Uni-Directional RF Attenuators - Input Port?

[killingtime]

Hi,

I've been looking at 5W/10W 30dB RF attenuators for use as an input to a spectrum analyzer. 3GHz.

Plenty of no-name brands on Ali and Eby in 30-40 USD range, but they don't get good reviews around power handling.

I assumed they were all bi-directional, but that's not the case. Those that use dedicated chip attenuation resistor networks have a specified direction for the power. Some sellers define this, others don't. Attached.

I'm wondering if people are using these the wrong way round, and that's the reason they're burning out.

It could also be that they're just badly made. Picture attached from an Amazon review.

Looking further afield to named brands like Bird, I see they offer bi-directional attenuators, but they cost many 100's. Bird attenuators make a new analyzer look cheap.

https://birdrf.com/Products/Test%20and%20Measurement/Attenuators/5-watts/5-6A-Series_6-GHz-Bi-Directional-Attenuators.aspx

So this leads to the question, how do you determine which end is the input on a uni-directional attenuator?

Both ends should measure 50 Ohms (if that's the nominal impedance of your system).

Thanks.

EDIT

This Weinschel attenuator is rated for 250W forward and 50W in reverse, but the input port is defined on the spec sheet.

https://www.rfmw.com/datasheets/weinschel/45xx58xx.pdf

[TimFox]

If the attenuator is competently designed, one would assume that the heatsinking is concentrated where the maximum dissipation occurs, but it's not clear from the Weinschel photo.
I don't see a model with that photo in the Weinschel link you posted.
This reminds me of another hypothetical question:
It is well known that a Tee connection has an equivalent Delta connection for three resistors, and you can't determine the internal connections in a black box from ohmmeter measurements.
However, if you slowly increase the voltage across one pair of connections until you get a discontinuous change in current and possibly smoke, then measure the resistances after disconnecting the voltage source, you can tell if the late, lamented box were Tee or Delta.

[killingtime]

However, if you slowly increase the voltage across one pair of connections until you get a discontinuous change in current and possibly smoke....

Ahhh, destructive testing. The one test to rule them all. Yes, all attenuators will look electrically identical from both ends, but I would imagine it's the power rating of the resistor network that makes them uni-directional.

Well, I posted a few questions and now have some replies.

Asked on Amazon (they sell attenuators on there) and was told by an RF 'Engineer' that all attenuators are bi-directional....I'm guessing here that he's only ever used decent kit, and the company he worked for never ordered from Eby or Ali. Which would explain the answer.

Went hunting on Ali, and found a seller that claims to be a manufacturer - RFteleworld. To my surprise, they understood my question and offered a coherent reply. Most of their (cheaper) 3G/4G stuff in the 2-50W range is uni-directional (male input and female output), but the 6G stuff in the 2-50W range is bi-directional. The 6G offerings are about 2x the price of the 3G, but still 10x less than anything branded like Bird.

Kind of important to know because most analyzers have a female panel mounted N port. If you hang an attenuator of this connector as I've seen many people do, then the uni-dir will be the wrong way round, and will burn out at rated value. Incidentally, none of this directional info is posted on the sales pages on Ali. You have to ask.

I've put an order in. Lets see what arrives.

[DavidAlfa]

Low power ones are usually bidireccional, but for high wattage they have a specific array configuration where initial elements are designed for more power than the last ones, so plugging it backwards will blow it.
Happened to me once, put 52dBm to the wrong side, lived for few seconds and suddenly VSWR skyrocketed, the attenuator became open-circuit.

[rfclown]

I stay clear of no-name attenuators. I buy used from eBay (or wherever) with names like Narda, Weinschel, Aeroflex, Bird, Mini-Circuits, etc. If they are marked "bi-directional" I pay attention to the marking. If not, I assume bi-directional.

[radiolistener]

I assumed they were all bi-directional, but that's not the case.

No, they are one direction.

I'm wondering if people are using these the wrong way round, and that's the reason they're burning out.

yes, I burned several attenuators just because didn't know that they are one direction and seller didn't notice it. Often sellers even don't know that they are one direction.


Burned 40 dB attenuator turns into ~42 dB attenuator with about 66.9 Ω on its output (instead of expected 50 Ω).

The input is still 50 Ω, because 40 dB attenuator load almost doens't affect it's input impedance, so I found that burned attenuator can be used as a nice dummy load

So this leads to the question, how do you determine which end is the input on a uni-directional attenuator?

The high power input connector is usually SMA-male, which is also often marked with "SU" letters on the connector.

On your second picture, the left connector is probably a high power input, at least all my Chinese attenuators have SMA-male as a high power input.

[max-bit]

Some attenuators, especially the higher power ones, are directional and the input and output are clearly marked.
This is due to the fact that the first stages of the attenuator have very low attenuation levels so that they do not emit great power