Global Sources
EE Times-Asia
Stay in touch with EE Times Asia
?
EE Times-Asia > Power/Alternative Energy
?
?
Power/Alternative Energy??

The right voltage supervisor improves system reliability

Posted: 04 Jul 2005 ?? ?Print Version ?Bookmark and Share

Keywords:voltage supervisor? circuit board? lattice semiconductor?

By Shyam Chandra, Marketing Manager
Lattice Semiconductor

Advanced integrated circuits such as communication processors achieve increased performance, added functionality and reduced power consumption when fabricated using the latest sub-micron technologies. This results in reduced operating core voltages. Increasingly, inter-device communication standards also require the use of a variety of I/O supply voltages. These factors result in multi-voltage devices with multiple power supply requirements. A typical microprocessor datasheet, in addition to specifying the core voltage of 1.2V and I/O voltages of 2.5V, 3.3V, etc., also specifies the voltage tolerance (for example, \3 percent for 1.2V supply, \5 percent for 2.5V and 3.3V). As long as the power supply voltages are within these tolerance limits, the processor operates as specified.

If the core supply voltage drops below the specified threshold, the processor begins to misinterpret instructions. If its I/O voltage drops below the signaling specification, the data transferred between the memory and the processor becomes corrupted, resulting in the processor misreading the instruction and consequently misinterpreting it.

The misinterpretation of instructions by a microprocessor results in unpredictable behavior. In some cases, the microprocessor could overwrite the on-board Flash memory, resulting in a failed circuit board. Imagine the circuit board failing just because it was extracted from its sub-rack slot!

Any ASIC/FPGA on the board can behave unpredictably under low voltage conditions. For example, if the power supply voltage drops below the limit for a networking ASIC, it might send a garbled packet, or might lose an internally buffered acknowledged packet, resulting in a corrupt message.

A voltage supervisor is used to prevent such unpredictable behavior.

What is a voltage supervisor?
A voltage supervisor is an IC that prevents the microprocessor from operating under low voltage conditions by holding the processor in reset condition. In some cases, the voltage supervisor can provide an early warning to the processor by interrupting its current instruction flow, resulting in a safe system shut down.

A typical voltage supervisor (Figure 1) contains a voltage comparator, a band-gap reference voltage source, and a voltage attenuator to set the monitoring voltage threshold. The output of the comparator can be used either to interrupt the processor or reset it. Figure 2 shows the architecture of a device that monitors multiple power supply voltages. These devices contain multiple comparators with individual attenuators to permit the simultaneous monitoring of different power supply voltages. The output of these comparators are logically combined to provide a single logic output to interrupt/ reset the processor.

Figure 1: Single power supply voltage supervisor circuit

Figure 2: Three power supply voltage monitoring circuit

Voltage supervisor accuracy
The circuit shown in Figure 1 assumes the use of ideal blocks: band-gap reference source (output voltage is always 1.25V), attenuator (its output voltage is exactly 1.25V when the input voltage is 3.135V), and an ideal comparator (without any offset, infinite gain, zero propagation delay, always toggles exactly when the monitored voltage is 3.135V).

In reality, though, the band-gap reference voltage changes with temperature, the output voltage of the attenuator varies from device to device, and there are inaccuracies with the comparator. Cumulatively, these inaccuracies contribute to variation in the threshold voltages for each device and across temperature and voltage. The accuracy specification of a supervisor is a measure of the variation of actual threshold with respect to the intended threshold across operating temperature as well as from device to device.

Figure 3 shows the effect of the accuracy of a voltage supervisor on its functionality. Consider the use of a voltage supervisor (Figure 1) to signal a power supply failure when the supply voltage drops below 3.3V!5 percent (3.135V). If the voltage supervisor accuracy is 2%, the supervisor can flag a power supply fault anywhere between 3.135!2 percent (3.072V) or 3.135 + 2 percent (3.2), as shown by points A & B.

Figure 3: Fault detection with supervisor accuracy of 2%

If a fault is flagged when the supply is at 3.2V, it means that the processor is prevented from operating even when it can operate reliably. But, if the fault is flagged when the voltage is at 3.07V, the processor would already be operating with a power supply voltage lower than its specified lower voltage level. The latter case is more serious, because at lower than threshold voltage the processor can be mis-executing instructions, defeating the very purpose of using a supervisor IC.

Compensating for voltage supervisor accuracy

To prevent the processor from operating at unacceptably lower voltage due to the supervisor's inaccuracy, the supervisor threshold should be selected so that the entire power supply fault detect range lies within the operating voltage range of the processor. As shown in Figure 4, if the supervisor threshold is set at 3.2V, the voltage range in which the supervisor can declare the power supply faulty is between 3.14V to 3.26V, avoiding the condition under which the processor is operating at a voltage less than its threshold (3.3V-5 percent).

Figure 4: Fault detection with supervisor with correct threshold

In Figure 4, the threshold value of the supervisor is set at 3.2V, which is calculated using the following equation:

VTSup= Vin * (1-VCktTol/100)/(1Asup/100) Equation 1

Where VTSup!Supervisor Threshold Vin!Power Supply Nominal Voltage VCktTol!Circuit Power Supply Tolerance Asup!Accuracy of the Supervisor

In this example, Vin at 3.3V,VCktTol!5 percent, and Asup!2%. Inserting these values in the equation above, VTSup = 3.3 * (1!(5/100)) / (1!(2/100)) = 3.2V

By selecting a Supervisor IC with a threshold of 3.2V or above, the processor will to held in reset when the power supply voltage is less than or equal to 3.3V!5 percent.

Power supply output voltage variation determines supervisor accuracy

The output voltage of a power supply can deviate from its typical voltage setting for various reasons, including the load current, temperature of operation and device-to-device variation. It is a common design practice to select the power supply with a power supply voltage variation lower than that of the circuit board operating voltage tolerance. For example, a power supply with an output voltage range of 3% is used to power devices with 5 percent voltage tolerance. The lowest voltage of the power supply will be 3.3!3 percent = 3.2V. The threshold voltage range of a supervisor with 2 percent accuracy is 3.14V to 3.26V. This means that, whenever the power supply output voltage is below 3.26V (due to increased dynamic current or temperature, for example) the supervisor might detect a fault and reset the processor, which will result in a system with intermittent failures.

Equation 2 is used to calculate the required supervisor accuracy for a given power supply output voltage range and the circuit board voltage tolerance:

Asup = (VCktTol!VSupRng) / (2!2 * VSupRng!VCktTol) Equation 2

Where VSupRng!Power Supply Output Voltage Range in % VCktTol!Circuit Power Supply Tolerance in % Asup!Accuracy of the Supervisor in %

Note: Equation 2 shows that the supervisor accuracy is independent of actual power supply voltages and only depends on the output voltage variation of the power supply and the circuit's operating power supply tolerance.

Figure 5 shows a graph of calculated supervisor accuracy vs. power supply output range. There are 2 lines corresponding to a circuit power supply tolerance of 5 percent and 3 percent. As the pointers on the graph illustrate, for a power supply with an output voltage range of 3 percent to be used in a circuit that can operate with a 5 percent variation of the power supply, the voltage supervisor used should have an accuracy of 1 percent.

Figure 5: Output voltage range vs. supervisor accuracy graph

In a multi-voltage circuit board, some devices might specify a voltage tolerance of 3%. If the supervisor accuracy is 1%, then, as the graph shows, the power supply output voltage range should be limited to 1%. So now it is clear that, for reliable system operation, supervisor accuracy is a very important factor. For example, the Lattice ispPAC-POWR1208P1 device offers 0.5% accuracy at room temperature (0.8% accuracy across industrial temperature range). This device can be a single-chip precision supervisor for up to 12 power supplies. For reliable system design, not only supervisor accuracy but also the delay in fault detection should be considered, as discussed below.

Supervisor fault detection delay
Fault detection delay is the duration from the time the power supply voltage drops below the threshold of the supervisor to the time the output of the supervisor toggles, indicating the fault. Figure 6 shows the 3.3V power supply output voltage during a fault as well as the voltage supervisor output toggling after detecting the power supply fault.

Figure 6: Effect of fault detection delay on board operation

Note: For the sake of simplicity, this discussion ignores the effect of fault detection accuracy. As Figure 6 shows, the longer the supervisor takes to report the fault, the lower will be the power supply voltage. For example, the power supply voltage is decaying at a rate of 1V per millisecond. The supervisor is set at a threshold of 3.3V!5 percent. Here are two scenarios:

Scenario 1: Fault detection delay is 1ms:
Because the power supply output voltage continues to drop, by the time the processor is reset its power supply voltage would be much less than the low voltage threshold (about 2V). This means that the processor (with the power supply specification of 3.3V \5 percent) was executing code until the supply reached 2V! Clearly, the purpose of the high precision supervisor is defeated.

Scenario 2: Fault detection delay is 50?s:
By the time the supervisor output is active, the processor voltage is reduced by about 50 mV from its threshold of 3.3V- 5%. Again, the processor operation is not guaranteed at this voltage.

Raise the threshold for reliable fault detection
Now, if the threshold was set 50mV above the 3.3V-5 percent level, the processor would be reset by the time the power supply crossed the operational threshold.

In this application, the fault detection delay of 1ms is unacceptable. But a fault detection delay of about 50?s requires the threshold to be set 50mV above the minimum operating power supply voltage threshold. Now it is clear that, for reliable operation, the supervisor threshold should consider not only accuracy but fault detection delay as well.

Monitoring for higher voltage also is important to avoid damaging devices due to excess voltage. In this case, the speed of over voltage detection is more important than the under voltage fault detection. For example, the ispPAC-POWR1208P1 device monitors 12 power supplies simultaneously and has a fault detection delay of 4?s.

The example above considers only one power supply voltage and uses a very accurate supervisor IC. In reality, the number of power supplies that the supervisor should monitor is frequently more than one. The supervisor should be able to monitor all supplies simultaneously for fault and should be able to detect power supply faults with minimum fault detection delay.

Other factors contributing to increased reliability
Other factors to consider for reliable power supply fault detection include:

Glitch filter!Power supplies are usually noisy during circuit board operation. The noise could be due to power supply output ripple, or to transient currents in the system due to device operation. This noise can result in random toggling of the supervisor output. To prevent this, supervisors have a glitch filter that generates a clean input to the threshold comparators.

Hysteresis!A small amount of hysteresis is added to the threshold comparators to prevent the outputs from toggling multiple times, due to power supply noise, when the power supply voltage is at its threshold.

An example of precise, simultaneous fault detection circuitry
Following, is a discussion of a power supply voltage monitoring circuit using the Lattice ispPAC-POWR1208P1 device. As noted before, this device provides a monitoring accuracy of 0.5 percent and fault detection delay of 4?s.

The Lattice Power1208P1 features 12 high precision analog power monitor inputs. Each analog input has programmable threshold (384 steps + power supply discharge detection) synchronous comparators with a fault detection precision of 0.5 percent. In addition, the 1208P1 device is equipped with on-chip programmable voltage references for supply monitoring, 4 noise-immune digital inputs and 4 open-drain digital outputs for system control interfacing, 4 programmable timers with an on-chip 1 MHz oscillator for delay control and a 16 macrocell Complex PLD (CPLD) to implement sequencing and control functions. The Power1208P1 is ruggedized to operate reliably in noisy power supply environments from 2.7V to 5.5V.

Each of the 12 power supply monitoring inputs of a Power1208P1 device has:

* Programmable threshold comparator

!384 steps between 0.68V to 5.95V and one 80mV threshold for power supply discharge detection

!0.5 percent threshold accuracy (0.9% max across process, supply voltage and temperature

* High-speed fault detection

!4?s fault detection without glitch filter

!32?s fault detection with glitch filter.

* Hysteresis

The same device can be used to monitor different power supply voltage levels because the input threshold is programmable. No additional components (resistors, capacitors or glitch suppressing inductors) are needed. Additionally, the threshold voltage level of each analog input can be individually programmed using the PAC-Designer software.

Power supply monitoring circuit using a Power1208P1
The circuit in Figure 7 implements complete power management of an ATM Port card. The backplane into which this card plugs provides 3.3V and 5 V. The required power management functions are:

Phase 1!Wait for the backplane supplies to stabilize

Phase 2!Sequence locally generated power supply voltages to individual devices

Phase 3!Once all supplies are stabilized, release reset to processor after delay

Phase 4!Monitor all supplies for faults and, in case of a supply fault, reset the processor; if power supply fault persists, initiate the power supply shutdown process (Phase 5)

Phase 5!Power supply shutdown sequence.

Figure 7: Power management circuit of an ATM port card

The POWR1208P1 device shown in Figure 7 monitors seven power supply voltages simultaneously. Depending on the phase of operation, power supply faults are handled in different ways. The on-chip CPLD is used to control power supply management functions. For example:

During Phase 1, the power supply voltages monitored are 3V and 5V from the backplane. The purpose is to ensure the backplane supplies are stable before other supplies on the circuit board are turned on, and not generate a reset to the processor because the backplane voltages are below their threshold.

During Phase 2, individual power supply voltages are monitored to enable proper sequencing of supplies. If this turn-on process is not complete within a specified time, all supplies are shut down.

During Phase 3, the precision monitoring for faults begins and, after all the supplies stabilize, the reset pulse is stretched to ensure CPU start-up. Precision fault monitoring continues in phase 4.

Phase 4 indicates that the circuit board is functioning properly and the precision fault monitoring of all the circuit board power supply voltages continues. If there is a fault either due to on-board power supply failure or to surprise card extraction, the CPU is reset and activated within 4?s, minimizing the likelihood that the CPU will misinterpret instructions due to low voltage operation.

Conclusion
For reliable power supply fault detection, the required supervisor accuracy should take into account both the power supply output voltage range and the device's operating power supply voltage tolerance. In addition to voltage supervisor accuracy, designers must take into account the supervisor's fault detection delay.

Designers can use the Lattice Semiconductor ispPAC-POWR1208P1 device for all multi-voltage circuit board power supply supervisor applications because it offers 12 high precision (0.5 percent), programmable thresholds with hysteresis, and a glitch filter supply voltage monitoring inputs with a fault detection delay 4?s. The high precision and fast fault detection ability of the device maximize power supply tolerance, enabling the use of lower cost power supplies while maintaining high system reliability.

Because each monitoring input's fault detection threshold is programmable, the same device can be used to monitor up to 12 different power supply voltages simultaneously. In addition, the on-chip CPLD enables the implementation of a complete power supply management function across all circuit boards. The PAC-Designer software-based design methodology makes the power management implementation even easier.

About the author
Shyam Chandra
is the marketing manager for the in-system programmable mixed signal products at Lattice Semiconductor Corp. Prior to joining Lattice, Chandra worked for Vantis and AMD in sales and applications and previously was a telecom design engineer with Indian Telephone Industries.




Article Comments - The right voltage supervisor improve...
Comments:??
*? You can enter [0] more charecters.
*Verify code:
?
?
Webinars

Seminars

Visit Asia Webinars to learn about the latest in technology and get practical design tips.

?
?
Back to Top