Calculating Er at other frequencies

the pcb material manufactures provide a permittivity value at one specific frequency.

If a bus on my PCB works on another frequency/data rate, then what Er should I use for impedance calculation? (for example my bus is 400MHz, the provided info is valid on 1MHz)

On a bus, there is no sine wave, and the "frequency" also changes depending on the data pattern. So what is the frequency of my bus, to select the Er to calculate the impedance?

Re: Er @ other frequency

For a data signal, it isn't the clock frequency or pulse repitition rate that matters - it is the edge rate. The data in a digital signal is all in the pulse edges. The pulse rate only determines how soon the edges get sent. Rise and fall times determine the true bandwidth of your data.

If you have exceptionally fast pulse edges, you should be using the Er measured at 1GHz or above. However, the variation in Er from 1MHz to 1Ghz probably is going to cause less impact on your controlled impedance than other manufacturing variables. Remember that the fab is only able to etch the trace width to +/- 5-10% accuracy at best. Most fabs will only guarantee controlled impedance to +/- 10%. The roughness of the trace edges, undercutting while etching, and variations in dielectric thickness from weave and lamination pressures and temperatures are all going to affect the finished impedance.

The bottom line is that the 1MHz Er value is probably good enough for the impedance calculations in your application. If the material you are going to use is capable of producing tighter manufacturing tolerances, the material vendor will supply Er values for other frequencies. Your fab should have the data sheet from his material vendor, or should be able to tell you exactly who makes the material he will be using. There's no way for you to calculate the dielectric constant for board material without having your fab make some test coupons and measuring it yourself at different frequencies.

Remember that the frequencies you are interested in for signal integrity are the equivalent frequencies calculated from the rise and fall times of your data pulses - that's the part of the signal you need to protect.

Er @ other frequency

thanx

if i calculate the Fkee for the 400MHz bus, its about 1...1.2GHz

And the 1MHz and 1GHz values are very different. so should i use 1GHz Er values?

I got datasheets, very detailed, but only this, about Er: 5.3@1MHz --@1GHz (missing data)

Re: Er @ other frequency

Yes, use the 1gHz value for Er.

Normally, you would expect about 0.1-0.2 difference between the 1mHz and 1gHz values for Er. So if it's 5.3 a 1mHz, I would expect about 5.1-5.2 at 1gHz.

There's an article that may help you a bit at :http://www.sigcon.com/Pubs/news/4_5.htm

Er @ other frequency

once i have seen a variation of 4.3...4.8 for FR4 between 1MHz and 1GHz in a spreadsheet about materials.

In protel, i can set only one Er, so if i want to use different values for the 100MHz ethernet and a 667MHz FSB, then its not too good. i will have to calculate manually at least one of them for trace width determination.

Added after 27 minutes:

i think i have found a solution:
1. i give Er@1MHz for the protel
2. i predistort all Z0s: Z0-->Z0*k
where k=sqrt( (Er(fx)/Er(1MHz) ) and Er(Fx) is from the article, you mentioned: Er(fx)=Er(1MHz)*(Fx/1MHz)^-0.006366 for FR4 material

Added after 3 minutes:

and how did you do that before?
calculating, or using test coupons? or just U used Er(1MHz) ?

Re: Er @ other frequency

For very critical applications I use test coupons.

For day-to-day applications, the FR4 my fab uses has an Er of 3.85 at 1GHz, and the vendor supplies test data for other frequencies when needed.

Don't believe spreadsheets that give general data for materials like FR4. FR4 is a class of materials, not just one. There are many Tg, Er, and loss tangent possibilities for FR4. The high Tg material that my fab uses is very stable up through 9GHz.

Re: Er @ other frequency

i asked them about test coupons, but they dont make tests, only measure width and tickness for controlled impedance. This is the highest quality PCB manufacturer in my country.

I think i will use the calculations. The results are realistic, i think, for example if i calculate Er@1GHz, its the same as in the datasheet.
And the Z0 difference is about 1.3 ohm