Mixer and Upconversion: Getting Your Signal to RF¶
Every spurious product your mixer generates becomes a filtering problem, a compliance risk, or both. Mixer selection is really spur management.
The upconversion stage is where your baseband signal meets the local oscillator and becomes an RF signal. It's also where unwanted frequency products are born — and where the trade-off between conversion efficiency, linearity, and spectral cleanliness defines the rest of your transmitter's filtering requirements.
What the mixer does
A mixer multiplies two signals — the baseband (or IF) input and the local oscillator (LO) — to produce sum and difference frequencies. For a transmitter upconverting a 100 MHz IF signal with a 2.4 GHz LO, the outputs are at 2.5 GHz (sum) and 2.3 GHz (difference). You want one of these — typically the sum for high-side injection — and need to suppress the other.
That's the textbook version. In practice, a real mixer produces far more than two output frequencies. Every mixer generates intermodulation products at frequencies mIF +/- nLO, where m and n are integers. A complete spur table for a double-balanced mixer can contain dozens of significant products, each with a power level that depends on the mixer's nonlinearity profile and the input power levels.
Mixer topologies
Passive double-balanced mixers. The workhorse of RF design. Two or four diodes (or FETs) in a balanced configuration. No DC power required. Conversion loss of 5-8 dB typically. Good linearity — IIP3 of +15 to +25 dBm is achievable. The main drawback is that they need substantial LO drive power, typically +7 to +13 dBm, which means the VCO or LO buffer needs to deliver real power. Available from DC to 40+ GHz.
Active mixers. Use transistors (typically Gilbert cell topology) to provide conversion gain instead of loss — typically +5 to +15 dB of gain. Require DC power (50-200 mW typically) but need much less LO drive, often 0 dBm or less. Lower IIP3 than passive mixers — typically +5 to +15 dBm. The gain relaxes downstream amplification requirements but the lower linearity means distortion becomes a concern at lower signal levels.
I/Q mixers (image-reject mixers). Two mixers driven by quadrature LO signals (0 and 90 degrees), with outputs combined through a hybrid coupler. Provides 20-40 dB of image rejection without external filtering. The rejection depends on amplitude and phase balance between the two paths — 25 dB is typical for integrated solutions, 35-40 dB is achievable with careful design or digital calibration. Critical for designs where the image frequency falls close to the desired output and can't be filtered effectively.
Sub-harmonic mixers. Use the second or third harmonic of the LO internally, allowing a lower-frequency LO to reach higher RF frequencies. A 2x sub-harmonic mixer targeting 28 GHz needs only a 14 GHz LO. Conversion loss is typically 2-4 dB worse than a fundamental mixer, but the simplification of the LO chain can be worth it.
Conversion loss vs. conversion gain
This is the first system-level trade-off in mixer selection. A passive mixer with 7 dB conversion loss means your signal drops from, say, 0 dBm at the IF input to -7 dBm at the RF output. Your driver amplifier and PA need to make up that loss plus deliver the final output power.
An active mixer with +10 dB conversion gain puts the RF signal at +10 dBm — significantly reducing the gain requirement of downstream stages. That might eliminate one amplifier stage entirely.
But the active mixer consumes DC power, has lower linearity, adds noise, and is generally less broadband than a passive alternative. For a wideband radar transmitter that needs to cover 2-18 GHz, a passive mixer is almost always the right choice. For a narrowband cellular transmitter at a fixed frequency, an active mixer that eliminates a gain stage may reduce overall power consumption and component count.
Spur analysis: the non-negotiable step
Spur analysis is where mixer selection moves from datasheet comparison to engineering judgment. Every combination of IF frequency, LO frequency, and mixer topology produces a unique spur spectrum. Some spurs fall far from the desired output and are easily filtered. Others fall in-band or close to the output and create real problems.
The spur table. Mixer datasheets provide spur tables showing the relative power level of each mIF +/- nLO product, typically referenced to the desired output. A well-designed double-balanced mixer suppresses even-order products (m or n even) by 25-40 dB. Odd-order products — especially 3x1 (third harmonic of IF mixed with LO fundamental) — are typically only 15-25 dB down.
Frequency planning around spurs. The IF frequency should be chosen so that the most problematic spurs don't fall in the output band or in protected spectrum. For a transmitter at 3.5 GHz, an IF of 350 MHz with a 3.15 GHz LO places the image at 2.8 GHz — 700 MHz from the desired output, easily filterable. But the 3x1 spur at 3.15 GHz + 3*350 MHz = 4.2 GHz might land in another allocated band.
Changing the IF by even 50 MHz can shift spurs enough to avoid a problematic collision. This is why frequency planning — choosing IF and LO frequencies — should happen before component selection, not after.
LO leakage. The LO signal itself appears at the mixer output, typically 20-40 dB below the desired signal in a double-balanced mixer. If the LO frequency is close to the output frequency (as in low-IF architectures), this leakage is difficult to filter and can cause regulatory issues.
Linearity and dynamic range
IIP3 (input third-order intercept point). The standard measure of mixer linearity. Higher IIP3 means the mixer can handle larger signals before intermodulation products become significant. For a transmitter, IIP3 matters when the input signal has a high peak-to-average power ratio (PAPR) — which is the case for OFDM waveforms used in LTE and 5G NR, where PAPR can reach 8-12 dB.
P1dB (1 dB compression point). The input power level at which the mixer's conversion gain or loss deviates by 1 dB from its small-signal value. Operating the mixer near P1dB generates significant distortion. General practice is to keep the input signal at least 10 dB below P1dB for linear operation.
The linearity budget. In a transmitter chain, each stage has a linearity spec. The mixer's linearity requirement is derived from the overall system ACLR or EVM requirement, working backwards through the gain stages. If the PA needs to see a signal with -40 dBc third-order products, and the driver adds no significant distortion, then the mixer needs to deliver that level of spectral purity at its output.
Port isolation
Mixer port isolation describes how much of each input signal leaks to the other ports. Three isolation specs matter:
LO-to-RF isolation. How much LO power appears at the RF output. Typically 25-40 dB for a double-balanced mixer. The LO leakage at the output must be filtered or must fall outside the transmit band.
LO-to-IF isolation. How much LO power feeds back to the IF input. Poor isolation here can cause the IF amplifier to saturate or the LO signal to re-enter the mixer through the IF path and create additional spurs.
RF-to-LO isolation. Prevents the modulated RF signal from pulling the LO frequency. Important when the LO and RF frequencies are close, as in direct-conversion transmitters.
Integrated I/Q mixers generally have better port isolation than discrete passive mixers because the balanced structure and on-chip routing can be tightly controlled.
Practical selection summary
For narrowband cellular or IoT transmitters below 6 GHz with moderate output power, an integrated I/Q active mixer is often the simplest path — low LO drive requirement, image rejection built in, and conversion gain that simplifies the chain.
For wideband, high-linearity applications (radar, electronic warfare, instrumentation) above 6 GHz, a passive double-balanced mixer with an external LO buffer amplifier gives the best combination of bandwidth, linearity, and spur performance.
For mmWave transmitters above 24 GHz, the choice narrows to whatever is available in MMIC form — and spur performance verification through simulation becomes critical because component-level spur data is often limited.
How do you approach frequency planning for your mixer? Do you start with the IF frequency and work outward, or do you start with available VCO/PLL options and back-calculate the IF?
This is Part 3 of an 8-part series on RF transmitter design. Previously: "VCO Selection." Next: "Filter Selection: Taming Spurious Emissions."