Issue link: https://iconnect007.uberflip.com/i/1116895
12 PCB007 MAGAZINE I MAY 2019 showed you the north and south poles on mag- nets in school, north and south poles on mag- nets, you can align all of these by putting a magnet on the top. If you switch it around, all of the poles switch around the other way. Alternating current does that all the time; it alternates up and down very quickly. All of these small, polarizable bits in the substrate flick back and forth all the time. Absorbing en- ergy and generating heat is exactly the same way that a microwave oven works. In that case, it's a water molecule that's vibrating back and forth, generating heat. But overall, generating heat is bad in two ways. First, you have to get rid of it, so in big server farms, there's a lot of heat to get rid of because you've generated a lot of extra heat. Second, it saps the signal, so if you put 100 in, 20 will be used for heat and only 80 will be used for the signal. You end up with a loss of signal. The property we talk about for that is called the dissipation factor, which is a number; it's dimensionless. The dissipation factor in a vac- uum is zero, more or less, so you have no loss. And this number continues to rise depending on the kind of material you have and the fre- quency. There has been a big push to get that number lower and on the substrate side. About the lowest that we know of so far are PTFE ma- terials, which are extremely difficult to polar- ize because the structure is pretty simple. It's a carbon backbone with fluorine atoms on either side. As fluorine is highly reac- tive the atoms get as far away from each other as they can, so they form a twisted helix and force themselves in position so that they can't vibrate, which is why PTFE has a very low loss. Materials like FR-4 are very polarizable; they contain moisture and all kinds of things that can be polarized. You end up with orders of magnitude difference in loss, so the drive has been towards that low side. The other big problem is copper. It has a certain roughness on one side, particularly to stick it to the substrates, you don't want the copper falling off. Typically, in modern circuitry, the cop- per thickness is 17 microns or one-half ounce, sometimes even less. For a 17-micron foil, the treatment on the back of it can be five or even eight microns, even half of the total thickness can be treated, which is very rough; it's like hills and valleys or a mountain range. The problem is as you reach higher and high- er frequencies, the electrons travel in the out- side of the conductor—not through the bulk of it. Even at relatively low frequencies, you find all of the electrons flowing to the outside of the conductor. One side is pretty smooth and flat while the other side is very mountainous and hilly. The electrons have to travel a further dis- tance on the lower side and through a higher resistance path. That compromises signal in- tegrity as well. As you reach a higher frequen- cy, you must have lower profiles on the copper down to submicron level, which is a bit of a challenge on the copper side. You must also consider the reinforcement in the substrate, which is formed from a resin and a reinforcement. PTFE will most likely be the lowest loss that we have because we normally use glass fibers. Glass fibers are used for re- inforcement because they're very stable, hold everything in place, and stop things from mov- ing around. They're pretty inert, so you can process them through wet chemistry and not have them dissolve or become damaged. The problem is that the electrical properties of the resin and glass fiber are quite different. Ventec International Group lab.