smooth voltage
ramping under all conditions of load capacitance and input
supply potential, the minimum timeout should be set to
accommodate these system variables. To do this, a rough
estimate of the maximum voltage ramp time for a
completely discharged plug−in card provides a good basis
for setting the minimum timer delay.
Due to the three−phase nature of the load current at
turn−on, the load voltage ramp potentially has three distinct
phases (compare Figures 3 and 4). This profile depends on
the relative values of load capacitance, input dc potential,
maximum current limit and other factors. The first two
phases are characterized by the two different slopes of the
current ramp; the third phase, if required for bulk
capacitance charging, is the constant−current charging at
I
MAX
. Considering the two current ramp phases to be one
period at an average di/dt simplifies calculation of the
required timing capacitor.

TPS2399DMT7G

Mfr. #:
Manufacturer:
ON Semiconductor
Description:
Hot Swap Voltage Controllers -48 V HOT SWAP 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