INSIGHTS

Engineering case studies

How we design, generate, and prove verified FPGA IP, written for the engineers who evaluate it. Every number in these pieces is a measured result.

A deep-space decoder that was wrong on every frame but the first

Case study · CCSDS FEC on silicon · 7 min

A CCSDS AR4JA LDPC decoder matched its golden model bit for bit in simulation across all nine code points, then lost the head of every frame after the first on real hardware. The cause was not in the decoder, and the vendor manual named it faster than a waveform would have.

Reproducibility gate81 / 81 runs MEASURED
Frames on silicon vs golden model675, 0 bit errors MEASURED
Read the case study

Updating a deployed FPGA over its own data link

Case study · Remote FPGA update · 7 min

A second FPGA on the board used to need a service cable for every firmware change. The update now runs over the 10-gigabit link the two chips already share: the image is written into the configuration flash beside the target and read back byte by byte, then one in-band command reboots the chip into it. Proving which copy it booted took a counter, not a flag.

Firmware programmed over the link9.54 MB, 0 bytes wrong MEASURED
Erase, program and verify146 sectors in 295 s MEASURED
Read the case study

Turning a second FPGA into an extension of the first

Case study · Multi-FPGA system · 8 min

Two large FPGAs on one board over eight 10-gigabit lanes. Neither chip alone had the room, so the second became an extension of the first: a compute node and four gigabytes of remote memory, each verified byte-exact on real silicon. Bringing up the link meant catching a reversed wiring record that every simulation had faithfully obeyed.

Remote decoder, model vs hardware0 errors MEASURED
Recovered lane, ten-minute soak6 trillion bits, 0 errors MEASURED
Read the case study

Why we automate the path from algorithm to silicon

Perspective · Our mission · 6 min

High-value radio and FPGA hardware sits idle because crossing from an algorithm to a working system is still done by hand. What we are building to change that, and how bit-exact validation makes AI-generated hardware trustworthy.

Wi-Fi modes over the air8 / 8 bit-exact MEASURED
RTL vs reference model0 LSB MEASURED
Read the perspective

A Wi-Fi 5 transceiver on a custom RFSoC board

Case study · Wi-Fi 5 on RFSoC · 8 min

A reference transceiver written for a vendor evaluation board, moved onto custom hardware over a cable loopback. Two failures passed every simulation and only showed up on silicon: a defect the build itself introduced, and one converter configuration field.

Format and rate combinations decoding24 of 24 MEASURED
Payload compared byte by byte0 errors in 192 B MEASURED
Read the case study

Bringing up Linux on a new RFSoC board in a day

Case study · RFSoC bring-up · 7 min

A partner trial board arrived with no operating system on it. An AI drove the whole bring-up, proving each layer of hardware before building the next, then built and installed Linux and booted it to a login prompt reachable over ssh.

Gigabit Ethernet on the board948 Mbps MEASURED
Processor memory pattern sweep0 errors MEASURED
Read the case study

Putting a quantum error-correction decoder on an FPGA

Case study · Quantum LDPC · 9 min

A first-principles re-build of IBM's Relay-BP decoder for the gross quantum code, generated from a model proven bit-exact at every layer. It lands on the published silicon envelope, and the genuine headroom is algorithmic: stop early when a decode stalls, which cuts worst-case latency without costing accuracy.

Worst-case iterations, early stop-32% MEASURED
Model vs RTL, every test0 LSB MEASURED
Read the case study

AI-driven FPGA resource optimization of a Viterbi decoder

Case study · Convolutional FEC · 8 min

A generated decoder, handed to an AI to shrink its FPGA footprint. AI folded the logic to a third, diagnosed that the lost clock was a micro-architecture problem not a folding one, swapped in a rotating read to recover 160 MHz, backed out of a measured dead end, and proved every change bit-exact.

Logic across the fold1,919 to 681 LUT MEASURED
In a full 802.11a receiver8/8 schemes, 0 errors MEASURED
Read the case study

An AI-generated Wi-Fi receiver, verified bit for bit

Case study · 802.11a receiver · 6 min

A complete 802.11a receiver went from a MATLAB reference design to bit-exact FPGA hardware no one wrote by hand, and recovered a MATLAB-generated waveform with zero bit errors. How every layer was checked against the standard.

End-to-end recovery0 errors MEASURED
Receive clock (Zynq-7010)162.7 MHz MEASURED
Read the case study

An AI-generated 5G receiver that reads real cells

Case study · 5G NR SSB receiver · 7 min

A 5G NR cell-search receiver went from a reference design to FPGA hardware no one wrote by hand, and read three real over-the-air cells to a consistent broadcast message. The detector is 38% smaller than an earlier HLS implementation.

Real cells decoded3 / 3 CRC clean MEASURED
Detector vs earlier HLS38% fewer LUTs MEASURED
Read the case study

An AI-generated LDPC decoder, clocked past a commercial IP

Case study · 5G NR LDPC · 9 min

A 5G LDPC decoder, generated as RTL from a Python algorithm, optimized from 221 to 463 MHz, past a paid commercial IP on the same FPGA. The hard part was the data dependency between layers, not where to add a register.

Clock vs commercial IP463 > 459 MHz MEASURED
Match to 3GPP referencebit-exact MEASURED
Read the case study

Twenty 5G LDPC decoders from one automated flow

Case study · 5G NR LDPC · 9 min

One parameterized flow generates, timing-closes, and bit-exact verifies twenty 5G LDPC decoder configurations. Tuning one design is a special case; automating the whole batch, and proving every one, is a method.

Configs at/above IP clock19 / 20 MEASURED
Match to 3GPP reference20 / 20 MEASURED
Read the case study