6G ISAC PCB: One Array, Two Jobs
6G ISAC PCB design at Upper 6 GHz: sensing starts with bandwidth, leakage sets the isolation budget, and one set of traces now answers two specifications.
Table of Contents
Integrated sensing and communication is the first 6G feature that changes the radio itself rather than the spectrum it runs on. One aperture transmits the communication waveform and listens for the echoes of that same waveform; the board that carries it is simultaneously a base station and a radar. Most of what is hard about that lives on the PCB, because the PCB is where the transmit and receive paths run closest together for the longest distance.
The trials now running in Upper 6 GHz — including field demonstrations that detect and track multiple drones from a modified cellular radio — are the first time this hardware has been built outside a laboratory. This page works through the board-level consequences in the order they bite: what bandwidth the sensing side needs, what the leakage budget looks like, why one routing set now answers two different specifications, and where 6–8 GHz sits for materials.
1. One Aperture, Two Specification Sets
A shared-aperture radio is three subsystems on one board, and each one specifies the layout differently.
| Subsystem | What it specifies | Board-level consequence | What breaks first |
|---|---|---|---|
| Communication TX | EVM, ACLR, PA efficiency | TX chain loss and the thermal path under the PA | A hot spot that moves a neighbouring RX block |
| Communication RX | Sensitivity, blocking, ACS | RX front-end placement and a clean ground away from TX | Desense from leakage, not from the air |
| Sensing RX | Echo dynamic range, channel phase stability | Isolation budget, calibration couplers, phase-stable routing | Slow phase drift read as a moving target |
The first two rows are a normal radio and the industry knows how to buy them. The third row is what ISAC adds, and it is specified in degrees and decibels that have to be agreed before layout starts.
The communication rows are familiar and the industry prices them well. The sensing row is where new boards get into trouble, for two reasons. First, its key numbers — isolation and channel phase — are distributed across the whole layout rather than attached to a component, so they are easy to lose between disciplines. Second, its failures are slow and quiet: a phase drift of a few degrees does nothing to the EVM and everything to the tracking filter downstream. Boards built for this job carry their own witnesses, which is the subject of the checklist at the end.
2. The Sensing Budget Starts with Bandwidth
Range resolution is c/2B, which makes bandwidth the only currency the sensing side trades in.
| Waveform bandwidth | Range resolution | Practical meaning |
|---|---|---|
| 100 MHz | 1.499 m | A cell the size of a person; drone against clutter stays marginal |
| 200 MHz | 0.749 m | Separates a small UAV from the structure it is near |
| 400 MHz | 0.375 m | Where a typical Upper 6 GHz allocation starts to be useful for small drones |
| 800 MHz | 0.187 m | Resolves rotor-scale detail; the waveform and the ADC become the limit |
Velocity costs nothing extra: a 10 m/s drone at 7 GHz sits at 467 Hz of Doppler, and velocity resolution is set by dwell time rather than bandwidth. Range is the dimension that has to be bought with spectrum.
This is the actual argument for Upper 6 GHz as an ISAC band. It is not that 7 GHz is magic; it is that the band offers continuous blocks wide enough for the middle rows of that table, in hardware that behaves far better than mmWave for the same aperture. Sub-6 allocations were too narrow for the sensing rows, and mmWave buys resolution at the price of every loss and assembly problem this site writes about. In between sits a band where the board is ordinary and the waveform is not.
3. Leakage Is the Whole Design
Put numbers on the radar link first, because the rest of the budget hangs off them.
| Antenna/board isolation | Leakage at the RX input | Above the echo (-109.4 dBm) | Above the noise floor (-91.0 dBm) |
|---|---|---|---|
| 50 dB | -20.0 dBm | 89.4 dB | 71.0 dB |
| 60 dB | -30.0 dBm | 79.4 dB | 61.0 dB |
| 70 dB | -40.0 dBm | 69.4 dB | 51.0 dB |
| 80 dB | -50.0 dBm | 59.4 dB | 41.0 dB |
Radar link: 30 dBm transmit, 20 dBi each way, a 0 m² small UAV at 300 m returns -109.4 dBm. Everything the cancellation chain does not remove sits on top of that return, so the isolation budget has to be written down before anyone starts routing.
The column that matters is the last one. Whatever the canceller fails to remove does not merely sit beside the echo — it sets the floor the echo has to rise above, and the dynamic range the receiver must survive includes not just the leakage but the same target at close range: the same 0 m² drone at 30 m returns -69.4 dBm, some 40 dB hotter than at 300 m, inside one receive window. The budget is therefore a table with named contributors — antenna, package, layout, digital — and the layout's share is the cheapest of them to buy. Partitioning, via fencing and ground discipline are set out in our RF/digital mixed-signal stackup guide.
4. One Routing Set, Two Specifications
The array is shared, so the inter-channel phase specification is shared too — but the two users read it differently.
| RMS channel phase error | EVM contribution | Angle error (D = 200 mm) | Cross-range error at 300 m |
|---|---|---|---|
| 1° | 1.75% | 0.034° | 0.18 m |
| 3° | 5.24% | 0.102° | 0.54 m |
| 5° | 8.73% | 0.170° | 0.89 m |
Left half: a communications engineer's view — degrees of RMS phase error are a modest EVM line item. Right half: the same error converted to where the sensor thinks the target is. One specification, two meanings, and only the second one scales with the job the board was bought for.
Communications converts phase error into modulation quality and stops there: 1° RMS is a 1.75% EVM line item, well inside budget. The sensor converts the same error into where the target is, and at 300 m the conversion is uncomfortable. Neither number is wrong; they are different unit systems for the same copper, and only the second one appears on the drawing.
| Quantity | Value | Read across |
|---|---|---|
| PTFE-class Dk temperature coefficient | -150 ppm/K | Material property, not a defect |
| Dk change over a 50 K swing | 0.75% | Small in material terms |
| Guided wavelength change | 0.38% | Halved by the square root |
| Phase drift, 100 mm line at 7 GHz | 4.0° | 0.68 m of cross-range error at 300 m |
| Phase spread of a 100 mm line, 6.425 → 7.125 GHz | -106° | Wideband sensing waveforms live across this whole span |
Row four is why calibration couplers are not optional hardware on an ISAC board: the drift is real, it is slow, and a sensor that cannot separate it from target motion will report the weather.
The drift row is the practical conclusion. A material that moves 4.0° on a 100 mm line over a 50 K swing — ordinary for a day outdoors — is indistinguishable from target motion unless the system measures its own reference. That is what the calibration couplers are for, and why a board that omits them has not saved cost but moved it to the customer's algorithm team. The behaviour underneath is the same Dk temperature dependence described in our Dk and Df performance page.
5. Where 6–8 GHz Sits for Materials
Between the sub-6 world where FR-4-class laminates are normal and the mmWave world where they are not, 6–8 GHz is a transition band, and the honest way to treat it is numerically.
| Laminate class | Dk / Df at 7 GHz | Dielectric loss | Per 100 mm | Where it lands |
|---|---|---|---|---|
| FR4-class | 4.40 / 0.0180 | 0.0370 dB/mm | 3.70 dB | RX noise and PA heat |
| Mid-loss hydrocarbon | 3.00 / 0.0035 | 0.0047 dB/mm | 0.47 dB | Balance of both |
| PTFE-class | 2.20 / 0.0009 | 0.0010 dB/mm | 0.10 dB | Feed and reference layers |
Dielectric term only, 50 Ω lines, first-order. Conductor and surface effects are the same physics as any band and are covered separately.
The 37× spread in that table does not argue for one material. It argues for asking where the loss lands: transmitter loss becomes heat under the PA, receiver loss becomes noise figure, and reference-distribution loss becomes phase that the calibration path has to track. Different layers of the same board carry different requirements, which is why hybrid stackups are normal at this band rather than an exotic choice — the mechanics are described in our FR-4 hybrid guide.
One more number belongs in the RFQ at this band: the phase spread of the feed across the full allocation. A 100 mm line sweeps -106° from 6.425 to 7.125 GHz, and wideband sensing waveforms live across that whole span. Group-delay consistency between channels, not just matched phase at one frequency, is what keeps a wide waveform focused.
6. The DFM Checklist for a Shared-Aperture Board
Related reading
- multi-band counter-UAS jammer PCB — band-dedicated chains, sealed-shell dissipation and co-site isolation
- GNSS anti-jamming PCB — CRPA arrays and multi-RF isolation on the receive side
- phased array PCB design and manufacturing — element pitch, feed networks and Dk control on the analogue side
- 5G mmWave PCB overview — the wider picture from 24 GHz to 40 GHz
- aerospace and defence high-frequency PCB — qualification and traceability for mission hardware
Eight lines, in the order they decide the outcome.
| # | Item to specify | Why it decides the outcome |
|---|---|---|
| 1 | The isolation budget as a table of contributors | Antenna, package, layout and digital each own decibels; unowned decibels are lost twice |
| 2 | Calibration and monitoring couplers, TX and RX | Phase drift is invisible to EVM and fatal to sensing; the couplers are its only witness |
| 3 | TX/RX partition with via fence dimensions | The 70 dB that layout contributes is cheaper than the same dB in the canceller |
| 4 | Which channels are phase-matched, over which band, at what temperature | Phase matching without a temperature range is a number without a meaning |
| 5 | PA decoupling and the power tree | Supply ripple lands in the EVM budget and, through heating, back into phase |
| 6 | LO/clock route length matching across the array | The sampling clock offsets that a communications link would absorb appear as range bias |
| 7 | Layer roles: loss-critical feed vs dimensionally critical structure | The two requirements rarely live on the same layer; hybrid stackups are normal |
| 8 | What the panel test verifies before assembly | Passive S-parameters and coupler flatness are checkable; sensing performance is not |
Line 2 is the one most often value-engineered away, and it is the only item on this list that keeps working after the board ships.
The pattern across all of it: an ISAC board is bought twice, once as a radio and once as a sensor, and the second purchase is the one nobody has a template for. The isolation budget, the couplers and the phase specification are how it gets written down. Transmit-side counter-drone hardware is a related but separate discipline, covered in our multi-band CUAS jammer page; receive-side resilience in GNSS anti-jamming; and the array itself in phased array PCB.
6G ISAC PCB — Frequently Asked Questions
What shared-aperture sensing adds to a communication board: the isolation budget, the bandwidth bill, and the phase specification that has to carry two meanings at once.
What does ISAC actually change at board level?
The radio keeps every specification it had, and gains a second receiver that listens for echoes of the board's own transmission. That adds three things a normal communication board never carried: an isolation budget between transmit and receive paths, calibration couplers that let the system separate its own drift from target motion, and channel phase specifications written in degrees rather than as an EVM line item. The aperture, the power tree and the clock distribution are shared, which is exactly why the additions are hard to retrofit.
Why is leakage the hard problem instead of sensitivity?
Because the interference is the board's own transmit signal, and it is enormously stronger than what the sensor is trying to hear. With 30 dBm transmit and 20 dBi on each path, a 0 m² drone at 300 m returns about -109.4 dBm. At 60 dB of isolation the leakage at the receiver input is -30.0 dBm — 79 dB above the echo and 61 dB above the noise floor of a 100 MHz channel. Antenna and layout isolation, and digital cancellation, have to add up against that table.
How much bandwidth does the sensing side actually need?
Range resolution is c/2B, so it is bought directly with spectrum: 100 MHz gives 1.50 m, 400 MHz gives 0.375 m, 800 MHz gives 0.19 m. Doppler is cheap — a 10 m/s drone is 467 Hz at 7 GHz — but velocity resolution comes from dwell time. That is the case for Upper 6 GHz: it offers the wide continuous blocks that sensing wants, in a band where the RF hardware is far more forgiving than mmWave.
Can the same traces carry both jobs?
Physically yes, and sharing them is the point of ISAC. Specification-wise they are two different contracts. Communications reads inter-channel phase error as a small EVM contribution: 1° RMS is 1.75%. The sensing path reads the same error as pointing: 0.034° of angle on a 200 mm aperture, 0.18 m of cross-range at 300 m. A routing set that passes the first check can fail the second, which is why the phase specification needs the sensing numbers written next to it, and why the array itself is treated properly in our phased array PCB guide.
Does 6–8 GHz need exotic laminates?
It needs the loss budget to be written per layer rather than assumed. At 7 GHz the dielectric term alone is 3.70 dB per 100 mm on a generic FR-4-class laminate, 0.47 dB on a mid-loss hydrocarbon, and 0.10 dB on a PTFE-class material — a 37× spread that lands partly as transmitter heat and partly as received noise. That is a comparison of loss terms, not a recommendation to swap materials: feed layers, control layers and structures each have their own requirement, and hybrid stackups are the normal answer. The behaviour behind the numbers is on our Dk and Df page.
What should an ISAC board RFQ contain that a normal RF board RFQ does not?
Four things. The isolation budget, as a table of who contributes which decibels. The calibration and monitoring couplers, with their own routing, because they are the only witness to phase drift. The channel phase specification with its band and temperature range attached. And the panel-test scope: passive S-parameters and coupler flatness can be verified before assembly, while sensing performance cannot — quoting the latter from an empty board is the mistake. Batch evidence for these builds is set out in our RF PCB quality control guide.
Specifying an ISAC or shared-aperture board?
Send the band, the waveform bandwidth, the transmit power and how the isolation budget is split. We will set the TX/RX partition, the coupler plan and the phase specification against the sensing budget, then quote the build with per-panel passive data.
Contact Us → WhatsApp: +86 137 6047 3650Request a PCB Quote
Upload your Gerber ZIP file and project requirements. Our engineering team will review your PCB material, stackup, impedance needs, surface finish, and production quantity before quoting.
Please prepare:
- Gerber files in ZIP format
- PCB material or stackup requirements
- Controlled impedance notes if available
- Prototype or batch production quantity