I have an open client slot for FPGA design consulting: fpgacpu.ca/gateforge/in... Over the last 9 years, I've built control loops for nanotech research, DSP accelerators for medical, custom 10 Gb/s sensor for a satellite, video processing, industrial control, and radiation exposure testing.

Here's a common workhorse for designs: a Pipeline FIFO Buffer. Any FIFO depth is allowed, not only powers-of-2. The input-to-output latency is 2 cycles. Can function as a Circular Buffer. The tricky bit in this FIFO design is the read output, which is registered for speed. fpgacpu.ca/fpga/Pipelin...

I have an open client slot for FPGA design consulting: fpgacpu.ca/gateforge/in... Over the last 9 years, I've built control loops for nanotech research, DSP accelerators for medical, custom 10 Gb/s sensor for a satellite, video processing, industrial control, and radiation exposure testing.

A Pipeline Half-Buffer: A single pipeline register with ready/valid handshakes. Decouples the input and ouput handshakes (no combinational path), but does not allow concurrent read/write like a full Pipeline Skid Buffer. fpgacpu.ca/fpga/Pipelin...

Here's the most important pipeline building block: a Skid Buffer. It decouples two sides of a ready/valid handshake to allow back-to-back transfers without a combinational path between input and output, thus fully pipelining the path. fpgacpu.ca/fpga/Pipelin...

As I struggled today with an FPGA CAD tool's IP code generation and management, it occurred to me that I haven't needed to write HDL-generating code for years now. Extreme modularization, along with generate blocks and parameters, has always sufficed, even for 20k+ line designs.

Bit twiddling hacks are "neat tricks" in software, but they are often branch-free and need no allocated memory, so they map naturally to hardware design. More listed here: fpgacpu.ca/fpga/reading...

Here is a module which extends the 1-to-N Differential Deserializer by packing two consecutive differential N-bit deserialized words into one differential 2N-bit word and extending the bitslip word-alignment logic to support this double-width word. fpgacpu.ca/fpga/Deseria...

When aligning multiple data channels, you need to pass them through variable-delay shift registers. This is where using an FPGA's built-in shift-register hardware is a must, since using LUTs+FFs ends up being the large majority of the deserializer area! fpgacpu.ca/fpga/Registe...

A pipeline serial to parallel converter. Reads in multiple serial words and signals when the last input word has been read-in and a new, wider output word is ready to be read-out, in the same cycle if necessary to receive an uninterrupted serial stream. fpgacpu.ca/fpga/Pipelin...

A deserializer for differential data, with a configurable 1:N ratio, SDR or DDR data rate, and independent positive and negative data polarity input delays and deserializers, which enable bit and word alignment training later on. fpgacpu.ca/fpga/Deseria...

An IDELAYCTRL instance for calibrating IDELAY2 blocks. This is the first step to high-speed interfaces on Series 7 AMD/Xilinx FPGAs. fpgacpu.ca/fpga/IDELAYC...

fpgacpu.ca now supports HTTPS! It's still a static site without cookies or Javascript or any inputs, but I was getting reports of problems with browsers, corporate networks, and Google not liking the absence of SSL. Enjoy more easily now!

I have an open slot for a client at this time. Let's discuss your FPGA design needs! fpgacpu.ca/gateforge/in... I'm a consultant doing primarily logic design on FPGAs and related full-stack software development for clients who need application-specific processing for their products.

I have an open slot for a client, starting *immediately*. Let's discuss your FPGA design needs! fpgacpu.ca/gateforge/in... I'm a consultant doing primarily logic design on FPGAs and related full-stack software development for clients who need application-specific processing for their products.

In one project, I had to first configure an external device before it could generate it's own source-synchronous clock for data and control. This Clock Switchover module detects when the device data clock becomes active and stable and handles clock selection and reset. fpgacpu.ca/fpga/Clock_S...

I have an open slot for a client, starting in the last week of July. Let's discuss your needs! fpgacpu.ca/gateforge/in... I'm a consultant doing logic design on FPGAs and related software development (from kernel up to application) for clients who need application-specific processing.

We can now build a saturating accumulator with variable upper/lower limits. I used this as the Integral term of a PID controller, where the limits are dynamically adjusted to limit integral wind-up, which makes the PID controller recover from saturation faster. fpgacpu.ca/fpga/Accumul...

Here's a Saturating Adder/Subtractor. The limits can be any signed numbers, even dynamically variable. One use is as the final stage of a PID controller, so we can guarantee the output never wraps around and flips sign suddenly, which would be catastrophic. fpgacpu.ca/fpga/Adder_S...

With an Subtractor that provides the carry-out and the overflow, we can create an Arithmetic Predicates module which computes all possible signed and unsigned comparisons of two numbers. This way we don't have to worry about getting RTL arithmetic right each time. fpgacpu.ca/fpga/Arithme...

And with those previous modules, we can now build an adder/subtractor module. It's much more complex than usual, but it also gives us all bit carries and the overflow bit, and is independent of Verilog's signed/unsigned arithmetic behaviour. fpgacpu.ca/fpga/Adder_S...

Next step in getting to saturating arithmetic: a Binary Carry-In calculator, which tells us which bits in a sum had a carry-in, which has many uses (overflow, predicates, vector registers, etc...) fpgacpu.ca/fpga/CarryIn...

Let's build up to some saturating arithmetic logic. Here's a Width Adjuster, which either extends (optionally signed) or truncates values. I often use it to normalize widths of misc. signals before packing/unpacking/multiplexing/arithmetic/etc... fpgacpu.ca/fpga/Width_A...

Here's a Pipeline Credit Gate. It let's handshakes complete only if there are available credits, which are added via external pulses and subtracted by each completed handshake. It can ensure that an operation is only considered complete after N items have been read out. fpgacpu.ca/fpga/Pipelin...

I've updated my "rules for ready/valid handshakes" to describe both a control implementation I made heavy use of in the past year, and to add discussion on the synchronization operation which underlies ready/valid handshakes, much like ready/valid handshakes underlie AXI4. fpgacpu.ca/fpga/handsha...

A Handshake Multiplier, which accepts and buffers the command, and will not accept a new command until the buffered copy has been accepted a set number of times (specified in the command) by the internal module. fpgacpu.ca/fpga/Pipelin...

I built myself a little linter out of Verilator, Icarus, and some greps. It finds all the little width mismatches, any combinational loops, typos, missing ports, etc... It's a LOT faster than synthesis, and nearly eliminates CAD warnings, making finding bugs later easier. fpgacpu.ca/fpga/verilin...

When writing very modular HDL code the module instantiations become the bulk of the program. Here is a script, not a real parser just a quick hack, which reads just the start of one of my Verilog modules and spits out a roughly formatted blank instantiation template. fpgacpu.ca/fpga/generat...

In theory, any combinational path between ready and valid should be avoided. In practice, that can bloat a design with skid buffers, and misses out on some useful control logic (like that Synchronizer). Some combinational ready/valid paths are acceptable, with care. fpgacpu.ca/fpga/handsha...

My favourite thing about elastic pipelines is how they distribute control. A Pipeline Synchronizer forces all inputs to only accept new inputs simultaneously. Thus a faster input is modulated without having to touch the central control logic. fpgacpu.ca/fpga/Pipelin...

Since I got asked about documentation yesterday, here's the Python script I use: It takes Verilog code and processes the comments (line comments only) as Markdown, and wraps the code in <pre> blocks. This generates a quick, neat HTML page that uses any CSS you want for style. fpgacpu.ca/fpga/v2h.py

Let's have a look at the guts of the CDC FIFO buffer. The nucleus is the ability to pass the read and write addresses across clock domains. I do it via toggle signals, which allows non-sequential addresses. This also means we are not limited to power-of-2 buffer depths. fpgacpu.ca/fpga/CDC_Wor...

Following up on the CDC FIFO Repacker: it is a specialization of this plain CDC FIFO Buffer. The main difference is the number of counters used to track where the data begins and end. It also does CDC of whole words without using Grey Codes. :) fpgacpu.ca/fpga/CDC_FIF...

Here's a CDC FIFO Repacker, a hardware module to pack/unpack arbitrary data without stalls and without introducing gaps in the data words, while crossing clock domains. The tradeoff I made here was to sacrifice area for simplicity and speed. fpgacpu.ca/fpga/CDC_FIF...

If a memory has dual use, the data sizes for each use case may not match. You can pre-compute the Least Common Multiple of the data sizes to make sure either use case can store an integer number of items. This simplifies addressing and multiplexing. (more on this later...) fpgacpu.ca/fpga/lcm_fun...

When you focus on synthesizable-only Verilog like I do, it's easy to forget that you can do arbitrary computations during elaboration to compute constants for synthesis. Here's an example: a helper function to compute the Greatest Common Divisor of two integers. fpgacpu.ca/fpga/gcd_fun...

For analog, there is Spice. It's hard to use, and fails to simulate in obscure ways. I'm a fan of CircuitJS1, which is so interactive, and portable: github.com/sharpie7/cir... And I just discovered CircuitShepherd, which is even more amazing (though for iOS/Mac only): www.circuitshepherd.com

Someone asked "what application areas within FPGA design have the most potential for consulting opportunities?" I've seen two primary application areas in my FPGA consulting practice. One is ordinary. One is esoteric. People usually need one or both when they need an FPGA.

A sketch of a fourth-order low-pass filter circuit using plain RLC design with floating gyrators to synthesize the inductors. This design makes LPFs easier to calculate. Floating gyrators enables many other active filter circuits not normally feasible. fpgacpu.ca/bass/floatin...

A fourth-order high-pass filter circuit. However, instead of the usual multiple-feedback or Sallen-Key designs I use a plain RLC design with conventional grounded gyrators to synthesize the inductors. An RLC filter's parameters are very easy to calculate. fpgacpu.ca/bass/gyrator...

When I look at the logs on my website, the clog2() Verilog-2001 function is the most searched query. This implementation deals with a quirk of Vivado to reduce warnings. It's useful to compute address and counter widths from other parameters. fpgacpu.ca/fpga/clog2_f...

I am currently in negotiations to fill my two open slots for new FPGA design clients. These will fill fast. Let's discuss your needs! fpgacpu.ca/gateforge/in...

I normally post about digital logic, but when I looked for analog circuits to generate *only* even harmonics, I couldn't find any. So I designed one. It's simple, but tricky: any extra non-linearity will add odd harmonics. If there is another solution, please let me know. fpgacpu.ca/bass/even_ha...

I'm writing an online book: "FPGA Design Elements", which provides a reference library of fundamental digital logic design elements. Think of it as a hardware analog to the C Standard Library ("libc") and its documentation. fpgacpu.ca/fpga/index.h... github.com/laforest/FPG... #FPGA