Re[4]: [SI-LIST] : Decoupling capacitor selection & plac

tderego@hadco.com
Thu, 30 Oct 97 08:20:21 EST

The EmCap material has a dielectric constant of ~50. I was wondering
about the transmission speed as well. I looked at what the capacitance
is at a given frequency based on an area of 1/6th wavelength away (or
what area would act like a lumped circuit) and I found that the
material acted slightly better than a 0.1uF or a .001uF 1206 SMT cap
impedance-wise.
10" x 10" Circuit Board
Impedance
Freq EmCap 0.01uF 1206 0.1uF 1206
1Khz 628.760 15915.494 1591.549
2.5Khz 251.504 6366.198 636.620
5Khz 125.752 3183.099 318.310
10Khz 62.876 1591.549 159.155
25Khz 25.150 636.620 63.662
50Khz 12.575 318.310 31.831
100Khz 6.288 159.154 15.915
250Khz 2.515 63.660 6.365
500Khz 1.258 31.828 3.181
1Mhz 0.629 15.909 1.588
2.5Mhz 0.252 6.350 0.627
5Mhz 0.126 3.150 0.301
10Mhz 0.063 1.526 0.135
25Mhz 0.025 0.474 0.148
50Mhz 0.047 0.104 0.329
100Mhz 0.093 0.541 0.683
250Mhz 0.233 1.667 1.724
500Mhz 0.465 3.425 3.454
1Ghz 0.931 6.896 6.911
2.5Ghz 2.327 17.273 17.278
5Ghz 4.654 34.554 34.557
10Ghz 9.308 69.114 69.115
25Ghz 23.270 172.787 172.788
50Ghz 46.540 345.575 345.575
100Ghz 93.080 691.150 691.150
I have much more work to do on these distrubuted capacitance planes.

Thanks!
Todd DeRego
HADCO

______________________________ Reply Separator _________________________________
Subject: Re: Re[2]: [SI-LIST] : Decoupling capacitor selection & plac
Author: etxaek@cubt.ericsson.se (Anders Ekholm) at SMTPLINK-HADCO
Date: 10/30/97 2:13 AM

It its very nice to get more low-inductance capacitance, but to get that
you are increasing the dilectric constant for the material (ceramic ~39).

Which gives us a very low transmission speed in the transmission-line formed
by the ground and power plane. That should give about 18ns/m.
I do not expect the capacitive plane to give all the charge I need
so I need decoupling capacitors still. To maintain a risetime of 1ns
the decoupling capacitors need to be placed closer to the drivers than
~1/18 m (5 cm). Normally we use risetimes of 0.4ns which would give
approximatly 2.5cm. If charge from decoupling capacitors outside of that
region is needed the signal risetime will slowdown.

Without detailed analysis I would not know if the new capacitive plane
is better or not, I get more low-inductance capacitance but I also get
a reduction in transmission speed in the capacitive plane.

It probably could be analysed by QUAD AC-Grade. Maybe someone at
QUAD could have a look at it.

Regards /Anders Ekholm
Ericsson Telecom




> From owner-si-list@silab.Eng.Sun.COM Wed Oct 29 14:16 MET 1997
> From: tderego@hadco.com
> Date: Wed, 29 Oct 97 07:43:46 EST
> Cc: si-list@silab.Eng.Sun.COM
> Subject: Re[2]: [SI-LIST] : Decoupling capacitor selection & placemen
>
> How does everyone feel about power/ground distributed capacitance?
>
> Zycon originally developed ZBC2000 - a thin FR4 (.002") power/ground
> core with a capacitance of ~500pf/sq in (usually 2 cores are used) -
> as a decoupling capacitor replacement or adjunct.
>
> HADCO and Polyclad have developed EmCap - a 0.004" Ceramic / FR4 epoxy
> power/ground core with 2750pf/sq in - as the next generation.
>
> I have been doing work to predict the ability of both these products
> to replace discrete capacitors in our customers boards and would like
> to hear some honest thoughts and feedback.
>
> FYI - I'm an EE working at HADCO. I work with the electrical issues in
> PWB - impedance, capacitance, layout, etc. - more or less in house SI.
>
> Thanks!
> Todd DeRego
>
>