Continue to Site

# Return Loss Measurement Basics

Status
Not open for further replies.

#### cmd

##### Member level 2
Hi,

The formula of return loss is given everywhere RL=10*log(Pi/Pr) and RL=-20*log(abs(reflection coef))
This formulas gives positive values about return loss,however the measurement devices give negative values they calculate RL=10*log(Pr/Pi) and RL=20*log(abs(reflection coef))
The question i have is which one is the general description of return loss? In which ares we use positive and negative description?
How calculate cable and antenna analyzers or network analyzers the return loss?
Do they use Pi and Pr values or s11 parameter?
An another question is about frequency domain measurement;
In application notes it is written that, for cable loss measurements the cable far end should be short circuited. My question is what will be if we terminate the far end with open circuit? Does it change for cable loss measurement?
Also it is written that, for distance to fault measurement the far end should we terminate with load. If i terminate the far end with open or short to see also the end of the cable too? Does it affect my distance to fault measurement acuracy?

Last edited:

The title says it all. A return LOSS is normally a positive number. If half the power reflects back, it is a 3 dB LOSS. If you return loss was -3 dB, it would imply a gain.

cmd

### cmd

Points: 2
Hi biff44,
Thank you for your reply. I know return loss is positive, but cable and antenna analyzers describe it with negative numbers. I want to know why is the description changed?
For example i uploaded my last return loss measurement and its distance to fault traces on a 25 m feeder cable with open circuited end

An another question is return loss value at the open circuited end is -3.59dB . In theory it should be 0 dB. Does this 3.59dB value describe the cable loss in 50 meter or?

Hi,
I made a distance to fault measurement on a bent feeder cable at different frequency ranges. First test was made at 700MHz-1GHz and the second was at 1700MHz-2GHz. The return loss values at high frequency range measured less than first measurement. But i think because of cable loss return loss values at 1700MHz-2GHz shoulde be less than first measurement. What is the point i didn't get?
First trace is at 1700MHz-2GHz and the second at 700MHz-1GHz ;

Even if RL should be, by definition, a positive number, sometimes it is expressed as a negative number (even in IEEE articles). I don't know the reason.

If you have a cable having "X" dB losses, its return loss will be 2*X dB (because you have to take into account a round-trip of the signal). You will have 0 dB return loss from an open-end (or shorted-end) cable if it is lossless (ideal).

Hi albbg,
Thank you for your reply. Did you look the last DTF traces i uploaded. First peak at 19.01m is a bend fault and the second peak at 24.88m is the end of the feeder cable. I think the measured RL values at 1700-2000MHz should be less than values at 700-1000MHz because cable loss is higher at higher frequencies. But traces says opposite of it. I didn't get this?

I think the measured RL values at 1700-2000MHz should be less than values at 700-1000MHz because cable loss is higher at higher frequencies.
It's a computed time domain plot. Without knowing the exact instrument specification and actual measurement and transformation parameters, it's likely to jump into conclusions. There's also indicated CAL:ON. What does calibration involve in this case?

cmd

### cmd

Points: 2
Hi FvM,

You are right, i should give some more details. I made this measurements with Anritsu Site Master S331D cable and antenna analyzer. I have a 25m long 1/2" feeder cable for my experimental purpose. At first i made some measurements at 700-1000MHz range. I calibrated the device with its original OSL kit.
In first experiment i left the far end of the cable open and measure the length of cable. The result was -3.59dB at 24.88m. The point i got here 3.59dB is the cable attenuation value for 50 m, because normally it should be 0 db with a open end. The cable attenuation is 3.59/50 = 0.0718dB per meter
Then i bent the cable harshly at 19. meter, terminated the end with a 50 ohm RF terminator and make distance to fault measurement. The fault was located with the device at 19.01 meter. Then i set the frequency range to 1700-2000MHz to analyze the effenct and difference of frequency level to the measurement. I calibrated the device for the new measuremnt and made dtf at the new frequency stage. The resulted traces are what i upluoaded.
I expected less Return loss values but the situation is the opposite of it. What do you think? I s there a problem with the measurement accuracy or do i think in wrong direction?

The question i have is which one is the general description of return loss? In which ares we use positive and negative description?
How calculate cable and antenna analyzers or network analyzers the return loss?
Return loss is ratio of total and reflected power and p is Pr/Pi. p is a positive value 0-1. It can not be above or below these values due to physical limitations. Expressed as VSWR is it a value 1 and up and expressed as dB is it less then 0 dB as less then 0 dB is less then 1. It results in dB values with a minus sign but it is still a positive amount of power that reflects back. If calculating in dB must also mathematical dB rules apply. A few exampels:
0 dB =1
-10 dB = +0.1
RL [dB] = 10 log (PI/PR) = PI[dBm]-PR[dBm]

I would expect a cable bend to give a higher reflection factor at higher frequencies. So it may compensate the attenuation.

Regarding positive or negative dB values, who said, that the negative values are actually designating return loss? Reflection factor |s11| is a negative dB number.

cmd

### cmd

Points: 2

You say that at higher frequencies returns more power from a bend fault,so it can compansate the attenuation at higher frequencies. That's why the return loss values at higher frequencies were measured more then lower freq. Did i understand right? But i think that reflection coefficient is independent from frequency.
It depends on impedance mismatch. Am i wrong?

Reflection factor as definition is not frequency related. A impedance mismatch can have frequency depending elements but it is no rule.
A common effect of a bent cable is that inner conductor is moved sideways and/or cable area is reduced which both increases parallel capacitive losses. This capacitance results in a frequency depending mismatch. A better analyze of its frequency related behavior do probably need a rather good definition of the actual cable bend and the whole cable. Error sources is for example external shield RF current that travels in both directions but not need to reacts in same way for the bend or changes in dielectric behavior in the bent area. A bend can also cause a opening in a form of a slits is created in outer shield which can create fun effects both inside and outside of the bent cable shield. If the bend causes internal shortcut in a certain point, is it likely that low frequencies have higher transmission loss.
It is common that higher frequencies reacts more for a certain amount of mechanical match/mismatch as relative size of this impedance element (in this case a deformation on a cable) compared to signal wavelength is bigger. It is not a rule but it is often so.

Hi E Kafeman,

Thank you for your reply! I uploaded the bent cable used in distance to fault measurement. I measured at the deformed point a return loss -9.92dB at 700-1000MHZ and -8.48dB at 1700-2000MHz. Do you think its normal? In which frequency range from these should be expected that return loss value is bigger?

I generally agree with Kafeman's explanation. In addition, you should consider, that higher frequencies give a better spatial resolution in cable test. As a result, small disturbances give a higher response. But I don't think that the behaviour can be predicted precisely.

My reflection is that no RF engineer would consider use a semi rigid coax cable with that bend or allow it to become that bent as it now is destroyed.
From my view are your measured values very believable but that opinion is not worth much. From the photo, seems a RL something in the span -2dB - -20dB as likely but a wider range is also accepted.
As FvM writes, a better prediction is not possible without a detailed study such as analyzing material stress and how bad inner conductor is misplaced in the bend. This analyze can be done with aid of X-ray or similar tools. But why try to do a complicated prediction within less error then 1 dB about something when you have a measurement tool? It is not worth that job on a cable that anyway is useless.

I can well imagine similar discontinuties in typical cable test applications. Consider e.g. a TV network cable that may have been damaged when pulling it into a pre installed conduit. These cables aren't installed by qualified RF engineers.

E Kafeman, I have a graduation project about fault location on coaxial cables. For the experimental study i needed create a fault on the coaxial cable. That's why i bent the cable and destroyed it. I should show that the device can locate RF faults on a feeder cable.
The measurement i made with different freq. ranges is because i wondered that what will be change if i made the same measurement with same frequency interval (300MHz) but at higher frequency range. i.e. Spatial resolution, rl values what would be affected
I know like FvM said, higher frequencies give a better spatial resolution. As you said the rl values are believable for a bent cable.The thing made be currios is with the high range measurement (1.7-2GHz) all return loss vaules in the trace became higher? The thing is that i want to know it.
The blue trace is at 0.7-1GHz
The green trace is at 1.7-2GHz

#### Attachments

• comparison.JPG
84.2 KB · Views: 87
Last edited:

You can try to connect defined discontinuities, either lumped elements, short stubs (e.g. a T-connector) or segments with known different characteristic impedance for test. Also frequency independant elements (e.g. resistors) can be used to calibrate the instrument.

I can well imagine similar discontinuties in typical cable test applications. Consider e.g. a TV network cable that may have been damaged when pulling it into a pre installed conduit. These cables aren't installed by qualified RF engineers.
Correct, cables get broken for a lot of reasons but when the fault is found with aid of suitable instruments, is further guesses and material inspection and analyzing normally not of interest. Nothing without exception of course. At aircraft accidents and similar can a deeper analyze be of interest of faulty cables to try to find if the problem existed before the accident.

---------- Post added at 18:03 ---------- Previous post was at 16:53 ----------

CMD, as you sure understand, no two bends and for different types of coaxial cables will be identical and bends is only a fraction of types of faults in a coaxial cable. As I described, frequency behavior is very depending what part of the the cable that have been deformed and type of internal deformation. Guesses what is most likely to happen in these two frequency bands and with variations of a few dB can not be very useful information and worth nothing even in a graduation project. Long time ago did I use reflectometer for localization of faults in twisted unshielded wires buried in ground (phone cables). It worked up to 10 km and with precision of around 10 meter. About same precision as I did got for optical wires but that kind of technology is hardly new. Modern faultfinding technology is often of type signature test and precise checking of long term variations. Frequency behavior is then a part of the information. With aid of these methods can faults such as crushed by a car, eaten by a rat, fire, over voltage, water leakage and similar things be easy differentiated. Very similar technology is also used for distance controlling healthy status of switches and transformers by listening at high voltage power lines with listening equipment maybe up to 100 km away. Several types of problem can be predicted and repaired before it have caused any serious error. See this as ideas if you want to extend your report with how to find and characterize errors in better details then partly bent, short & open. A easy alternative or complement can be to start to search for full and half wavelength resonance at 30 meter and from that calculate amount of impedance error at the point of interest.

cmd

### cmd

Points: 2
E Kafeman,
Thank you for your helpfully reply. My graduation project is mainly about time domain reflectometry and frequency domain reflectometry methods for fault locating on RF and Microwave coaxial cables. I made some experiments about TDR with RG6 and RG58 coaxial cables and about FDR with a 1/2" feeder cable as you had seen before.
TDR is easy to understand but in frequency domain it is a little bit complicated. Network analyzers should bu used. Magnitude information is not enough. It should be measured also phase information to get time domain. iFFT algorithm is used to transform the data to time and then to distance.
I have also some problems to understand about phase measurement. The electromagnetic sinusoidal waves sourced to the cable and reflected from discontinuites to the receiver at the beginning. The round trip give a phase delay and also discontinuites change the phase of signal. How the time info calculated from this phase info.

Status
Not open for further replies.