5.3 Amplifier Stability, Neutralisation, and Intermodulation Distortion

5.3 Amplifier Stability, Neutralisation, and Intermodulation Distortion

An RF power amplifier is intended to function purely as a controlled transconductance amplifier, linearly converting an input voltage into an amplified output current. However, high-gain, high-power RF circuits are inherently prone to instability, parasitic self-oscillation, and intermodulation distortion (IMD). Achieving clean, reliable, and compliant operation requires an in-depth understanding of feedback dynamics, neutralisation techniques, and non-linear distortion mechanisms.

---

1. Amplifier Instability, Feedback Mechanisms, and Self-Oscillation

Any amplifier will transform into an oscillator if positive feedback of sufficient amplitude and phase is transferred from its high-power output port back to its sensitive input port. This condition is defined by the Barkhausen Stability Criterion:

$$|A \cdot \beta| \ge 1 \quad \text{and} \quad \arg(A \cdot \beta) = 0^\circ \pmod{360^\circ}$$

where $A$ is the open-loop voltage gain of the active device and $\beta$ is the feedback transfer function between output and input.

Internal Device Feedback Capacitance

The primary internal feedback path within active amplifying devices is the inter-electrode or junction capacitance connecting output to input:

• Thermionic Triodes: Anode-to-grid capacitance ($C_{ag}$ or $C_{gp}$, typically $5\text{ pF}$ to $15\text{ pF}$ in power triodes).
• Bipolar Transistors: Collector-to-base capacitance ($C_{cb}$ or $C_\mu$, typically $10\text{ pF}$ to $100\text{ pF}$).
• Field-Effect Transistors: Gate-to-drain reverse transfer capacitance ($C_{gd}$ or $C_{\text{rss}}$, typically $2\text{ pF}$ to $20\text{ pF}$ in LDMOS).

The Miller Effect and Tuned-Plate Tuned-Grid (TPTG) Oscillation

In an inverting amplifier stage with voltage gain $A_v = -|A_v|$, the feedback capacitance ($C_f$) appears at the input terminals multiplied by the gain—a phenomenon known as the Miller Effect:

$$C_{\text{in(Miller)}} = C_f \cdot (1 - A_v) = C_f \cdot (1 + |A_v|)$$

At radio frequencies, the plate or collector load is a tuned parallel resonant circuit. If the output tank is tuned slightly higher in frequency than the operating frequency, the load impedance presented to the plate/collector becomes inductive ($Z_L = R + jX_L$).

With suitable phase relationships, feedback through $C_f$ can make the tuned input present negative conductance: the active stage supplies energy to the input resonator instead of damping it. Its value depends on the complete device and circuit model.

Represent the input resonator loss by parallel conductance $G_{\text{loss}}>0$ and the feedback contribution by $G_{\text{fb}}<0$. At a frequency where susceptances cancel, oscillation can start when:

$$G_{\text{loss}}+G_{\text{fb}}<0$$

Equivalently, $|G_{\text{fb}}|>G_{\text{loss}}$. For parallel resistance models this means $|R_{\text{negative}}|<R_{\text{loss}}$; the comparison reverses for a series-resistance model. In valve circuits, coupling between resonant plate and grid circuits can produce tuned-plate tuned-grid (TPTG) oscillation.

External Stray Coupling Mechanisms

In addition to internal inter-electrode capacitance, self-oscillation can be triggered by external electrodynamic coupling:

• Common Ground Impedance: Circulating RF currents traversing chassis metal or ground traces shared by both input and output circuits create common-mode voltage drops that feed back into the input.
• Stray Magnetic Flux Linkage: Unshielded toroidal or air-core solenoidal inductors in the input and output networks can couple inductively if mounted in the same physical orientation.
• Power Supply Rail Coupling: High RF currents leaking through inadequate RF chokes or defective bypass capacitors on the DC high-voltage rail can modulate the bias circuitry.

---

2. VHF Parasitic Oscillations and Suppressor Design

While an amplifier is designed to operate on HF bands ($1.8\text{ MHz}$ to $30\text{ MHz}$), modern power valves and transistors possess significant gain up into the VHF spectrum ($100\text{ MHz}$ to $300\text{ MHz}$). At these VHF frequencies, physical component lead lengths—connecting the valve socket to capacitors, or transistor leads to the PCB—exhibit small parasitic inductances ($L_{\text{lead}} \approx 10\text{ nH}$ to $50\text{ nH}$).

Mechanism of Parasitic Resonances

These parasitic lead inductances combine with internal device capacitances and stray capacitances to form unintended VHF resonant circuits at frequencies completely unrelated to the operating band (typically between $70\text{ MHz}$ and $180\text{ MHz}$). Because the tuned HF input and output networks act merely as large capacitances or open circuits at VHF, the amplifier easily satisfies the Barkhausen criterion and oscillates violently at VHF.

Destructive Consequences of Parasitic Oscillation

Parasitic oscillation is destructive and often catastrophic:

• Generates massive high-voltage RF peaks that arc across air-variable tuning capacitors, destroying their dielectric plates.
• Punches through delicate gate oxide layers in MOSFETs or melts emitter wire bonds in BJTs.
• Causes sudden, erratic spikes in quiescent plate/collector current with zero RF drive applied, rapidly overheating the device.
• Radiates wideband hash and spurious emissions across VHF public safety, aircraft navigation ($108 - 137\text{ MHz}$), and FM broadcast bands.

Parasitic Suppressors

To eliminate VHF parasitics, a parasitic suppressor is inserted directly at the plate (anode) or collector terminal of the active device.

          +---[47 ohm non-inductive R]---+
          |                              |
Plate ----+                              +----> HF tank
          |                              |
          +--------[small L]-------------+

A classic parasitic suppressor consists of a low-value, non-inductive power resistor ($22\ \Omega$ to $100\ \Omega$, typically $47\ \Omega$ carbon composition or metal film) connected in parallel with a small inductor ($2$ to $5$ turns of heavy copper or Nichrome resistance wire, $L \approx 50\text{ nH}$ to $100\text{ nH}$) wound directly across the resistor body:

• Operation at HF ($1.8 - 30\text{ MHz}$): The inductive reactance of the small coil is extremely low:

$$X_L = 2\pi f L = 2\pi \times (14\text{ MHz}) \times (80\text{ nH}) \approx 7.0\ \Omega$$

Because $7.0\ \Omega\ll47\ \Omega$, most HF current follows the coil. The resistor still dissipates power. For ideal $L$ in parallel with $R$, $\Re\{Z\}=RX_L^2/(R^2+X_L^2)$, about $1.02\ \Omega$ here; resistor heating is $I_{\text{total,RMS}}^2\Re\{Z\}$. Its power rating must cover this loss.

• Operation at VHF ($100 - 200\text{ MHz}$): At VHF parasitic frequencies, the inductor's reactance rises dramatically:

$$X_L = 2\pi \times (150\text{ MHz}) \times (80\text{ nH}) \approx 75.4\ \Omega$$

At $150\text{ MHz}$ the parallel combination has an effective series resistance of about $33.9\ \Omega$, much greater than at 14 MHz. Current divides between both branches, and the resistor adds damping to the parasitic loop. Select values and ratings for the actual layout, then verify stability; this network does not guarantee a particular loop Q or suppression by itself.

In solid-state circuitry, VHF parasitic suppression is achieved by placing ferrite beads directly over the gate or base leads, inserting low-value non-inductive series damping resistors ($1\ \Omega$ to $10\ \Omega$), and incorporating input swamping networks.

---

3. Neutralisation Theory and Practical Circuits

In amplifiers where device internal feedback capacitance cannot be ignored (such as thermionic triodes in grounded-cathode configurations, or single-ended tetrodes at VHF), the amplifier must be neutralised.

The Neutralisation Principle

Neutralisation introduces an RF feedback path whose current is equal in magnitude and opposite in phase to the unwanted feedback current at the input:

$$I_{\text{neut}} + I_{\text{feedback}} = 0$$

The antiphase current cancels the targeted capacitive feedback. Other coupling paths can remain, so a neutralisation null is not complete proof of stability.

Bridge Neutralisation Topologies

1. Plate/Grid Neutralised Bridge: A center-tapped tuned tank provides balanced, equal-and-opposite RF voltages ($+V_{\text{out}}$ and $-V_{\text{out}}$) relative to ground. A small variable capacitor ($C_n$, the neutralising capacitor) is connected from the $-V_{\text{out}}$ node back to the grid. When $C_n = C_{ag}$, bridge balance is achieved.
2. Capacitive Bridge Divider: In modern circuits, tapping an inductor is avoided. Instead, a capacitive voltage divider in the plate tank creates an antiphase voltage ratio that cancels feedback.
3. Cross-Neutralisation (Hazeltine Circuit) in Push-Pull Stages: In a push-pull amplifier, balanced $180^\circ$ antiphase voltages already exist naturally at the two plates/collectors. Neutralisation is accomplished simply by connecting two small cross-coupled capacitors:

• $C_{n1}$ connected from Plate 1 to Grid 2.
• $C_{n2}$ connected from Plate 2 to Grid 1. Because the voltages at Plate 1 and Plate 2 are naturally symmetrical and out of phase, cross-neutralisation provides near-perfect wideband cancellation.

Neutralisation Alignment and Safety

Neutralisation is service work for a qualified person following the equipment manufacturer's procedure. A passive cold feed-through test examines capacitive coupling without operating the valve:

1. Isolate stored and supplied energy: Disconnect all equipment power, discharge stored energy with rated equipment and verify the required safe condition before access. Do not rely on an HV switch or bleeder alone.
2. Configure the cold circuit: With power isolated, break the anode and screen DC current paths as required by the circuit while retaining the RF networks. Leave heaters and operating supplies off for this passive test. Merely reducing their voltage to zero can leave a rectifying current path that invalidates the null.
3. Use a low-level test source: Couple a signal generator to the input at the specified test frequency and connect a suitably protected sensitive detector to the output/load. This tests passive feed-through, not normal exciter drive into a hot valve.
4. Find minimum coupling: Tune the RF networks and adjust the neutralising component for minimum output indication while checking the specified tuning range. A perfect zero is not required as proof of alignment.
5. Restore only under the service procedure: Remove the test source, isolate again before changing connections and restore covers/interlocks. Powered checks, if prescribed, use the manufacturer's limited-power method and verify operating stability separately.

Never apply the generic rule “remove anode voltage, keep the valve hot and apply drive” to every amplifier. Some zero-bias triodes can suffer excessive grid dissipation in that condition. A coupling minimum alone does not establish electrical safety or stability under power.

---

4. Intermodulation Distortion (IMD), Linearity, and Spectral Purity

When an RF amplifier operates in linear modes (such as Single Sideband voice or multi-carrier digital transmissions), any departure from a strictly linear transfer characteristic produces distortion products.

Mathematics of the Non-Linear Transfer Function

A practical amplifier's transfer characteristic can be expressed as a Taylor series expansion relating output voltage ($v_{\text{out}}$) to input voltage ($v_{\text{in}}$):

$$v_{\text{out}} = a_1 v_{\text{in}} + a_2 v_{\text{in}}^2 + a_3 v_{\text{in}}^3 + a_4 v_{\text{in}}^4 + a_5 v_{\text{in}}^5 + \dots$$

• $a_1 v_{\text{in}}$ (Linear Gain): The desired amplified replica of the input signal.
• $a_2 v_{\text{in}}^2, a_4 v_{\text{in}}^4$ (Even-Order Non-Linearity): Generates harmonic frequencies ($2f_0, 4f_0$) and DC rectification terms. Because these even harmonics fall at twice or four times the operating frequency, they are located far away from the fundamental passband and are easily stripped away by the amplifier's tuned tank or low-pass filter.
• $a_3 v_{\text{in}}^3, a_5 v_{\text{in}}^5$ (Odd-Order Non-Linearity): Generates odd harmonics ($3f_0, 5f_0$) AND in-band intermodulation mixing products. Because odd-order mixing products fall directly adjacent to the fundamental signals, they cannot be filtered out by any external tuned circuit or low-pass filter!

The Two-Tone Test and Third-Order IMD

The standard industry and amateur method for evaluating amplifier linearity is the Two-Tone Test. Two equal-amplitude, non-harmonically related sinusoidal audio or RF tones ($f_1$ and $f_2$) are injected simultaneously into the transmitter input (for example, $f_1 = 14.200\text{ MHz}$ and $f_2 = 14.202\text{ MHz}$, with tone separation $\Delta f = 2\text{ kHz}$).

When these two tones pass through the cubic non-linear term ($a_3 v_{\text{in}}^3$), trigonometric expansion generates new spectral frequencies:

$$v_{\text{in}}^3 = \left( A \cos(2\pi f_1 t) + A \cos(2\pi f_2 t) \right)^3$$

Expanding this expression yields terms at the original frequencies ($f_1, f_2$), third harmonics ($3f_1, 3f_2$), and crucially, the Third-Order Intermodulation Distortion (IMD3) products:

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

Calculating the frequencies for our $14.200\text{ MHz}$ and $14.202\text{ MHz}$ test tones:

$$2f_1 - f_2 = 2(14.200\text{ MHz}) - 14.202\text{ MHz} = 28.400 - 14.202 = 14.198\text{ MHz}$$ $$2f_2 - f_1 = 2(14.202\text{ MHz}) - 14.200\text{ MHz} = 28.404 - 14.200 = 14.204\text{ MHz}$$

Notice where these products land: $14.198\text{ MHz}$ is located exactly $2\text{ kHz}$ below $f_1$, and $14.204\text{ MHz}$ is located exactly $2\text{ kHz}$ above $f_2$! In Single Sideband voice communications, these spurious mixing products manifest as splatter—unintelligible buckshot and interference spreading dozens of kilohertz into adjacent channels.

Higher-order terms generate Fifth-Order IMD products ($3f_1 - 2f_2$ and $3f_2 - 2f_1$), which fall further out at $14.196\text{ MHz}$ and $14.206\text{ MHz}$.

          Signal Spectrum in Two-Tone Test

                f1       f2
                ||       ||
                ||       ||
        IMD3    ||       ||    IMD3
         |      ||       ||     |
  IMD5   |      ||       ||     |   IMD5
   |     |      ||       ||     |    |
---|-----|------||-------||-----|----|----
 14.196 14.198 14.200  14.202 14.204 14.206  (MHz)

Third-Order Intercept Point ($IP_3$)

In the weakly nonlinear region, fundamental output power rises $1\text{ dB}$ for each $1\text{ dB}$ increase of input power per tone. Third-order product voltage amplitude scales as input voltage cubed, so its power scales as input power cubed: it rises $3\text{ dB}$ per $1\text{ dB}$ input increase. The IMD products therefore rise faster than the wanted signals.

The theoretical point at which the extrapolated fundamental power line intersects the extrapolated third-order IMD power line is defined as the Third-Order Intercept Point ($IP_3$):

$$OIP_3 = P_{\text{tone}} + \frac{\Delta \text{IMD}_3}{2}$$

where $P_{\text{tone}}$ is the power of one of the test tones (in $\text{dBm}$) and $\Delta \text{IMD}_3$ is the suppression in decibels of the third-order product below the test tone. A higher $IP_3$ indicates superior dynamic range and linearity.

Decibel Standards: $\text{dB PEP}$ vs. $\text{dBc}$

A widespread source of confusion and commercial specification manipulation involves how IMD ratios are referenced:

1. Decibels Relative to Carrier / Tone ($\text{dBc}$): The amplitude of the IMD product is measured relative to the amplitude of one of the two test tones.
2. Decibels Relative to Peak Envelope Power ($\text{dB PEP}$): The amplitude of the IMD product is measured relative to the composite Peak Envelope Power (PEP) of the two combined tones.

The Mathematical 6 dB Difference

Consider two equal-amplitude sinusoidal tones, each having a peak voltage of $V_m$ across a load resistance $R$. The average power of each individual single tone is:

$$P_{\text{single-tone}} = \frac{V_m^2}{2R}$$

When the two tones align in phase at the peak of their modulation envelope, their instantaneous voltages add constructively: $V_{\text{PEP}} = V_m + V_m = 2V_m$. The resulting Peak Envelope Power is:

$$P_{\text{PEP}} = \frac{V_{\text{PEP}}^2}{2R} = \frac{(2V_m)^2}{2R} = \frac{4 V_m^2}{2R} = 4 \cdot P_{\text{single-tone}}$$

Expressing this fourfold power ratio in decibels:

$$10 \log_{10}(4) = +6.02\text{ dB} \approx +6\text{ dB}$$

Peak Envelope Power is exactly $6\text{ dB}$ higher than the power of either individual tone. Therefore, an IMD product measured relative to Peak Envelope Power ($\text{dB PEP}$) will always appear $6\text{ dB}$ better (more suppressed) than when measured relative to a single test tone ($\text{dBc}$):

$$\text{IMD (in dB PEP)} = \text{IMD (in dBc)} - 6\text{ dB}$$

For example, an amplifier whose third-order distortion products are $30\text{ dB}$ below one tone ($-30\text{ dBc}$) can be marketed as having an IMD rating of $-36\text{ dB PEP}$. High-quality amateur linear amplifiers strive for at least $-30\text{ dBc}$ ($-36\text{ dB PEP}$) to maintain spectral cleanliness.

Over-Driving, Flat-Topping, and Automatic Level Control (ALC)

The primary operational cause of severe IMD splatter in amateur stations is amplifier over-driving, where excessive RF drive from the transceiver forces the amplifier's active devices into saturation. When a valve or transistor reaches saturation, its output voltage can swing no further, cleanly chopping off the peaks of the RF waveform—a condition known as flat-topping.

Flat-topping transforms the smooth sinusoidal peaks into truncated trapezoids, causing a massive explosion of high-order odd IMD products (5th, 7th, 9th, and 11th orders) that splatter across $50\text{ kHz}$ of spectrum.

To prevent flat-topping, transmitters incorporate Automatic Level Control (ALC) circuitry:

• An ALC detector monitors an output level or device current against a set limit.
• Its control signal reduces gain or exciter drive; external ALC polarity and range must match both units. Grid current is normal in some valve operating classes and is not a universal clipping threshold.
• ALC has finite response time. It limits drive but cannot guarantee freedom from transient overdrive or repair an already clipped envelope.

---

5. Diagnostic Guide: RF Amplifier Instability and Distortion

---

6. Worked Engineering Examples

Worked Example 5.3.1: Two-Tone IMD Calculations and Decibel Conversion

An amateur linear amplifier is tested using a two-tone generator at frequencies $f_1 = 7.100\text{ MHz}$ and $f_2 = 7.103\text{ MHz}$ (tone separation $\Delta f = 3\text{ kHz}$). Each tone produces an individual average output power of $P_{\text{tone}} = 250\text{ W}$ into a $50.0\ \Omega$ dummy load. An RF spectrum analyser measures the third-order IMD products at an absolute power of $P_{\text{IMD3}} = 0.25\text{ W}$.

1. Calculate the exact frequencies of the third-order and fifth-order IMD products:

$$\text{IMD3 (lower)} = 2f_1 - f_2 = 2(7.100) - 7.103 = 14.200 - 7.103 = 7.097\text{ MHz}$$ $$\text{IMD3 (upper)} = 2f_2 - f_1 = 2(7.103) - 7.100 = 14.206 - 7.100 = 7.106\text{ MHz}$$ $$\text{IMD5 (lower)} = 3f_1 - 2f_2 = 3(7.100) - 2(7.103) = 21.300 - 14.206 = 7.094\text{ MHz}$$ $$\text{IMD5 (upper)} = 3f_2 - 2f_1 = 3(7.103) - 2(7.100) = 21.309 - 14.200 = 7.109\text{ MHz}$$

1. Calculate the Peak Envelope Power (PEP) of the two-tone transmission:

$$P_{\text{PEP}} = 4 \cdot P_{\text{tone}} = 4 \times 250\text{ W} = 1000\text{ W PEP}$$

1. Calculate the third-order intermodulation distortion level in $\text{dBc}$ (relative to either tone):

$$\text{IMD}_3(\text{dBc}) = 10 \log_{10} \left( \frac{P_{\text{IMD3}}}{P_{\text{tone}}} \right) = 10 \log_{10} \left( \frac{0.25\text{ W}}{250\text{ W}} \right) = 10 \log_{10}(0.001) = -30.0\text{ dBc}$$

1. Calculate the third-order intermodulation distortion level in $\text{dB PEP}$:

$$\text{IMD}_3(\text{dB PEP}) = 10 \log_{10} \left( \frac{P_{\text{IMD3}}}{P_{\text{PEP}}} \right) = 10 \log_{10} \left( \frac{0.25\text{ W}}{1000\text{ W}} \right) = 10 \log_{10}(0.00025) = -36.02\text{ dB PEP}$$

This confirms the direct $6\text{ dB}$ difference: $-30.0\text{ dBc} - 6.02\text{ dB} = -36.02\text{ dB PEP}$.

Worked Example 5.3.2: Output Third-Order Intercept Point Calculation

Using the data from Worked Example 5.3.1, calculate the Output Third-Order Intercept Point ($OIP_3$) of the amplifier in $\text{dBm}$ and in Watts.

1. Convert single-tone power ($250\text{ W}$) into $\text{dBm}$:

$$P_{\text{tone(dBm)}} = 10 \log_{10} \left( \frac{250\text{ W}}{0.001\text{ W}} \right) = 10 \log_{10}(250000) = +53.98\text{ dBm}$$

1. Calculate $OIP_3$ using tone suppression $\Delta \text{IMD}_3 = 30.0\text{ dB}$:

$$OIP_3(\text{dBm}) = P_{\text{tone(dBm)}} + \frac{\Delta \text{IMD}_3}{2} = +53.98\text{ dBm} + \frac{30.0\text{ dB}}{2} = 53.98 + 15.0 = +68.98\text{ dBm}$$

1. Convert $OIP_3$ back to power in Watts:

$$P_{OIP3(\text{W})} = 10^{\frac{68.98 - 30}{10}} = 10^{3.898} \approx 7907\text{ W} \approx 7.91\text{ kW}$$

An $OIP_3$ of $+68.98\text{ dBm}$ ($7.91\text{ kW}$) represents outstanding engineering linearity for a $1\text{ kW}$ PEP amateur linear power amplifier.