Upconverter vs Downconverter

An upconverter and a downconverter are the same mixing identity read in opposite directions. That is the whole distinction, and it is enough to decide which assembly you are buying. What people then argue about — high-side or low-side, image rejection, conversion loss — is the RF / IF / LO plan, not a property of “up” versus “down”. This page sits in the signal chain section because the converter is a plan in a box.

The conclusion

Buy a downconverter when the signal arrives at RF and the next stage cannot process that RF. The output is an IF. The mixer identity is |RF − LO| = IF. The image is a second RF, 2 × IF away, and it must be rejected before the mixer.

Buy an upconverter when the signal already exists at IF or baseband and must leave at RF. The output is an RF. The identity is IF + LO = RF (or |LO − IF| on the other sideband). The unused sideband is the twin of the image, and it must be rejected before the antenna or the next mixer.

Do not buy one and wire it backwards. A mixer die is often bilateral. A converter module is a set of filters, matches, LO routing and spurious traps built for one direction. Reversing the ports keeps the die and discards the plan.

High-side versus low-side is a packaging of the same 2 × IF geometry. It is not a third kind of converter.

Side by side

Criterion	Downconverter	Upconverter
Signal in	RF from the antenna or previous RF stage	IF or baseband from the modulator or previous IF
Signal out	IF	RF
Mixing identity	|RF − LO| = IF	IF + LO = RF, or |LO − IF| = RF
Unwanted product	Image at RF, 2 × IF from the wanted RF	Unused sideband at RF, 2 × IF from the wanted RF
Where you filter it	RF preselector, before the mixer	RF filter, after the mixer and before the antenna
Typical place in the chain	Receiver front end; analyser converter	Transmit chain; block upconverter; test source
LO’s job	Select which RF becomes this IF	Place the IF at the wanted RF
What “gain” usually means	Conversion loss, sometimes followed by IF gain	Conversion loss, sometimes followed by RF gain

Conversion loss, LO-to-RF isolation, LO-to-IF isolation, input P1dB and IIP3 are mixer numbers. They show up on both assemblies. A module that hides the mixer behind an embedded LO adds that LO’s phase noise and harmonics to the specification; you are then buying an assembly, the same way a YTO-plus-driver is an assembly.

The unused product is the same shape

On a downconverter the unused input is called the image. On an upconverter the unused output is called the other sideband. Both sit 2 × IF from the wanted RF, on the opposite side of the LO.

Worked downconversion, 10 GHz RF, 1 GHz IF, high-side LO:

• LO = 11 GHz
• Wanted IF = 1 GHz
• Image at 12 GHz also yields 1 GHz

Worked upconversion, 1 GHz IF, 11 GHz LO:

• Wanted RF = 12 GHz (sum) or 10 GHz (difference)
• The product you did not choose is the unused sideband, 2 GHz away

The hardware response is a filter on the RF side. In receive, that is a preselector — YIG if the RF walks, cavity if it does not. In transmit, that is a bandpass on the wanted RF, often a cavity because the band is known and power is about to rise. YIG versus cavity is that choice.

What actually differs in the box

Direction changes three practical things.

Filter order. A downconverter filters RF, then mixes, then filters IF. An upconverter filters IF, then mixes, then filters RF. Swap those and the image or the unused sideband has already been translated into the band you meant to keep.

LO leakage. On a downconverter, LO–RF isolation decides how much of the LO escapes toward the antenna. On an upconverter, LO–RF isolation decides how much of the LO sits next to the wanted RF, IF away, and has to be filtered before a power amplifier. The isolation number is the same kind of number; the place it hurts is not.

Drive and limiting. A receiver downconverter is usually protecting a mixer from the antenna. A transmit upconverter is usually delivering a tone a later amplifier can use. Sphere-based YIG filters in the RF path still limit around 0 to +10 dBm in either direction; a transmit chain will hit that sooner.

Embedded LO versus external LO

A converter brick with a built-in oscillator is convenient and opaque. The phase-noise plot, the lock time and the spurious lines are then properties of that oscillator, not of mixing as such. If the embedded source is a VCO, you have bought the VCO side of the source comparison. If you supply the LO, you can put a YTO or a DRO on that port and the converter becomes a mixer-plus-filters again.

For a wideband receiver the usual split is an external YIG-tuned oscillator into a downconverter, with a tracking preselector on the RF. For a fixed SATCOM block upconverter the usual split is an embedded DRO or a locked VCO and a cavity on the RF. The converter type did not force those choices; the frequency plan did.

A radar front end is this page plus a waveform: the receive path is a downconverter, the exciter path is often an upconverter, and both share an LO purity problem. That assembly is radar RF components.

Direction changes the integration problem

An upconverter and a downconverter can use the same mixer topology, yet the surrounding hardware sees different risks. A downconverter accepts whatever the antenna and preselector deliver. Its first priorities are usually image rejection, input survival, gain distribution and preventing a strong off-channel signal from desensitising the receiver. An upconverter begins with a controlled IF and must deliver a clean RF product to a gain chain. Its priorities become unwanted sideband rejection, LO feedthrough, drive level into the power amplifier and emissions at the antenna connector.

Write the desired and unwanted mixer products down before placing filters. For an upconverter, a filter after the mixer must preserve the intended RF while rejecting the other sideband and LO leakage; a later amplifier can magnify all three. For a downconverter, filtering in front of the mixer often has to reject an image or blocker before it produces a false IF. An IF filter after the mixer still matters, but it cannot undo a product that has already folded into the wanted channel. The frequency converter guide is useful when the module is more elaborate than a bare mixer.

Level planning also changes direction. In a receiver, a little conversion loss can be costly because it follows the antenna and adds to the noise figure. In a transmitter, the mixer is often followed by enough gain that loss is tolerable, while linearity and spurious content bind first. Do not infer a transmit-safe RF path from an input P1dB number measured for a receiver use case. Include the highest IF drive, the gain at every later stage, and the output filter’s power rating in the same calculation.

The LO must be evaluated at its operating port. A shared LO may feed both an upconverter and a downconverter, making its phase noise and reference behaviour a system property. If the output is coherent with another radio, verify reference distribution, phase repeatability after retune and the effects of LO leakage paths in the installed layout. The RF, IF and LO explainer gives the frequency-plan checks; the final test should use the actual cables, filters and amplifier state rather than an isolated brick.

For acceptance, inject a known IF, sweep the LO across the intended band and measure the wanted RF, the unused sideband, LO leakage and broadband spurs. Reverse that test for a downconverter by injecting RF and examining the actual IF passband. A direction label is only the beginning; those measurements show whether the integration matches the label.

Document the reference planes with the results. A conversion-loss figure measured at the module connector cannot by itself describe the gain, emissions or sensitivity at an antenna connector after cables, filters and amplifiers. Naming the input and output planes for every measurement keeps an upconverter’s transmit budget and a downconverter’s receive budget comparable as the assembly changes.