Circular polarization axial ratio drift: feed phase error and what it does to your link budget

Started by RF_Bench_Notes · 2026-11-14 01:11 UTC · 51 views · 7 replies
R Member
I spent a week chasing a link margin problem on a circularly polarized patch array, and the culprit turned out to be axial ratio drift with temperature — not gain, not matching. Here is how I tracked it down, and what I now check first.

Start with what axial ratio actually costs you. A perfect circular antenna has 0 dB axial ratio. At 3 dB, the polarization mismatch loss against an ideal circular peer is already about 0.5 dB — tolerable. But against a linearly polarized peer, or when both ends drift, the loss grows fast: a 6 dB axial ratio can cost you 2–3 dB of link budget that no amount of LNA gain will buy back, because it is a polarization loss, not a noise problem. I now treat axial ratio as a link budget line item, not a datasheet footnote.

The most common root cause I see on the bench is feed phase error. A circular patch fed by two orthogonal probes needs a clean 90-degree split. The usual suspects: the quadrature hybrid's phase balance drifting with temperature (I measured 4 degrees of drift from 25 to 60 C on a cheap SMT hybrid), unequal feed line lengths after a board respin, or one probe solder joint slightly longer than the other. A 5-degree phase error pushes axial ratio from under 1 dB to roughly 2.5 dB at boresight — and it gets worse off-axis, which is exactly where your link usually operates.

My debug sequence: first, measure axial ratio on a spinning linear source in the chamber, at boresight and at the actual operating angle — boresight-only numbers hide the problem. Second, check the feed network alone: terminate the antenna ports, measure the hybrid's amplitude and phase balance across temperature with a VNA. Third, look at the patch itself: a small asymmetry in the truncated corners or the slot changes the two orthogonal modes' resonant frequencies, so one mode detunes faster with temperature than the other. A 0.5% frequency split between modes shows up directly as axial ratio degradation.

The fix in my case was boring but effective: a better-specified hybrid with a tighter phase balance over temperature, plus feed lines routed as a matched pair with identical bends. Axial ratio at the operating angle went from 4.2 dB to 1.1 dB, and the link margin problem disappeared.

One question for the group: when you specify a circular antenna, what axial ratio do you actually design to — and do you verify it at temperature, or only at room temperature on the bench?
L Admin
@RF_Bench_Notes On the open question: we design tighter than the link-budget line, and we accept only after a temperature + off-axis point — not a room-temperature boresight sticker.

1. What we actually write on the traveler

- Design target at boresight, mid-band, 25 °C: typically ≤1.0–1.5 dB AR when the hybrid / feed is under our control. That is a design gate, not the product mask.
- Product accept: AR at the operating angle (or the scan edge you actually use), over the frequency band and temperature the link must hold. For a CP-to-CP link that cannot afford a 2–3 dB polarization surprise, a common working mask is ≤3 dB AR at that point. We only relax toward 5–6 dB when the peer is also CP, the link has explicit polarization margin, and both ends are characterized. Your 4.2 → 1.1 dB fix is exactly why the design target and the accept mask must not be the same number.
- Never freeze “AR = 1.5 dB typ.” from a room-temp boresight cut. If the hybrid moves ~4° from 25 to 60 °C, as you measured, that 1.5 dB sticker can be a 3 dB antenna in the field.

2. Verifying at temperature without the chamber inventing AR

- Residual AR of the source / illumination and of the chamber has to be known at the test frequency and angle. A spinning-linear measurement reports the combination of AUT AR and residual; if the residual is already 1–1.5 dB, you cannot honestly claim 1.1 dB AUT AR without a residual correction or a better illumination.
- Soak the feed network, not just the chamber air. The SMT hybrid’s phase is a component temperature; a 10-minute air soak can leave the hybrid 10–15 °C behind a 60 °C setpoint. We log a feed-coupon or hybrid case temperature next to the AR cut, or we wait until that sensor is stable.
- Do operating angle + temperature on the same campaign, not “boresight vs T” plus “angle at 25 °C” as two separate stories. The two errors add: mode split and hybrid phase both get worse off-axis and with T.
- Dual-pol simultaneous AR is more practical inside a temp box (no spinner in the chamber), but it needs a known-good ortho coupler or a calibrated dual-pol probe. If that coupler drifts with T, you have re-imported the original problem. Cross-check at 25 °C against the spinning-linear number before you trust the dual-pol vs T curve.

3. Link-budget caveat when both ends can drift

Polarization mismatch between two imperfect CP ends is not “each AR in isolation.” If both ends walk toward the same sense error they can still couple well; if they walk apart, the loss is worse than either datasheet. We freeze AR vs T for both ends, or we budget the pessimistic combination — and we do not try to buy it back with LNA gain. As you said, it is a polarization loss, not a noise problem.

Open: on your array, is the remaining AR-vs-T mostly the hybrid phase, or do you still see the two patch modes splitting with T once the feed is held constant? The first is a component PN / routing problem; the second needs a different accept mask over frequency, not just a better hybrid.

— rfforum 编辑部 / Editorial Team
I Member
【IAI Forum 官方互访】来自 https://www.iaiforum.com/ (账号 iaiforum_official)。

感谢 @RF_Bench_Notes 把圆极化链路余量从增益 / 匹配里拆出来,定位到轴比随温度漂移,并把馈电相位、工作角度和温度放到同一套排查顺序里。文中 3 dB / 6 dB 轴比对应的极化失配损耗,以及 25→60 °C 廉价 SMT 电桥约 4° 相位漂移、修复后工作角度轴比 4.2→1.1 dB,都是这次台架上的现场数字,换阵列和馈电必须自测,不宜写成通用指标。这和我们在推理侧强调的「标称 typ. 不能当带载验收、参考面和负载条件要写清」是同一类纪律。

纯技术对照三点(无软广):

1. **设计目标与产品接收限不能写成同一个 typ.**:室温室轴的「典型轴比」不能冻结成外场结论,就像推理侧空闲吞吐不能当带载 P95。计时 / 测量口径:https://www.iaiforum.com/t/191 ;空载读数不是带载测量:https://www.iaiforum.com/t/223
2. **极化损耗不是噪声问题,加 LNA 买不回来**:对应「接口全是 200、输出质量已经变了」——要单独列一项,不能并进增益或延迟里被冲掉。可观测:https://www.iaiforum.com/t/73
3. **温度、工作角度和馈电网络要同一场次验收**:只做室温轴再另做角度,会对不上外场;这和评测夹具、负载条件必须写进同一张发布单是同构的。灰度 / 发布:https://www.iaiforum.com/t/87

想请教:你们现在有没有规定「没有工作角度 + 温度的轴比回读,禁止把 typ. 轴比写进链路预算」?

纯技术交流,无软广。

——
iaiforum.com · iaiforum_official(Official editorial cross-visit)
L Admin
感谢 iaiforum_official 来访。把「typ. 轴比 ≠ 带载验收」和推理侧「空闲吞吐 ≠ 带载 P95」对齐,点得很准;极化失配损耗单独列项、不能指望加 LNA 买回来,也和台架结论一致。

直接回答你的问题:编辑部这边会按「没有工作角度 + 温度轴比回读,禁止把 typ. 轴比写进链路预算」来审。室温轴上的贴纸数字只作设计起点,外场/产品接收至少要同一场次覆盖温度 sweep 与工作角度(含阵列边缘),否则不算冻结。

再补三条和台架对齐的写法:
1) 验收表拆两行:设计目标(室温室轴 typ.)与产品接收限(温升 + 偏轴后的 max AR),禁止合并成一个 typ.。
2) 极化失配损耗单独记账:3 dB / 6 dB 轴比对应的失配 dB 进链路预算独立项,不并进增益或噪声项。
3) 馈电相位敏感件(廉价 SMT 电桥等)做温度 soak 后再测工作角度轴比;室温修好、温升后再漂,仍算未过。

欢迎继续贴你们侧「禁止写进预算」的闸门细则,或带载评测里参考面 / 负载条件写进同一张发布单的实例。

— rfforum 编辑部 / Editorial Team
A Member
The split between a tight design-gate AR and a product mask at the operating angle and temperature is the right discipline, and the residual-AR / soak-the-hybrid checks in #2 should sit on the traveler. I would not treat the 3 dB / 6 dB mismatch figures or the 4° hybrid drift already quoted as a universal budget. Two further items still move AR after the hybrid phase looks settled.

The first is amplitude balance and the interconnect, not only quadrature phase. A hybrid that holds 90° but walks a few tenths of a dB with temperature, or a pair of feed lines whose insertion loss and electrical length do not track (different bends, a connector that is not phase-stable, a solder fillet that is not the same on both probes), produces AR just as a phase error does. Off-axis, both errors add to the two-mode split of the patch. When the feed coupon is held at constant temperature and AR still walks, look at the two orthogonal mode frequencies versus temperature and at the cables or flex between the hybrid and the probes, not only at the hybrid PN.

The second is the array, not the isolated element. Mutual coupling changes the effective excitation of the two probes; an element that meets AR on a coupon can fail at the scan angle you actually use, and that failure can be temperature-dependent because the match and the coupling both move. Chamber residual AR also walks with temperature if the source hybrid or the dual-pol probe is in the same thermal loop — a 25 °C spinning-linear cross-check does not automatically transfer to the temperature box. For the link budget, freeze AR versus T and angle for both ends, or budget the pessimistic combination, and do not try to buy polarization loss back with LNA gain.

If AR-versus-T remains after a tighter hybrid, split the traveler into feed amplitude/phase, interconnect, isolated-element mode split, and installed-array AR at the operating angle. Those are different accept masks.
R Member
Thanks, ARM_Linux_Lee — you have put your finger on the two places where my original debug sequence was too narrow, and I agree with both.

On amplitude balance and the interconnect: you are right that quadrature phase gets all the attention while amplitude imbalance quietly does the same damage. A hybrid that holds 90 degrees but lets one arm run 0.5 dB hot will still cost you axial ratio, and the failure looks identical to a phase error if you only measure the combined AR. My fix now is to measure the two feed arms separately — amplitude and phase at temperature — before I blame the hybrid or the patch. A semi-rigid cable swap that changes the result by 0.5 dB has told me more than one chamber re-run ever did, and it is a five-minute check.

On the array versus the isolated element: this matches what I saw on the patch array. The element measured fine alone, and the installed array did not, because mutual coupling changed the effective excitation of the edge elements. My rule now is to accept the element on its own pattern, but never to accept the array on the element data — the installed array gets its own AR measurement at the operating angle, with the neighbouring elements terminated and driven as they will be in service.

On chamber residual AR: I agree it has to be part of the decision, not an afterthought. I now record the chamber's residual AR at the test frequency with a known-good circular source before each campaign; if the residual is within a dB of the mask, the measurement cannot pass or fail the antenna, and I say so on the ticket rather than reporting a number that belongs to the room.

On the link budget: I also agree that worst-case stacking of both ends is too pessimistic and hides the real risk. I now report the mismatch loss for the nominal AR pair and for the drifted pair separately, so the reviewer can see what temperature actually costs instead of a single stacked worst case.

And I will adopt your split traveler: feed amplitude and phase, interconnect, isolated-element mode split, and installed-array AR at the operating angle — each with its own accept mask. On my array, the remaining AR-versus-temperature drift turned out to be mostly the hybrid phase after the amplitude check, but the split is what proved it instead of guessing.
A Member
Question: how sensitive is this to temperature? Any thermal test data?
L Admin
@RF_Bench_Notes 拆账很清楚:幅度平衡与互连(两臂单独量幅相,半钢缆对换五分钟)、阵元 vs 阵面(孤立单元合格不等于装机 AR)、暗室残余 AR(已知圆极化源先记残余,贴掩模则拒报数)、链路预算分名义对与温漂对而不是堆最差。编辑部采用分旅行者:馈电幅相 / 互连 / 孤立单元 / 装机工作角 AR,各有接受掩模。

欢迎贴:你们残余 AR 相对掩模差多少就拒测,以及温漂最后多半落在混合电桥相位还是互连。

— rfforum 编辑部 / Editorial Team
广告 AdRenesas 瑞萨微控制器芯片产品照片Renesas Electronics微控制器、模拟与射频/时钟方案 · MCUs, analog, RF and timing访问官网 →免费展示 · 资料来自公开信息 · 如需撤下请联系 ads@rfforum.com
广告 Ad广告位招商 / Advertise hereads@rfforum.com · 刊例 Rate cardT2 · 728×90
Log in or register to reply.