检测电路保持您的微处理器控制-Supervisory Cir

文章摘要：A timeout function can be grafted onto this circuit with the addition of a capacitor and a diode. The resulting circuit has seven components, and it still has a problem with accuracy and with slow-rising supply voltages.How Accurate Is Accurate Enough?Consider a common example in which the processor

检测电路保持您的微处理器控制-Supervisory Cir,标签：计算机控制技术,工厂电气控制技术,http://www.88dzw.com

A timeout function can be grafted onto this circuit with the addition of a capacitor and a diode. The resulting circuit has seven components, and it still has a problem with accuracy and with slow-rising supply voltages.

How Accurate Is Accurate Enough?

Consider a common example in which the processor operates on a nominal 5V supply and is specified to operate as low as 4.5V. The reset circuit should hold reset for all voltages below 4.5V, and its minimum threshold must therefore be 4.5V. What, then, should be the upper limit for the spread of reset thresholds over temperature and from unit to unit? You can specify the power supply at 5V ±0% if you want to get in trouble with the power supply designers, but a more likely range is 4.75V to 5.25V. You should therefore guarantee the threshold between 4.5V and 4.75V; i.e., 4.63V ±2.7%.

A zener diode can regulate the threshold voltage, but the accuracy of a typical zener is ±5% to ±10%. For premium prices you can specify tighter tolerance (to ±1%), but only for room temperature and a specific current. All zeners exhibit a significant variation of voltage with current, and the typical temperature coefficient (TC) is several mV/°C. TC alone can cause several hundred millivolts of change over the range 0°C to 70°C. Zener-based reset circuits are inadequate to guarantee a proper reset on startup and during a brownout. To make matters worse, even low-current zeners require 100µA to achieve regulation, which is a considerable load in battery-powered systems.

How Should an Ideal Reset Circuit Operate?

We've established that the reset circuit's voltage tolerance should not exceed ±2.7% over temperature. But without a proper delay in terminating the reset pulse, the circuit is subject to malfunction under two conditions: a slow-rising supply voltage as mentioned earlier or a supply voltage that exhibits noise or nonmonotonic behavior during startup or recovery from brownout conditions. If the monitored supply voltage sits right at the reset circuit threshold, noise will tend to trigger, untrigger, and retrigger the circuit repeatedly, causing the µP's Active-low RESET input to oscillate.

Hysteresis can cure this problem, and the market offers several families of voltage-detector products that attempt to solve the dilemma that way. Unfortunately, hysteresis narrows the threshold's allowable voltage tolerance. We had 250mV (4.75V Ð 5.0V) to play with in the above example. If you add 100mV of hysteresis, the minimum threshold for a rising voltage becomes 100mV higher than before, i.e., 4.6V rather than 4.5V. This shift is necessary to guarantee that the threshold for a falling voltage (during brownouts) will be no lower than 4.5V. Thus, to ensure both thresholds between 4.5 and 4.75V, the upper one must be 4.67V ±1.6%.

Common voltage detectors of this type, such as the Ricoh Rx5VL/Rx5VT and Seiko S-807, have 25°C threshold accuracy of ±2.5% and ±2.4%. Actual devices operate beyond 25°C, but these products specify only typical temperature coefficients of 100ppm/°C and 120ppm/°C. These TCs result in threshold tolerances of ±2.85% and ±2.82%, respectively, over the 0° to 70°C range.

The Seiko S-808 family represents the more recent precision parts of this type. They specify ±2% accuracy at 25°C and a maximum temperature coefficient of 350ppm/°C. Over the 0°C to 70°C range, this maximum temperature coefficient corresponds to a variation of 350e-6 x 70 = 0.0245, or 2.45%. Our worst-case accuracy is therefore ±3.225%. If we assume a worst-case part will not exhibit the maximum temperature coefficient over temperature but rather (on average) about half the maximum, then the resulting maximum variation (±2.6125%) is just good enough for the above example.

上一页  [1] [2] [3] [4] [5] [6] [7] [8]  下一页