IC Layout Using Magic
Simple Inverter Tutorial


Magic is an interactive system for creating and modifying VLSI circuit layouts. With Magic, you use a color graphics display and a mouse to design basic cells and to combine them hierarchically into larger structures. Magic has built-in knowledge of layout rules - as you are editing, it continuously checks for rule violations. It has a built-in circuit extractor: from the layout, the extractor creates an output file suitable for use with circuit simulation tools such as Spice (for detailed, mixed-signal simulation) or IRSIM (for digital simulation).

Starting Magic, Text and Graphics Windows

Magic must be run from a workstation that supports the Unix X Windows system (although it can be easily compiled for Linux at home. Please contact me if you need help in doing this). Once you login to the Unix machine that runs Magic and setup your environmental variables (using source magic.cshrc), Magic is started by the shell command:
prompt> magic [-Ttechnology] [file without .mag extension]
where technology is the name of the fabrication technology that the design is intended for, and file is the name of the file that you edit. You could type any file name for file, and Magic would start editing that file. If a file with this name does not exist, Magic will create it. If the technology option is omitted, the layout is done in a default technology. Fortunately, I have set up the default technology as AMI C5N 0.5u which is the technology we plan on using for this class. Therefore, just invoking magic is sufficient. If you wish to learn how to set up the default technology, feel free to contact us.

After starting Magic, a graphics (paint) window pops up, and a ">" prompt appears in the text (shell) window. Magic starts editing the file.

Magic is not a menu-driven program: the layout shows up in the graphics window, but all commands must be entered in the text window. The graphics window is an ordinary X window, and can be moved and resized using the mouse. With a two-button mouse (on a PC with X-server), pressing both buttons at the same time is equivalent to pressing the middle button (which is available on Unix systems). The graphics window is where you draw your layout. To work in this window, move the cursor to the window and click the left mouse button. This makes the graphics window active. The mouse should always point to the graphics window.

The text window is the same window from which you started Magic. It is convenient to position this window under the graphics window so that both windows are visible at the same time. All commands you want to invoke should be typed in the text window. All error messages are also displayed in this window. Commands can be invoked in Magic in three ways:

Before trying to use any command, make sure you have the graphics window active and that the mouse cursor is pointing to the active graphics window. The commands show up in the text window, but affect only the active graphics window.

Scalable Design, Lambda and the Grid

Magic uses what is called scalable or "lambda-based" design. In scalable design, layout items are aligned to a grid which represents a basic unit of spacing. For a particular technology, lambda represents an actual distance (e.g., lambda = 0.3 um for the AMI C5N 0.5u SCMOS technology). Once again, If you plan on setting up Magic to run at home, I highly suggest talking to us, so we can let you know what files to modify and also where to add the technology file. However, in Magic all designs are done in terms of lambda, not in terms of actual distances. The paint window is a layout surface on which items can be placed with a resolution of one lambda.

Magic has a grid which can be set to an arbitrary multiple of lambda. The default value of this grid is one lambda by one lambda. The grid of n lambda by n lambda can be displayed using the :grid n command. The g macro toggles the grid on and off.


The Cursor and the Tools

The mouse cursor is used to select, point to, and manipulate objects on the graphics window. The shape of the cursor indicates which of the four available tools is currently being used. Pressing the space bar macro toggles between the 4 tools available:

Box Tool - this is the default tool and is indicated by a crosshair cursor. It is used to position the box and to paint and erase; this is the tool used for all basic drawing tasks. The use of the box tool is described below in the Basic drawing section and in more detail in the Magic Tutorial #1: Getting Started  and Magic Tutorial #2: Basic Painting and Selection .
  • Wiring Tool  is indicated by an arrow cursor and is used for advanced drawing tasks such as wiring and plowing. See the Wiring section below and the more detailed Magic Tutorial #3: Advanced Painting
  • Netlist Tool is indicated by a box cursor. See Magic Tutorial #7: Netlist and Routing
  • Rsim Tool is indicated by a hand cursor. See Magic Tutorial #11: Using RSIM with Magic
  • The box tool is sufficient for basic editing. The purpose of the box tool is to specify a rectangular area of
    the layout for editing. The left and right mouse buttons are used to position the box. If you click on the left mouse button, the box will move so that its lower left corner is at the cursor position. If you click on the right mouse button, the upper right corner of the box will move to the cursor position, but the lower left corner will not change. The two mouse clicks are sufficient to position the box with arbitrary size anywhere on the screen.

    Some Useful Global Commands

    Here is a short list of useful global commands.  These commands can be typed in at any time.

    :help command prints out a brief description of all the commands or the specified command.

    :load circuit-name loads circuit-name into the window; if circuit-name doesn't exist, Magic creates a new empty circuit.

    :save circuit-name saves all the changes to the circuit.

    :view or v fits the window with everything drawn thus far..

    :grid or g toggles a grid.  Useful for lining up various cells, wires, etc.

    :zoom amount  zooms in and out by a factor of amount,  i.e. :zoom 2 zooms in twice as much, and :zoom 0.5 zooms out twice as much. The z macro zooms out to fit the box on the paint window. The Z macro zooms in the same as the :zoom 2 command

    :macro displays all current macros

    :quit quits Magic.


    Cells, Paint and Layers

    In Magic, a circuit layout is a hierarchical collection of cells. Each cell contains three things: colored shapes, called paint, that define the circuit's structure; textual labels attached to the paint; and subcells, which are instances of other cells.

    The two basic layout operations are painting and erasing. They can be invoked using the :paint and :erase
    commands, or using the mouse buttons.

         :paint layers (paints rectangular regions, specified by the box)

         :erase layers (deletes the specified layers from the region under the box)

    In each of these commands layers is one or more names separated by commas. In Magic there is one paint layer for each kind of conducting material (polysilicon, ndiffusion, metal1, etc.), plus one additional paint layer for each kind of transistor (n-transistor, p-transistor, etc.), and, finally, one further paint layer for each kind of contact (pcontact, ndcontact, m2contact, etc.).

    The easiest way to paint and erase is with mouse buttons. To paint, position the box over the area you'd like to paint, then move the cursor over an existing color and click the middle mouse button (i.e. click both the left and the right mouse button at the same time on a two-button mouse). To erase everything in an area, place the box over the area, move the cursor over a blank spot, and click the middle mouse button. While you are painting, white dots may occasionally appear and disappear. These are design rule violations and will be explained in Design Rule Checking.

    Here's a legend of relevant colors/layers:

     
    Name of layer 
    What the layer represents 
    p or poly or polysilicon or red  polysilicon (gate areas)
    green or ndiff or ndiffusion  n-diffusion (n+ source/drain areas)
    brown or pdiff or pdiffusion  p-diffusion (p+ source/drain areas)
    blue or metal1 or m1 metal layer 1
    purple or metal2 or m2  metal layer 2
    pw or pwell p-well
    nw or nwell n-well
    nwc or nwcontact metal1 to n-well contact (n-tub tie)
    pwc or pwcontact metal1 to p-well contact (p-tub tie)
    ndc or ndcontact metal1 to n-diffusion contact 
    pdc or pdcontact metal1 to p-diffusion contact 
    polycontact metal1 to poly contact 
    m2contact metal2 to metal1 contact (via)
    nfet or ntransistor n-transistor 
    pfet or ptransistor p-transistor 
     
    Some layers are created by crossing two layers. For example drawing poly over ndiff (or vice versa) will produce an n-transistor. Contacts are made by placing the box over the region of the contact and by painting the appropriate contact. For example, to create a contact between ndiffusion and metal1 layers, place a box over an overlap between the two layers and type :paint ndc command.

    For a complete list of layers use the Magic command :layers.


    Basic drawing
    The box and cursor are used to select things on the layout window. The left and right mouse buttons are used to position the box:
  • click on the left mouse button to position the lower left corner of the box;
  • click on the right mouse button to set the upper right corner of the box.
  • After drawing a box, you can paint layers using command :paint layer or by clicking the middle button as explained in the Cells, Paint and Layers section above. To select a certain piece of the layout, move the cursor over the block you want to select, and type s, which is a macro for :select. A thin white outline is left around the block to show it has been selected. Using the macro s repeatedly at the same spot will toggle through electrically connected layers. This is a useful method to make a quick check of how things are connected in your layout design. The macro S (macro for :select more ) is just like s except that it adds on to the selection, rather than replacing it. For a synopsis of all the options to the :select command, type :select help.  You can also select an entire area of stuff. Place the box over the area you want selected, and type :select area or macro a. If you want to know what's selected, type :what .

    The following commands can be used to maneuver the selected part of the layout:

    :move (macro t), and macros q, w, e, and r (to move by 1 lambda in different directions);
    :stretch (macro T), and macros Q, W, E, and R;
    :upsidedown
    :sideways
    :clockwise
    :copy (macro c)
    :delete (macro d)
    :erase
    The best way to learn these commands is to try them out. Often you may want to restore things after you've made mistakes using:
    :undo (macro u)
    :redo (macro U)
    Make sure that you delete any paint that you don't want. Painting over something with another color just adds more layers, your original paint will still be there underneath.

    You can use the "." (dot) macro to repeat the last command.

    The easiest way to move around the graphics window is to use the "," (comma) macro, which centers the window around the cursor position, together with the :zoom n command, (or z and Z macros) to zoom in and out.

    Basic drawing is explained in more detail in the Magic Tutorial #1: Getting Started and the Magic Tutorial #2: Basic Painting and Selection.


    Labels

    In order to make your layout readable, and to prepare your layout for extraction and simulation, you need to label it. First decide where you want your label. Then left click and right click at that spot. You should end up with a small cross (+). Then type

    :label labelname
    Label names can be anything, but it is a good idea to use the labels that will correspond to circuit signals and make the layout readable. In particular, for digital simulation using the IRSIM simulator (and some other applications that can use Magic outputs), it is necessary to label the supply wires (nodes) as Vdd and GND

    Note: Some layout editors will use the Vdd! and GND! for power supply rails. The ! means that the node is global. However, when using Magic and IRSIM together, the ! is not recommended, because it prevents IRSIM from recognizing where you want to have your supply voltage. Hence your circuit would operate without voltage and not work, even though it passed gemini.


    Wiring

    Using the Box Tool to paint every little piece of a layout may not be the most efficient. The Wiring Tool allows you to extend existing paint, and to make contacts between layers. Once you have set down a few boxes of paint, you can connect them together with "wires."

    Switch to the Wiring Tool, typing either :tool or space-bar macro until you get there. Now left click on the paint that you would like the wire to start from. Magic will automatically select the width of the wire to be the largest box that can fit in the paint you just clicked on. Then right click where you would like the wire to end. The Wiring Tool only paints in horizontal and vertical lines. Try to paint a couple of diagonals and see what happens.

    Sometimes it's necessary to make a contact between different layers. When different layers cross, they are insulated from each other. So you need to explicitly tell Magic to make a contact between the layers. To do this, first make sure that you are using the Wiring Tool (arrow cursor). Left click on one layer where you want the contact to be made. Then middle click (i.e. press both left and right button at the same time on a two-button mouse) on the layer you want to connect to. The layers don't have to be overlapping to begin with, but they must be immediately next to each other.

    Wiring and other advanced drawing techniques are explained in more detail in the Magic Tutorial #3: Advanced Painting (Wiring and Plowing).


    Design Rule Checking

    Design Rule Checking is an automatic feature of Magic. Magic knows certain rules that your layout should satisfy so that the IC can be fabricated without errors. In general, design rules specify how far apart various layers must be, or how large various aspects of the layout must be for successful fabrication, given the tolerances and other limitations of the fabrication process. As you lay out the circuit, any time you place wires too close together, paint a block too narrow, or make any other rule violations, Magic lets you know immediately by splattering small white dots around the area of concern. Once the error has been corrected the dots will disappear. For example, paint a pair of metal1 wires fairly far apart, select one, and move it closer to the other. The white dots will appear once the wires are less than 3 lambda apart. In general, you can find out what design rule is violated by clicking the cursor on the white-dots area and by typing :drc why or macro y.

    Design rule checking is explained in more detail in the Magic Tutorial #6: Design-Rule Checking.
    MOSIS Scalable CMOS (SCMOS) Design Rules specifies the complete set of design rules defined by the MOSIS VLSI fabrication service.
     


    Extraction for Simulation

    The built-in Magic extractor computes from the layout the information needed to run simulation tools such as Spice or IRSIM. The information includes the sizes and shapes of transistors, and the connectivity, resistance, and parasitic capacitances of nodes. Both capacitance to substrate and several kind of internodal coupling capacitances are extracted. Magic's extractor is both incremental and hierarchical: only part of entire layout must be re-extracted after each change, and the structure of the extracted circuit parallels the structure of the layout being extracted.

    The command

    :extract
    produces  a separate .ext file for each .mag file in a hierarchical design. If your Magic layout is just a single cell named cell-name.mag, the extraction output will be placed in the file cell-name.ext.

    If your layout includes subcells, to extract all the edited subcells, use the command:

    :extract all
    or to extract just the selected (current) cell type
    :extract cell-name
    The output will be placed in the file cell-name.ext.

    Extraction is explained in more detail in the Magic Tutorial #8: Circuit Extraction.
    The output .ext file is used to generate the netlist file suitable for simulation. Spice simulation based on the file extracted from layout is explained in the Spice page.


    A Step-by-Step Example: Layout of a CMOS Inverter

    Here is a step by step example of how to layout a CMOS logic inverter shown below:

    The inverter consists of an NMOS transistor M1 and a PMOS transistor M2. The channel width W and the channel length L of the two devices will be set at 3.0um and 6.0um, respectively. The sizings of of each transistor allow designers to custom the current-drive for each gate based on fan-in, fan-out, and other influential factors. Note that the source and the body (p- substrate) of the NMOS are connected to ground (GND node), while the source and the body (n-well) of the PMOS are connected to the positive supply (Vdd node) even though they are not shown. This is because for most introductory circuit classes, the three port model is used (shown), however, when we are dealing with VLSI design, we must utilize the four-port model which accounts for substrate connections which are vital to preventing the body-effect. We assume that the target fabrication technology is AMI C5N 0.5u CMOS process. The technology name is scn3me_subm. The unit length lambda for this technology is lambda=0.3um. Therefore, the width for both the NMOS and PMOS transistors in lambda is 10 lambda and 20 lambda, respectively (e.g. 3.0um/0.3um/lambda = 10 lambda).

    In this example, we use basic Magic drawing command to layout the inverter. Remember that you can undo just about everything by typing :undo or u.

    Start Magic from the Unix prompt:

    prompt> magic inverter
    Expand the graphics window over the screen, but so that you can still see the text window prompt. Remember that the graphics window must be active and the cursor must always point to the graphics window. Zoom out and show the lambda grid so that it is easier to see what you are doing.
    :zoom 0.5
    :grid 1
    The order in which layers are placed in the graphics window is not important, and so the layout steps described here are by no means unique. Since the device channel width W and length L are specified, we can start by painting the device active areas: p-diffusion for PMOS and n-diffusion for NMOS.

    The PMOS transistor has the channel width W=6.0um, which is equal to 20 lambda in the selected technology. The channel length L=0.6um equals 2 lambda. The design rules specify that we need at least 4 lambda for the source and drain contacts, plus at least one lambda between the poly gate and the source/drain contacts. Using left and right mouse click, make a box 20 lambda high and 2+4+4+1+1=12 lambda wide. The command:

    :paint pdiff
    paints the PMOS active area (p-diffusion). The NMOS transistor has the channel width W=3.0um, which is equal to 10 lambda, and the channel length L=0.6um (2 lambda). Make a 10 lambda (height) by 12 lambda (width) box under the p-diffusion area you already painted. You may want to align the left edge of the box with the left edge of the p-diffusion area. The command:
    :paint ndiff
    paints the NMOS active area (n-diffusion). A design rule is that n-diffusion must be at least 12 lambda away from p-diffusion. If you placed the n-diffusion area closer to the p-diffusion area, white dots will appear indicating design rule violation. To move the n-diffusion area, point to the area, type macro s to select the area, and use macros q, w, e, or r to move the area until the white dots disappear. At this point, your layout should look something like this:

    Next, paint horizontal, 6-lambda metal1 wires that will serve as Vdd and GND. Although design rules allow for minimal widths of 4 lambda for m1, making power and ground "rails" wider helps current flow since these wires are used extensively and are connected in many places. Align the top edge of the Vdd wire with the top edge of the p-diffusion. Make this box about 25 lambda wide. The command:

    :paint metal1
    paints the metal1 (Vdd) wire. Then, select the Vdd wire using the s macro, position the cursor to the lower left corner aligned with the left edge of the Vdd wire and the bottom edge of the n-diffusion area, and type c. This will copy the metal1 wire to where the GND wire should be.

    To make a contact between the Vdd wire and p-diffusion (source of the PMOS), make a box over at the overlap between the metal1 and pdiffusion areas and type the command:

    :paint pdc
    To make a contact between the GND wire and n-diffusion (source of the NMOS), make a box over the overlap between metal1 and n-diffusion layers. Note that design rules specify that the minimum contact area is 4 lambda by 4 lambda. Type the command:
    :paint ndc
    At this point, your layout should look like this:

    The next step is to paint the n-well area where the PMOS is located. Actually, this step is not necessary, because Magic knows that the n-well is needed for the specified technology, and would automatically generate the required layer for fabrication. Nevertheless, it is a good practice to place the n-well layer by hand. Place a box extending at least 5 lambda above and below the p-diffusion, and as wide as the metal1 wires. The command:

    :paint nwell
    paints the n-well for the PMOS. At this point, we can also add contacts between the GND wire and the p-substrate (body of the NMOS), as well as between the Vdd wire and the n-well (body of the PMOS). Place a 4 lambda by 4 lambda box over the metal1 wire, but do not overlap the PMOS source contact you already made. The command:
    :paint nwc
    paints the n-well to metal1 contact. If the contact is too close to the pdiff to metal1 contact, white dots will appear. Select the n-well-to-metal1 contact you just created and use Q or R macros to move it away until the white dots disappear. Repeat the same to create a metal1 (GND) to p-substrate contact. Type:
    :paint pwc
    The layout should look like this:

    The body contacts (such as the ones you just created) should always be located as close as possible to the device source contacts to minimize the possibility of latch-up that plagues CMOS circuits. It is also a good practice to put as many body contacts as possible.

    The next step is to make drain contacts, connect the drains of the NMOS and the PMOS, and lay out the output node using metal1. The steps are the same as the steps you used to create the Vdd and GND wires and the source contacts. Note that the simplest way to paint a box is to simply click the middle button (i.e. both the left and the right mouse button on a two-button mouse) over the area already painted with the desired layer.

    You should now have a layout that looks approximately like this:

    The next step is to paint and connect the gates of the NMOS and the PMOS. Position a 2-lambda wide horizontal box to overlap the middle n-diffusion area by at least 2 lambda on both sides. Type

    :paint poly

    Notice how the area of poly-to-ndiffusion overlap changes to green/red stripes. This is the channel of the NMOS. Do a similar poly box over the p-diffusion. Correct the size or position of the poly areas if you have white dots indicating design rule violations. Connect the poly areas and make the input node. The layout should look like this:

     

    The final step is to put labels on the important signal nodes. Left click and than right click on the poly, close to the left edge. You should see a small yellow + at that spot. Then type:

    :label in

    to label the poly as the input node of the inverter. Similarly, label the Vdd wire as Vdd, the GND wire as GND, and the output wire as out. The final layout should look like this:


     

    The grid has been turned off using the g macro, and the circuit diagram is shown again for easy comparison with the layout.

    Save the layout you created:.

    :save
    Extract the layout to create the output file that will be used with a simulator:
    :extract
    To exit Magic, type:
    :quit

    Cell Hierarchies

    The concept of cells is essential for efficient and structured layout of VLSI circuits. The basic idea is quite simple: if you created a layout for a building block (such as an inverter) that can be used in another design or is going to be used many times in the current design, then it is useful to simply include this layout of the building block as a cell. Every time you create and save a circuit in Magic, you have effectively created a cell. Each cell can be included in the hierarchy of a new circuit. The cells you created or the cells created by others in a library can then be put and connected together in a new circuit. Here are basic commands related to working with cells.

    Start a new circuit. Then,

    :getcell name -- includes a cell that you want in the new circuit.
    The included cells are shown as "black boxes" with only the instance names showing. No details of the included cell are shown. This is known as the unexpanded form.

    You can select and manipulate each cell as you could a box of paint. In order to "connect" cells together, you need to know where the "connection points" of a cell are. Thus, expand the cell and paint wires between appropriate points of the cells. The following macros are used to expand or unexpand cells:

    x – expands a selected cell, all details are visible.
    X – unexpands a selected cell, all details are hidden.
    If you included a number of pre-designed identical cells and you realize that you need to make one small change to all of them, you can edit just one cell by itself and the changes will be reflected in all the other cells. To edit the cell, type:
    :load cell-name
    Magic will load that cell (and all of its sub-cells) into the window. Once all changes have been made, :save your changes and :load the circuit you were working on. All changes will be reflected in your circuit's cells.

     A more detailed explanation of the cell hierarchies can be found in the Magic Tutorial #4: Cell Heirarchies.


    Magic Tutorials, Summary of Commands, and Complete Manual Pages

    Magic comes with a set of tutorials that go through commands, macros, and mouse functions. You may want to go first through the tutorials labeled with *. Please do not print the tutorials or the manual pages on the lab printers - these are large files that can best utilized on-line.

    * Magic Tutorial #1: Getting Started
    * Magic Tutorial #2: Basic Painting and Selection
    * Magic Tutorial #3: Advanced Painting (Wiring and Plowing)
    * Magic Tutorial #4: Cell Heirarchies
    Magic Tutorial #5: Multiple Windows
    * Magic Tutorial #6: Design-Rule Checking
    Magic Tutorial #7: Netlists and Routing
    * Magic Tutorial #8: Circuit Extraction
    Magic Tutorial #9: Format Conversion to CIF and Calma
    Magic Tutorial #10: The Interactive Router
    Magic Tutorial #11: Using RSIM with Magic

    Summary of Magic commands
    Complete Magic manual pages



    Prepared by James Stine, ECE Department. Last updated: January  28, 2000.