SMA Connectors: Frequency Limits, Torque and Reflections

What an SMA connector is for

The SMA connector is a threaded coaxial connector with a 50 ohm interface. It is the usual choice for bench RF work: test cables, adapters, attenuators, evaluation boards and the ports of many instruments. It is small, cheap to use and screws together firmly, so it suits a setup that you connect and disconnect many times.

It is a precision-limited part, not a perfect one. Every mated pair is a small discontinuity in the 50 ohm line. It adds a little loss and reflects a little of the wave. Most of the time that is negligible. In a careful S-parameter measurement it is the kind of error you must understand, because it shows up directly in the trace.

Frequency limits

SMA connectors are commonly specified up to 18 GHz. Treat this as a rating of the connector family, not as a promise for every part. A given cable, adapter or terminator has its own datasheet, and the chain is only as good as its weakest part.

Above the rated range the connector can still pass a signal, but the line no longer behaves as a clean 50 ohm line. Reflections grow and become less repeatable. Higher-order modes can also appear in the coaxial structure at high enough frequency. For measurements near the top of the range, the connector stops being a minor detail and becomes part of the result.

3.5 mm connectors mate mechanically with SMA connectors. That is a statement about fit. It does not give the combination a higher limit than the SMA part, so read the rating of each part before you trust the pair near the upper end of your band.

Why torque matters

The reflection of a connector depends on how well the two contact surfaces and the outer conductors meet. That depends on how hard you tighten the nut.

A common tightening torque for SMA connectors is 5 in-lb, about 0.56 N*m. A torque wrench sets this value each time. Hand tightening varies from one connection to the next. Each connection then has a slightly different reflection and a slightly different loss, even though nothing else has changed.

For one-off checks this does not matter. For a calibrated measurement it does, because the repeatability of the connection sets the floor of your accuracy. If you re-connect the same part and the trace changes, you are measuring your fingers. A torque wrench removes that variable.

Losses add in dB

Cables and connectors in series each have a loss in dB. The total is the sum:

This is true because dB is a logarithm of a power ratio, and the ratios of stages in series multiply. Adding logarithms is the same as multiplying the ratios. It means you can budget a test setup by adding one line per part: both cables, every adapter and every connector pair.

Return loss is a different number. It describes the reflection, not the power that is lost on the way through. A connector with a return loss of 30 dB reflects only 0.1 % of the power, so it passes 99.9 % of it. That is 10 log10(0.999), about -0.004 dB of loss from the reflection. The loss from the reflection is tiny. The problem is not the lost power but the phase of the reflected wave, which is the next section. For the difference between the two numbers, see Return Loss vs Insertion Loss.

Two reflections make a ripple

Return loss in dB converts to a reflection coefficient magnitude:

For RL = 30 dB this gives |Γ| = 0.0316. The power reflected is |Γ|², which is 0.001, or 0.1 %.

Now put two such connectors on one line. Each sends back its own small wave. The wave from the second connector travels an extra distance, there and back, so it arrives at the first connector with a phase delay of

where d is the distance between the connectors, v is the wave speed in the line and f is the frequency. Because φ grows with f, the two reflections go in and out of phase as you sweep. The sum swings between two limits:

• In phase: |Γ| = 0.0316 + 0.0316 = 0.063, which is about -24 dB.
• In opposite phase: |Γ| = 0, so the reflections cancel.

In S11 this is a ripple. It sits between nearly nothing and -24 dB, even though each connector alone is at -30 dB. The ripple repeats every time φ changes by 2π, so the spacing in frequency is v / (2d). A longer line between the connectors gives a faster ripple.

This is a worst-case picture. It ignores the loss of the line between the connectors and the smaller multiple reflections. It is accurate enough to explain what you see.

Worked example

Two connectors, each with a return loss of 30 dB:

1. Each has |Γ| = 10^(-30/20) = 0.0316.
2. The worst case adds the two magnitudes: 0.0316 + 0.0316 = 0.063.
3. In dB: 20 log10(0.063) = about -24 dB.
4. As a VSWR, this is (1 + 0.063) / (1 - 0.063), about 1.13.

Two good connectors can therefore give an S11 that is 6 dB worse than one of them at its peak, with the ripple moving as you change the frequency or the cable length. To convert between these forms, use the VSWR, Return Loss and Mismatch Loss Calculator.

In plain words · for students and beginners

SMA connectors in plain words

An everyday picture

Think of shouting across a valley and hearing a faint echo. A connector is like a small wall in the path of the signal. Most of the signal goes through, and a very small part bounces back.

Now put up two small walls. You hear two faint echoes. Sometimes they arrive together and sound louder. Sometimes one cancels the other. Move one wall a little, and the loudness changes.

The picture breaks down in one place. A real echo is a sound you can hear after a delay. In RF, you measure the sum of the echoes at one point, and the sum changes with frequency. That change is the ripple.

The same example in simple steps

Take two connectors. Each one sends back a very small part of the signal.

1. A return loss of 30 dB means the echo is tiny. Its size is 0.0316 compared with the original, which is 0.1 % of the power.
2. If the two echoes arrive together, they add: 0.0316 + 0.0316 = 0.063.
3. That is about -24 dB. This is the worst case.
4. If the two echoes arrive opposite, they cancel, and the sum is close to zero.
5. As the frequency changes, the echoes go from together to opposite and back. The trace on the screen goes up and down. This is the ripple.

The chart shows how return loss in dB and VSWR describe the same thing. A bigger return loss in dB means a smaller echo and a VSWR closer to 1.

Mistakes beginners make

• Tightening by hand and calling it the same. Your hand does not give the same force each time. Use a torque wrench at 5 in-lb (about 0.56 N*m) for measurements you want to repeat.
• Assuming a good connector has no effect. Even 30 dB is a small echo, not zero. Two of them can show up as ripple.
• Mixing up loss numbers. Insertion loss is what the signal loses on the way through. Return loss is how big the echo is. They are not the same number.
• Adding losses as percentages. Add dB values for parts in series. Do not add percentages.
• Trusting the 18 GHz figure for everything. It is a rating for the connector family. Check the datasheet of each part.

Remember

• SMA is a 50 ohm connector, commonly specified up to 18 GHz.
• Use a torque wrench at about 5 in-lb (0.56 N*m) so each connection is the same.
• Losses in series add in dB, and two 30 dB connectors can combine into a ripple of up to about -24 dB in S11.

Related pages on RF View

• VSWR, Return Loss and Mismatch Loss Calculator
• Return Loss vs Insertion Loss: What Each Number Tells You