English
Language : 

MIC384_05 Datasheet, PDF (12/22 Pages) Micrel Semiconductor – Three-Zone Thermal Supervisor
MIC384
shutdown mode, and the interrupt status will not be retained.
Since entering shutdown mode stops A/D conversions, the
MIC384 is incapable of detecting or reporting temperature
events of any kind while in shutdown. Diode faults require
one or more A/D conversion cycles to be recognized, and
therefore will not be reported either while the device is in
shutdown (see "Diode Faults" above).
Fault_Queue
Fault queues (programmable digital filters) are provided
in the MIC384 to prevent false tripping due to thermal or
electrical noise. The two bits in CONFIG[4:3] set the depth
of Fault_Queue. Fault_Queue then determines the number
of consecutive temperature events (TEMPx > T_SETx,
or TEMPx < T_HYSTx) which must occur in order for the
condition to be considered valid. There are separate fault
queues for each zone. As an example, assume the part is
in comparator mode, and CONFIG[4:3] is programmed with
10b. The measured temperature in zone one would have to
exceed T_SET1 for four consecutive A/D conversions before
/INT would be asserted or the S1 status bit set. Similarly,
TEMP1 would then have to be less than T_HYST1 for four
consecutive conversions before /INT would be reset. Like
any filter, the fault queue function also has the effect of de-
laying the detection of temperature events. In this example,
it would take 4 x tCONV to detect a temperature event. The
depth of Fault_Queue vs. D[4:3] of the configuration register
is shown in Table 4.
C ONFIG[4:3]
00
01
10
11
* Default s etting
F ault_Queue Depth
1 convers ion*
2 convers ions
4 convers ions
6 convers ions
Table 4. Fault_Queue Depth Settings
Interrupt Generation
Assuming the MIC384 is in interrupt mode and interrupts
are enabled, there are seven different conditions that will
cause the MIC384 to set one of the status bits, S0, S1, or
Micrel
S2, in CONFIG and assert its /INT output. These conditions
are listed in Table 5. When a temperature event occurs, the
corresponding status bit will be set in CONFIG. This action
cannot be masked. However, a temperature event will only
generate an interrupt signal on /INT if it is specifically enabled
by the interrupt mask bit (IM =0 in CONFIG). Following an
interrupt, the host should read the contents of the configura-
tion register to confirm that the MIC384 was the source of the
interrupt. A read operation on any register will cause /INT to
be de-asserted. This is shown in Figure 5. The status bits
will only be cleared once CONFIG has been read.
Since temperature-to-digital conversions continue while /INT
is asserted, the measured temperature could change between
the MIC384’s assertion of /INT and the host’s response. It
is good practice for the interrupt service routine to read the
value in TEMPx, to verify that the over-temperature or under-
temperature condition still exists. In addition, more than one
temperature event may have occurred simultaneously or in
rapid succession between the assertion of /INT and servic-
ing of the MIC384 by the host. The interrupt service routine
should allow for this eventuality. Keep in mind that clearing
the status bits and deasserting /INT is not sufficient to allow
further interrupts to occur. TEMPx must become less than
T_HYSTx if the last event was an over-temperature condition,
or greater than T_SETx if the last event was an under-tem-
perature condition, before /INT can be asserted again.
Putting the device into shutdown mode will de-assert /INT
and clear the status bits (S0, S1, and S2). This should not
be done before completing the appropriate interrupt service
routine(s).
Polling
The MIC384 may either be polled by the host, or request the
host’s attention via the /INT pin. In the case of polled opera-
tion, the host periodically reads the contents of CONFIG to
check the state of the status bits. The act of reading CON-
FIG clears the status bits. If more than one event that sets
a given status bit occurs before the host polls the MIC384,
only the fact that at least one such event has occurred will be
apparent to the host. For polled systems, the interrupt mask
bit should be set (IM = 1). This will disable interrupts from the
MIC384 and prevent the /INT pin from sinking current.
MIC384
12
September 2005