Near Field Notes

Videos

The notes, in motion

One video per topic, newest first. Point at a card for a silent preview.The card in the middle of the screen plays a silent preview. Open one to watch it whole.

41 videos, 41 with a note to read after.

4:18 · With sound

PCB materials

Do you need Rogers for RF?

Read Note 30

Transcript

Do you need Rogers for RF?

At 2.4 GHz, a 3 cm trace on FR-4 still delivers 92%.

At 77, only 21.

So radar boards use Rogers, but only for one thin layer on top.

Here's what each board material is good and bad at, and how hybrids work.

A board is glass cloth soaked in resin, with copper foil outside.

Cores come cured. Prepreg is the sticky layer that glues them together when it's pressed hot.

Change the resin and the glass, and 4 numbers change.

Dk sets the signal's speed, and how wide a 50 Ω trace is.

Df is how much of the signal turns into heat.

Z-expansion is how much the board swells when it's soldered.

And thermal conductivity is how well it pulls heat away.

Standard FR-4 is cheap and everywhere.

Its Df is about 0.018. It swells 4.5% when soldered, and lasts 2 minutes at 288 °C.

High-Tg FR-4 handles lead-free soldering. 30 minutes at 288.

But high Tg doesn't mean low loss. Isola's 370HR has a Df of 0.025.

Low-loss laminates like Megtron 6, 7 and 8 drop Df to 0.004, then 0.0012.

They're made like FR-4, and they're the standard in servers. But they cost about 3.5 times as much, and need smooth copper to pay off.

Rogers RO4350B is a ceramic-filled hydrocarbon, not PTFE. Its Df is 0.0037, and it's processed almost like FR-4.

5 boards at JLCPCB cost $186, instead of $4. The laminate itself is about 4 times FR-4.

PTFE laminates like RO3003 have the lowest loss. Df 0.001.

And its dielectric constant barely moves with heat. From −40 to 85 °C, a 5 cm radar feed drifts 87° of phase on FR-4. On RO3003, just 1.

But PTFE is soft, it swells, and drill smear on it can't be cleaned off.

Polyimide bends, so it's used for flex and rigid-flex. But it soaks up almost 1% water.

Phones use LCP for mmWave antennas. It's low loss, and flexible.

Many LED bulbs put their LEDs on aluminum. I simulated a 1 W LED.

On FR-4, it runs 25 °C above the heatsink. With thermal vias, 11. On aluminum, 3.

Power modules go further, with ceramics like silicon nitride and aluminum nitride. Up to 170 W/m·K, where FR-4 is 0.4.

But they're brittle, and expensive.

I simulated one 50 Ω trace, 10 cm long, on each material.

At 10 GHz, FR-4 loses 3.9 dB, and RO3003, 1.3. At 77, FR-4 loses 25. Under 1% arrives.

Now the hybrid. Here's TI's 77 GHz radar board.

Only the top 0.1 mm is Rogers. Layers 2 to 6 are FR-4. Rogers is 7% of the board.

I simulated the field around the antenna feed. 83% sits in the Rogers, and 17 in the air. None gets past the copper plane.

So the signal sees nearly all Rogers. The FR-4 underneath just holds the board together.

Stripline is different. Its field goes above and below.

Low-loss on one side only lands halfway. 14 dB over 10 inches at 16 GHz, against 8 with low-loss on both.

Hybrids have rules. Keep the stack symmetric, or it warps.

Bonding PTFE takes 371 °C. FR-4 starts breaking down at 345. So PTFE hybrids are glued with FR-4-class prepreg, at about 180.

FR-4 swells 4 times more than RO4350B in thickness, and that pulls on the vias.

And each mix needs its own drilling and cleaning recipe, so ask your fab first.

Other tricks mix in the same way.

Smooth copper on the fast layers. On the same Megtron 6, standard foil lost 37% more at 10 GHz.

Flat glass, or the design turned about 10°, so a pair doesn't skew over the weave.

Copper coins pressed in under hot parts. About 400 W/m·K.

Rigid-flex: FR-4 where parts sit, polyimide where it bends.

And thin, high-Dk cores as power planes, for up to 127 times the capacitance.

FR-4 for most boards. Low-loss where signals are fast. And Rogers or PTFE only on the layers the RF field touches.