PTAT Voltage

In the world of precision electronics, temperature is a persistent adversary, causing the fundamental properties of semiconductors to drift. This thermal instability presents a significant challenge for creating reliable and stable circuits. The voltage across a transistor junction, for instance, naturally decreases as it gets warmer—a predictable behavior described as Complementary to Absolute Temperature (CTAT). This article explores an elegant solution born from a deep understanding of this "flaw," demonstrating how to turn this temperature dependence on its head by engineering its exact opposite: a voltage that is Proportional to Absolute Temperature (PTAT).

In the chapter on ​​Principles and Mechanisms​​, we will delve into the semiconductor physics that explains why CTAT behavior occurs and uncover the clever engineering tricks used to generate a pure PTAT voltage from a pair of transistors. Following this, the ​​Applications and Interdisciplinary Connections​​ chapter will demonstrate the true power of this concept: how the artful combination of PTAT and CTAT voltages leads to the creation of ultra-stable bandgap voltage references, a cornerstone of modern analog and digital systems. Through this exploration, we will uncover the real-world limitations and the profound connection between this circuit and the fundamental constants of physics.

Imagine you are trying to build the world’s most precise clock. You have chosen the finest gears and the most stable materials, but you discover, to your dismay, that your workshop's temperature fluctuations—the morning chill, the afternoon sun—are causing the metal components to expand and contract, throwing your clock's timing off. The world of precision electronics faces a similar, and in some ways more profound, challenge. The very heart of modern electronics, the semiconductor, has a "personality" that changes dramatically with temperature. The task in circuit design is not to eliminate this temperature dependence, but to understand it and, with clever engineering, turn it from a flaw into a feature.

Let's look at our main character: the bipolar junction transistor, or BJT. For it to operate, we need to apply a small forward voltage across its base-emitter junction, a voltage we call $V_{BE}$. This voltage is like the key that opens a gate, allowing a much larger current to flow through the transistor. You might expect that for a given, constant current, you would always need the same $V_{BE}$. But that’s not what happens. As the transistor heats up, something remarkable occurs: the voltage required to maintain the same current decreases.

This behavior isn't random; it's a direct consequence of the physics of semiconductors. The voltage $V_{BE}$ has a strong, predictable, and nearly linear negative temperature coefficient. For every degree Celsius the temperature rises, $V_{BE}$ drops by about 2 millivolts. We call this a ​​Complementary to Absolute Temperature​​, or ​​CTAT​​, voltage.

Why does this happen? The answer lies in a beautiful tug-of-war within the material. The flow of current in a semiconductor depends on charge carriers (electrons and holes) having enough energy to overcome an energy barrier. As you increase the temperature, you are essentially giving every particle in the system more thermal energy, described by the term $k_B T$, where $T$ is the absolute temperature and $k_B$ is the Boltzmann constant. These more energetic carriers can cross the junction barrier more easily. Furthermore, the heat also creates more charge carriers to begin with, a phenomenon governed by the semiconductor's bandgap energy and described by the temperature-dependent saturation current, $I_S$. The combined effect is that the junction becomes "more willing" to conduct at higher temperatures. To keep the current constant, we must therefore reduce the "push" we are giving it—that is, we must lower $V_{BE}$. A detailed analysis reveals that the temperature coefficient, $\frac{dV_{BE}}{dT}$, is a negative value determined by fundamental constants, the bandgap voltage of silicon, and the initial operating point.

For a typical silicon transistor operating at room temperature, this coefficient is around $-2.0 \text{ mV/K}$. This may seem small, but in a high-precision instrument, a 20-degree change could cause a 40 mV shift—an eternity in the world of analog circuits.

So, we have a voltage that reliably goes down with temperature. This seems like a problem. But what if we could be alchemists? What if we could take this troublesome temperature dependence and create its exact opposite—a voltage that goes up with temperature? A voltage that is ​​Proportional to Absolute Temperature​​, or ​​PTAT​​. If we could create such a voltage, we could add it to our CTAT voltage in just the right proportion, and the two opposing trends would cancel each other out, leaving us with a perfectly stable voltage.

This is not alchemy; it's exquisite engineering. The secret lies not in one transistor, but in two. Let's look again at the equation governing the transistor's behavior:

$V_{BE} = V_T \ln\left(\frac{I_C}{I_S}\right)$

Notice that pesky term $V_T = k_B T / q$. This is the ​​thermal voltage​​, and it is directly, beautifully proportional to absolute temperature $T$. The logarithm term, $\ln(I_C/I_S)$, contains all the complicated, messy temperature dependencies we discussed before. The genius move is to find a way to get rid of the logarithm and isolate the pure $V_T$.

How? By taking a difference. Consider two transistors, Q1 and Q2. Their base-emitter voltage difference is:

$\Delta V_{BE} = V_{BE1} - V_{BE2} = V_T \ln\left(\frac{I_{C1}}{I_{S1}}\right) - V_T \ln\left(\frac{I_{C2}}{I_{S2}}\right) = V_T \ln\left(\frac{I_{C1}}{I_{C2}} \cdot \frac{I_{S2}}{I_{S1}}\right)$

Now we have a choice of two clever tricks to simplify this expression.

​​Method 1: Play with Geometry.​​ Let's build our two transistors on the same piece of silicon, but make one physically larger than the other. Specifically, let's say the emitter area of Q2 is $N$ times the area of Q1. Since the saturation current ($I_S$) is proportional to this area, we have $I_{S2} = N \cdot I_{S1}$. Now, let's force the same collector current ($I_{C1} = I_{C2}$) through both transistors. Look what happens to our equation:

$\Delta V_{BE} = V_T \ln\left(\frac{I_{C1}}{I_{C1}} \cdot \frac{N \cdot I_{S1}}{I_{S1}}\right) = V_T \ln(N)$

And there it is! By taking the difference between two transistors operating at the same current but with different sizes, we have created a voltage, $\Delta V_{BE}$, that is equal to the thermal voltage $V_T$ multiplied by a constant, $\ln(N)$. Since $V_T$ is proportional to absolute temperature, our $\Delta V_{BE}$ is a perfect PTAT voltage. From a design perspective, an engineer can choose the area ratio $N$ to generate a desired PTAT slope. For instance, to create a specific PTAT voltage of $21.5 \text{ mV}$ at room temperature, one would need to design the transistors with a very specific ratio of current densities, which is achieved by setting a precise area ratio.

​​Method 2: Play with Currents.​​ What if we use two identical transistors? In this case, $I_{S1} = I_{S2}$. This time, we'll force their collector currents to operate at a fixed ratio, say $I_{C1} = M \cdot I_{C2}$. The equation for the voltage difference becomes:

$\Delta V_{BE} = V_T \ln\left(\frac{M \cdot I_{C2}}{I_{C2}} \cdot \frac{I_{S1}}{I_{S1}}\right) = V_T \ln(M)$

Once again, we have produced a pure PTAT voltage. This demonstrates a beautiful unity in the underlying physics: whether we create an asymmetry in the device's geometry or in its operating current, the result is the same. We have successfully conjured a voltage that rises linearly with temperature.

Now we have our two dancers. In one corner, we have $V_{BE}$, the CTAT voltage, which gracefully bows as the temperature rises. In the other, we have $\Delta V_{BE}$, our engineered PTAT voltage, which stands taller as it gets warmer. It's time to bring them together to create a perfectly stable reference voltage, $V_{REF}$.

The idea is simple: we create a weighted sum of the two.

$V_{REF} = V_{BE} + G \cdot \Delta V_{BE}$

Here, $G$ is a scaling factor, typically set by a ratio of resistors in the circuit. Our goal is to make the total change in $V_{REF}$ with temperature equal to zero. Mathematically, we want $\frac{dV_{REF}}{dT} = 0$.

$\frac{dV_{REF}}{dT} = \frac{dV_{BE}}{dT} + G \cdot \frac{d(\Delta V_{BE})}{dT} = 0$

Since we know $\frac{dV_{BE}}{dT}$ is a negative constant (our CTAT slope) and $\frac{d(\Delta V_{BE})}{dT}$ is a positive constant (our PTAT slope), we can always find a positive gain $G$ that satisfies the equation:

$G = -\frac{dV_{BE}/dT}{d(\Delta V_{BE})/dT}$

By choosing the right resistor ratio, we can lock our two opposing dancers into a perfect balance, creating a reference voltage that, in theory, does not change with temperature at all. If, due to a manufacturing flaw, our scaling factor $G$ is slightly too large, the positive PTAT trend will overpower the negative CTAT trend, and the final reference voltage will have a small, residual positive temperature coefficient.

When you perform this cancellation and calculate the value of the resulting stable voltage, something magical appears. The resulting $V_{REF}$ is very close to 1.205 volts for silicon. This number is no accident. It is the extrapolated ​​bandgap voltage​​ of silicon at absolute zero. The cancellation process has, in effect, revealed a fundamental property of the material itself. This is why these circuits are called ​​bandgap voltage references​​, and they are one of the most elegant and essential building blocks in all of analog electronics.

Is our story finished? Have we achieved perfection? Not quite. Nature is always a bit more subtle and interesting than our simple models. If you were to build this circuit and precisely measure its output voltage over a wide range of temperatures, you would not see a perfectly flat line. Instead, you would see a gentle parabola, a "bowing" shape, which is flat only at the specific temperature for which you designed the cancellation.

Where does this curvature come from? It comes from the fact that we cheated a little. Our model of $V_{BE}$ as a perfectly straight line decreasing with temperature was an excellent first-order approximation, but it wasn't the whole truth. A more precise physical model of $V_{BE}$ reveals that it contains not just a linear term in $T$, but also higher-order terms, most notably a term proportional to $T\ln(T)$.

Our PTAT voltage, $G \cdot V_T \ln(N)$, is relentlessly linear—it's a perfect straight line. You cannot use a straight line to cancel a curved line at every single point. You can make them tangent at one point—meaning their values and their slopes match—but you cannot eliminate the difference in their curvature. The non-zero second derivative of the $V_{BE}(T)$ function, originating from that $T\ln(T)$ term and other non-linearities, is what remains after our first-order cancellation. This residual curvature, $\frac{d^2V_{REF}}{dT^2}$, is the mathematical fingerprint of the bowing shape we observe in reality.

This isn't a failure; it's a window into deeper physics. It tells us that the relationship between energy, temperature, and current in a semiconductor is a rich, non-linear phenomenon. While simple models allow us to achieve remarkable feats of engineering, the true behavior of the universe always holds more complexity and beauty for us to discover. The humble PTAT voltage, born from a clever trick with two transistors, not only gives us the stable references that underpin our digital world but also serves as a reminder of the elegant and intricate dance of physics happening inside every piece of silicon.

We have explored the elegant physics that gives rise to a voltage Proportional to Absolute Temperature—the PTAT voltage. In essence, we have learned how to build a very precise electronic thermometer. But what if our goal is the exact opposite? What if we need a voltage that is profoundly indifferent to the thermal chaos of the world around it? What if we need an unwavering yardstick, a perfect and stable ruler against which all other signals in a circuit can be measured, whether that circuit is in a scorching desert or the frigid vacuum of space?

This is where the true genius of the PTAT voltage is revealed. Its greatest application is not in being a thermometer itself, but in serving as the key ingredient to create its perfect antithesis: the temperature-independent voltage reference. This is a story of turning a predictable change into an unshakable stability, a beautiful dance of opposing forces that lies at the heart of nearly every modern electronic device.

Nature has already given us a voltage that changes reliably with temperature: the base-emitter voltage, $V_{BE}$, of a bipolar junction transistor (BJT). As we've seen, this voltage is "Complementary to Absolute Temperature" (CTAT)—it decreases almost linearly as temperature rises. Now, we have engineered its counterpart, the PTAT voltage, which increases linearly with temperature. An imaginative physicist or engineer, upon seeing two opposing, predictable trends, would immediately ask: can we make them fight each other to a standstill?

The answer is a resounding yes. The core principle of a ​​bandgap voltage reference​​ is to sum these two opposing voltages. We take the falling CTAT voltage, $V_{BE}$, and add to it a carefully scaled version of our rising PTAT voltage.

$V_{REF} = V_{CTAT} + K \cdot V_{PTAT}$

By choosing the scaling factor $K$ just right, we can ensure that the rate at which the CTAT term falls with temperature is perfectly canceled by the rate at which the scaled PTAT term rises. The result is a combined voltage, $V_{REF}$, whose derivative with respect to temperature is zero, at least around our desired operating temperature. We have achieved a first-order cancellation, creating a stable voltage out of two unstable ones.

This is more than just a mathematical trick; it has an elegant practical implementation. The PTAT voltage itself is typically generated as the difference in base-emitter voltages, $\Delta V_{BE}$, between two transistors operating at different current densities. This is often accomplished with the help of an operational amplifier, whose feedback loop works tirelessly to enforce the exact conditions—namely, equal collector currents—needed to produce a clean $\Delta V_{BE}$ signal. How is this voltage then scaled by the magic factor $K$? In a marvel of integrated circuit design, this is typically achieved using a simple ratio of two resistors. By passing a current proportional to $\Delta V_{BE}$ through a second resistor, we generate the scaled voltage. The scaling factor $K$ becomes a function of this resistor ratio, for instance $K \propto R_2/R_1$. This is a profoundly clever choice, because on a silicon chip, it is far easier to manufacture two resistors with a very precise ratio than it is to create a single resistor with a precise absolute value. The ratio remains stable even if the absolute resistor values drift with manufacturing variations or temperature, a testament to the robust principles of analog design.

So we have performed this delicate balancing act. We have cancelled the temperature dependencies and produced a stable voltage. What is its value? Is it arbitrary? Here, the story takes a turn from clever engineering to profound physics. For a reference built with silicon transistors, this stable voltage invariably turns out to be around $1.22$ volts. This number is not an accident of the circuit's design; it is a whisper from the quantum world of the material itself.

If we trace the mathematics of the $V_{BE}$ voltage back to its origins in semiconductor physics, we find that it contains terms that depend on temperature and, buried within, a term related to the ​​bandgap energy​​ of silicon, $E_g$. The bandgap is the fundamental quantum-mechanical property that defines a semiconductor—the energy required to break an electron free to conduct electricity. The entire purpose of the PTAT circuitry is to create a set of temperature-dependent terms that precisely cancel all the other temperature-dependent terms in the $V_{BE}$ expression. When the dust settles, what remains? At an extrapolation to absolute zero ($T=0$ K), where all thermal effects cease, the reference voltage $V_{REF}$ becomes equal to the bandgap energy of silicon divided by the elementary charge, $E_{g0}/q$. The circuit, through its elegant cancellation, has managed to "excavate" this fundamental physical constant and present it as a stable, macroscopic voltage. It's a breathtaking connection, where a practical electronic component becomes an instrument for measuring a cornerstone of solid-state physics.

Is this remarkable principle confined to the world of Bipolar Junction Transistors? Fortunately, no. The fundamental idea is more universal. Modern electronics are dominated by Metal-Oxide-Semiconductor (MOS) transistors, the building blocks of the digital processors in our computers and phones. While the device physics is different, the underlying statistical mechanics are related. One can construct a PTAT voltage using two MOS transistors operating in their "subthreshold" or "weak inversion" regime. By forcing the same tiny current through two transistors of different sizes (specifically, with different width-to-length ratios, $W/L$), a voltage difference appears between them that is, once again, proportional to the thermal voltage $V_T$. This MOS-based PTAT voltage can then be used in exactly the same way to cancel the temperature dependence of another voltage, creating a stable reference entirely within a standard CMOS process. This universality ensures that the principle of temperature compensation remains a vital tool across the entire landscape of microelectronics.

Beyond voltage references, the PTAT principle finds direct application in creating stable ​​biasing currents​​ for other analog circuits and, by its very nature, in creating on-chip ​​temperature sensors​​. A simple PTAT current or voltage can be digitized to provide a direct reading of the chip's temperature, a critical function for performance management and thermal protection in high-performance processors.

Of course, our story of perfect cancellation is an idealized one. Real-world engineers spend much of their time wrestling with the messy imperfections that deviate from this beautiful ideal.

• ​​The Flawed Amplifier:​​ The operational amplifier we use to generate the PTAT voltage is not perfect. It has a small, intrinsic DC error known as the ​​input offset voltage​​, $V_{os}$. This offset voltage gets added directly to the $\Delta V_{BE}$ term. The result is that this small error is then amplified by the very same resistor ratio we used for scaling, leading to a significant and temperature-dependent error in the final output voltage.

• ​​The Unsteady Power Supply:​​ What happens if the circuit's power supply voltage, $V_{DD}$, isn't perfectly stable? A good reference should be immune to this. However, due to the non-ideal behavior of transistors (an effect quantified by the "Early voltage") and the finite resistance of the current sources providing power, some of this supply variation can "leak" through and appear as an error in the output. The ability of the reference to reject this interference is a key metric known as the Power Supply Rejection Ratio (PSRR).

• ​​The Disobedient Resistors:​​ Our cancellation scheme relied on the scaling factor $K$ being constant. But the resistors used to set this scaling have their own temperature coefficient; their resistance changes slightly with temperature. This non-ideality means our PTAT term is not perfectly linear, spoiling the cancellation and leaving a residual "bow" or curvature in the $V_{REF}$ vs. temperature plot. This has led to the development of more advanced, higher-order compensation techniques to cancel this curvature as well, a constant battle for ever-greater stability.

Let's imagine we could vanquish all these imperfections. We build our circuit with a perfect op-amp, perfectly stable resistors, and a rock-solid power supply. Is there a limit to how "quiet" and stable our reference can be? Yes. The limit is set by fundamental physics.

Electric current is not a smooth, continuous fluid. It is composed of a stream of discrete particles: electrons. The random, statistical arrival of these individual charge carriers gives rise to an irreducible noise known as ​​shot noise​​. This noise is present in the collector currents of our BJTs. When we build our reference voltage from the $V_{BE}$ of these transistors, this microscopic current noise is translated into a tiny, ceaseless flicker in the output voltage. This "noise floor" represents the ultimate limit of precision for our bandgap reference. We can calculate its magnitude, and it depends on fundamental constants like the elementary charge $q$, the Boltzmann constant $k_B$, and the absolute temperature $T$. Here, at the very edge of performance, the world of practical circuit design confronts the quantum and statistical nature of reality itself.

From a simple observation about transistor physics to a universal engineering principle, the PTAT voltage has taken us on a remarkable journey. It has shown us how to turn a predictable flaw into a source of unparalleled stability, how to build a circuit that reveals a fundamental constant of nature, and how the quest for perfection ultimately runs into the fundamental graininess of the universe.