Matthias Függer
10 references under this name, matched as it is written. Somebody else may write under it too, and the same person may appear here spelled another way.
Matthias Függer
Matthias Függer, of the ECS Group at TU Wien and later CNRS and LSV at ENS Paris-Saclay. Ten records, which makes him one of the most heavily represented authors in this corpus, and the reason is that two of its central lines run through him.
**The involution line, which is about whether a binary model can be faithful to a circuit at all.** fugger2016unfaithful proves that no continuous-time binary value-domain model captures glitch propagation faithfully, by showing that every bounded single-history channel gets short-pulse filtration wrong in one direction or the other. fuegger2014faithful answers it with involution channels, delay functions that are their own inverse and have no lower bound, and shows they solve short-pulse filtration precisely when physical circuits do. najvirt2015experimental takes that to silicon: SPICE in two technologies down to near sub-threshold supply, and a purpose-built UMC-90 ASIC read by on-chip amplifiers. fugger2019faithful is the journal statement, maier2021composable makes the model composable and answers the multi-input-gate limitation its predecessor named, and ohlinger2021involution is the tool. Six records, one argument, carried from an impossibility theorem through a repair to a measurement and a piece of software.
**The metastability line, mostly with Christoph Lenzen.** bund2020synchronizerfree removes the synchroniser from a digital link controller; tarawneh2017metastability confines the corruption metastability causes to a subset of computations and trades reliability for latency; fugger2022fast is clock frequency adaptation for voltage droop tolerance. These sit beside friedrichs2018metastabilitycontaining and bund2026codes in the corpus, which are Lenzen's without him.
**What is worth noticing about the body of work as a whole.** It is unusual in this corpus for proving impossibility results and then not stopping. The 2016 theorem would be a natural place to conclude that the boolean abstraction cannot be repaired; instead the same group identified which assumption carried the obstruction, removed it, built a model without it, validated that model against fabricated silicon, made it composable, and shipped a tool. The corpus holds few programmes that complete in that way, and this one is the clearest.
elshehaby2023susceptibility on transient faults in QDI circuits is the one member outside both lines.