VCO Pulling
VCO pulling is what happens when a strong nearby signal couples energy directly into a voltage-controlled oscillator's own resonant tank — through substrate, bond wires, or package parasitics rather than through any mixer or filter — and drags the oscillator's instantaneous frequency around in sync with that aggressor, corrupting the local oscillator at its actual source rather than downstream of it.
Every impairment this series has covered so far assumes the local oscillator itself is a stable, well-defined frequency and asks what goes wrong afterward — image rejection, IQ imbalance, and reciprocal mixing all take the LO's frequency as a given and examine how it interacts with the rest of the signal chain. VCO pulling attacks that assumption directly, at its source. A voltage-controlled oscillator's frequency is set by its resonant tank, and that tank has no inherent ability to distinguish its own intended tuning energy from any other energy at a nearby frequency that manages to couple into it — through the chip substrate, through bond wires running close to a strong aggressor, or through package-level parasitic coupling. When that coupled aggressor is strong enough and close enough in frequency to the VCO's own tuning, it doesn't just add noise; it actively pulls the oscillator's instantaneous frequency toward itself, in direct proportion to the aggressor's strength and inverse proportion to how far away in frequency it sits.
The classic real-world aggressor is the chip's own power amplifier, which is exactly why this problem is often most severe in a transceiver that uses the same VCO to generate both its transmit and receive local oscillator signals. A PA driving significant RF power at a frequency related to the VCO's own output is an extremely strong, extremely close aggressor by construction — it sits on the same chip, often physically near the VCO, and operates at exactly the frequency most likely to couple efficiently back into the tank. The result is a local oscillator whose frequency wobbles in time with the PA's own output envelope, imposing unwanted FM or PM directly onto the LO signal that then gets used for every subsequent mixing operation in the chain, with no way for a downstream fix to separate the LO's correct frequency component from the pulled aggressor's influence after the fact.
Unlike reciprocal mixing, which degrades a desired signal's noise floor specifically in the presence of an interferer, VCO pulling corrupts the LO signal itself, which means it propagates to every mixer the LO drives rather than being confined to one signal path. A receiver with a pulled LO doesn't just lose sensitivity to one specific weak signal near one specific interferer — every mixing operation downstream of that LO, on both the I and Q paths, inherits the exact same frequency and phase wander, because they all share the single corrupted LO source. This is also why pulling is treated as a layout and isolation problem as much as a circuit-design one: the fix is rarely a cleverer VCO topology alone, but careful physical separation, shielding, and substrate isolation between the VCO and whatever aggressor couples into it, alongside differential tank structures specifically chosen for their improved common-mode rejection of coupled interference.
Pulling's severity depends heavily on the ratio between the VCO's own tank quality factor and the coupled aggressor's strength, which is why a higher-Q tank is pulled less by the same aggressor than a lower-Q one. A tank with a higher quality factor stores energy more efficiently at its own designed frequency and is correspondingly more resistant to having an external signal's energy redirect its oscillation — the tank's own resonance dominates more strongly over any coupled interference. This is one more reason transceiver designers push for the highest achievable tank Q beyond the phase-noise benefits already covered elsewhere in this series: a cleaner tank is also a more pulling-resistant one, for the same underlying physical reason.
VCO pulling is the fourth and final independent way a local oscillator can degrade a receiver's or transmitter's performance covered in this series, and it is deliberately the one that happens earliest in the signal chain. Image rejection, IQ imbalance, and reciprocal mixing all assume a correctly-generated LO and ask what happens next; pulling asks whether the LO was ever correctly generated in the first place, under the specific stress of a strong nearby aggressor like the chip's own PA. A transceiver design that solves the other three problems perfectly can still fail in the field if its VCO pulls badly under real transmit conditions — which is exactly why pulling sensitivity is measured and specified separately, often under worst-case PA output power, rather than assumed away as a secondary concern.
| Quantity | Reciprocal mixing | VCO pulling |
|---|---|---|
| What's corrupted | The mixed-down IF signal | The LO signal itself, at its source |
| Root cause | LO phase-noise skirt + interferer | Coupled aggressor energy into the tank |
| Scope of damage | One signal path, near one interferer | Every mixer the LO drives |
| Primary fix | Cleaner LO / frequency synthesizer | Physical isolation, higher tank Q |
VCO pulling closes out this series' set of independent local-oscillator and receiver impairments: image rejection and IQ imbalance concern how a correctly-generated LO's mixing products are handled, reciprocal mixing concerns how a correctly-generated but imperfectly pure LO interacts with a strong interferer, and pulling concerns whether the LO was generated correctly in the first place under real operating stress. A transceiver has to get all four right, because each one is capable of being the actual limiting factor on real-world performance regardless of how well the other three were solved.