7.1 RF Mixers, Frequency Translation, and Intermodulation

7.1 RF Mixers, Frequency Translation, and Intermodulation

In modern radio frequency communications, the ability to translate signals seamlessly from one frequency band to another is the cornerstone of transmitter and receiver design. Whether shifting a baseband voice channel up to the HF, VHF, or UHF spectrum for transmission, or translating an incoming microvolt signal from an antenna down to a fixed intermediate frequency (IF) for sharp filtering and amplification, the fundamental operational block responsible for this frequency translation is the RF mixer.

While linear circuits—such as ideal Class A amplifiers and passive filters—preserve the exact frequency content of an input waveform, frequency translation strictly requires a non-linear or time-varying process. Understanding the physics of non-linear mixing, the operational characteristics of balanced mixer topologies, and the mechanisms of intermodulation distortion is essential for mastering advanced radio engineering.

---

1. Principles of Non-Linear Mixing and Frequency Translation

Frequency mixing is mathematically described as the time-domain multiplication of two distinct signals: a Radio Frequency (RF) signal at frequency $\omega_{\text{RF}} = 2\pi f_{\text{RF}}$, and a Local Oscillator (LO) signal at frequency $\omega_{\text{LO}} = 2\pi f_{\text{LO}}$.

The Mathematical Multiplication Identity

Consider two pure sinusoidal voltages applied to an ideal multiplier with scaling constant $k$:

$$v_{\text{RF}}(t) = V_{\text{RF}} \cos(\omega_{\text{RF}} t)$$ $$v_{\text{LO}}(t) = V_{\text{LO}} \cos(\omega_{\text{LO}} t)$$

Multiplying these two time-domain signals yields:

$$v_{\text{out}}(t) = k \cdot v_{\text{RF}}(t) \cdot v_{\text{LO}}(t) = k V_{\text{RF}} V_{\text{LO}} \cos(\omega_{\text{RF}} t) \cos(\omega_{\text{LO}} t)$$

Applying the standard trigonometric product-to-sum identity:

$$\cos(A) \cos(B) = \frac{1}{2} \left[ \cos(A + B) + \cos(A - B) \right]$$

we obtain:

$$v_{\text{out}}(t) = \frac{k V_{\text{RF}} V_{\text{LO}}}{2} \left[ \cos\left((\omega_{\text{RF}} + \omega_{\text{LO}})t\right) + \cos\left((\omega_{\text{RF}} - \omega_{\text{LO}})t\right) \right]$$

The resulting output spectrum contains no energy at either original input frequency ($f_{\text{RF}}$ or $f_{\text{LO}}$). Instead, it consists purely of two new sideband frequencies:

1. The Sum Frequency: $f_{\text{sum}} = f_{\text{RF}} + f_{\text{LO}}$ (used in transmitter heterodyne up-conversion).
2. The Difference Frequency: $f_{\text{diff}} = |f_{\text{RF}} - f_{\text{LO}}|$ (used in superheterodyne receiver down-conversion).

Why Non-Linearity is Essential

In a strictly linear time-invariant (LTI) network, the output contains only the frequencies present at the input; amplitudes and phases may change, but new spectral lines cannot be created. To generate new frequency components, the active circuit device must possess a non-linear transfer characteristic.

The transfer function of any physical semiconductor diode, bipolar transistor, or field-effect transistor can be modeled using a Taylor series expansion relating output current $i(t)$ to input voltage $v(t)$:

$$i(t) = a_0 + a_1 v(t) + a_2 v^2(t) + a_3 v^3(t) + a_4 v^4(t) + \dots$$

When a composite voltage consisting of the sum of the RF and LO signals ($v(t) = v_{\text{RF}}(t) + v_{\text{LO}}(t)$) is applied to this non-linear device, the quadratic term ($a_2 v^2(t)$) expands as follows:

$$a_2 v^2(t) = a_2 \left( v_{\text{RF}}(t) + v_{\text{LO}}(t) \right)^2 = a_2 \left[ v_{\text{RF}}^2(t) + 2 v_{\text{RF}}(t) v_{\text{LO}}(t) + v_{\text{LO}}^2(t) \right]$$

The cross-product term $2 a_2 v_{\text{RF}}(t) v_{\text{LO}}(t)$ is precisely the mathematical multiplication required to yield the sum ($f_{\text{RF}} + f_{\text{LO}}$) and difference ($|f_{\text{RF}} - f_{\text{LO}}|$) frequencies. However, the squared individual terms ($v_{\text{RF}}^2(t)$ and $v_{\text{LO}}^2(t)$) and higher-order terms ($a_3 v^3, a_4 v^4$) generate direct-current (DC) offsets, second harmonics ($2f_{\text{RF}}, 2f_{\text{LO}}$), third-order intermodulation products, and higher-order spurious responses.

---

2. RF Mixer Topologies and Circuit Architectures

Practical RF mixers are classified by their degree of electrical balance and whether the active elements provide conversion gain or loss.

Unbalanced Mixer

An unbalanced mixer utilizes a single non-linear device (such as a single point-contact diode, Schottky diode, or JFET) with both the RF and LO signals coupled into the same input terminal or adjacent gate/source terminals.

• Operation: The device conducts non-linearly on the combined waveform.
• Disadvantages: There is no isolation between ports. The high-level LO signal leaks directly into the RF and IF ports. LO power leaking out of the RF port can radiate from the station antenna as an illegal spurious emission, while LO power entering the IF port severely overloads sensitive intermediate-frequency amplifier stages.
• Application: Restricted to primitive microwave detectors or low-cost experimental circuits.

Single-Balanced Mixer

A single-balanced mixer employs two matched active devices (diodes or FETs) connected to a center-tapped $180^\circ$ broadband transformer (balun).

• Operation: A $180^\circ$ input hybrid applies the LO in opposition to two matched diodes and the RF in phase. The outer diode terminals join at the IF output; diode orientation and hybrid balance cancel direct LO feedthrough there while retaining the mixing product.
• Performance: The LO-balanced arrangement isolates LO from RF and suppresses LO at IF, commonly by about $20\text{ dB}$ to $30\text{ dB}$.
• Limitations: Only one input port (either RF or LO) is balanced; the other port still leaks into the IF output.

Double-Balanced Diode Ring Mixer (DBM)

The double-balanced diode ring mixer is the workhorse of high-performance amateur HF/VHF transceivers. It consists of four carefully matched Schottky barrier diodes arranged in a continuous ring (quaternary quad), driven by two broadband transmission line center-tapped transformers ($T_1$ for RF, $T_2$ for LO).

Diode ring, with each arrow running from anode to cathode:
 A --D1>|-- B --D2>|-- C --D3>|-- D --D4>|-- A

T1 secondary: A -- winding -- CT_RF -- winding -- C
T2 secondary: B -- winding -- CT_LO -- winding -- D

RF input -> T1 primary       LO input -> T2 primary
IF output: CT_RF to CT_LO    CT_LO -> ground

• Commutator Switching Action: Strong LO drive switches one pair of diodes into forward conduction and the other pair off. The pairs exchange roles every half-cycle, reversing the RF contribution at the IF output. For the ring connections above, the conducting pair meets at RF-winding end A on one half-cycle and end C on the next.
• Mutual Port Isolation: The matched diodes and balanced windings cancel direct RF and LO feedthrough at IF. The desired product survives because the LO reverses the RF connection each half-cycle. Common-mode means equal polarity, not equal and opposite. Isolation is finite in a practical mixer; good balance can provide tens of decibels of rejection.
• Even Harmonic Suppression: Ideal double balance suppresses IF products having an even RF or LO harmonic index. Real imbalance leaves residual products, including the $2\times2$ half-IF response discussed below. This does not mean cancelling every product of even total order: the wanted $1\times1$ product must remain.
• Bilateral Operation: A passive diode ring can be used for either upconversion or downconversion, within the specified frequency ranges of its RF, LO and IF ports.

Active Gilbert Cell Mixers

An active Gilbert cell mixer (named after Barrie Gilbert) utilizes a cross-coupled differential transistor topology (four switching transistors configured as a differential quad, driven by a lower differential transconductance pair).

• Operation: The input RF signal is converted into a linear differential current by the lower transistor pair, which is then commutated by the upper four cross-coupled transistors driven by the differential LO.
• Conversion Gain: Active Gilbert-cell designs can provide conversion gain, commonly around $10\text{ dB}$ to $15\text{ dB}$ in the examples considered here. This reduces the gain required from following IF stages; the receiver gain and noise budgets determine whether another amplifier is needed.
• Monolithic Integration: Gilbert cells require no bulky ferrite wound transformers; they are implemented entirely on silicon using bipolar or CMOS processes, making them the standard architecture inside integrated transceiver ICs and software-defined radio (SDR) front-ends.
• Dynamic Range Trade-Off: Active mixers exhibit significantly lower third-order intercept points ($IIP_3 \approx 0\text{ dBm}$ to $+10\text{ dBm}$) and higher noise figures than high-level passive diode ring mixers.

---

3. Mixer Performance Metrics and Linearity

Quantifying mixer performance requires evaluating several critical figures of merit that dictate receiver dynamic range and transmitter spectral purity.

Conversion Loss and Conversion Gain

• Conversion Loss ($CL$): In passive diode ring mixers, conversion loss is the ratio of RF input power to available IF output power, expressed in decibels:

$$CL\text{ (dB)} = 10 \log_{10}\left( \frac{P_{\text{RF(in)}}}{P_{\text{IF(out)}}} \right)$$

For a well-designed Schottky diode ring, typical conversion loss ranges from $6.0\text{ dB}$ to $8.5\text{ dB}$. This loss arises from diode forward junction resistance, transformer core losses, and the fundamental theoretical limit of converting sinusoidal energy into infinite Fourier switching sidebands ($2/\pi \approx -3.92\text{ dB}$ minimum theoretical loss for a square-wave commutator).

• Noise Figure ($NF$): In a passive diode ring mixer, the noise figure is practically equal to its conversion loss ($NF \approx CL$). Thus, an $8\text{ dB}$ conversion loss mixer has an $8\text{ dB}$ noise figure.

1 dB Compression Point ($P_{1\text{dB}}$)

At low RF input power levels, a mixer's IF output power tracks the RF input power linearly with a $1:1$ slope on a decibel plot ($1\text{ dB}$ increase in RF input yields a $1\text{ dB}$ increase in IF output).

As the RF input power approaches the LO switching threshold, the RF voltage begins to modulate the diode conduction intervals. The diodes spend less time in full saturation or cutoff, and the conversion gain begins to drop. The $1\text{ dB}$ Compression Point ($P_{1\text{dB}}$) is the input RF power level at which the IF output power deviates from the ideal linear response by exactly $1\text{ dB}$:

$$P_{1\text{dB(input)}} \approx P_{\text{LO}} - 6\text{ to }10\text{ dB}$$

Operating a mixer near or above its $P_{1\text{dB}}$ causes severe non-linear distortion, gain compression, and cross-modulation.

Third-Order Input Intercept Point ($IIP_3$)

When two strong, closely spaced signals at frequencies $f_1$ and $f_2$ enter a mixer alongside the LO, device third-order non-linearities ($a_3 v^3$) generate in-band third-order intermodulation distortion (IMD3) products:

$$f_{\text{IMD3}} = 2f_1 - f_2 \quad \text{and} \quad 2f_2 - f_1$$

Because third-order products grow at a $3:1$ rate ($3\text{ dB}$ increase in product power for every $1\text{ dB}$ increase in two-tone input power), their power level eventually overtakes the fundamental linear response ($1:1$ rate) on a mathematical plot. The Third-Order Input Intercept Point ($IIP_3$) is the theoretical, extrapolated input power level at which the third-order intermodulation product power equals the fundamental output power:

$$IIP_3\text{ (dBm)} = P_{\text{in}} + \frac{\Delta P}{2}$$

where $P_{\text{in}}$ is the input power per tone (in $\text{dBm}$), and $\Delta P$ is the difference (in $\text{dB}$) between the fundamental output tone and the third-order distortion product measured at the output.

Standard Mixer Drive Levels

The level number specifies the nominal LO drive in dBm. Higher-barrier or series-stacked diode junctions permit stronger RF signals but require more LO drive. The following are representative performance examples, not a definition of the internal construction:

---

4. Spurious Mixing Products and the IF Diplexer Network

In addition to the desired sum and difference frequencies, non-linear mixing generates a vast constellation of spurious responses governed by the general equation:

$$f_{\text{spurious}} = |m f_{\text{RF}} \pm n f_{\text{LO}}|$$

Here $m$ and $n$ are harmonic indices. Their sum describes the combined mixing order, but receiver distortion tests usually hold LO drive fixed and classify the RF response by its dependence on RF input level. Thus a $2\times2$ spur is second order in RF input, despite having an index sum of four. Low-index responses deserve particular attention when planning receiver frequencies.

The Half-IF ($2 \times 2$) Spurious Response

The half-IF response occurs when the second harmonic of an unwanted RF signal mixes with the second LO harmonic ($m=2,n=2$). It is a second-order RF spurious response at fixed LO drive:

$$f_{\text{IF}} = |2 f_{\text{interferer}} - 2 f_{\text{LO}}|$$

This condition occurs when a strong interfering off-channel signal enters the receiver at exactly half the intermediate frequency away from the local oscillator:

$$f_{\text{interferer}} = f_{\text{LO}} \pm \frac{1}{2} f_{\text{IF}}$$

One solution lies midway between the wanted signal and the LO; the other lies on the opposite side of the LO. The nearer interferer can be difficult for a broad RF preselector to reject before it reaches the mixer.

Internal Birdies

Birdies are unmodulated carrier signals heard in a receiver's audio output in the absence of an incoming antenna signal. They are generated internally when high-order harmonics of the receiver's local oscillator, synthesiser reference clock, or micro-controller clock cross-modulate within the mixer or leak directly into the intermediate frequency strip.

The IF Diplexer Network: Controlling Reflected Products

Diode-ring mixer performance is normally specified in a broadband $50\ \Omega$ system. Reactive loads can reflect unwanted products back into the mixer, where they mix again and change conversion loss or distortion. Good loading at the desired IF and significant unwanted-product frequencies helps reproduce the specified performance.

                          desired IF
Mixer IF o----+----[Low-pass arm]----[Matched IF buffer]
              |                              |
              |                         Roofing filter
              |
              +----[High-pass arm]----[50-ohm resistor]
                      higher-frequency products

A narrow crystal filter has a specified termination impedance, which is not necessarily $50\ \Omega$. Its stopband can be strongly reflective. Connecting it directly to the mixer can therefore disturb the intended loading. A matched IF buffer or resistive pad can reduce the effect of those reflections; its input match must cover the relevant frequencies.

The Diplexer Solution

A complementary low-pass/high-pass diplexer separates frequency ranges:

• Low-pass arm: Passes the desired difference-frequency IF into the matched following stage. Its load must provide the intended impedance throughout the frequencies passed by this arm.
• High-pass arm: Sends higher-frequency products, such as the sum frequency in the worked example, to a non-inductive $50\ \Omega$ termination.

The high-pass resistor does not absorb a reflection that remains within the low-pass arm's frequency range. A low-pass/high-pass pair feeding a narrow reflective filter directly therefore does not guarantee a broadband match. A bandpass diplexer requires a complementary rejection path rather than simply relabelling the low-pass arm. Check the complete network's impedance and loss over its designed range; no universal IIP3 improvement or insertion-loss figure follows from the block diagram alone.

---

5. Comprehensive Comparison of Mixer Topologies

---

6. Worked Engineering Examples

Worked Example 7.1.1: Mixer Output Frequencies and Half-IF Spurious Calculation

An amateur superheterodyne receiver is tuned to receive a $14.150\text{ MHz}$ signal on the 20-meter band. The receiver uses a fixed intermediate frequency of $f_{\text{IF}} = 9.000\text{ MHz}$ with high-side local oscillator injection.

1. Calculate the required Local Oscillator frequency ($f_{\text{LO}}$):

$$f_{\text{LO}} = f_{\text{sig}} + f_{\text{IF}} = 14.150\text{ MHz} + 9.000\text{ MHz} = 23.150\text{ MHz}$$

1. Calculate the resulting sum frequency produced at the mixer output:

$$f_{\text{sum}} = f_{\text{sig}} + f_{\text{LO}} = 14.150\text{ MHz} + 23.150\text{ MHz} = 37.300\text{ MHz}$$

(In the illustrated receiver, the diplexer can route this $37.300\text{ MHz}$ product to its terminated high-pass arm while passing the $9.000\text{ MHz}$ IF.)

1. Calculate the receiver's Image Frequency ($f_{\text{image}}$):

$$f_{\text{image}} = f_{\text{LO}} + f_{\text{IF}} = f_{\text{sig}} + 2 f_{\text{IF}} = 14.150\text{ MHz} + (2 \times 9.000\text{ MHz}) = 32.150\text{ MHz}$$

1. Find the off-channel interferer between the wanted signal and LO that produces a half-IF ($2\times2$) response. The equation has two roots, $23.150\pm4.500=18.650$ or $27.650\text{ MHz}$; select the nearer one:

$$|2 f_{\text{spur}} - 2 f_{\text{LO}}| = f_{\text{IF}} \implies |f_{\text{spur}}-f_{\text{LO}}| = \frac{f_{\text{IF}}}{2}=4.500\text{ MHz}$$ $$f_{\text{spur}} = f_{\text{LO}} - 4.500\text{ MHz} = 23.150\text{ MHz} - 4.500\text{ MHz} = 18.650\text{ MHz}$$

Verification: If an $18.650\text{ MHz}$ signal enters the mixer:

• Second harmonic of the interferer: $2 \times 18.650\text{ MHz} = 37.300\text{ MHz}$
• Second harmonic of the LO: $2 \times 23.150\text{ MHz} = 46.300\text{ MHz}$
• Mixing product: $46.300\text{ MHz} - 37.300\text{ MHz} = 9.000\text{ MHz} = f_{\text{IF}}$! The interfering signal translates directly into the IF passband.

Worked Example 7.1.2: Calculating $IIP_3$ from Two-Tone Laboratory Measurements

A laboratory test is conducted on a high-level (Level 17) double-balanced diode ring mixer. Two equal-amplitude RF tones at $f_1 = 14.000\text{ MHz}$ and $f_2 = 14.020\text{ MHz}$ are injected into the RF port at a power level of $P_{\text{in}} = -10.0\text{ dBm}$ per tone. A $+17\text{ dBm}$ LO signal is applied at $23.000\text{ MHz}$.

At the IF port, the fundamental desired difference tones ($8.980\text{ MHz}$ and $9.000\text{ MHz}$) are measured at $-16.5\text{ dBm}$. The third-order intermodulation products ($2f_1 - f_2$ and $2f_2 - f_1$ translated to IF) are measured at $-82.5\text{ dBm}$.

1. Calculate the conversion loss ($CL$) of the mixer:

$$CL = P_{\text{in}} - P_{\text{out(fund)}} = -10.0\text{ dBm} - (-16.5\text{ dBm}) = 6.5\text{ dB}$$

1. Calculate the intermodulation suppression ($\Delta P$):

$$\Delta P = P_{\text{out(fund)}} - P_{\text{out(IMD3)}} = -16.5\text{ dBm} - (-82.5\text{ dBm}) = 66.0\text{ dB}$$

1. Calculate the Third-Order Input Intercept Point ($IIP_3$):

$$IIP_3 = P_{\text{in}} + \frac{\Delta P}{2} = -10.0\text{ dBm} + \frac{66.0\text{ dB}}{2} = -10.0 + 33.0 = +23.0\text{ dBm}$$

The extrapolated $IIP_3$ is $+23.0\text{ dBm}$ (about $200\text{ mW}$). At the tested $-10.0\text{ dBm}$ per tone, the IMD3 products are $66.0\text{ dB}$ below the desired outputs. The intercept compares linearity; it is not an actual distortion-free input-power rating.