One of the key specifications for a comparator is propagation delay - the amount of time it takes for a signal propagate from the comparator input to the output. Whether as a threshold detector in a battery powered application, or on a high speed signal processing board, the comparator's propagation delay is often the first parameter engineers want to know when selecting a suitable device. Unfortunately, the propagation delay specification can be and is often vague or misleading in terms of the information it does or doesn't reveal. This is because there are several factors that influence the propagation delay specification, but these factors are usualy not discussed in the data sheet. Factors that affect propagation delay are: the way it is measured, the amount of overdrive, the supply voltage, the output driver supply voltage, capacitive loading, the common mode voltage, the configuration (inverting or non-inverting), the edge on which one measures (rising or falling) and temperature.
Measuring Propagation Delay
To understand propagation, one must first look at what is being measured. Let's assume an ideal comparator (no offset voltage). The comparator basically compares the two input signals and trips the output when one input signal exceeds the other. But the output does not change instantaneously; there is a delay as the signal makes its way (propagates) through the internal circuitry before reaching the output. It is important to note that propagation delay is defined as the point where the output reaches 50% of the output value, not the full value. This designation, combined with output loading, is one of several factors that can cause your circuit's measured delay time to be longer than the expected delay time.
Propagation delay is usually a typical specification, meaning the value given is not production tested or guaranteed. Due to process and fab variations, in addition to normal statistical variations, the typical value can have quite a wide range. Averaging all values will give a value that is near the expected value, but an individual IC may have a measured propagation delay specification that is really not that typical at all. When there is a guaranteed specification, it may include a note that upon further inspection reveals that the device is sample tested, guaranteed by correlation or guaranteed by design. The popular industry-standard, LT1016 dual 10ns comparator provides such an example. Below is the propagation delay specification from the data sheet with guaranteed maximum numbers. Upon further inspection, note 4 statesthat " tPD cannot be measured in automatic handling equipment with low values of overdrive. The LT1016 is sample tested with a 1V step and 500mV overdrive. Correlation tests have shown that tPD limits shown can be guaranteed with this test if additional DC tests are performed to guarantee that all internal bias conditions are correct."
Linear Technologies latest high speed comparators provide guaranteed specifications in the data sheet. The LTC6752 2.9ns CMOS output comparator's propagation delay specification is shown below. Note 8 simply tells the signal step size (150mV).
The LTC6754 1.8ns LVDS output comparator also has a guaranteed propagation delay specfication. Due to its speed, LVDS outputs were provided to ease the digital interface clocking requrements.
Kick into Overdrive
One of the factors that influences propagation delay is the amount of overdrive that is applied to the comparator; the higher the overdrive, the faster the propagation delay. So looking at propagation delay without knowing the amount of overdrive can be misleading. ATE customers are aware of this and will often ask for a dispersion plot as shown below, which shows propagation delay as a function of overdrive. Dispersion is also a typical value, but when combined with the typical propagation delay provides a tighter range for the expected propagation delay values one would expect when using the comparator. In some cases, it may be advantageous to use a comparator with a slower propagation delay but a tight dispersion, than to use a device with a slightly faster propagation delay but a wide dispersion. The plot below shows the propagation delay vs. input overdrive for the LT1719 single 4.5ns 3V/5V comparator.
Turn Up the Juice
Another factor that influences the propagation delay specification is the power supply voltage. The graphs below show how the propagation delay changes as the supply voltage changes for the LTC6752 single 2.9ns comparator and the LT1719. The amount of variation depends on the supply voltage or range where the device was optimized, but with single supply systems, in general a lower supply voltage typically translates to a slower propagation delay. Note for the LT1719, the graph reveals that the propagation delay is little changed as the positive supply changes as long as there is a negative supply voltage on VEE.
More Volts at the Output
Some comparators have a separate supply pin for the output drivers and output logic levels. Similar to the supply pin, the output driver supply voltage affects the speed of the propagation delay. Generally speaking, the higher the output driver voltage, the faster the propagation delay. The graph below highlights this relationship.
Can You Carry That Load?
When measuring propagation delay, the load at the comparator output is not consistent from manufacturer to manufacturer, and is often not consistent within the same manufacturer. Electrical table capacitive loading is typically in the 10pF to 20pF range, but stray capacitance and heavy capacitive loads can have a significant effect on the propagation delay.
Uncommon Common Mode Voltage Changes
The comparator's input common mode voltage can play a role in the propagation delay. This affect can be very pronounced for rail-to-rail input comparators that consist of a PNP pair and an NPN pair that are active over different input common mode ranges. The graphs below show how the propagation delay changes as the common mode voltage changes for several comparators. In the first graph, the falling edge data (in red) reveals a 13% change in the propagation due to this effect. The change is much less pronounced for the rising edge data. Some comparators exhibit a step in the common mode voltage rather than a spike at this transition point. In figure 10, the propagation delay increases as the common mode nears the rail, causing a slight increase in the delay.
To Invert or Not Invert, that is the Question
Topology can play a role in the propagation delay. The comparator can be thought of as an amplifier running open loop and without the linear output stage. Like and amplifier, it can be configured for non-inverting or inverting configuation. The scope photo below shows the propagation delay for the LT6700/3 micropower 18ns comparator family using the inverting and non-inverting configurations. From the graph, we see that the rising edge noninverting propagation delay is about 24µs and the falling edge is approximately 20µs. For the inverting configuration, the falling edge delay is 40µs and the rising ege is 10µs.
Pick Your Edge
Sometimes the propagation delay is similar for the both rising edge and falling edge, and other times it is skewed. When the numbers are not identical, the best specification is typically shown on the front page of the data sheet. The graph below, though used above in a different example, is a good one to observe to see the difference between rising and falling edge delays. It is important to make sure you compare the same propagation delay edge when looking at two comparators.
Can Somebody Turn Up the Heat?
Temperature is the last specification that we mention for propagation delay variations. Though not guaranteed and production tested, one can often find graphs in the data sheet that show the temperature relationship between the two parameters. Sometimes the relationship is fairly linear; other times, not at all; it really depends on the comparator design. Below are a couple of graphs that show the relationship for the LTC6752 and the LT1719 high speed comparators. Note how the change over temperature can be more than 20% in some cases, and in other cases be a fairly small percentage of the room temperature value.
Armed and Ready to Choose Wisely
Hopefully, this has demystified some of the subtleties of propagation delay measurements. Linear Technology offers a wide range of comparators, grouped by high speed (≥500ns propagation delay), micropower (≤110µA typical supply current), application specific and high temperature.