We’ve created a special domain for legal issues in this country. If it looks like a lawyer might have to get involved, the rest of us steer clear and know better than to try to make sense of things. Even if we understand the words, we might not understand what they mean when they’re used together in a paragraph or document. And, even if we do, it still might not make sense.
But we don’t seem to mind: whatever… leave it to the lawyers. As long as they’re happy…
This includes – perhaps more than anything – patents. There, language is everything. “Blue” might not just mean “blue,” and “2+2”, in the hands of a skilled lawyer, just might mean something other than “5”. Or including “5”. Jointly and/or severally. Or something.
As a result, engineers will sometimes look at the document describing their invention and not recognize it at all. It’s a world in which I’ve worked in the past, writing up patent language, originally for stuff I was involved in and, later, for others. It gave me a sense of the elements you approach as a lay person and the elements you completely leave to the lawyers.
In a perverse way, sorting through such stuff preps you for the craziness that can happen if a patent actually gets disputed. And that’s the key: you can put anything you want in a patent. If the examiner allows it, then it’s patented. Heck, people have been able to patent genes that they merely found, without really inventing them. Which is like Kepler patenting the ellipse because he found it in planetary orbits. But it’s not really tested unless someone disputes it: that’s when you find out what can really be patented.
There was some triumphant press activity the other day in one such dispute – and it wasn’t about who won or lost the patent infringement case. It was about definitions of words or phrases. Which gives an interesting opportunity to peer into how some of this stuff goes down.
Now… before I get into any detail here, let me state loud and clear that I have no intention of providing my opinion as to who is right or wrong either with respect to the definitions or the overall case. I’m simply not that stupid. But I thought it interesting to look at how things get argued and how the judge can resolve things. I will say that I was pretty impressed with the judge’s ability to cut through a lot of tangle on some possibly tough technical issues.
That aside, let’s start with some basics for those of you who haven’t found yourself embroiled in a patent write-up. There are, more or less, two main components to a patent. There’s a description, referred to as the “specification” (or “spec”), and then there are the “claims.”
The spec is a more or less human-readable narrative. You could say that, done well, it shows why there’s a problem that needs to be solved, shows how lots of other attempts by people have failed, and then lays out the invention, along with examples of ways of implementing the invention (so-called “embodiments”) – in enough detail that someone could actually go do it.
That last bit is key, because the original intent behind the patent law was that, in exchange for a limited monopoly, you show everyone else how you work your magic; they can then build on that – within the limits of your monopoly rights.
What you invent has to be new (or “novel”) and non-obvious. There has to be a spark of creativity. But non-obvious to whom? None of our technology, not even the obvious stuff, would be obvious to Uncle Maynard in the trailer back in the holler.
So the standard tends to be “a person of normal skill in the art” (or something like that). I can’t think of how many times I wrote the patent equivalent of “the details are obvious and left to the reader” using some variant of, “the details can be implemented in a manner straightforward to one of normal skill in the art in light of the description provided herein.” (That last bit is important: without it, it would seem like your invention was obvious – so you have to say that some follow-on implementation detail is obvious, but only once you understand the invention, which is itself non-obvious.)
The deal with the spec is that you want to describe the technology in the broadest possible terms. Once you’ve turned it in, you can’t go change it. If your invention works with incandescent light bulbs and you write it up that way, then, if you find later on that it also works with CFLs, it’s too late.
Claims are different, however. This is where the rubber hits the road. This is where you say specifically what you’re patenting. Everything in the claims has to be supported in the spec, although the spec may be much broader than what the specific claims claim. And that’s on purpose: while you can’t change the spec in the future, you may be able to add claims in the future. So you want the spec to cover possible future claims.
The language of the claims is likely to tie you in knots. You may well want a map to go with it. And, when you’re writing them, the lawyer will probably not want you to draw up anything looking like a map. Because that would be interpreting the claims. Which they will want you to avoid at all costs.
You see, if you actually try to be broad and vague and expansive in your claims, and yet, in your emails or notes or whatever, you write down specifically what you mean (say, in response to an email, you say “A “light-emitting source” is just lawyer-speak for “your basic incandescent light bulb”), then, later on, when your lawyer is trying to argue in court that CFLs should be covered, their lawyer will point to the email, which they got through discovery, and say, “No, see? It says right here that they intended to cover only incandescent.”
And your lawyer will be very unhappy with you.
This whole process is called “claims construction.” Which is itself confusing because it sounds like “building claims,” which is what you’re doing when writing them. But it’s actually a different “construction,” meaning the act of construing (which I noticed doesn’t show up in the dictionary unless you pursue the etymology, in which case you find that “construct” comes from the past participle of the Latin “construere”, “to construe”, meaning to “build together”).
So, in our modern times, this isn’t about building claims, it’s about construing, or interpreting, them. And you don’t want to do that when you write them; you want to put that off until the last possible moment (even forever).
Which brings us to our current case study, wherein Kilopass has sued Sidense for patent infringement (and other related things). These guys both make OTP memory cells, but they do it differently. The question is, are they different enough, or does one have patent protection broad enough to include the variation of the other?
In the process of construing the meaning of the claims, they have to construe the meaning of the words used in the claims. Should be easy, right? Wrong. An example claim is shown below, taken verbatim from the judge’s write-up, and with disputed terms in bold:
“1. A programmable memory cell useful in memory array having column bitlines and row wordlines, the memory cell comprising:
a transistor having a gate, a gate dielectric between the gate and over a substrate, and first and second doped semiconductor regions formed in said substrate adjacent said gate and in a spaced apart relationship to define a channel region there between and under said gate, the gate being formed from one of said column bitlines;
a row wordline segment coupled to the second doped semiconductor region of the transistor, said row wordline segment connected to one of said row wordlines; and
a programmed doped region formed in said substrate in said channel region when said memory cell has been programmed.”
Almost every non-trivial word is in dispute. And this is only one example.
So I thought it would be interesting to go through the list of terms that were agreed and disputed and, for the latter, look at the different interpretations and how they were resolved by the judge.
A couple of key decision processes became clear in the judge’s comments. First, both parties agreed that the terms, which show up in three patents, should have the same meanings in all three patents. That might be obvious, but it gets to the second point: these definitions are made in the context of the patents. In some cases, as we’ll see, the judge may say the equivalent of, “Everyone knows what that means, let’s move on because I see no reason not to.” In other cases, specific language in the spec or other arguments may create a definition for use in this case that might not apply to other cases.
The following terms were agreed to early: “memory array”, “floating”, “programmable memory array”, “memory cell which has been programmed”, and “breakdown”. Later, it became evident that there was also agreement on: “spaced apart relationship”, “gate dielectric”, “programmed”, and “laterally separated by a distance D”.
“Semiconductor” wasn’t quite so easy, but the two definitions were resolved between the parties, and the judge accepted that definition. They also agreed on the definition of “a person of ordinary skill in the art.”
All of the agreed definitions are shown in the first table.
Table 1. Agreed Terms (click to enlarge):
Then come the tough things, which I’ll go through in more detail. I don’t want to focus too much on who prevailed or didn’t, so I’ll use “plaintiff” and “defendant” (as the judge does) below, but, for the sake of completeness, the plaintiff is Kilopass and the defendant is Sidense. And I shouldn’t have to say this, but I must… As hard as I’m trying not to take sides or pretend to be a judge or a lawyer, I am still necessarily interpreting what the judge wrote up; someone else might disagree with my interpretation. And I’m OK with that.
“Bitline/column bitline” and “wordline/row wordline”
Plaintiff: all four mean the same thing: ”a line that connects to one terminal of each memory cell in a memory array.”
Defendant: “bitline” and “column bitline” mean “the line that connects the memory cell to the sensing circuit during the read operation”; “wordline” and “column wordline” [sic; I think the judge meant “row wordline” there in her order, from which I’ve extracted this…] mean “the line connected to the memory cell, which is selected by the row addresses.”
The judge focused on a couple of issues. The first is that, by the plaintiff’s argument, there are four terms that mean the same thing, and she wasn’t ok with defining terms as identical if they weren’t, in fact, identical. And there was no dispute during arguments that, although these may look the same in some regards, the bitline and wordline are different because they’re orthogonal to each other.
On the other hand, the defendant’s definitions included not only the internal structure of the items, but also elements from their relationship to external structure (e.g., “sensing circuit”) as well as how the thing was used (e.g., “selected by the row addresses”). Because the legal dispute has only to do with the internal structure, the judge didn’t like including external or operational elements.
Judge’s decision: “bitline” and “column bitline” mean “a line orthogonal to the row wordline that connects to a terminal of each memory in a memory array”; “wordline” and “row wordline” mean “a line orthogonal to the column bitline that connects to a terminal of each memory cell in a memory array.”
“Transistor” (Yes, seriously)
Plaintiff: should use ordinary meaning; from one internet source: “a solid-state semiconductor device that can regulate electric current flowing through it.”
Defendant: “a MOS transistor structure.” [You can tell in the document that the judge pronounces “MOS” as “moss”, not “em oh ess” because of the use of “a” instead of “an”. Kind of like the situation where “SQL” may be pronounced “sequel,” requiring “a,” or “ess cue ell,” requiring “an”… Ah, English…]
This seemed to be a key issue, suggesting that the defendant wanted the plaintiff’s claims construed only in the context of an MOS transistor. They proposed this because, each time a transistor was described in the claims, the description included elements found only in an MOS transistor.
The plaintiff, on the other hand, said that this was “tautological” and limited the invention to “one embodiment.”
Another part of the disagreement was whether the structures in these memory cells were MOS transistors connected in unusual ways, or whether, because of the unusual connections, they were therefore not MOS transistors anymore. Oi! The plaintiff argued that the odd configurations meant that they weren’t MOS transistors; the defendant said that this had nothing to do with the definition of an MOS transistor, but was just about how the MOS transistor was used or connected.
This was one of the longer discussions by the judge, and she seemed to zero in on a couple of concepts. First, because the transistor references included structural descriptions, she said that the “transistor” definition couldn’t then include those structural descriptions: those descriptions acted to limit the kind of transistor being discussed, meaning that the full definition of “transistor” could be broader.
The other tricky bit is that “transistor” is used differently in the three patents. In two of them, only structural elements are used; in one, there is mention of how it’s used. Since they agreed to use the same definition for all three patents, the final definition has to work in all three. So having a definition that made no mention of how the transistor was used – that is, structure only – wouldn’t fly.
Judge’s decision: don’t equate “transistor” with “MOS transistor.” She deferred to the “ordinary and customary” meaning, which she got from a dictionary of scientific terms: “An active component of an electronic circuit consisting of a small block of semiconducting material to which electrical contacts are made, and which may be used as an amplifier, detector, or switch to control the flow of current.”
Plaintiff: “a basic unit of memory.”
Defendant: originally, “a semiconductor memory device comprised of a single MOS transistor that stores data” [This definition highlights the common way “comprise” is abused – including by me in the past. It is actually, originally, equivalent to “includes” in this context, as in, “… device comprising a single…” Just a bit of pedantry on my part]. This was later changed to, “a single MOS transistor that stores data.”
The judge had trouble with defining a memory cell as identical to the transistor from which it’s built, with no rationale given for why two different terms are then needed. Plus there’s the tie to MOS which got tossed in the prior decision on “transistor.”
In the end, the judge decided that there was no need to define this term, citing a prior case that ruled that “in some cases, the ordinary meaning of claim language as understood by a person of skill in the art may be readily apparent even to lay judges, and claim construction in such cases involves little more than the application of the widely accepted meaning of commonly understood words.” In other words, “duh!”
Judge’s decision: let’s move to the next item.
“Programmed doped region”
Plaintiff: “a doped region in a channel region of a substrate under the gate caused by breaking down the gate dielectric.”
Defendant: originally, “a highly doped (n+ or p+) region formed in the substrate of a programmed memory cell as a result of programming.” Later changed to, “a highly doped n+ (or p+) region formed in the channel region of the oppositely doped substrate of a [programmed] memory cell as a result of programming.”
The issues here boiled down to whether the region had to be a “highly doped n+ (or p+) region” and whether the channel had to be in an “oppositely doped substrate.” The defendant said that the spec limited the context; the judge didn’t seem to agree. She did refine the definition, however, since the term “programmed” was already agreed to.
Judge’s decision: “a doped region formed in the channel region of the substrate of a memory cell as a result of programming.”
“Doped semiconductor region”
Plaintiff: “a semiconductor region that contains a level of impurity.”
Defendant: originally, “a region of semiconductor material, to which impurities are added to create one side of a p-n junction.” Later changed to, “the source or drain structure of a MOS transistor structure.”
The MOS aspect of the latter definition got axed due to the earlier decision.
Judge’s decision: “a region of semiconductor material to which impurities are added to modify the electron or hole concentration in that semiconductor.”
“Gate” and “channel region”
Plaintiff: “gate” means, “an electrode region that is above the channel region”; “channel region” means, “a region under the gate in the substrate where the channel is located.” [Sounds circular to me…]
Defendant: “gate” means, “the MOS transistor electrode which controls conductivity of the channel region”; “channel region” means, originally, “a region of the substrate beneath the gate defined by the space between first and second doped semiconductor regions, whose electrical conductivity is controlled by the gate.” Later changed to add “In a MOS transistor, …” to the beginning.
This got adjudicated based upon the early issues with MOS focus and defining things in terms of operation (instead of internal structure).
Judge’s decision: “gate” means, “an electrode region that is above the channel region”; “channel region” means, “a region of the substrate beneath the gate defined by the space between the first and second doped semiconductor regions.”
These are all summarized in the second table.
You may now return to normal engineering thinking.
Table 2. Disputed Terms (click to enlarge):