Wednesday, May 4, 2022

Microsoft Excel table (part of 4G specs) unearthed by OPPO may render Nokia patent non-standard-essential: Mannheim trial

Yesterday (Tuesday) was the first courtroom clash between Nokia and OPPO in Mannheim (a FRAND hearing had already taken place in Munich). The patent-in-suit in that first case is EP2981103 on an "allocation of preamble sequences" (originally declared essential to 4G and later also to 5G). Given that the same court had ordered an injunction over this patent against Daimler in 2020 (which never got enforced), and that the Federal Patent Court of Germany had rendered a preliminary opinion according to which it was valid (and the EPO upheld another patent from the same family in an opposition proceeding), I thought the technical merits were going to be clear. But no.

To my surprise, OPPO has developed a more elaborate non-infringement argument than Daimler. To be fair, the smartphone giant had the benefit of being able to learn from what had happened in the Daimler case, but even if all other things were equal, OPPO--which is a major SEP holder itself by now--would almost always outperform any automaker by virtue of a far deeper understanding of the relevant technology and of standard-setting processes.

Before yesterday's Mannheim trial, I re-read the 2020 judgment in the Daimler case as well as parts of the nullity case file, which I had obtained from the Federal Patent Court. Against that background combined with what was said yesterday, I firmly believe this is a case in which Nokia would simply have to drop the patent from an ITC case, and even the vast majority of U.S. district judges would presumably throw it out on summary judgment (otherwise, the Federal Circuit would enter JMOL). By U.S. standards, it's not even close: the patent might simply be found invalid for indefiniteness, but at any rate, no court could arrive at an intellectually honest claim construction that would result in an infringement finding. The problem in Germany is that the courts don't follow the Markman procedure and firstly put down in writing an interpretation of the disputed claim limitations. In more than ten years of watching German patent infringement cases, I've seen a proper (by U.S. standards) claim construction only once--in Mannheim, but from a different panel (Presiding Judge Andreas Voss ("Voß" in German), in an Apple case over slide-to-unlock if I recall correctly). What is normally done there is that claim construction and infringement analysis become amalgamated without the court performing the important mental exercise of formally resolving a dispute over the interpretation of a claim limitation by writing up new claim language that would address the relevant question. And that shortcoming of the German approach is Nokia's (only) chance here, as a more systematic approach would simply render the patent non-standard-essential.

Not only didn't I expect to reach this conclusion before the trial started but I even agreed with Presiding Judge Dr. Holger Kircher's introductory remarks on the infringement (here, essentiality) allegations. He discussed three different non-infringement contentions by OPPO, all of which take aim at the final part of the claim language:

"ordering the sequence in each subset according to their cubic metric values, wherein the sequences of adjacent subsets are ordered with alternating decreasing and increasing cubic metric values" (emphases added)

The patent is all about arranging number sets called preambles. In other words, it's a sorting patent. These are--more specifically--the two issues I've already mentioned:

  • The sort criterion--in Excel, that would be the "Sort by" column--is called "cubic metric value" or just "CM value." OPPO doesn't affirmatively say that the specifications of the 4G and 5G standards don't contain CM values, though they did present an alternative set of CM values calculated by an expert witness. Judge Dr. Kircher said that the patent specification doesn't prescribe a particular formula with which to derive the CM values, and that's why OPPO's non-infringement argument is irrelevant unless and until OPPO officially--and under the German equivalent of Rule 11--disputes that the numbers in the standard are CM values.

    I do understand the court's perspective on the pleading standard. However, if a number-sorting patent doesn't even define the very type of number by which a set is to be sorted, that to me represents a potential case for an indefiniteness argument. I'd have to re-read the patent document to be 100% sure, but at this point I strongly suspect that this patent lacks a clear boundary and would, therefore, be indefinite by (strict) U.S. standards.

  • The reason for which I believe the patent doesn't map to the 4G/5G specs is that "decreasing and increasing" are mathematically clear directions (numbers go down or up), which depending on the segment of a chart that you look at would give you roughly the shape of a V (decreasing, then increasing) or a reverse V (the other way round). But an Excel table that is part of the specifications of the standards shows that in a few instances, there are apparently microscopic bumps in the road.

    The problem is now that those bumps are literally microscopic: there is a table in the specs of the standard that has a limited degree of precision, which is why you don't see bumps but just consecutive numbers that--up to the ninth post-point digit--are identical.

    Those identical consecutive numbers were already raised by Daimler, and I would agree with the court that an occasional plateau doesn't break a decreasing or increasing order. I would also agree with the court that it doesn't mean too much that neither the claim language nor the specification of the patent provide instructions as to how to deal with pairs of identical numbers. However, OPPO explained that the numbers aren't really identical. The appearance of equality results from the limited degree of precision of the output, but the specifications of the standard are not just a text document: there's also an Excel table that's part of it, and the numbers in the table are unique--and those numbers break the "decreasing and increasing" order that would be required to infringe the patent.

    I've looked at the 2020 Nokia v. Daimler judgment again, and in that one, Excel is mentioned only in connection with the accounting that Nokia demanded from Daimler (in order to calculate a damages claim). So, if Daimler never pointed the court to the fact that the actual numbers in the relevant Excel table are irreconcilable with the claim language, then OPPO has presented hard evidence of non-essentiality and is entitled to a different outcome. Potentially, they could get an injunction stayed rather quickly by the appeals court on that basis.

    From my own usage of Excel, I know exactly what the problem is: internally, Excel stores floating-point numbers that can be extremely precise; but normally the column width doesn't suffice for numbers with many post-point digits. In that case, Excel does and does not round: it does round in the sense of displaying a rounded number, but it doesn't actually round internally until you use the ROUND() function in a formula. I have indeed used the ROUND function on some occasions. A good example is Value-Added Tax: if you have multiple items with, say, 19% VAT, you arrive at many numbers with more post-point digits than the maximum of two that you can put on an invoice (example: if the net invoice amount is 325.99, 19% VAT amounts to 61.9381, but an invoice would round this to "61.94" and then arrive at a bottom-line amount of 387.93). Now, if you have multiple such items and round each line in this case, you really store 61.94 in that example, while in the other case you still have greater internal precision (albeit invisible), which in the sum of many lines could result in a wrong aggregate amount (just a rounding error, but still enough to get into trouble in the event of a tax audit).

    The specification of the standard is not like a law where only a text document is relevant. If the standard-setting organization provides an Excel table with exact numbers, then those numbers are the ones that matter, and the text document merely provides a first overview.

    I don't know whether OPPO (or previously Daimler) made that argument, but as someone who back in the 1980s reverse-engineered an operating system and BASIC interpreter and did some other coding over the years, I know how computers sort floating-point numbers. The way they do it is that they run a loop and compare the digits in the same position (for instance, the first post-point digits of both numbers), starting at the highest level. If one number is greater than the other, that determines the result. If the numbers in that position are identical, the loop continues until it finds a difference--and if it never does, it returns "equal." So, a number-sorting algorithm doesn't simply stop after an arbitrary number of digits. It makes use of all the data it has available.

    In a U.S. patent infringement case, the court would now have to construe "decreasing and increasing" as OPPO would give it the plain and ordinary meaning while Nokia's position would be that the numbers just have to be "materially" decreasing and increasing, with a level of tolerance that has scope for microscopic bumps. Now, "materially" or "largely" wouldn't be acceptable language, neither in a claim as granted by a patent office nor in a construction of a disputed claim limitation by a court or the ITC. Therefore, Nokia would have to propose some rather complicated language that would define a level of tolerance (such as "up to 0.000000001") for deviations from the "decreasing and increasing" rule. In that case, the numbers in the standard would be consistent with the (new) claim language, but the exercise of attempting to rewrite the claim language would make it clear that this is not what Nokia's patent actually covers. There wouldn't even be any support for that in the description. That's why I said further above that a sequential approach--first claim construction, then infringement analysis--would make Nokia lose the case. Nokia can win this only if one conflates the two steps and says that "decreasing and increasing" doesn't have to be applied with absolute mathematical precision.

OPPO's FRAND defense was discussed in a sealed courtroom. I'll try to find out what happened, but I don't have much hope given that the parties probably have a strict NDA in place. The resolution of the dispute will have to be a license agreement, as Nokia will sooner or later prevail on some SEP (and there is every indication that OPPO is prepared to pay FRAND royalties), but contrary to what I thought before, I no longer believe that EP'103 entitles Nokia to a 4G/5G injunction.

Finally, I'd like to show you two pictures taken before the court was officially in session. I'm sorry that I missed the moment when the judges had briefly taken off their masks, but I really did want to show you the Mannheim Regional Court's Second Civil Chamber for two reasons: there are many interesting cases pending before that division (not only those Nokia-OPPO but also numerous Ericsson-Apple cases, and maybe the most absurd patent-related antitrust case ever, Deutsche Telekom v. IPCom), and Judge Dr. Kircher never or only very rarely speaks at conferences and webinars. In the pictures below you see Presiding Judge Dr. Kircher in the middle, and his side judges are Judge Boettcher ("Böttcher" in German) to his left and Judge Elter (the judge rapporteur in this case) to his right:

Share with other professionals via LinkedIn: