Coupler directivity, not coupling, is what fixes your reflected-power accuracy

Started by SI_Practice_Han · 2026-11-17 14:59 UTC · 12 views · 3 replies
S Member
We lost half a day to a reflected-power reading that was wrong by 6 dB, and the cause was directivity, not coupling. Writing it up because the arithmetic is easy and almost nobody checks it before blaming the amplifier.

The setup: 20 dB coupler, reflected port into a power sensor, calibrating a 3:1 load. Coupling was fine, insertion loss was fine, and the reading said the load was better than 20 dB return loss when it was actually about 4 dB.

The mechanism: a coupler with 20 dB directivity leaks forward power into the reverse port at -20 dB relative to the forward wave. When the true reflected wave is -4 dB (return loss 4 dB), the leakage is 16 dB below it, which is a small vector error. When the true reflection is -20 dB (return loss 20 dB, VSWR about 1.2), the leakage is the same amplitude as the signal, and the sum can be anything between -26 dB and -14 dB depending on relative phase. That is exactly the "6 dB error" we saw: the reading was dominated by the leakage and had nothing to do with the load.

The numbers worth carrying in your head: if you want plus/minus 1 dB accuracy on return loss, the leakage must sit at least 15 dB below the reflected signal, so you need directivity about 15 dB better than the return loss you are trying to measure. A 20 dB directivity coupler therefore measures down to about return loss 20 dB before it starts lying, and 30 dB is where it becomes a real instrument.

Four things I do now, in order:

1. Measure directivity itself, don't trust the datasheet. Terminate the coupled port and the output port with precision loads, drive the input, and read the reverse port. Whatever you see is directivity plus the residual of your loads. If it reads -22 dB and the datasheet says -25 dB, the difference is usually your load, not the coupler.

2. Check the directivity at the frequency you care about, not at the band centre. Directivity usually falls at the band edges, and the fall is often 6 to 10 dB from centre to edge on a broadband coupler.

3. When the load is good, stop using the coupler. Put a reflectometer or a bridge in front of it, or reverse the coupler and measure the reflected wave as the forward wave of the reversed path - the directivity now works in your favour because the wanted signal is 20 dB stronger than the leakage.

4. Log both the forward and reflected readings and compute the ratio, so that source level drift cancels. Our error survived three technicians because each of them read the reflected power alone.

One more trap: thermal drift of the coupler moves its directivity. A coupler sitting at 50 C after an hour of key-down can lose 2 to 3 dB of directivity, which is again a silent 1 dB error on a good load. If you are making a pass/fail decision at 20 dB return loss, keep the coupler away from the heat sink and measure directivity warm.

Discussion: how do you handle directivity correction on a production line, where you cannot afford a full reflectometer per station? Do you correct with a table measured at one frequency, or do you simply keep the DUT return loss 10 dB above the coupler's directivity and accept the loss of coverage?
R Member
We've hit this exact pitfall. Ended up switching vendors - sometimes it's a supply issue, not a design issue.
L Admin
Thank you, SI_Practice_Han — the 6 dB reflected-power error that was directivity, not coupling, is exactly the kind of "the amplifier is fine, the coupler was lying" story this board should keep.

Editorial keepers (soft promo = 0):
1. For about ±1 dB accuracy on return loss, leakage needs to sit ~15 dB below the reflected signal, so directivity must be about 15 dB better than the return loss you are trying to measure. A 20 dB-directivity coupler is already in trouble near 20 dB RL.
2. Measure directivity at the frequency you care about (band edges often drop 6-10 dB), with precision loads, and again after the coupler is warm. Log forward and reflected together so source drift cancels.
3. When the DUT is a good load, stop trusting the coupler alone — a bridge, a reflectometer, or a reversed coupler puts the wanted wave above the leakage.

Open: on a production station, do you correct directivity with a one-frequency table, or do you simply keep the DUT return-loss limit ~10 dB above the coupler's directivity?

— rfforum 编辑部 / Editorial Team
S Member
A 20 dB directivity on a 20 dB coupler is close to worst case, because the leakage term lands at the same order as the coupled signal when the reflection is high. A sanity check that costs nothing: terminate the coupled port, measure reverse isolation, and compare with the datasheet. If measured isolation is about 20 dB better than coupling, directivity is fine. If it is only 5 or 6 dB better, the reading is leakage dominated. After that experience we started subtracting the directivity error term in software below 10 dB return loss, which matched the VNA within 0.5 dB.
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.