Hi am designing a Wilkinson power divider on a FR4 pcb. I need to know an accurate loss tangent for this material to use on the simulation.

 

I found the following links,

https://www.mgchemicals.com/downloads/tds/tds-500-series.pdf
https://www.farnell.com/datasheets/1644697.pdf

The first one gives a dissipation factor of 0.015 at 1GHz and the second gives a value of 0.014 at 1 MHz. It seems that the two are in agreement so I would use 0.015 as the loss tangent. This is a high loss tangent, you can expect a dielectric loss of about 1.8 dB for a quarter wavelength TEM transmission line at 1 GHz.

What frequency are you designing your Wilkinson at? FR4 is not a good dielectric for high frequencies. The loss tangent becomes nonlinear with high frequencies and the fiberglass weave will cause issues for impedance control. Not to mention the coefficient of thermal expansion is large (if you care about performance over temperature). Personally, I would not use this material over 1GHz.

 

Hi! Thanks for the reply! Am working at 2.16Ghz, I know FR4 it's not suitable for this frecuency but am limited to it by the cost and availability in my country. By tunningthe parameters with HFSS I have 0.4db of insertion loss, but the width calculated for the impedance of the microstrip it varies 0.4 mm (0.015 inches) from the simulated, and this depends of the loss tangent. Do you know if this results are good to use? Thanks a lot!

 

Loss tangent for FR4 is not tightly controlled - in fact, it isn't even loosely controlled. About all you can do is to take an average of all the loss tangents from all the various vendors and design with that.

If your design depends on knowing the loss tangent, then you are just wasting your time, because you will never know it.

 

It is hard to tell without any simulation data or a layer stack-up, could you provide some more information? A width of 0.4mm is not unheard of for producing a matched line. and 0.4 dB loss is reasonable though this will vary once fabricated. In my limited experience I have seen 10 mil to 30 mil being used for transmission lines.

The loss tangent should not be effecting your impedance which means that the lossless and low-loss line assumptions are violated. And as SLK001 mentioned it will be time consuming to design when it is dependent on the loss tangent.

Generally to avoid this issue one would use a less lossy substrate but since that is not possible in your case, look into how sensitive your design is to the loss tangent. If it is acceptable to meet your design criteria then it may be okay to proceed. One might expect the loss tangent to vary from 0.015 to 0.025. I would put more margin since the materials are not usually characterized past 1 GHz and it varies with frequency.

If it is in your budget, you could order a few test boards and characterize them to find suitable geometry and design values.

Based on the material alone your design will not be optimized, and you may not be able to get a very good design. What requirements are you trying to meet? What manufacturer are you using (some give out more information than others) to fabricate the PCB?

 

Thank you what you are telling me is very useful! It is hard to explain the result without showing you the simulation. The Wilkinson divider is part of a phase shiffter for a smart antenna, all made in FR4. Since the divider provides a 3db attenuation on each output my criteria is having less than 0.5db. I made two designs, one straight power divider and one circular.
For the straight one I obtain 3.048db on the outputs , and for the circular 3.46db.
Both have a 1.748 mm width for the lambda/4 TL (zo*√2) and the effective lambda/4 length is 19.48mm.
Adjusting the circular design the best I get is that 3.48db and aprox -38db for isolation between ports and -29db for the input reflection coefficient. But the width of the lambda/4 ends up in aprox 1.2 mm, and the length in 17mm.
Am not sure if those variation are ok.

 

For GHz range operations the easy solution is Rogers Duroid(TM) TC350™ with 2.0 pF/cm2 at the price of $115 for 18”x12” plate and a loss factor of tgδ=0.0015 (Q=667). To ensure better natural cooling, copper cladding is produced with matted surface, thus larger surface area was obtained and better cooling, but on the other hand Focault effect was worsened, thus the heat production increased. Duroid-6035HTC (high temperature material, with 2.2 pF/cm2 and tgδ=0.0013 (Q=770). The cladding was glossy, but still costly - $240 for the same size. Duroid-5880® material with 1.7 pF/cm2 and tgδ=0.0004 (Q=2500). The cladding was also glossy but its cost reached $325.

 

RE: jrLasota - if You are living in the World where pre-last century Postal services are not available from abroad, thus to buy anything from Rogers is plainly impossible, may try to use the high quality teflon plate and squeeze the copper foil between two side-planks made of the same teflon. Rogers grand science was to invent how to attach the copper to teflon, as it normally isnt possible because of record-low adhesion.

PS, actually, which is this land not attached to postal services system??? Just for to know.
PPS Dont feel insulted, I myself was forcefully lived a certain time in country like that, so I know well what means arrest for trial to sent a letter abroad and what means a "wooden rouble" with official course 1rouble=1dollar whilst practical course is 5000roubles=1dollar. And I remember well the faces of executed to death people whose only sin was they owned a one to five dollars (namely - state unfriendly "valuta" operations). However I realize the similar orders today may happen only in Iran, N-Korea and that seems all. Everywhere elsewhere a guy push his bank card data into puchase website and then after month (or somewhere a day) waits when postman will whistle him out to undersign the parcel.

 

PPPS - actually, isnt it cheaper in non-mass production cases to organize the divider with means of stub of coaxial cable instead of pcb line? 2,4 GHz is still in the senses of most the coaxials, and sub-mm accuracy is still good enough for practical results.

 

RE: jrLasota - if You are living in the World where pre-last century Postal services are not available from abroad, thus to buy anything from Rogers is plainly impossible, may try to use the high quality teflon plate and squeeze the copper foil between two side-planks made of the same teflon. Rogers grand science was to invent how to attach the copper to teflon, as it normally isnt possible because of record-low adhesion.

PS, actually, which is this land not attached to postal services system??? Just for to know.
PPS Dont feel insulted, I myself was forcefully lived a certain time in country like that, so I know well what means arrest for trial to sent a letter abroad and what means a "wooden rouble" with official course 1rouble=1dollar whilst practical course is 5000roubles=1dollar. And I remember well the faces of executed to death people whose only sin was they owned a one to five dollars (namely - state unfriendly "valuta" operations). However I realize the similar orders today may happen only in Iran, N-Korea and that seems all. Everywhere elsewhere a guy push his bank card data into puchase website and then after month (or somewhere a day) waits when postman will whistle him out to undersign the parcel.

Hi! Thanks for the data. Am not offended at all haha. Am from Argentina, you can look up us dollars vs Argentinian Pesos and you will see my problem haha, and it really is a pain in the a** the troubles you need to go through when you buy from outside .

 

Aa-ow, I realize. Tomorrow the Argentinian she-scientist will come to us to Riga city for some years to work shoulder to shoulder. She was only which said 5000 Eur salary is not small money. All other candidates was willing more.
So... You have an option of squeezed three layer teflon platelets and copper film or coaxial precision cut-offs. By the way, probably the CAT-5 style winkeled wire pair as well, but with that I have experience under 1 MHz. Works brilliantly. Hmmm, but in case of mass product may meditae about quartz, sapphyre or (probably) just the glass platelets used to vacuum sputtering the copper layer. Actually, not sputter but thermal vacuum deposition. The vacuum must be rather near to 10^-7 or better (Torr) and cuivette of boat form must be filled with high purity of electrotechnical purity copper. Wolframium spiral must heat it up with few volts and thousand of amps until 1200 C about, and when it evaporates then keep it few seconds and job is done. Places where copper must not be deposited may be secured by means of Positivus-20 aerosol of etched by ordinary Ferrum Chloride afterwards. If to do the tiny drilling into the ends then as mechanical contact as even soldering with low temperature metal is possible, to fix the cabling.

 

1) Lit a bit more of I explained - tan(delta) may be measured qualitatively just putting that chosen material (without copper layer) in microwave cooker. If warms - bad. If not warms - good. If warms strongly - dangerous.
2) About vacuum depositing tech - the platelet before vacuuming MUST be very well degreased - allways when I do this job I love the amylacetate for it`s strong effect. But afterward anyway the ethanol 99,3% is mandatory, and those 0,7% must never contain any oils. Sorry to tell, the most alcohol recordist country of World the ussr once from all hundreds or thousands of spiritus plants only one was capable to produce a clean enough ethanol. And even those not in all stock. Thus, let check a quality by ultraclean glass platelet, drop a few droplets, wait whilst its completely dried and look in reflected light - if see anything, spiritus is not clean enough for the task. But simple destillation will sweep off the impurities, thus dont be fooled that distillation of 99,3% is sensless. It always gives a good result. Without a good degreasing that copper will make an unsatisfactory adhesion to the glass.

 

Got simple table of tan(delta) for different plastics at 1 MHz
Non-polar types: polystirol 2E-4; PE 3E-4; teflon-4 1...2E-4; PP 2...3E-4, quartz 0.6E-4 at 3 GHz; glass No7059=36E-4 at 10 GHz
Polar types: polycarbonate 100E-4; PET 250E-4; cellulose triacetate 30E-4
paper impregnated with PE 30E-4; polyethylene teraftalate (=lavsan) 31E-4; polyimide 30-34E-4
These was old russian book found data. Thus, for a while the best even imaginable material is fused quartz.
Better data may found at http://www.rfcafe.com/references/electrical/dielectric-constants-strengths.htm