WaveDrom

Rendering Beautiful Waveforms from Plain Text

Aliaksei Chapyzhenka
Jonah Probell

www.wavedrom.com

Abstract
WaveDrom is a tool for rendering digital timing diagrams, and other technical visualization, as SVG
or PNG images from an intuitive plain text language. It is easy to learn, easy to use, and browser-
based. WaveDrom is free open source software.

http://wavedrom.com/

This paper provides simple and complex examples of code and its resulting graphics. Also provided
are guidelines on integrating WaveDrom into source code, diffing waveforms, parameterizing
waveforms, describing and visualizing constraints, expressing design intent, and generating
assertions.
SNUG 2016

Table of Contents
Introduction ............................................................................................................................................................................... 3

Problems with current practices....................................................................................................................................... 3

Inspiration .................................................................................................................................................................................. 4

Exchange format (WaveJSON)............................................................................................................................................ 5

Image rendering ....................................................................................................................................................................... 5

Waveforms in code comments ........................................................................................................................................... 8

Diffing waveforms ................................................................................................................................................................... 9

Parameterization ................................................................................................................................................................... 10

Representing constraints in WaveJSON ....................................................................................................................... 11

Generating assertions .......................................................................................................................................................... 13

Other users ............................................................................................................................................................................... 16

Conclusion ................................................................................................................................................................................ 16

Page 2 WaveDrom: Rendering Beautiful Waveforms from Plain Text

SNUG 2016

Introduction
Digital timing diagrams (“waveforms”) are the type of diagram that best describe how electronic
signals do, or should, behave. Chip designers naturally draw waveforms on whiteboards, and
routinely view waves when debugging simulations. When called upon to write a specification, we
know our word processing options. However, there has never been a simple, widely used tool for
rendering waveforms. That is until now.

WaveDrom is a tool for rendering waveforms as SVG or PNG images.1 It defines a language that uses
textual symbols, which are easy to create with a keyboard, to represent the types of information
that waveforms present. WaveDrom follows the software development trend of running
applications in browsers for automatic portability across operating systems. The WaveDrom editor
(online and offline version) is WYSIWYG. The effect of changes to text describing waveforms is
immediately rendered, without a compilation step. Furthermore, since WaveDrom waveforms are
described as text, they can be embedded in source code comments, in source control metadata, and
diffed between versions.

WaveDrom was first created by this paper’s author, Aliaksei Chapyzhenka. It is available online, as
desktop application (Windows, Linux, OSX), and open source on github.2 WaveDrom is well
documented online with a fun interactive tutorial.3 Others have contributed ideas, and WaveDrom
has been extended for:
● drawing schematic diagrams
● generating register field documentation
● a translator to TikZ for TeX / LaTeX typesetting4

This paper intends to stimulate ideas for other extensions and other uses.

Problems with current practices

Waveforms are a common, well understood, industry standard way to express hardware behavior.
When done well, they effectively communicate ideas visually, in a readable way.5

Some readers of this paper draw waveforms as ASCII text. Some draw diagrams manually within a
word processor’s graphics tools. Some copy-paste lines and hexagons in drawing programs. Some
cleverly use cell outlining features in a spreadsheet, and some will even write Verilog with #delay

1 Scalable Vector Graphics (SVG) is a vector-based graphics format recognized by many document editing
programs. Portable Network Graphics (PNG) is a raster graphics format, recognized by many document
editing programs. PNG is comparable to the Joint Photographic Experts Group (JPEG) format.
2 https://github.com/drom/wavedrom
3 http://wavedrom.com/tutorial
4 TikZ is a description format for vector graphics rendered in TeX and LaTeX documents. Describing timing

diagrams in TikZ is not intuitive. See http://www.texample.net/tikz/examples/timing-diagram/.

5 Tufte, Edward. Envisioning Information, 1990, and

Tufte, Edward. The Visual Display of Quantitative Information, 2001.

Page 3 WaveDrom: Rendering Beautiful Waveforms from Plain Text

SNUG 2016

statements, run a simulation with a trace file, and take screenshots of the traces in a waveform
viewer GUI. Various commercial and freeware tools are available for download and installation on
PCs to produce waveforms, but none is widely used.

All such methods have important limitations:

1. Not everyone is a visual artist, especially at work, and especially during routine
work
2. It takes time to make a good drawing
3. Different authors draw with different and inconsistent styles
4. Drawing waveforms directly intermingles presentation and content
5. It is difficult to version control, change control, and compare waveforms
6. Waveforms are not interactive
7. It is difficult to assemble or concatenate multiple waveforms
8. Waveforms are not parameterized

As a result, designers avoid drawing waveforms where they could be helpful. When designers draw
waveforms, they require editors and technical writers to redo the drawings for clarity.

WaveDrom resolves all of these limitations. Furthermore, because WaveDrom uses a textual
language it can serve as a modeling language that computers can parse. Designers at several
companies use WaveDrom descriptions as part of hardware design entry, design capture, or test
case entry. Furthermore, some designers parse simulation traces to generate WaveDrom
waveforms for selected signals.

Inspiration
The concept of using textual (markup) language as a source for diagram rendering engine is not
new. Many domain specific language exist to drive or hint diagram rendering process. There is full
spectrum of language approaches.

One approach is to provide some domain specific computer language that can be utilize standard
computer language processing tools like lexer and parser, and processed than as abstract syntax
tree. For example DOT to describe graphs or PlantUML to describe UML diagrams.

Another approach is to employs the fact that block of monospace font text can have meaningful
alignment that can be understood by the computer program. The ASCII graphics type descriptions
like ditaa or shaape rely on fact of 2D alignment to render sophisticated block diagrams with
desired spacial relationship of it’s parts, and certain color, shape, size effects.

Some of the diagram languages employ standard mark languages (typically XML) to capture
structure and content. For example: CML (Chemical Markup Language) For graphic rendering of the
molecular structure of chemical compounds, or MathML (Mathematical Markup Language) to
describe mathematical notations. JSON is new emerging standard for data object representation
replacing XML in many areas.

Page 4 WaveDrom: Rendering Beautiful Waveforms from Plain Text

SNUG 2016

Exchange format (WaveJSON)

WaveJSON is the compact exchange format used by WaveDrom to render waveforms.6 It is based on
the JSON language syntax. WaveJSON is an Embedded Domain Specific Language.7 It uses the
industry-standard JSON format with additional semantics embedded into the JavaScript object
structure syntax.8

WaveJSON satisfies the following criteria for a useful waveform exchange format:
● Text format
○ Plain text editor can be used
○ Waveform code can be inserted into HDL or software code as comment
○ Code can be inserted into wiki, markdown, literate, or Asciidoctor documents9
● Machine readable
○ JSON data object notation language
○ Has JSON schema: https://github.com/wavedrom/wavedrom-schema that can be
used for data validation.
● Domain specific language (DSL)
○ Great expressive power within the domain
○ Exhibits minimum redundancy of domain specific knowledge definition
○ Is an extension of a general purpose language (i.e. JavaScript)
● Open Source format and tools
○ WaveDrom : WaveJSON → SVG rendering engine in JavaScript
https://github.com/drom/wavedrom
○ WaveDromEditor : WaveJSON editor for Desktop and Browser
https://github.com/wavedrom/wavedrom.github.io

Image rendering
The first and most natural use for WaveDrom is rendering of images for publishing purposes.10
Here is a code example:

6 The WaveJSON format specification is at https://github.com/drom/wavedrom/wiki/WaveJSON

7 Mernik, Marjan, Jan Heering, and Anthony M. Sloane. "When and how to develop domain-specific languages."
ACM computing surveys (CSUR) 37.4 (2005): 316-344.
Knuth, Donald Ervin. The texbook. Vol. 1993. Reading, Massachusetts: Addison-Wesley, 1986.
Spinellis, Diomidis. "Notable design patterns for domain-specific languages."Journal of systems and software
56.1 (2001): 91-99.
8 JavaScript Object Notation (JSON) is an open standard format that uses human-readable text to transmit

data objects consisting of attribute–value pairs. More information at http://json.org/.

9 Knuth, Donald E. "Literate programming." CSLI Lecture Notes, Stanford, CA: Center for the Study of Language

and Information (CSLI), 1992 1 (1992).

Asciidoctor diagram extension at https://github.com/asciidoctor/asciidoctor-diagram
10 The online editor is available at http://wavedrom.com/editor.html.

The desktop application is available at https://github.com/wavedrom/wavedrom.github.io/releases.

Page 5 WaveDrom: Rendering Beautiful Waveforms from Plain Text

SNUG 2016

{signal: [
{wave: '01010101010101'}, // toggling
{wave: '0.1.0.hl'}, // dot(.) holds a value, h/l for high and low
{wave: '222xxx345.6789'}, // multi-bit, X, 345 = colors, 6789 == x
{wave:''}, // blank line

// text
{wave: '2.2.2.2.2.2.2.',
data:["abcdefg","hijk","lmnop","qrs","tuv","wx","yz"]},
{wave:''},

// names and clocks

{name: "posclk", wave: 'pPp...........' }, // capital letters for arrows
{name: "negclk", wave: 'n.N..n........' },
{name: "divclk", wave: 'lplpl.h.l.h.pl' },
{wave:''},

// fun
{name: "Barak", wave: '01.zx=ud.23.45'},

// gaps
{name: "gaps", wave: '01|022|0|x|.22' },

// arrows with nodes and edges

{name: "arrows", wave: '0n0....2..x2..',
node: '.a........d' },
{ wave: '1.0.10..x0....',
node: '....b...c' },
],
edge:['a~>b glitch',
'c<~>d I found the bug!',
],
}

It renders this waveform:

Page 6 WaveDrom: Rendering Beautiful Waveforms from Plain Text

SNUG 2016

The online tutorial has more examples and more sophisticated examples.

WaveDrom render produces SVG or PNG. Standard skins can be used, but custom skins are
supported for specific stylistic formatting. WaveDrom allows text to have XML style attributes.11
WaveDrom has a two panel editor that renders WaveJSON source into SVG image in realtime. The
panels can be rotated to show the source and waveform horizontally or vertically. A user menu is in
the lower right.

11 Package available at https://github.com/drom/tspan

Page 7 WaveDrom: Rendering Beautiful Waveforms from Plain Text

SNUG 2016

The instant feedback of real time rendering gives a designer an ability to draw the intended
waveform and correct mistakes very quickly.

The following source code describes a DDR DRAM access:

{ signal: [
{ name: "CK", wave: "P.......", period: 2 },
{ name: "CMD", wave: "x.3x=x4x=x=x=x=x",
data: "RAS NOP CAS NOP NOP NOP NOP", phase: 0.5 },
{ name: "ADDR", wave: "x.=x..=x........",
data: "ROW COL", phase: 0.5 },
{ name: "DQS", wave: "z.......0.1010z." },
{ name: "DQ", wave: "z.........5555z.",
data: "D0 D1 D2 D3" }
]}

Below is the rendered waveform.

Waveforms in code comments

The following Verilog module code includes a WaveDrom waveform in a code comment:

Page 8 WaveDrom: Rendering Beautiful Waveforms from Plain Text

SNUG 2016

module reducer(big, little)

input [7:0] big;
output [2:0] little;

/* Waveform of example expected behavior

{signal: [
{name: 'big',
wave: '=======',
data: '0x0 0x1 0x5 0x10 0x3F 0x80 0xFF'},
{name: 'little',
wave: '=======',
data: '0 0 3 4 5 7 7'},
],}
*/
always @(*) begin
little = 0;
case(big)
8’h02: little = 1;
8’h04: little = 2;
8’h08: little = 3;
8’h10: little = 4;
8’h20: little = 5;
8’h40: little = 6;
8’h80: little = 7;
endcase
end
endmodule

The waveform code can be rendered for purposes of documentation. It would look like this.

Diffing waveforms
Though waveforms are the best way to display timing information in two dimensions, even when
placed side by side it is very difficult to visually compare one to another to identify changes or
differences. However, by representing waveforms as text, it is much easier to identify the difference
between visual waveforms.

Page 9 WaveDrom: Rendering Beautiful Waveforms from Plain Text

SNUG 2016

Similarly, when a simulation or logic analyzer dumps a large amount of trace data as text, it is
difficult to understand the diff in context. As long as the trace dump is in WaveDrom format, or can
be easily converted to WaveDrom format by a script, then rendering the waveform as an image and
looking at the point of the difference makes it much easier to understand.

Parameterization
WaveDrom is built on JavaScript. Therefore, it is extensible. The example below is a waveform for a
parameterizable memory block. It can return data in the same cycle, or with some latency. A
JavaScript functions makes the read latency parameterizable, based on the variable rdlatency.

(function (o) {
var rdata = 'x.';
rdata += '.'.repeat(o.rdlatency);
rdata += '=..x';
rdata += '.'.repeat(4 - o.rdlatency);
return {
signal: [
{name: 'clk', wave: 'p.........'},
{name: 'addr', wave: 'x=........', data: 'A<sub>0'},
{name: 'rd', wave: '0.1..0....'},
{name: 'rdata', wave: rdata, data: 'D<sub>0'}
]
};
})({rdlatency: 1}) /* <-- change this */

Page 10 WaveDrom: Rendering Beautiful Waveforms from Plain Text

SNUG 2016

Representing constraints in WaveJSON

WaveJSON has several features that can be used to describe behavioural constraints in the (human |
computer) (readable | writable) language. Therefore, these features can facilitate human-machine
interaction:
● [signal[].wave] property represents one character per cycle, so it is possible to represent
alignment of several signals.
● Vertical bar (|) character represents gaps in the timing.
● [signal[].node] property allows placing anchors aligned with [signal[].wave] in time.
● [edge] property describes dependencies between the nodes with optional textual attributes.

Consider the following code.

{signal: [
{name: 'addr', wave: 'x=|', node: '.a..', data: 'a0'}, {},
{name: 'ack', wave: '0.|1', node: '...b'},
{name: 'data', wave: 'x.|=', node: '...c', data: 'd0'}
],
edge: ['a|->b [2:5]', 'b-c']
}

That yields the following waveform.

Several system constraints can be drawn and analyzed from this description by human or
computer. For example:
1. one may expect ack to rise in 2 to 5 cycles after addr is asserted.
2. The data signal should carry some payload, and be asserted on the same cycle.

If the designer describes that expected behavior, then a computer can use it to:

Page 11 WaveDrom: Rendering Beautiful Waveforms from Plain Text

SNUG 2016

● Create a test case, and generate testbench code that behaves like the description, varying
parameters on the edges
● Assert the constraint during simulation or for formal analysis
● In the case of a mismatch (case failure), a computer can back-annotate the waveform,
showing the difference with the actual behaviour, and use it for reporting.

A report may contain multiple failing cases that look like this:
{signal: [
{name: 'addr', wave: 'x=x..', node: '.a..', data: '345'}, {},
{name: 'ack', wave: '0.10.', node: '..b'}, {},
{name: 'data', wave: 'x..=x', node: '...c', data: '825'}
],
edge: ['a--b expected [2:5] got 1', 'b--c expected 0 got 1'],
head: {tock: 123}
}

rendered form:

The report may have a timestamp or cycle number in a waveform header or footer section as a
reference to the simulation time.

Specific edges may contain more failure information.

Below is code to represent a common protocol violation on an AMBA AXI read transaction.12 Text
such as this could be generated by protocol checkers as part of verification IP packages:
{ signal: [
{ name: "CLK", wave: 'p..........', },
{ name: "AR_VALID", wave: '01.0.......', },
{ name: "AR_READY", wave: '0.10.......', },
{ name: "AR_ADDR", wave: 'x=.x.......',
data: "0x0000", },
{ name: "R_VALID", wave: '0....1....0' },
{ name: "R_READY", wave: '1.....01...',

12 See http://www.arm.com/products/system-ip/amba-specifications.php

Page 12 WaveDrom: Rendering Beautiful Waveforms from Plain Text

SNUG 2016

node: '......a....', },
{ name: "R_DATA", wave: 'x....====x.',
data: "D0 D1 D2 D3",
node: '.......b...', },
],
edge:['a~>b protocol violation: data change when READY is low'],
}

Below is the rendered waveform.

Generating assertions
It is possible for a user to express constraints using WaveJson in a way that they can both be
rendered visually and translated, such as with a Perl script, into standard SystemVerilog assertions.
There are an infinite number of ways to express constraints in WaveJson that make different trade-
offs between readability and extensibility. Users should choose their format based on the range of
types of assertions they intend to create, and what display format they find most readable.

The authors of this paper are not recommending any particular syntax for assertion expression.
Creating a script to perform a translation to SystemVerilog assertions is left as an exercise to the
reader. Below are some possible examples.

The following diagram is rendered from the following WaveDrom code (color added for emphasis).

Page 13 WaveDrom: Rendering Beautiful Waveforms from Plain Text

SNUG 2016

{ signal: [
{name: 'REQ', wave: '010', node: '.ab'},
{name: 'GNT', wave: '0|1', node: '..c'}
],
edge: ['a|>b ##[1]',
'a|>c ##[1:4]']
}

A simple translation of the WaveDrom code yields a property assertion for each edge such as the
following SystemVerilog assertions.

assert property (REQ |-> ##[1] ~REQ); // a|b

assert property (REQ |-> ##[1:4] GNT); // a|c

The translation uses the state, 1 or 0, of the wave corresponding to each node value to determine
the state of the assertion implication (~ inversion). The translator uses SystemVerilog assertion
syntax in the WaveDrom edge text. Other implementations could use any syntax appropriate for
describing assertions. This opens a new level of assertion expression possibilities.

In another example, the following WaveDrom code

{signal: [
{name: 'addr', wave: 'x=|', node: '.a..', data: 'a0'}, {},
{name: 'ack', wave: '0.|1', node: '...b'},
{name: 'data', wave: 'x.|=', node: '...c', data: 'd0'}
],
edge: ['a|->b ##[2:5]', 'b-c']
}

yields

assert property (addr |-> ##[2:5] ack);

assert property (ack |-> data);

The rendered diagram looks like this.

The dot character (‘.’) in WaveJson means repetition of the value of the previous cycle, which
naturally maps to the SystemVerilog $stable construct.

Page 14 WaveDrom: Rendering Beautiful Waveforms from Plain Text

SNUG 2016

{ signal: [
{ name: "READ", wave: "0.1..", node: '=.a..' },
{ name: "DATA", wave: "=....", node: '=..b.' },],
edge: ['a|->b']
}

One particular translation of the code above finds that the b node corresponds to a . in the DATA
wave. Since the a node represents a transition to 1 in the READ wave, the following SystemVerilog
assertion results.

assert property (READ |=> $stable(DATA);

The assertion waveform is rendered below.

The translator converts the following WaveJson code

{signal: [
{name: 'full', wave: '1..', node: '.a..'},
{name: 'write', wave: '1..', node: '.b..'},
{name: 'read', wave: '0..', node: '.c..'},
{name: 'wptr', wave: 'x=.', node: '..d.'},
],
edge: ['a&->d', 'b&->d', 'c&->d']
}

to the following SystemVerilog assertion.

/* full && write && !read |=> $stable(wptr) */

Note that node d is the right hand side node for each edge, implying that the state of its signal
(wptr) is the target of the assertion. The assertion waveform is rendered below.

Page 15 WaveDrom: Rendering Beautiful Waveforms from Plain Text

SNUG 2016

Other users
The web version of WaveDrom is used by tens to hundreds of users at each of: Intel, AMD, Apple,
Synopsys, Analog Devices, Nvidia, Cisco, National Semi, TI, ARM, Nvidia, HP, Marvell, Hynix,
Infineon, Altera, Avago, Qualcomm, Volkswagen, Ford, and others.

Many others outside of the semiconductor industry use or have adapted WaveDrom.

Tutorial by Dave Vandenbout: http://www.xess.com/blog/timing-diagrams-made-easy/

Hackaday comparison to other waveform drawing applications:

http://hackaday.com/2015/05/25/need-timing-diagrams-try-wavedrom/

PSHDL Verilog coding game: http://waves.pshdl.org/

Yosys synthesis Verilog code verification game:

http://www.clifford.at/yosys/nogit/YosysJS/snapshot/demo03.html

Coroutine Co-simulation Test Bench (COCOTB) simulation trace to waveform usage:

http://potential.ventures/beautiful-documentation.html

BeagleBone BeagleLogic logic analyzer result display: http://beaglelogic.github.io/webapp/

A simple function to add wavedrom waveforms into an ipython notebook:

https://github.com/witchard/ipython-wavedrom

A Trac plugin: https://github.com/jun66j5/trac-wavedromplugin

Asciidoctor Diagram extension: https://github.com/asciidoctor/asciidoctor-diagram

What will you do?

Conclusion
Current practices for drawing waveforms do not support a fast, easy, error-free workflow.
WaveDrom does so with an intuitive plain text language for describing waveforms. A free, open
source editor renders clear diagrams with a consistent format. WaveDrom text is easy to embed in
source code comments. Furthermore, WaveDrom text allows for diffing waveforms. Waveforms can
also be parameterized. Waveforms becomes a source for describing and for visualizing constraints.
Edges within waveforms naturally indicate relationships, and can be translated to other expression
syntaxes such as SystemVerilog assertions. A lot of companies already use WaveDrom in their
design work. Shouldn’t you?

Page 16 WaveDrom: Rendering Beautiful Waveforms from Plain Text