Tag Archives: power integrity

Istvan Novak on power integrity

A couple years ago my pal Bob Thomas over at Apple told me how Istvan Novak over at Sun Microsystems figured out a clever way to keep RF from radiating out the edge of the board. For years, engineers have put power and ground planes or two ground planes on the top and bottom of the board, so it makes an enclosure, a metal can that keeps the signal traces inside from radiating. Those same engineers noticed that when the edges of the boards were open, RF would leak out of the edges and cause problems with signal integrity, power integrity and EMC (electromagnetic compatibility).

Istvan-Novak_DesignCon-2014

Istvan Novak, here at the DesignCon show in 2014, figured out you should stitch resistors or RCs around the edge of a PCB so the RF would die when it hits it’s characteristic impedance.

So these same engineers would stitch hundreds of vias between the top and bottom ground planes, all around the edge of the PCB (printed circuit board). If they had a top-side power plane instead of two ground planes, they would stitch hundreds of decoupling caps all around the edge of the board.

Istvan figured out a problem with this scheme. See, if you have a via or a decoupling cap at the edge of the board, that looks like a dead short. Remember that RF reflects off a dead short. So what would happen is that the RF radiating out of the signal traces would hit the vias and then bounce back inside the board and wreak even more havoc with signal integrity.

Istvan figured out that you don’t want to just stitch vias, you want to figure out the characteristic impedance of the two planes, which you can think of as a big fat transmission line. Then you stitch resistors all around the edge of the board. When the RF hits the resistors, it dies with no reflections, since it just hit its characteristic impedance. If you have a power and ground plane, you leave the decoupling caps, but add a resistor in series to each cap. That way there is no dc power loss from all the resistors. Istvan patented this at Sun, but he insists he is not the sole person to see this. He mentioned several people that have also worked on this problem.

All this bouncing RF also raises hell with your power integrity. See, Bob Thomas described the RF radiating out of the board edge. When I mentioned Istvan’s trick to Howard Johnson, the famous signal integrity consultant, he said that what was going on is that the power planes were resonating and the resistors were adding damping. OK, but I knew my pal Bob is no slouch, and what he described seemed to be right.

howard-johnson

Howard Johnson, shown here in repose at his Signal Hill Ranch in Washington State sees the RC stitching as stopping power plane oscillation.

Bob-Thomas_Cisco_Apple_HP

Bob Thomas, a pal from my HP consulting days back in the 90s, was at Cisco when he told me Istvan’s trick.

Accordingly I arranged to meet with Istvan at the DesignCon show here in Silicon Valley. He was part of a panel discussion run by my old pal Martin Rowe, over at EDN and EETimes. So after the panel I put Istvan on the spot. Who was right, my brilliant pal Bob Thomas, who says it was killing leaking RF, or brilliant consultant Howard Johnson, who said it stopped the planes from oscillating?

Istvan smiled and said “They are both right!” He explained that if RF is pumping out the edge of the board or hitting a dead short and bouncing back inside, well then the planes are oscillating, they are intimately related. I really love this signal integrity stuff, or in this case, power integrity. Istvan also pointed out that these days you need so many power planes, you don’t get one big plane you can stitch all around the edge. For this he says you bring the power and ground planes close together, and close means like a 1-mil (0.001 inch) spacing. That raises the capacitance up and makes the transmission line formed by the plans lossy, which keeps them from oscillating and radiating RF.

Best of all, Istvan was nice enough to write me a follow up note:

“To get the basics about terminating planes, you can specifically look at “Reducing Simultaneous Switching Noise on Power Planes by Dissipative Edge Termination,” EPEP’98, October 25-27, 1998. I just realize that there is no link any more from my webpage to this paper, but you can get from this direct link:

http://www.electrical-integrity.com/Paper_download_files/EPEP98_DET.pdf

“As I mentioned during our brief chat, this paper and the subsequent patents, were not the first on the subject. One earlier paper is referenced in my EPEP98 conference paper as:

“G. Lei, R. Techentin, B. Gilbert, “Power distribution noise suppression using transmission line termination techniques,” Proceedings of the 5th Topical Meeting on the Electrical Performance of Electrical Packaging, October 28-30, 1996, pp. 100-102.”

“If you Google the plane termination subject, you will find other papers and other patents as well. One other thing worth mentioning: like every new solution, these inventions have their optimum time when they are needed and it makes sense to use them.

“The plane termination technique was very useful in the late 90s and early 2000s when many boards had large contiguous power and ground planes, prone to strong resonances. However, as system density continues to grow, we are now forced to chop up the power plane layers into many smaller puddles. Under these circumstances using edge termination becomes less attractive. If resonances are still making problems, a better way of using very thin laminates. See for instance:

http://www.electrical-integrity.com/Paper_download_files/DC02_HP-TF2.pdf

So thanks to Istvan, and Bob Thomas and Howard Johnson for making our power integrity more solid and reliable. I have a video about this as PCB202, and will back link to that as soon as it is posted.