[SI-LIST] Re: Dielectric constant for FR4 at above 1Ghz frequency
- From: "Scott McMorrow" <scott@xxxxxxxxxxxxx>
- To: chris.mcgrath@xxxxxxxx
- Date: Tue, 02 Dec 2003 12:34:56 -0800
Chris,
Several comments:
First, FR4 has easily been shown to work well in applications of greater
than 10 Gbps, depending upon the driver and receiver type, trace length,
whether the system is equalized, and whether the encoding is binary,
PAM4 or some other encoding scheme.
Second, in my experience, most fab houses have no clue what the actual
Er and loss tangent of the material they use is. (The fabricators that
are exceptions to this, also have a technical edge, since they
understand the electrical properties of the materialthey use.)
Additionally, they will rarely specify a different Er for different
laminate types within the stackup, as Dave Hoover showed in his post
below. And it is rare for material vendors to actually specify the Er
of there material at more than 1 or possibly 2 frequency points, all
usually uninteresting. These numbers are often suspect, since the
actual high frequency measurements may not have done on your particular
laminate type, and internal inductance may not have been de-embedded
from the final Er calculation. A measurement at 5 GHz and 10 GHz is
generally necessary to obtain the Er that most closely correlates to
actual TDR impedance measurements at the incident edge of the pulse.
Finally, anyone trying to really push the limits on FR4 ought to make
measurements or have measurements made for your particular fabricator's
production process, stackup and laminates used. There is quite a bit of
variation from fabricator to fabricator. There are also variations when
a board is moved from a prototype fabrication line to a production
fabrication line, since different equipment and process are often used.
Best regards,
scott
--
Scott McMorrow
Teraspeed Consulting Group LLC
2926 SE Yamhill St.
Portland, OR 97214
(503) 239-5536
http://www.teraspeed.com
>1. Frequencies greater than 1 GHz are perfectly fine to run on FR4(-6)
>material and we have done a number of 2GHz designs with FR4. As in many
>areas of design, the criteria for use depends on your application. =20
>
>2. When trying to determine the Er (vs. frequency) of the material for
>use in simulation, the best method is to ask your fab house unless you
>are dealing with a specific manufacturer like Nelco. Either way, the
>fab house will have this information as it relates to the manufacturers
>they regularly use.
>
>
>
>
>
>>-----Original Message-----
>>From: Fasig, Jonathan L. [mailto:fasig.jonathan@xxxxxxxx]=20
>>Sent: Wednesday, November 26, 2003 11:14 AM
>>To: si-list@xxxxxxxxxxxxx
>>Subject: [SI-LIST] Re: Dielectric constant for FR4 at above=20
>>1Ghz frequency
>>=20
>>=20
>>For what it's worth, Nelco has nice tables of dielectric constant at=20
>>1MHz, 1GHz, 2.5GHz and 10GHz for several laminated thicknesses along=20
>>with the constructions (e.g. 2 layers of 2116). Reference=20
>>the tables at=20
>>http://www.parknelco.com/parknelco/resources.htm and take the=20
>>link for "Typical Dielectric Property Tables". You will need=20
>>to register for a=20
>>password first.
>>=20
>>Isola has some similar information for their FR406 and FR408=20
>>products at=20
>>http://www.isolalaminatesystems.com/isolaus.nsf/main?OpenFrame
>>set . See the spreadsheets under Processing Guide for the=20
>>respective laminates.
>>=20
>>jf
>>--------------------------------------------------------------
>>----------
>>=20
>>Jonathan Fasig Email: fasig.jonathan@xxxxxxxx=09
>>Mayo Foundation=20
>>4001 41st Street NW=20
>>MSC Sn 2-132 Phone: (507) 538-5464
>>Rochester, MN 55901 Fax: (507) 284-9171
>>=20
>>=20
>>-----Original Message-----
>>From: David Hoover [mailto:dhoovy@xxxxxxxxxxxxx]=20
>>Sent: Tuesday, November 25, 2003 10:46 PM
>>To: alexh1@xxxxxxxxxxxxx; si-list@xxxxxxxxxxxxx
>>Subject: [SI-LIST] Re: Dielectric constant for FR4 at above=20
>>1Ghz frequency
>>=20
>>=20
>>Alex, (it's been awhile) (below is my $.02 from a fabricator)
>>=20
>>In the case of FR-4 we typically see a series of variables
>>that gives us the PCB dielectric properties.
>>For example:
>>=20
>>If we look at the 5 most common (E-Glass) fiberglass styles=20
>>used in the
>>U.S.: 1060, 1080, 2113, 2116, and 7628 These yield different=20
>>thicknesses within the PCB. The 1060 yields the thinnest all=20
>>the way up to the 7628 which is (one of ) the thickest. The=20
>>Dk of the pure resin of FR-4 epoxy is ~3.6 and the pure=20
>>(E-glass) fiberglass is ~6.6 (@ 1 MHz . That's all I've found=20
>>on the raw materials so far). It's the resin to glass ratio=20
>>that gives us the variance.
>>=20
>>We typically find:
>>Prepreg Thickness Resin Er
>> Style Percentage (@ 1 GHz)
>>1060 ~.002" 69 3.63
>>1080 ~.003" 62 3.80
>>2113 ~.004" 55 4.00
>>2116 ~.005" 52 4.08
>>7628 ~.007" 45 4.32
>>=20
>>Those values above are used to manufacture the various FR-4=20
>>cores. Each is basically like a building block used to=20
>>achieve various thicknesses. The thinner ones typically cost=20
>>more than the thicker ones. Some PCB fab shops or OEMs have=20
>>rules where they require the use of two plys minimum. A key=20
>>note to bring up is that even though the 1060 has the lowest=20
>>Er, it would not be a good idea to build an entire PCB out of=20
>>that prepreg. The resin content is too high and the pressed=20
>>PCB would flow resin all over the edges of the lamination=20
>>templates onto the lamination press. The thinner
>>material(s) also has the greatest amount of dimensional=20
>>movement within the PCB. An entire PCB built of 1060 would=20
>>cost alot and would experience large amounts of material=20
>>shrinkage when pressed. For cost and dimensional movement=20
>>reasons we minimize the amount of plies by choosing those=20
>>different styles of prepreg to build the various dielectric=20
>>thickness. (This allows the pad to hole ratio used today to=20
>>be maintained)
>>=20
>>There is beginning to be a few more choices of fiberglass=20
>>styles. Nelco has introduced the "SI" style fiberglass which=20
>>has a Dk of 4.4. When used to make prepreg, it yields a=20
>>substantially lower Er product. If you factor in lower Dk=20
>>resin blends (like the "-13") you can achieve a fairly low Er=20
>>material (when compared to standard FR-4).
>>=20
>>Hitachi and a few others have blends coming on line that have=20
>>fairly low Er values. These are all thermoset materials.=20
>>(Meaning they cure into a hard rigid state) The thermoplastic=20
>>materials (like PTFE
>>based) have Er's that get very low but sometimes are very=20
>>difficult to build multilayer PCBs with.
>>=20
>>We should start seeing more and more folks starting to test=20
>>their materials at various frequencies soon. This is driven=20
>>by issues such what Stuart, Andy, and Tom are discussing.
>>=20
>>As far as manipulating resins properties to achieve less=20
>>loss, I havn't encountered that yet. I have heard of R and D=20
>>efforts by the laminate manufacturers working on this, but I=20
>>have not heard of any production ready releases.
>>=20
>>Kind Regards,
>>=20
>>David Hoover
>>Sr. Field Application Engineer
>>Multilayer Technology, Inc.
>>=20
>>San Jose, Ca
>>http://www.multek.com
>>=20
>>
>>
------------------------------------------------------------------
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.org
List archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
