Calculating FR4 permittivity for another frequency

[lindworm]

Hi guys,

I've designed a pcb for an 868 MHz RF project and need the permittivity (aka dielectric constant) for the frequency I'm using it at.
The datasheets of the FR4 manufacturers usually only contain the permittivity at 1MHz, is there any way to calculate the permittivity for another frequency?

If you have a rule of thumb instead of a formula thats fine too.

Kind regards

lindworm

[free_electron]

Thats like saying i want to calculate the speed of my car on gravel, instead of on asphalt...

First of all : what car ? There is no such thing as FR4 ... FR4 is a class of fire retardant materials. There are hundreds of different FR4 substrates. Even Rogers, Nelco or Isola RF substrates are FR4...

You need to know specifically what your board house is using and you need to know the prepreg thickness and layer stack. Once you have that info you can dig up the data for what you are going to build.

[lindworm]

Hi free_electron,

I do have the data I need (it's SHANGHAI NANYA SN-L4 FR4 NY1140) and I know that FR4 differs a lot between manufacturers, products and even batches.
My problem is that I know the specific Material used and it's dielectric constant at 1 MHz.
However, I couldn't find a way to figure out the dieelectric constant in the frequency range I design for.

Some datasheets for other base materials have three or four values, depending on the frequency, but most datasheets only list the permittivity at 1 MHz.

Kind regards

lindworm

[free_electron]

Ouch, then it becomes difficult... Call them ?
Or, have a proto run done with a coupon for impedance co trol and test it

[andersendr]

Have you looked at the Wiki?  http://en.wikipedia.org/wiki/FR-4  It looks as if it does not change a hole lot with frequency.  The 1GHz is listed.

[JorgeCarbajal]

4.25 is good to 1GHz

[free_electron]

No. not correct. Fr4 materials can range from 2.5 to well over 6. you need to talk to the manufacturer..
FR4 tells you nothing. itis purely an indicator for flame resistance. has NOTHING todo with dielectric constant.
Even rogers or Nelco materials are classiefied as FR-4 and those are specialRF materials with exceptional behavior.

The pre-preg or laminate manufacturer needs to give you the numbers.

there is other things at play : the weave density of the material , the fiberglass used ( there are two kindo fiberglass each having different properties ).  If you are desigining something in the RF field like a stripline or a tuned antena on pcb you need to know what you are dealing with.

now, there is alo going to be an impact depeding on the lamination pressure and temperature. for 800MHz it may not be a real problem , but scoot into the 2Ghz domain and you got trouble...

[lindworm]

Thanks for all your answers, luckily one of the manufacturers I contacted was able to give me a guesstimate based on their experience that their specific material (with 4.6 at 1MHz) will have a permittivity of 4.2 to 4.4 at 1 GHz.

So I'll design it with 4.3 and cross my fingers.

I also talked to my professor who said it's not extremely important at 1 GHz and that designing for over 6 or 7 GHz is where the real fun starts ;-)

He also told me that we could still modify the pcb a little with a CNC and cut some copper off or match the antenna to the PCB trace, as we have all the necessary instruments at my uni.

Thank you all for your answers!

lindworm

[ejeffrey]

I also talked to my professor who said it's not extremely important at 1 GHz and that designing for over 6 or 7 GHz is where the real fun starts ;-)

It depends a lot on your application.  If you are making the front end of a spectrum analyzer and you want no ripple from DC to 12 GHz and you care about ripple, dispersion, and board-to-boar repeatability then you care a lot.  I used some ECL chips and a mini circuits balun to build a ~4 GHz divide by 64 counter to make an optical PLL and it worked fine on a hand etched board that I layed out in about 30 minutes and printed on some cheap pre-sensitized copperclad board with an unknown dielectric constant.  It works great -- but the input signal is single frequency and I don't care about the amplitude since the divide-by-N only measures level crossings.