Trace or U.FL jumper

When does a U.FL jumper beat a trace to the antenna?

Past about 2.0 cm at 5.5 GHz on a common 4-layer board. Over 10 cm there, a trace delivers 60% of the radio’s power and a coax jumper between two U.FL connectors 83%. On a 2-layer board at 915 MHz, the trace wins up to 13.3 cm.

The trace loses the same share of its power in every centimeter, so its wave shrinks all the way to the antenna.

The jumper pays a small toll at each of its two connectors. In between, at 5.5 GHz, its coax loses a fifth as much per centimeter as the trace.

Two routes, one radio

10 cm at 5.5 GHz on the 4-layer board

The wave along each route, scaled to 1 where it leaves the radio’s match, and below it the power left. The jumper’s U.FL mates sit at its two ends.

The numbers

Power left at 5.5 GHz, as a share of what leaves the radio’s match. The jumper’s 100 mm row is before its second mate.
Distance from the radio	Trace	Jumper
0 mm	100%	96%
20 mm	90%	94%
40 mm	82%	92%
60 mm	74%	90%
80 mm	67%	88%
100 mm	60%	86%
At the antenna	60%	83%

The board and its two routes

130 × 56 mm, 4 layers

A Wi-Fi radio sits at the left end of the board and a chip antenna at the right. Between them are a processor, its DDR and a buck converter, with a USB-C port on the bottom edge.

The radio’s signal leaves its matching network and has 10 cm to cross. One way is a 50 Ω trace on the top layer, routed up along the board’s edge, around the busy middle.

The other is a U.FL receptacle at each end, J1 and J2, with a 100 mm coax jumper plugged between them. It goes straight over the DDR.

One board, two ways across

The same board with the trace, then with the jumper. The route is drawn over each in its color.

The numbers

Part	What it is
U1	Wi-Fi radio, QFN-48, with its match
ANT1	Chip antenna, 10 × 3.2 mm, in a copper keep-out
U2, U3	Processor and its DDR
U4, L2	Buck converter and its inductor
J3	USB-C receptacle
J1, J2	U.FL receptacles, to Hirose’s U.FL-R-SMT-1 drawing
Trace	0.373 mm wide, 100 mm long, on the top layer over the layer 2 plane
Jumper	1.13 mm coax, 100 mm between plug centers

A trace’s width is set by how far its plane sits under it. On JLCPCB’s common 4-layer stack the plane is 0.21 mm down, so 50 Ω is 0.37 mm wide. On a 1.6 mm 2-layer board the plane is the bottom layer, 1.53 mm down, and 50 Ω is 2.84 mm wide.

Either way most of the field runs through the FR-4: 87% of its energy on the 4-layer board, 89% on the 2-layer one.

The jumper’s cable is 1.13 mm across. Its field stays inside the shield, in FEP, a plastic that loses far less than FR-4.

Three 50 Ω lines, cut across

To one scale. Copper in gold, each line’s signal conductor in its route’s color. The 4-layer section shows only the top of the board’s core.

The numbers

Trace Z₀ and εeff from the 2D field solve, 1 oz copper. Coax from Taoglas’s cable sheet.
Line	Conductor	Dielectric	εr	Z₀	εeff
4-layer trace	0.373 mm	0.2104 mm, 7628 prepreg	4.4	49.8 Ω	3.19
2-layer trace	2.838 mm	1.53 mm, FR-4 core	4.5	49.8 Ω	3.36
1.13 mm coax	0.24 mm center	0.68 mm FEP, 0.90 mm shield	2.1	50 Ω	2.1

A trace loses in every centimeter

Copper and FR-4

A trace gives up the same share of its power in each centimeter. In decibels that adds up along the run:

L_trace=αd

=0.223 dB/cm×10 cm=2.23 dB

That is the 4-layer board at 5.5 GHz. Decibels turn back into the share that arrives:

PP₀=10^−L/10

=10^{−2.23 dB/10}=60%

Most of the loss heats the FR-4, and that part grows in step with frequency:

α_FR-4=8.686×πfcε_effqtan δ

=8.686×π × 5.5 GHz3.00 × 10⁸ m/s×3.19×0.87×0.018=14.1 dB/m=0.141 dB/cm

The copper’s part grows only with the square root of frequency, so above 1.8 GHz the FR-4 takes the larger share on the 4-layer board.

Loss in each centimeter

Copper
FR-4

A 50 Ω trace on each board, at each band.

The numbers

1 oz copper, Sq 3.2 µm, loss tangent 0.018, no solder mask.
Band, board	Copper	FR-4	Total, dB/cm	FR-4’s share
915 MHz, 4-layer	0.031	0.023	0.054	43%
915 MHz, 2-layer	0.005	0.025	0.030	83%
2.4 GHz, 4-layer	0.054	0.063	0.116	54%
2.4 GHz, 2-layer	0.009	0.066	0.075	88%
5.5 GHz, 4-layer	0.082	0.141	0.223	63%
5.5 GHz, 2-layer	0.014	0.148	0.161	91%

482

Trace loss

0.223 dB/cm at 5.5 GHz

Field marks. On the 4-layer board at 5.5 GHz the trace loses 0.223 dB/cm: 0.141 in the FR-4 and 0.082 in the copper. Over 10 cm that is 2.23 dB, and 60% arrives.

Look-alike. Skin effect. Thin copper gets the blame for trace loss, but on the 4-layer board copper leads only below 1.8 GHz. On the 2-layer board the FR-4 takes 83% or more at all three bands.

Range. From 0.054 at 915 MHz to 0.256 dB/cm at 6.5 GHz on the 4-layer board, and 0.030 to 0.190 dB/cm on the 2-layer one. Rogers measured FR-4 microstrip on 0.76 mm at 0.154 dB/cm at 5 GHz. That falls between this page’s 0.206 dB/cm over 0.21 mm of FR-4 on the 4-layer board and 0.147 dB/cm over 1.53 mm on the 2-layer one.

Habitat. Cheap 4-layer boards, where the plane sits one thin sheet of prepreg under the top layer and a 50 Ω trace is narrow.

483

A jumper pays at its connectors

Two mates, then the cable

A jumper’s loss comes in two parts: a fixed toll at its two mates, and its cable, which loses little per centimeter. Linx’s figures for its U.FL jumpers grow with the square root of frequency, so both parts do:

L_jumper=(A+Bd)f

=(0.153 dB+0.0195 dB/cm×10 cm)×5.5=0.82 dB

That is a 10 cm jumper at 5.5 GHz, and 83% of the power arrives. With no cable at all, the two mates alone would cost 0.36 dB. The jumper starts behind and catches the trace within a few centimeters.

Loss against length

Trace
Jumper
Linx, 100 mm

5.5 GHz on the 4-layer board. The circle where the lines cross is the break-even length.

The numbers

5.5 GHz, 4-layer board. The Linx row is its 100 mm figure, 0.34√f dB, against the fit’s 0.82 dB.
Length	Trace loss	Jumper loss	Trace delivers	Jumper delivers
0 cm	0.00 dB	0.36 dB	100%	92%
1 cm	0.22 dB	0.41 dB	95%	91%
2 cm	0.45 dB	0.45 dB	90%	90%
4 cm	0.89 dB	0.54 dB	81%	88%
6 cm	1.34 dB	0.63 dB	73%	86%
8 cm	1.78 dB	0.73 dB	66%	85%
10 cm	2.23 dB	0.82 dB	60%	83%
Linx, 100 mm		0.80 dB		83%

484

Jumper loss

0.36 dB at the mates, then 0.046 dB/cm

Field marks. Linx rates its U.FL-to-U.FL 1.13 mm jumpers at 0.34√f dB for 100 mm, 0.56√f for 200 mm and 0.73√f for 300 mm, with f in GHz. A straight line through the three meets zero length at 0.153√f, the two mates, and climbs 0.0195√f per cm, the cable.

Look-alike. The cable’s own rating. Taoglas rates 1.13 mm cable at 4.6 dB/m at 5 GHz, so 10 cm looks like 0.46 dB. The whole jumper loses 0.78 dB there, because of its mates.

Range. I-PEX’s chart for its 100 mm MHF I jumper reads 0.42 dB at 2.4 GHz and 0.67 dB at 5.8 GHz, below the Linx line’s 0.54 dB and 0.84 dB. With I-PEX’s figures the jumper wins past 1.2 cm at 5.5 GHz on the 4-layer board.

Habitat. Boards where the antenna sits far from the radio. u-blox’s JODY-W2 manual puts it plainly: “If traces longer than few centimeters are needed, use a coaxial connector and cable to reduce losses.”

485

Where the lines cross

Every frequency, both boards

Set the two losses equal and solve for the length where neither route wins:

d_even=Afα−Bf

=0.153 dB × 5.50.223 dB/cm − 0.0195 dB/cm × 5.5=2.0 cm

Shorter than that, the trace delivers more. Longer, the jumper does. A lower band, or more FR-4 between the trace and its plane, pushes the length out.

Break-even length

4-layer
2-layer

From 900 MHz, the bottom of the 915 MHz band, to 7 GHz. A run longer than its board’s line goes to the jumper. The circles are 915 MHz, 2.4 GHz and 5.5 GHz.

The numbers

Break-even length with Linx’s jumper. Shorter runs go to the trace, longer ones to the jumper.
Frequency	4-layer	2-layer
900 MHz	4.2 cm	13.5 cm
915 MHz	4.1 cm	13.3 cm
2.4 GHz	2.8 cm	5.4 cm
5.5 GHz	2.0 cm	3.1 cm
7 GHz	1.8 cm	2.7 cm

486

Break-even length

2.0 cm at 5.5 GHz, 13.3 cm at 915 MHz

Field marks. On the 4-layer board the jumper wins past 4.1 cm at 915 MHz, 2.8 cm at 2.4 GHz and 2.0 cm at 5.5 GHz. On the 2-layer board, past 13.3 cm, 5.4 cm and 3.1 cm.

Look-alike. One length for every board. The same 915 MHz radio breaks even at 4.1 cm on the 4-layer board and 13.3 cm on the 2-layer one. The 2-layer board’s trace sits over more FR-4, so it is 7.6 times as wide, and its copper loses 6 times less.

Range. From 1.8 cm at 7 GHz on the 4-layer board to 13.5 cm at 900 MHz on the 2-layer one.

Habitat. 915 MHz radios on 2-layer boards, where a trace wins up to 13.3 cm. Wi-Fi radios on common 4-layer boards, where the jumper wins once the antenna is more than a couple of centimeters away.

487

Shielding, hands and cost

What loss leaves out

The jumper’s coax is shielded, so it can cross parts the trace has to avoid. In Fig. 25a the trace detours along the board’s top edge, clear of the processor, its DDR and the buck. The jumper runs straight across the DDR and past the buck.

Fast digital parts and ports make noise in a radio’s band. Intel measured the 2.4 GHz noise floor about 20 dB higher beside a USB 3.0 hard drive, and about 25 dB higher at a notebook’s USB 3.0 port. A route that picks up that noise brings it to the antenna.

The catch is the cable. The receptacles go on with the other parts, but the cable is plugged in by hand. Hirose, I-PEX and TE test their mates to 30 cycles. At 1,000 pieces, two receptacles and a cable cost $1.26 to $3.33 at Digi-Key, as listed on 5 November 2026. That price is for a 50 mm cable, half the length of the board’s. The trace costs only its copper.

Choosing a route

Loss, noise and the link

Length and band settle most cases. When they don’t, what lies along the route and how much range you need decide.

Key to the RF route

Trace or jumper?

1aThe run is shorter than its board’s line in Fig. 25eTrace
1bLongerGo to 2
2aThe straight path crosses DDR, a buck or USB 3Jumper
2bThe path is clearGo to 3
3aThe link needs the decibels the trace losesJumper
3bIt can spare themTrace

The board in Fig. 25a ends at 2a: 10 cm at 5.5 GHz, past the DDR. Its jumper saves 1.41 dB over the trace. Range in free space grows as 10^1.41 dB/20=1.18, so that is 18% farther.

Route it yourself

Three bands, two boards, 1 to 30 cm

Choose a band, a board and a length. Each line is the share of the radio’s power that reaches the antenna. The jumper’s starts lower, after its mates, and falls more slowly.

Where the lines cross is the break-even length of Fig. 25e.

Band

Board

Length of the run

10 cm

Power delivered against length

At 10 cm the jumper delivers more

Trace
Jumper

Trace delivers: 60%
Jumper delivers: 83%
Break-even: 2.0 cm

Power that reaches the antenna, as a share of what leaves the radio’s match. The circle is where the lines cross.

The numbers

Power delivered, trace / jumper.
Band, board	2 cm	5 cm	10 cm	20 cm	30 cm	Break-even
915 MHz, 4-layer	98% / 96%	94% / 95%	88% / 93%	78% / 89%	69% / 85%	4.1 cm
915 MHz, 2-layer	99% / 96%	97% / 95%	93% / 93%	87% / 89%	81% / 85%	13.3 cm
2.4 GHz, 4-layer	95% / 93%	87% / 91%	76% / 88%	58% / 82%	45% / 77%	2.8 cm
2.4 GHz, 2-layer	97% / 93%	92% / 91%	84% / 88%	71% / 82%	60% / 77%	5.4 cm
5.5 GHz, 4-layer	90% / 90%	77% / 87%	60% / 83%	36% / 75%	21% / 67%	2.0 cm
5.5 GHz, 2-layer	93% / 90%	83% / 87%	69% / 83%	48% / 75%	33% / 67%	3.1 cm

Conditions

Board: 130 × 56 mm, 1.6 mm thick. The 4-layer case uses JLCPCB’s JLC04161H-7628 stack: 0.2104 mm of 7628 prepreg under the top layer, εr 4.4. The 2-layer case puts the trace 1.53 mm over the bottom layer, εr 4.5.
Trace: 50 Ω microstrip, 1 oz copper, no solder mask: 0.37 mm wide on the 4-layer board and 2.84 mm on the 2-layer one, both 49.8 Ω by the 2D solve. Loss tangent 0.018 and copper roughness Sq 3.2 µm, the Hammerstad roughness model.
Jumper: Linx’s CSI-UFFR U.FL-to-U.FL 1.13 mm jumpers, as the straight-line fit of their 100, 200 and 300 mm figures in the jumper’s formula above. The opening figure splits the mates’ share evenly between its two ends.
Bands: 915 MHz, 2.4 GHz taken at 2.45 GHz, and 5.5 GHz.
Trace loss: A 2D field solve of each cross-section: copper by incremental inductance, FR-4 by the share of the field’s energy in it.
Opening: The trace’s wave is the field in the FR-4 under its center line, from an openEMS run of the 10 cm 4-layer line at 5.5 GHz. That run’s S21 is −2.21 dB, 60% delivered, and its field falls 0.221 dB/cm against 0.223 dB/cm from the 2D solve. The power line below it follows that fitted fall. Its wavelength on the board is 29.9 mm. The jumper’s wave uses the coax’s wavelength from an openEMS run of the cable, 36.5 mm, and the Linx loss.
Video: The video used an earlier run of the 2D solve that read copper loss about 13% low. Its 61% is 60% here, and its 2.1, 3.0 and 14 cm are 2.0 cm, 2.8 cm and 13.3 cm.
Renders: Blender. Parts from their datasheet outlines, the U.FL receptacles from Hirose’s drawing.

Simulated only. Nothing on this page has been measured on a board yet.

Limits

Linx rates these jumpers from DC to 3 GHz. Above that its square-root law is carried past its range. I-PEX’s 5.8 GHz figure is lower.
TE allows up to 0.24 dB for each of its U.FL-style mates to 6 GHz, 0.48 dB for two, more than the Linx line’s 0.36 dB at 5.5 GHz.
Each mate’s reflection is inside the jumper’s figures, not modeled on its own.
Solder mask over the trace would add a little loss.
Noise pickup isn’t simulated. Intel’s figures are near-field probe readings in the 2.4 GHz band, not noise on a trace.
Bends and vias aren’t counted. A trace that changes layers pays at each via too, as Note 17 shows.

Files

post/site.pyTrace loss on both boards, the jumper’s line and every break-even length on this page
post/tradeoff.pyThe Linx and I-PEX jumper fits
sim/stack.pyThe two stack-ups and their 50 Ω widths
sim/fields.pyThe openEMS runs of the 10 cm line and the coax at 5.5 GHz
render/site.pyThe board renders in Fig. 25a
render/board.pyThe board, its parts and both routes

Sources

Like this note?

Rev 1 · 8 Nov 2026

Rev 18 Nov 2026First published

Each fix gets a line here, with the reader's name if they want credit.

Did it answer your question?

Want more on this?

Report a problem

What should the next note be? Ask for one