RF, IF and LO Explained

Every heterodyne radio has three frequencies, and most arguments about mixers, filters and oscillators are arguments about how those three are placed. This page names them and shows why the image is a planning problem. The rest of the signal chain section is what you do with the plan.

The three names

RF is the frequency at the antenna — the thing the system is trying to receive or transmit.

LO is the local oscillator: a tone the designer puts next to the RF so a mixer can translate it. In a receiver the LO is usually the part that tunes. Moving the LO is moving the radio.

IF is what comes out of that translation — low enough, or convenient enough, that the rest of the chain can filter, amplify or digitise it.

The identity for a downconverting mixer is:

|RF − LO| = IF

That is also the identity for downconversion versus upconversion. The same ports, the opposite direction.

None of the three names is a component family. An LO can be a YIG oscillator, a VCO or a DRO. An IF can be 70 MHz or 1.2 GHz. RF is just “wherever the antenna is”.

Why there is always an image

A mixer produces an IF from any input that differs from the LO by the IF. There are always two such inputs: one above the LO and one below it. One is the wanted RF. The other is the image.

The spacing is forced:

|RF − image| = 2 × IF

On high-side injection (LO above RF) the image sits above the LO. On low-side injection it sits below. Both put the image 2 × IF from the wanted signal. After mixing, the two products are the same frequency. An IF filter cannot separate them. The only chance is a filter at RF — a preselector or an image-reject mixer architecture — before the mixer.

That is why “we will filter it in the IF” is not a plan.

A worked placement

Take a 10 GHz RF, a 1 GHz IF and high-side injection.

• LO = RF + IF = 11 GHz
• Image = LO + IF = 12 GHz
• Image − RF = 2 GHz = 2 × IF

The preselector must pass 10 GHz and reject 12 GHz. A YIG filter with a 30–50 MHz 3 dB width does that easily; a 5% cavity at 10 GHz is 500 MHz wide and still has to be checked. Move the IF down to 100 MHz and the image is only 200 MHz away. The same cavity may no longer be enough, and the YIG filter’s skirt — set by stage count — becomes the specification that matters.

Low-side injection on the same RF and IF puts the LO at 9 GHz and the image at 8 GHz. The filter problem is identical in width and on the other side of the RF. Choose the side that puts the image somewhere easier: off the antenna band, away from a known jammer, or onto a slope you already have.

The IF is a trade, not a quality

IF choice	What it helps	What it costs
Higher IF	Image farther from RF; easier preselect	Harder IF filter, ADC or second conversion
Lower IF	Easier digitiser and channel filter	Image closer; sharper RF filter required
High-side LO	Image above the RF	LO must reach RF + IF; possible harmonic issues
Low-side LO	Image below the RF; sometimes a cheaper LO band	Possible in-band image if the RF span is wide
Zero IF / direct conversion	No analog IF; no classical image at 2 × IF	I/Q imbalance, DC offset, a different image (the opposite sideband)

Qualitative planning table. A dual-conversion radio is this table applied twice, with a first IF chosen to make the first image easy and a second IF chosen to make the channel filter easy.

Wideband microwave receivers often pick a first IF high enough that a tracking YIG preselector or a simple fixed filter can kill the image, then convert again to a second IF the ADC likes. That is two LO plans, two images, and two places to spend isolation. It is also why a single “the LO” conversation is usually incomplete.

The LO is a phase-noise port

The mixer multiplies the RF by whatever the LO actually is — including the LO’s phase noise. A strong unwanted signal offset by Δf from the wanted RF mixes with the LO noise at that same Δf and lands in the IF as if it were on-channel. That is reciprocal mixing. No IF filter removes it.

Two hardware responses, and they do different jobs:

• A preselector attenuates the interferer before the mixer, which attenuates the reciprocal-mixing product with it, at the cost of insertion loss.
• A quieter LO reduces the noise density at Δf. A tracking YIG preselector cannot fix a noisy oscillator.

Which offset Δf matters is a system question: adjacent-channel, image, or a known blocker. Quote phase noise there, not at the datasheet’s favourite marker.

Common planning mistakes

• Treating IF filtering as image rejection. It is not, and the identity |RF − image| = 2 × IF is why.
• Picking the LO side after the preselector is bought. High-side and low-side put the reject band on opposite sides.
• Ignoring the second conversion. Dual-conversion radios have two images. The pretty first-IF plan can dump a second image onto something you care about.
• Specifying the LO as a frequency only. The LO is also a noise source at every offset an interferer can occupy.
• Forgetting that the preselector and a YTO both settle. A tracking pair has one time budget, not two.

Where this plan turns into a component choice: YIG oscillator vs VCO vs DRO for the LO, YIG filter vs cavity for the preselector, and radar RF components for the same chain under a waveform constraint.

A plan needs a receiver environment

Frequency arithmetic tells you where the image lands; the receiver environment tells you whether it matters. List the wanted RF span, the instantaneous channel width, the largest expected in-band signal and the largest plausible out-of-band signal. Add known transmitters, duplex partners and local emissions rather than using only a generic “blocker” value. For each one, translate it through the proposed LO and mark whether it lands in the desired IF, on an image response, near a harmonic or outside the IF filter altogether.

This exercise usually reveals that IF selection is a filtering decision. A high first IF moves the image farther from RF and can make a tracking preselector practical. A low second IF may make a sharp channel filter and ADC easier. Dual conversion is not automatically better; it adds another LO, another image and another set of spurious products. It earns its complexity only when the first IF and second IF each solve a different rejection problem.

The LO side is a second design lever. With high-side injection, the image lies on one side of the wanted RF; with low-side injection it lies on the other. If one side contains a quiet spectrum and the other contains a strong service, the choice can eliminate a demanding filter before it is purchased. Use the same table when comparing an upconverter and a downconverter: the direction changes where the unwanted products leave the module, but it does not remove the need to track them.

Close the plan with an acceptance measurement. At several points across the RF band, inject the desired signal and then inject signals at the calculated image and prominent spur locations. Record conversion gain, image rejection, LO leakage and any false IF responses with the real IF filter installed. If a high-level blocker is part of the operating case, repeat it while observing noise-floor rise to check reciprocal mixing. That is the practical connection between a line on a frequency spreadsheet and the phase-noise plot on an LO datasheet.

Keep the plan reviewable

The most useful frequency plan is small enough for another engineer to challenge. Give every conversion its own row: wanted RF range, LO range, IF range, injection side, image range, expected desired level and largest unwanted level. Add rows for the second conversion, reference-related spurs and any harmonic products that have a credible path through the hardware. A plan that cannot fit on one review sheet is often hiding an assumption in a filter or module data sheet.

For every row, state which component is responsible for the rejection. The antenna or preselector may handle an image; a diplexer may separate mixer products; an IF filter may protect the ADC; a quiet LO may be needed where filtering cannot help. This makes gaps obvious. If the same unwanted response is assumed to be rejected by two different blocks, it probably has no real owner. If no block owns it, changing the IF or LO side early is cheaper than repairing the problem after the layout is frozen.

Use the same document during debug. When an unexpected IF tone appears, calculate backwards from its frequency through each LO and compare it with the planned images, harmonics and reference spurs. The exercise avoids random filter changes and produces evidence for whether the cause is a frequency-plan error, inadequate isolation or a non-linear stage. A disciplined plan is therefore not paperwork; it is the fastest route from a false response on the bench to the part of the chain that created it.