Steve - I used to have the luxury of being able to control all 3 inductances. I agree with you that most PCB designers can only control 2 of the 3 because they cannot control the package design. This is where the chip/package design houses have to understand things from the system perspective and make sure they leave adequate inductance budget for the PCB designers. =20 I'll do what I can to make that happen at this chip/package company. :) Regards, Larry Smith Altera Corporation -----Original Message----- From: steve weir [mailto:weirsi@xxxxxxxxxx]=20 Sent: Tuesday, October 25, 2005 3:24 PM To: Larry Smith; christopher.mcgrath@xxxxxxxxx; ludovic.levieil@xxxxxxxxxxx; si-list@xxxxxxxxxxxxx Subject: Re: [SI-LIST] Re: Power plane coupling Larry, as usual you offer very good advice. I would just comment a little=20 bit about the difference between optimization, and realization. Given the chance to independently control all three: a. IC package design,=20 b. Board stack-up, and c. Bypass capacitor selection, and count I agree=20 that portioning in about equal amounts likely optimizes the=20 economics. However, a board designer is pretty much stuck with only b) and=20 c) as the available variables. So to me the issue is one of realization: 1. Can I come up with a PDS that actually works as needed? 2. What is the least cost way to get there? To answer 1) I find that it is useful to take the target inductance you=20 have broken into three pieces, and evaluate the attachment from the IC to=20 any given proposed location of the supporting planes in the stack-up. If=20 this is more than half of the target, we are facing thin planes and/or a lot of capacitors. If it is over two thirds, it may be difficult if not impossible to realize a working design at any cost. After that, it is an=20 iterative process of solving for each plane pair, and the resulting bypass=20 network. If the package attachment strips the budget to where lots of caps=20 and thin dielectric are needed, then I will use those to get through the day as needed. Best Regards, Steve. At 01:59 PM 10/25/2005 -0700, Larry Smith wrote: >Chris - As you and Steve have mentioned, the best side of the board for >placement of decoupling capacitors is a function of inductance. There >are three important inductances involved, each of which play a different >roll. > >1) The capacitors are attached to the PCB with a mounting inductance >that is closely related to the via length from the cap pads to the >power/ground plane pair. The mounting inductance is an important factor >in determining the resonant frequency and Q of the mounted capacitor. >For this reason, you want the power/ground plane pair to be as close as >possible to the front or back surface of the PCB. In the stackup under >discussion in this thread, that would be layers 2, 3, 4 and 5 (or N-1 >through N-4 where N is the number of layers). I like to use a matrix of >different valued capacitors, each with a similar mounting inductance, to >establish a low and flat impedance from about 1MHz to 100MHz for a bank >of capacitors attached in this way. > >2) The power/ground plane pair is the conduit that carries power from >the bank of capacitors (above) to the vicinity of the load. Istvan >mentioned the value of thin power plane dielectric (Vcc to Gnd) to >maximize plane capacitance. Thin dielectric also has superior >inductance properties because the power plane spreading inductance is >proportional to the dielectric thickness. The second important >inductance in our problem is the horizontal power plane spreading >inductance. > >3) Charge that is stored in the capacitors flows vertically through the >mounting inductance, horizontally through the plane spreading inductance >and once again goes vertical in the vicinity of the load. A >checkerboard pattern of power and ground vias is often used underneath a >load (i.e. BGA package) to form a low inductance path from the power >planes to the power consumer. This is the third inductance in our >problem. > >These three inductances should be closely balanced in the power >distribution system because any one of them can dominate. I have seen >cases where hundreds of power/ground vias underneath the BGA load have >far less parallel inductance than the horizontal power plane spreading >inductance, even when the vias carried current all the way from the >backside of the board. Thin power plane dielectric could be very >valuable for this system. But if there are not very many power and >ground vias in the BGA pattern, there is no point in using thin power >plane dielectric. > >The sum of inductances 2 and 3 gives a series inductance for the bank of >decoupling capacitors mounted on the board and gives an upper limit to >the number of decoupling capacitors that are useful. The equivalent >inductance of all the parallel decoupling capacitors should be >calculated (mounting inductance divided by the number of caps). In a >well designed power distribution system, about 1/3 of the inductance >should be allocated to the BGA vias, 1/3 for the power planes and 1/3 >for the parallel mounting inductances. If one of these three >inductances significantly dominates, it establishes the performance of >the PDS and there is no point in putting additional cost into the other >two inductances. > >So, to answer your question on which side of the board to place the >caps, you need to look at the three inductances of the system. > >Regards, >Larry Smith >Altera Corporation=3D20 > > > >-----Original Message----- >From: steve weir [mailto:weirsi@xxxxxxxxxx]=3D20 >Sent: Friday, October 21, 2005 4:11 PM >To: christopher.mcgrath@xxxxxxxxx; Larry Smith; >ludovic.levieil@xxxxxxxxxxx; si-list@xxxxxxxxxxxxx >Subject: Re: [SI-LIST] Re: Power plane coupling > >Chris, > >You always want to think in terms of minimizing demon inductance. So, >first=3D20 >we want to put the plane(s) that support our lowest impedance >requirements=3D20 >closest to the device they feed. Not every plane can be in the top of >the=3D20 >stack, so this will compromise some supplies. > >The placement of capacitors should also be selected to minimize the >total=3D20 >inductance between the caps and the devices they serve. For a = plane=3D20 >towards the bottom of the board, this means the caps go on the bottom. >We=3D20 >are already burned by the long vias from the device to a plane near the=3D20 >opposite side of the board. If we place the caps on the same side as >the=3D20 >part, then the attachment vias for the caps will be just as long, a very > >bad thing. > >Spreading inductance is another issue. It is part of the price we pay >for=3D20 >the reality that it is essentially impossible to put discrete caps of >any=3D20 >size right on the die pads. The further we move away in any axis, the=3D20 >higher a penalty we pay. The natural log behavior of spreading >inductance=3D20 >and resistance is such that once we are in for a dime, we might as well >be=3D20 >in for a dollar. > >Regards, > > >Steve. >At 01:32 PM 10/21/2005 -0700, Mcgrath, Christopher wrote: > >Larry, > > > >I have a related question that popped into my head when I read your > >response. > > > >If you expand this example so that there were a number of layers >between > >VCC1 and VCC2 which included both ground and signal layers so that the > >thickness of the board begins to become significant with respect to the > >placement of decoupling capacitors, is the preferred location for > >decoupling of a device placed on the top side of the PCB on the top >side > >or bottom side of the board IF the voltage being decoupled is VCC2. > > > >At issue is whether it is better to place the decoupling capacitor > >closer to the voltage it is decoupling or closer to the device that is > >actually receiving the power. > > > >I believe that the path with the least inductance and maximum > >effectiveness would be to place the decoupling capacitor on the top >side > >of the board right next to the device. While the spreading inductance > >should be very similar in both cases, the loop inductance would be > >slightly less when the cap is placed on the top of the board. =3D3D20 > > > >Any thoughts? > > > >Thanks, > >Chris > > > >=3D3D20 > > > > >-----Original Message----- > > >From: si-list-bounce@xxxxxxxxxxxxx > >[mailto:si-list-bounce@xxxxxxxxxxxxx] On > > >Behalf Of Larry Smith > > >Sent: Friday, October 21, 2005 12:05 PM > > >To: ludovic.levieil@xxxxxxxxxxx; si-list@xxxxxxxxxxxxx > > >Subject: [SI-LIST] Re: Power plane coupling > > > > > >Ludovic - I like this power plane stackup sequence, particularly if >it > > >is on the top or bottom surface of the PCB. > > > > > >The power planes will be highly coupled to ground by discrete > >decoupling > > >capacitors mounted on the surface of the board. There are probably > > >100's of uF that are trying to maintain a constant voltage between >VCC1 > > >and Gnd, also between VCC2 and Gnd. But the internal plane to plane > > >capacitance is on the order of 1nF, not much compared to the external > > >capacitance. At 100MHz, the 1 nF plane-to-plane impedance is about > > >1/(2*pi*100e+6*1e-9) =3D3D3D3D 1.59 Ohms. This is not strong compared =3D >to >=3D3D > >the > > >impedance of the PDS which is probably in the mOhms. The impedance > > >division insures that there will not be substantial noise coupled >from > > >one power plane to the other in this stackup. But as Istvan has > > >commented in another note on this thread, this might not be best for >a > > >sensitive analog supply or PLL circuitry. Further filtering should >be > > >used for those supplies. > > > > > >Noise above 100 MHz usually gets onto a power plane because of > > >transmission line return current. I like your stackup because the > >power > > >planes are surrounded by Gnd planes. You have an opportunity for > > >transmission lines to reference only ground planes throughout the >rest > > >of the stackup. This keeps the return current noise off the power > > >planes and the power plane noise off the transmission lines. Skin > > >effect in solid ground planes greatly attenuates magnetic fields from > > >penetrating through the planes at 1 MHz and above. > > > > > >Noise below 100 MHz is usually caused by current transients from the > > >loads. A well designed PDS will be below target impedance from some > > >corner frequency (50 to 100 MHz) all the way down to DC. The noise > > >coupled between power planes below this corner frequency is >diminished > > >because the impedance of the plane-to-plane capacitance diminishes at > > >lower frequency. This stackup puts you well on the way towards good > > >power and signal integrity in your product. > > > > > >Regards, > > >Larry Smith > > >Altera Corporation > > >(Sun Microsystems was very good for me, but it was time to move on.) > > > > > >-----Original Message----- > > >From: si-list-bounce@xxxxxxxxxxxxx > >[mailto:si-list-bounce@xxxxxxxxxxxxx] > > >On Behalf Of Ludovic Levieil > > >Sent: Thursday, October 20, 2005 1:12 AM > > >To: si-list@xxxxxxxxxxxxx > > >Subject: [SI-LIST] Power plane coupling > > > > > >Hello All, > > >In my current board design I have the following stack up: > > > > > > ....... > > >---------------- GND (solid plane) > > >------ ----- --- VCC1 (splitted plane) > > >--- ----- ------ VCC2 (splitted plane) > > >---------------- GND (solid plane) > > > ....... > > > > > >4 mils separate GND and VCC planes > > >5 mils separate VCC1 and VCC2 planes > > > > > >Both VCC planes are splitted in different power domains and I am > > >wondering=3D3D3D20 > > >: > > > - if having two coupled VCC planes is good/acceptable >when=3D3D3D20 > > >thinking about noise ?? > > > - if there is a problem in having one power domain on on > >plane=3D3D3D20 > > >overlapping at least two power domains on the other plane ?? > > > > > >Thanks > > > > > >Ludovic Levieil=3D3D3D20 > > > > > >------------------------------------------------------------------ > > >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 FAQ wiki page is located at: > > > http://si-list.org/wiki/wiki.pl?Si-List_FAQ > > > > > >List technical documents are available at: > > > http://www.si-list.org > > > > > >List archives are viewable at: =3D3D3D20 > > > //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 > > > =3D3D3D20 > > > > > > > > > > > >------------------------------------------------------------------ > > >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 FAQ wiki page is located at: > > > http://si-list.org/wiki/wiki.pl?Si-List_FAQ > > > > > >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 > > > > > > >------------------------------------------------------------------ > >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 FAQ wiki page is located at: > > http://si-list.org/wiki/wiki.pl?Si-List_FAQ > > > >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 > > > > > > >------------------------------------------------------------------ >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 FAQ wiki page is located at: > http://si-list.org/wiki/wiki.pl?Si-List_FAQ > >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 > ------------------------------------------------------------------ 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 FAQ wiki page is located at: http://si-list.org/wiki/wiki.pl?Si-List_FAQ 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