Dear Neo, Hendrik Bode, in his seminal text, "Netowrk Analysis and Feedback Amplifier Design", D. Van Nostrand Co., 1945, goes into a long discussion of the relations between the real and imaginary parts of any causal linear network function (complex electric permittivity is an example of such a function). He devotes a whole chapter to the general relations, and then another chapter on the "Graphical Computation of Relations Between Real and Imaginary Components of Network Functions" -- in other words, he teaches you how to just LOOK at a curve of lass tangent and predict the shape of the dielectric loss curve. I included some parts of that discussion in my book, "High-Speed Signal Propagation: Advanced Black Magic", pp. 105-115. The most important, and practically useful, part of the discussion I think appears on page 108, equation 2.88, where I point out that for a low-loss material, a constant loss tangent equal to "tan theta" (that's the form of worst-case specification for most dielectric materials) implies an exponential decay in the dielectric constant, with a *very tiny* exponential decay constant equal to omega (frequency) raised to the power of (-2)(theta/pi). Example: Let the loss tangent be one percent: "tan theta" = 0.01 For small angles, working in units of radians, theta approximately equals "tan theta", so we have theta = 0.01 Assume you have Er = 4.00 at 1 GHz. Suppose that you want to know Er at 5 GHz. If the loss tangent is flat over a broad range extending well to either side of both 1 GHz and 5 GHz, then the dielectric constant at 5 GHz must be de-rated according to the rule: Er[5 GHz] = Er[1 GHz] * (5GHz/1GHz)^^(-2)(theta/pi) [EQ 2.88] (where the symbol "^^" means, "raised to the power of" Evaluating the term (5GHz/1GHz)^^(-2)(theta/pi) = (5)^^(-2)(0.01/3.1415926) = (5)^^(-0.0063662) = 0.9898 Multiplying that term times Er[1 GHz] gives: Er[1 GHz] * 0.9898 = (4.00)*0.9898 = 3.959 Why this works is explained in Bode's book (if you read the whole thing) in a fairly simple way. That was his genius. Bode points out, and this may be familiar to you if you are a student of electrical engineering, that flat, fixed phases are associated with each of the following network functions: |H1(f)| = f -- phase equals +90 degrees (a differentiator) |H2(f)| = 1 -- phase equals 0 degrees (flat function - a constant) |H3(f)| = 1/f -- phase equals -90 degree (an integrator) In each case, the slope of the function (power of f) is proportional to the phase: |H1(f)| = f^^1 -- phase equals +90 degrees (a differentiator) |H2(f)| = f^^0 -- phase equals 0 degrees (flat function - a constant) |H3(f)| = f^^-1 -- phase equals -90 degree (an integrator) ---------------------------- If you are familiar with the skin effect, you know that it works like this: |Hskin(f)| = f^^(-1/2) -- phase equals -45 degree (an integrator) Note that the skin-effect relation seems to fit "in between" H2 and H3 in my table. ---------------------------- Bode goes on to show how you can generalize his relation for any rational power of frequency, so that: |f^^a| must have a phase in degrees of (90 degrees times a) In your case, the "loss tan" for a low-loss material represents an angle given in units of radians. If we modify Bode's rule for radians we get this: |f^^a| must have a phase in radians of (pi/2 times a) If you supply the angle theta = (pi/2 times a), then the constant "a" must equal (2/pi)*theta, which is where I got the rule for [EQ 2.88]. Rule [EQ 2.88] works we as an approximation for any low-loss material that has a fairly flat loss tangent. From a simple specifiation of dielectric constant and loss tangent at one frequency, you may use this "constant loss tangent assumption" to evaluate the dielectric constant at every frequency, in a way that is guaranteed to produce a causal network function. I hope this brief note helps. Let me know if I've just made the subject more, rather than less, confusing. Best regards, Dr. Howard Johnson, Signal Consulting Inc., tel +1 509-997-0505, howie03@xxxxxxxxxx www.sigcon.com -- High-Speed Digital Design seminars, publications and films (I'll be teaching in Portland May 3-6th) -----Original Message----- From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] On Behalf Of Neo Sent: Tuesday, March 30, 2010 7:24 PM To: List Si Subject: [SI-LIST] Relationship between loss tangent and dielectric constant's frequency dependence Hi, This email's title is a bit long but it is exactly a question annoying me. I'm trying to find out whether there is a proven relationship between a dielectric material's real part and imaginary part. Imaginary part is dielectric loss tangent. And real part is the dielectric constant. For a material like FR4 or Rogers, their dielectric constants all change over frequency. And their loss tangents are also different. Is there any inner relationship between the value of loss tangent and how the dielectric constant (real part) changes over frequency? Thanks,Neo ------------------------------------------------------------------ To unsubscribe from si-list: si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field or to administer your membership from a web page, go to: //www.freelists.org/webpage/si-list For help: si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field List technical documents are available at: http://www.si-list.net List archives are viewable at: //www.freelists.org/archives/si-list Old (prior to June 6, 2001) list archives are viewable at: http://www.qsl.net/wb6tpu ------------------------------------------------------------------ To unsubscribe from si-list: si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field or to administer your membership from a web page, go to: //www.freelists.org/webpage/si-list For help: si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field List technical documents are available at: http://www.si-list.net List archives are viewable at: //www.freelists.org/archives/si-list Old (prior to June 6, 2001) list archives are viewable at: http://www.qsl.net/wb6tpu