What we take on

Five kinds of engagement

  • Datapath and interconnect architecture Bandwidth and latency budgets across a whole design before anything is written: port widths, clock domains, buffer depths, arbitration policy, and where the rate will actually be lost. Delivered as a document with the arithmetic in it, not a block diagram.
  • Host interface and driver work PCI Express integration, DMA descriptor design, and the driver on the other side of it — Linux, DPDK or Windows. Usually the fastest way to recover bandwidth from a design that trains a link and then underuses it.
  • Timing closure and CDC audit A design that fails intermittently is usually a crossing or a constraint, not luck. We audit crossings and resets against a formal model, produce the constraints, and close the timing. This is the engagement we are asked for most often before a tape-out.
  • Board bring-up Converter links, transceivers and memory interfaces on new hardware, including the parts that only fail on the third board or at the corner of the temperature range.
  • Design rescue An inherited design, no author available, a deadline. We start with measurement rather than opinion and tell you in the first week whether it is repairable and what it will take — including when the answer is that it is not.

How it works

First conversationTechnical, free, with the engineer who would do the work
ScopingFixed-price study for anything we cannot scope from a call
ContractingFixed price per milestone, or time and materials
IP ownershipWork product is yours; our existing cores stay licensed
ReportingWeekly, with measurements rather than percentages
We decline work we are not equipped to do well, and we say so at the first conversation rather than at integration. A wrong-fit engagement costs us more in support and reputation than it earns in fees.

Signal processing instead?

Algorithm-to-RTL work, modem and FEC implementation and radio front-end design are our sister company's.

AlfaRF design services

Describe the problem, not the solution.

What is measured, what was expected, and what has already been tried. That is enough for a useful first answer.