Cascode Current Mirror

In analog electronics, the creation of a stable, constant current source is a foundational requirement for countless high-performance circuits. However, simple single-transistor current sources are imperfect, suffering from effects like channel-length modulation that cause their output current to vary with voltage. This limitation, quantified by a finite output resistance, hinders the precision of circuits like amplifiers and data converters. How, then, can we engineer a current source that behaves closer to the ideal, resisting any change in current regardless of external conditions?

This article delves into the cascode current mirror, an elegant and powerful solution to this very problem. First, under "Principles and Mechanisms," we will dissect how stacking two transistors creates a "shielding" effect that multiplies output resistance, exploring the trade-off this creates with voltage headroom and examining the ingenious "wide-swing" variant that mitigates this compromise. Following that, in "Applications and Interdisciplinary Connections," we will see how this fundamental technique is applied to build high-gain amplifiers, design superior op-amps, stabilize voltage references, and even how its theoretical elegance translates into the physical geometry of chip layout.

In our journey to understand the world of electronics, we often encounter the need for something remarkably stable: a source that provides a constant electric current, regardless of what it's connected to. Imagine a garden hose that delivers exactly one gallon per minute, whether it’s spraying into the open air or trying to push water through a narrow nozzle. In electronics, this is the job of a ​​current source​​, a fundamental building block for nearly all analog circuits, from amplifiers to digital-to-analog converters.

A simple approach is to use a single transistor. We can set its gate voltage to a fixed value and hope it delivers a steady current. But nature is subtle. The current flowing through a real transistor, a Metal-Oxide-Semiconductor Field-Effect Transistor (MOSFET), for instance, is not perfectly constant. It has a slight, almost imperceptible dependence on the voltage across it. This phenomenon, known as ​​channel-length modulation​​, is a bit like our garden hose letting a little more water through if the pressure at the nozzle drops. For a Bipolar Junction Transistor (BJT), a similar phenomenon is called the ​​Early effect​​. We quantify this imperfection with a parameter called ​​output resistance​​, denoted as $r_o$. A perfect current source would have an infinite output resistance; it would resist any change in current no matter how the output voltage varies. Our simple transistor, with its finite $r_o$, is a good start, but for high-precision applications, we need to do much, much better.

How can we fight against this pesky channel-length modulation? How can we make our current source behave as if its output resistance is nearly infinite? The answer lies in a wonderfully elegant and powerful idea: the ​​cascode​​ configuration.

The idea is simple: we stack a second, identical transistor on top of our original current-defining transistor. Let's call the bottom one $M_1$ and the top one $M_2$. The output is now taken from the top of the stack, at the drain of $M_2$. At first glance, this might seem to just complicate things. But the top transistor, $M_2$, plays a profound role: it acts as a voltage ​​shield​​.

Think of it this way. The job of the bottom transistor, $M_1$, is to set the current. It does this best when the voltage across it, $V_{DS1}$, is held perfectly still. The job of the top transistor, $M_2$, is to absorb any and all voltage fluctuations from the output terminal and prevent them from reaching $M_1$. Imagine $M_2$ as a vigilant guard standing between the chaotic outside world (the changing output voltage) and the quiet, stable sanctum of $M_1$.

When the output voltage at the drain of $M_2$ changes, $M_2$ adjusts its own voltage drop to compensate. It leverages its own characteristics to keep the voltage at its source (which is also the drain of $M_1$) remarkably stable. Since the voltage across $M_1$ is now shielded from the outside world, its current, which was previously susceptible to channel-length modulation, becomes extraordinarily constant. The cascode doesn't eliminate channel-length modulation; it simply renders it harmless by stabilizing the conditions around the primary transistor.

This shielding effect can be described with mathematical beauty. To see how dramatic the improvement is, we can use a small-signal model, which tells us how the circuit responds to tiny changes. For a single transistor, the output resistance is simply its own intrinsic resistance, $r_{o1}$.

Now consider the cascode. We are looking into the drain of the top transistor, $M_2$. Let's trace the path of a small change. Any voltage change at the output has to contend not just with $M_2$'s own output resistance, $r_{o2}$, but also with its active nature. The top transistor $M_2$ acts to oppose changes at its source. This "fighting back" is the essence of its transconductance, $g_{m2}$.

A detailed analysis, like the one performed in, reveals a stunning result. The total output resistance of the two-transistor stack is not just the sum of the individual resistances. It is approximately:

$R_{out} \approx g_{m2} r_{o1} r_{o2}$

Let's pause to appreciate this. The term $g_m r_o$ is the intrinsic voltage gain of a single transistor, a number that is often in the range of 20 to 100. The cascode configuration multiplies the resistance of the bottom transistor, $r_{o1}$, by the entire intrinsic gain of the top transistor, $g_{m2}r_{o2}$! We have taken a modest resistance and amplified it by a huge factor. This "resistance multiplication" is the secret behind the cascode's power. If a simple mirror has an output resistance of $50 \, \text{k}\Omega$, a cascode version can easily exceed $5 \, \text{M}\Omega$, an improvement of a hundredfold. This is the kind of clever trick that makes circuit design such a fascinating art.

As is often the case in physics and engineering, there is no free lunch. The spectacular increase in output resistance comes at a price: ​​voltage compliance​​, or what we often call ​​headroom​​.

For a transistor to operate correctly as a current source, it must be in its "saturation" region. This requires a certain minimum voltage drop across it, known as the ​​overdrive voltage​​, $V_{ov}$, for a MOSFET, or the ​​saturation voltage​​, $V_{CE(sat)}$, for a BJT. A single-transistor current source needs only one such voltage drop. But our cascode has two transistors stacked in series. To keep the entire stack working properly, we must provide enough total voltage to satisfy the minimum requirement for both transistors.

At a bare minimum, the total voltage needed across the cascode stack is the sum of the minimums for each transistor, or about $2V_{ov}$. This means the output voltage of our current source cannot drop below this value. This "lost" voltage range reduces the operational window, or "swing," of the circuit it's part of. In a world of shrinking battery voltages and low-power electronics, every fraction of a volt counts. This trade-off between high output resistance and available voltage headroom is a central challenge in analog design.

This brings us to the next chapter of our story. Engineers, faced with this trade-off, asked a simple question: can we do better? Can we get the magnificent output resistance of the cascode without paying such a high price in voltage headroom? The answer, born of ingenuity, is the ​​wide-swing cascode​​.

The problem with a simple, "self-biased" cascode is that its biasing scheme is inefficient. In such a design, the gate voltage for the cascode transistor is often generated by a simple diode-connected stack, which forces the voltage at the intermediate node—the drain of the bottom transistor—to be a full $V_{GS} = V_{th} + V_{ov}$. But to keep the bottom transistor saturated, it only needs a minimum voltage of $V_{ov}$. The extra threshold voltage, $V_{th}$, is wasted headroom.

The wide-swing cascode employs a more intelligent biasing scheme. It uses a dedicated circuit to generate a precise bias voltage for the cascode gate. This voltage is crafted to do one thing perfectly: hold the voltage at the drain of the bottom transistor at the absolute minimum required for saturation, which is $V_{ov}$. No more, no less.

The result is that the minimum required output voltage for the entire stack becomes exactly $2V_{ov}$, the theoretical floor. Compared to the standard cascode's requirement of $V_{th} + 2V_{ov}$, we have reclaimed an entire threshold voltage worth of headroom!. This allows designers to build high-performance circuits that can operate from much lower supply voltages, a critical advantage in modern electronics.

The elegance of the cascode extends even to its more subtle, real-world behaviors. For instance, in an NMOS cascode built on a common substrate, the top transistors have their source terminals "floating" at some voltage above ground. This creates a non-zero source-to-bulk voltage, which triggers the ​​body effect​​, subtly increasing the transistor's threshold voltage. While this is a static and predictable effect on the reference side of a current mirror, on the output side it can be more pernicious. Small fluctuations in the output voltage can feed back to the intermediate node, modulating the body effect and slightly degrading the mirror's precision. It's a reminder that in the real world, every component is interconnected in subtle ways.

But perhaps the most beautiful and surprising property of the cascode emerges at high frequencies. One might think of transistors and their parasitic capacitances as just that—resistors and capacitors. But when arranged in a cascode, they conspire to create something entirely unexpected.

Imagine a rapidly rising voltage at the output. This changing voltage pushes a tiny current through the parasitic capacitance between the top transistor's gate and drain. This current is injected into the very sensitive internal node between the two transistors. The top transistor, seeing its source voltage rise, gets partially "turned off," actively resisting the flow of current from the output. This opposition to a change in current is the very definition of an ​​inductor​​.

Amazingly, at high frequencies, the cascode current source—a circuit made of silicon, with no coils of wire—begins to behave like an inductor! This can lead to resonance, causing the output impedance to peak at a specific frequency, a phenomenon known as "inductive peaking". This is a profound illustration of how simple elements, when combined in a clever structure, can give rise to complex and emergent behaviors. It is in discovering these hidden connections and unexpected transformations that we find the true beauty and unity of physics at play in the world of engineering.

After our journey through the fundamental principles of the cascode current mirror, you might be left with a feeling of satisfaction, but also a question: "This is a clever trick, but what is it for?" It is a fair question. A principle in physics or engineering is only as powerful as the problems it can solve. The beauty of the cascode connection is not just in its elegant structure, but in its profound and widespread utility. It is not merely a single circuit; it is a powerful strategy—a way of thinking—that has become a cornerstone of modern electronics. Let us now explore how this simple idea of "stacking" transistors unlocks extraordinary performance across a vast landscape of applications, from the heart of amplifiers to the physical silicon of the chip itself.

At the core of electronics lies the need to make small things big. We are constantly faced with signals—the faint whisper from a distant radio antenna, the minuscule electrical pulse from a neuron—that are too weak to be useful on their own. We need to amplify them. The voltage gain of a simple amplifier, in its essence, can often be described by a wonderfully simple relationship: it is the product of the transistor's ability to convert voltage into current (its transconductance, $g_m$) and the total resistance at its output ($R_{out}$).

$A_v \approx -g_m R_{out}$

If we desire a colossal gain, we need a colossal output resistance. But how does one create a very large resistor on a tiny integrated circuit? Building a passive resistor with such a high value would consume an enormous amount of precious silicon area. Here, the cascode current mirror enters not as a mirror, but as a so-called "active load." By connecting a cascode current source as the load for our amplifying transistor, we present the amplifier with an incredibly high dynamic resistance. The cascode's trick of shielding its main current-setting transistor from voltage variations means it stubbornly refuses to change its current, which, by Ohm's law ($R = V/I$), is the very definition of a high resistance. The result? Gains that are hundreds or even thousands of times larger than what a simple load could provide.

However, as is so often the case in nature, there is no such thing as a free lunch. This remarkable gain comes at a cost, a fundamental trade-off that every analog designer must navigate: voltage headroom. To perform their magic, each transistor in the cascode stack needs a certain minimum voltage drop across it to remain in its active, "saturated" region. By stacking two transistors, we effectively double this voltage requirement. For an amplifier swinging between a power supply voltage $V_{DD}$ and ground, the maximum and minimum output voltages are squeezed inwards, reducing the available "swing". For example, in a typical design, we might lose two overdrive voltage drops at the top and two at the bottom, shrinking a 3.3 V potential swing down to just 2.5 V. This eternal battle between achieving high gain and maintaining a wide output swing is a central drama in the world of amplifier design. Fortunately, clever refinements to the cascode, such as the "wide-swing" cascode, have been developed to cleverly bias the transistors and reclaim some of this lost headroom, giving designers more freedom.

From simple amplifiers, we graduate to one of the most versatile and powerful building blocks in all of electronics: the operational amplifier, or op-amp. The ideal op-amp amplifies the difference between its two inputs while completely ignoring any signal that is common to both. This ability to reject common signals—like electrical noise picked up by a pair of wires—is measured by the Common-Mode Rejection Ratio (CMRR). A high CMRR is not a luxury; it is the defining feature of a high-quality differential amplifier.

The secret to a high CMRR lies in the "tail" of the input differential pair. This is a current source that biases the two input transistors. An ideal tail source would be a perfect, unshakable provider of current, no matter what the common-mode input voltage does. Any imperfection, any finite resistance in this tail source, allows the common-mode voltage to "leak" through and be amplified, degrading the CMRR.

This is another perfect job for the cascode. By implementing the tail current source with a cascode structure instead of a simple single-transistor source, we dramatically increase its output resistance. This makes the tail source behave much more like the ideal. The improvement is not subtle. As a straightforward analysis shows, switching to a cascode can boost the CMRR by a factor proportional to the intrinsic gain ($g_m r_o$) of the transistors themselves—a factor that can easily be in the thousands. It is a textbook example of how a single, well-placed circuit technique can elevate the performance of an entire system by orders of magnitude.

The cascode multiplies output resistance by a factor of roughly $g_m r_o$. This raises a tantalizing question: if we can multiply it once, can we multiply it again? Can we get even closer to an infinitely high resistance? The answer is a resounding yes, and the technique is as beautiful as the cascode itself. It is called "gain boosting" or a "regulated cascode."

The idea is to add a small auxiliary amplifier into the cascode structure. This helper amplifier's job is to watch the voltage at the sensitive node between the two cascode transistors and actively adjust the gate voltage of the upper transistor to hold that node perfectly steady. This creates a powerful local feedback loop. The result is that the output resistance of the standard cascode, which was already huge ($R_{out} \approx g_m r_o^2$), is now multiplied again by the gain of the auxiliary amplifier, $A_{gb}$. The output resistance becomes truly astronomical: $R_{out} \approx A_{gb} (g_m r_o^2)$.

This technique is the engine behind modern, ultra-high-gain op-amps, like telescopic and folded-cascode architectures. By using gain-boosting on the active load, designers can achieve open-loop gains well into the millions from a single stage. This pursuit is a wonderful illustration of the iterative nature of science and engineering—a good idea (the cascode) inspires an even better one (the regulated cascode), pushing the boundaries of what is possible.

So far, we have seen the cascode in its role as a performance-enhancer for amplifiers. But it has another, equally critical role: as a guardian of stability. In virtually every electronic system, from your phone to a satellite, there are circuits called ​​voltage references​​. Their job is to produce a single, rock-solid voltage that serves as a reliable anchor for the rest of the circuitry. A good reference voltage must be immune to changes in temperature and, crucially, to fluctuations in the power supply voltage.

The ability of a circuit to ignore variations in its power supply is called the Power Supply Rejection Ratio (PSRR). Imagine trying to measure a delicate object on a ship tossing in a stormy sea; it's impossible unless your measuring scale is somehow isolated from the ship's motion. A bandgap voltage reference faces the same challenge. The cascode current mirror is the isolation mechanism.

When used as an active load within a bandgap reference, the cascode's extremely high output impedance acts as a formidable barrier between the power supply and the sensitive core of the reference circuit. It effectively says to the supply voltage, "You can jump and jitter all you want, but you will not disturb the currents flowing here." By replacing a simple current mirror with a cascode, designers can dramatically improve the PSRR, ensuring the reference voltage remains stable even when the supply is "noisy". This application showcases the cascode not in an offensive role (to create gain) but in a defensive one (to reject interference).

Our discussion has remained in the world of circuit diagrams and equations. But these transistors are real, physical objects fabricated on a slice of silicon. And in the real world, nothing is perfect. Across the surface of a silicon wafer, there are subtle, unavoidable gradients in temperature, chemical concentrations, and material thickness during fabrication.

For a current mirror to work accurately, its two paired transistors must be as identical as possible. If they are placed far apart on the chip, these process gradients mean one might have a slightly different threshold voltage or current factor than the other, ruining the mirror's precision.

Here, we see a stunningly beautiful connection between electrical engineering and pure geometry. To solve this problem, designers use a technique called ​​common-centroid layout​​. For a four-transistor cascode mirror (M1/M2 for the mirror pair, M3/M4 for the cascode pair), the four devices can be arranged in a 2x2 grid. The optimal arrangement is a cross-coupled pattern, such as placing M1 and M2 on one diagonal, and M3 and M4 on the other.

Why does this work? The center of mass, or centroid, of the M1/M2 pair is now at the exact geometric center of the 2x2 square. The centroid of the M3/M4 pair is at the very same spot. Now, any linear process gradient across the layout—say, hotter on the left than the right—will affect M1 and M2 in equal and opposite ways, and the errors will average out to zero! It is a triumph of symmetry, a physical embodiment of the cancellation of errors. It shows that brilliant circuit design extends beyond the schematic, all the way down to the artful placement of atoms on a chip.

From boosting the gain of an amplifier, to sharpening the precision of an op-amp, to anchoring the stability of an entire system, and finally to the geometric elegance of its own physical layout, the cascode principle is a testament to the power of a simple, profound idea. It is a recurring motif in the symphony of analog design, a beautiful example of how we can stand on the shoulders of one device to reach a higher level of performance.