The option to re-sequence is offered as a response to each type of fault. The re-sequence operation is not performed until the retries are exhausted for a given fault. During the re-sequence operation:
- The faulty rail and its Fault Shutdown Slaves (FSS) rails are immediately disabled (Stop Immediately) or disabled according to the configured shutdown sequence (Stop With Delay)
- After the faulted rail and its FSS rails' Enable pins are deasserted and all of these rails have turned off (the voltage has fallen below POWER_GOOD_OFF), the UCD91xxx waits for a programmable delay time before starting the resequence. The delay time is configured by the "Time between Resequences" byte in the MISC_CONFIG command, (see Section 26.42). While waiting for the rails to turn off, if any rail voltage does not fall below 12.5% of nominal output voltage within a time period defined by TOFF_MAX_WARN_LIMIT, a TOFF_MAX_WARN warning occurs. Software can configure whether to abort the resequence when the TOFF_MAX_)WARN warning occurs. For more information, see the MISC_CONFIG command (Section 26.42). It is configurable to ignore the POWER_GOOD_OFF and TOFF_WARNING status of the rail when resequencing, (see Section 26.42). This is mainly for those rails whose EN signals are not controlled by UCD.
- After the resequence delay time, the faulty rail and its FSS rails are sequenced on according to the start-up sequence configuration. The resequence attempt can repeat 1-4 times or unlimited times to bring the faulty rail and its FSS rails to regulation. The maximum resequence times is configured in the Section 26.42. If a resequence operation is successful – that is, all of the rails that were resequenced maintain in normal operation (power-good and no retries) for 1 second – the resequence counter is reset. (If a resequenced rail is not configured to turn on during the resequence, the device does not wait for that rail to come on before declaring the resequence operation is successful.) If the rails are resequenced the maximum number times and they fail to reach normal operation, a device reset is required to reset the resequence counter. Rails can be commanded off and then on in an attempt to get them back to normal operation.
This resequence operation is straight-forward if there is only one set of faulty rail and FSS rails. If two or more sets of rails require resequence at the same time, the device behavior is always conservative. For example, if a set of rails is already on its second resequence, and the device is configured to resequence three times, and another set of rails enters the resequence state, the second set of rails are only resequenced once so that the total resequence time is 3. Another example – if one set of rails is awaiting all of its rails to shutdown so that it can resequence, and another set of rails enters the resequence state, the device now waits for all rails from both sets to shutdown before resequencing.
If any rails at the resequence state are caused by GPI fault response, the whole resequence is suspended until the GPI fault is physically clear.