4
\$\begingroup\$

Based on this article, fr-4 can usually be used up to 7 GHz. Above 10 GHz losses become too high and a special high frequency (HF) material should be used (e.g. Rogers).

Advantages of a special material:

  • control of Er over a broad range of frequencies and from batch to batch,
  • lower dielectric losses at HF.

  1. Can anyone please share experience (give a broad guidance) at what frequencies an HF material should be used and when it is a must? Rogers, as a most common, is of a particular interest.
  2. How much is approximately lost due to dielectric losses (leakage) in Fr-4?

Appendix 1 enter image description here
Figure: Comparison of losses FR-4 vs Rogers. Source - the same article.

Appendix 2

My particular situation is a PCB with a 1 GHz square signal (thus, 5 GHz harmonics and higher should be present) and I am contemplating whether to buy an expensive Rogers or not. I can tolerate an insertion loss not greater than 10% (~0.5 dB) per inch within the 1-5 GHz range. The FR-4 the foundry provides does not have specs above 100 MHz - so, I can either use it at my own risk or use Rogers (which has specs in the needed range).

Appendix 3

Answer like "see the datasheet" are not very welcome. I've already done.

\$\endgroup\$
  • 2
    \$\begingroup\$ Yeah, but if we should not point you to the datasheets, what should we point you to? Your question is really one of the kind "I know I can tolerate a \$\tan \delta\$ of \$x\$ at frequency \$f\$", and that is what the datasheets would answer. So I'm not sure what you're asking. \$\endgroup\$ – Marcus Müller Feb 14 '17 at 12:33
  • \$\begingroup\$ @MarcusMüller 10x. I added to the question: "I can tolerate an insertion loss not greater than 10% (~0.5 dB) within the 1-5 GHz range." \$\endgroup\$ – Sergei Gorbikov Feb 14 '17 at 12:38
  • 2
    \$\begingroup\$ So you can actually take that value and look it up in your datasheets. What is your question? \$\endgroup\$ – Marcus Müller Feb 14 '17 at 12:42
  • \$\begingroup\$ I know a guy that uses FR4 for most demanding circuits. But he is not somebody that is doing things "by the way". lea.hamradio.si/~s53mv \$\endgroup\$ – Marko Buršič Feb 14 '17 at 12:43
  • 1
    \$\begingroup\$ I didn't mean to "defeat" you! I really think you've got a good question! I really also think you might want to make clearer what you're actually asking by editing your question. It's still, for me, not very clear what the question is that you can't answer from the datasheet. \$\endgroup\$ – Marcus Müller Feb 14 '17 at 12:56
4
\$\begingroup\$

FR-4 really only means the material is flame retardent; that may seem an oversimplification, but there are many varieties of FR-4 that are specified to 10GHz.

As a couple of examples, see Isola 370HR; this is a good general purpose laminate that I have successfully used with 5Gb/sec signalling rates. Above that, the variation of dielectric constant with frequency starts to make ISI troublesome over any reasonable distance.

For higher speeds, I have used 408HR as it has a very flat dielectric constant across frequency (to say nothing of a much lower loss tangent); there are numerous FR-4 type laminates with quite high performance for the simple reason you cite: Cost. I have used this particular laminate with signals (differential) of 10Gb/sec.

There are numerous suppliers, and all have high speed offerings.

Interestingly, many flexible materials have superior performance in terms of loss tangent.

Exotic materials have their place, but more usually they are found in microwave equipment.

\$\endgroup\$
  • \$\begingroup\$ There are also nowadays a lot of intermediate steps that have better performance than basic FR-4, but lower cost (and better processing characteristics) than Rogers. Tachyon from Isola; Megtron from Panasonic; a bunch of choices from Nelco; ... \$\endgroup\$ – The Photon Feb 14 '17 at 16:53
  • \$\begingroup\$ It's also worth adding that trace length vs bandwidth is a trade-off. If you have 5 Gb/s signals going 30 cm, you might need a better material than if you have 10 Gb/s signals going 5 cm. \$\endgroup\$ – The Photon Feb 14 '17 at 16:55
  • \$\begingroup\$ @ThePhoton 10x for your valuable comment (esp. for mentioning other suppliers). If you, please, copy and paste it as an answer, I'd be able to +1. \$\endgroup\$ – Sergei Gorbikov Feb 21 '17 at 8:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.