Dr.Y.Narasimha Murthy Ph.D yayavaram@yahoo.

com

MOS and BiCMOS Circuit design Process

.But the key element is a set of design rules which forms the communication link between the designer (specifying requirements) and the fabricator (who materializes them). Design rules are used to produce workable mask layouts from which the various layers in silicon will be formed or patterned. Among the design rules Lambda based rules are important. They are straightforward and relatively simple to apply. However, they are 'real' and chips can be fabricated from mask layouts using the lambda-based rule set. Correct and faster designs will be realized if a fabricator's line is used to its full advantage and such rule sets are needed not only to the fabricator but also to a specific technology.

MOS LAYERS : MOS design is aimed at turning a specification into masks for processing silicon to meet the specification. We have seen that MOS circuits are formed on four basic layers-n-diffusiondiffusion, polysilicon, and metal, which are isolated from one another by thick or thin(thinox) silicon dioxide insulating layers. The thin oxide (thinox) mask region includes n-diffusion, pdiffusion, and transistor channels. Polysilicon and thinox regions interact so that a transistor is formed where they cross one another. In some processes, there may be a second metal layer and also, in some processes, a second polysilicon layer. Layers may deliberately joined together where contacts are formed. It is also clear that the basic MOS transistor properties can be modified by the use of an implant within the thinox region and this is used in nMOS circuits to produce depletion mode transistors. The BiCMOS technology is developed by including the bipolar transistors in this design process by the addition of extra layers to a CMOS process.

STICK DIAGRAMS : A stick diagram is a diagrammatic representation of a chip layout that helps to abstract a model for design of full layout from traditional transistor schematic. Stick diagrams are used to convey the
layer information with the help of a color code .For example, in the case of nMOS design, green

color is used for n-diffusion, red for polysilicon, blue for metal, yellow for implant, and black

Dr.Y.Narasimha Murthy Ph.D [email protected]

for contact areas. Monochrome encoding is also used in stick diagrams to represent the layer information. The monochrome encoding chosen is shown in figure(a).

Fig.(a) .Monochrome encoding

Fig.(b). Encodings for a Double Metal CMOS Process

Dr.Y.Narasimha Murthy Ph.D [email protected]

The layout of stick diagrams faithfully reflects the topology of the actual layout in silicon. The color encoding is compatible with color terminals, printers, and plotters having quite simple color palettes. Using color workstations, the mask areas are usually color filled while pen plotters
produce color outlines only.

Fig. Encodings for a simple metal nMOS process(color).

Dr.Y.Narasimha Murthy Ph.D [email protected]

Stick diagram for n-MOS transistor is Shown above. The two parallel rails indicate VDD and

Gnd. nMOS Design Style : To understand the design rules for nMOS design style , let us consider a single metal, single polysilicon nMOS technology. The layout of nMOS is based on the following important features. n-diffusion [n-diff.] and other thin oxide regions [thinox] (green) ; polysilicon 1 [poly.]-since there is only one polysilicon layer here (red); metal 1 [metal]-since we use only one metal layer here (blue); implant (yellow); contacts (black or brown [buried]). A transistor is formed wherever poly. crosses n-diff. (red over green) and all diffusion wires (interconnections) are n-type (green). When starting a layout, the first step normally taken is to draw the metal (blue) VDD and GND rails in parallel allowing enough space between them for the other circuit elements which will be required. Next, thinox (green) paths may be drawn between the rails for inverters and inverterbased logic as shown in Fig. below. Inverters and inverter-based logic comprise a pull-up structure, usually a depletion mode transistor, connected from the output point to VDD and a pulldown structure of enhancement mode transistors suitably interconnected between the output point and GND.This is illustrated in the Fig.(b). remembering that poly. (red) crosses thinox (green) wherever transistors are required. One should consider the implants (yellow) for depletion mode transistors and also consider the length to width (L : W) ratio for each transistor. These ratios are important particularly in nMOS and nMOS- like circuits.

Dr.Y.Narasimha Murthy Ph.D [email protected]

Fig . nMOS stick layout design style

Dr.Y.Narasimha Murthy Ph.D [email protected]

CMOS Design Style : The CMOS design rules are almost similar and extensions of n-MOS design rules except the implant (yellow) and the buried contact (brown). In CMOS design Yellow is used to identify ptransistors and wires, as depletion mode devices are not utilized. The two types of transistors 'n' and 'p',

are separated by the demarcation line (representing the p-well boundary) above which all p-type devices are placed (transistors and wires (yellow). The n-devices (green) are consequently placed below the demarcation line and are thus located in the p-well as shown in the diagram below.

Diffusion paths must not cross the demarcation line and n-diffusion and p-diffusion wires must not join. The 'n' and 'p' features are normally joined by metal where a connection is needed. Their geometry will appear when the stick diagram is translated to a mask layout. However, one must not forget to place crosses on VDD and Vss rails to represent the substrate and p-well connection respectively. The design style is explained by taking the example the design of a single bit shift register. The design begins with the drawing of the VDD and Vss rails in parallel and in metal and the creation of an (imaginary) demarcation line in-between, as shown in Fig.below. The n-transistors are then placed below this line and thus close to Vss, while p-transistors are placed above the line and below VDD In both cases, the transistors are conveniently placed with their diffusion paths parallel to the rails (horizontal in the diagram) as shown in Fig.(b). A similar approach can be taken with transistors in symbolic form.

Dr.Y.Narasimha Murthy Ph.D [email protected]

Fig. CMOS stick layout design style (a,b,c,d)

The n- along with the p-transistors are interconnected to the rails using the metal and connect as shown in Fig.(d). It must be remembered that only metal and poly-silicon can cross the demarcation line but with that restriction, wires can run-in diffusion also. Finally, the remaining interconnections are made as appropriate and the control signals and data inputs are added as shown in the Fig.(d).

Dr.Y.Narasimha Murthy Ph.D [email protected]

Design Rules and Layout : The design rules are formed to translate the circuit design concepts , (usually in stick diagram or symbolic form) into actual geometry in silicon. The design rules are the effective interface between the circuit/system designer and the fabrication engineer. The design rules also help to provide a reliable compromise between the circuit/system designer and the fabrication engineer. In general the circuit designers expect smaller layouts for improved performance and decreased silicon area. On the other hand, the process engineer like those design rules that result in a

controllable and reproducible process. In fact there is a need of compromise for a competitive circuit to be produced at a reasonable cost. One of the important factors associated with design rules is the achievable definition of the process line. For example, it is found that if a 10: 1 wafer stepper is used instead of a 1: 1 projection mask aligner; the level-to-level registration will be closer. Design rules can be affected by the maturity of the process line. For example, if the process is mature, then one can be assured of the process line capability, allowing tighter designs with fewer constraints on the designer. The simple and well known design rules that are widely used in the design of multiproject chips are 'lambda ()-based' design rules developed by Mead and Conway . Lambda-based Design Rules : In this Lambda base design rules all paths in all layers will be dimensioned in units and subsequently can be allocated an appropriate value compatible with the feature size of the fabrication process. These design rules are such that, if correctly obeyed, the mask layouts will produce working circuits for a range of values allocated to . For example, can be allocated a value of 1.0m so that minimum feature size on chip will be 2 m (2). Design rules, also, specify line widths, separations, and extensions in terms of . Design rules can be conveniently set out in diagrammatic form as shown in Fig.(a) for wires , and the Fig.(b) for extensions and separations associated with transistor layouts.

Dr.Y.Narasimha Murthy Ph.D [email protected]

Fig.(a). Design rules for wires (n-MOS and p-MOS)

Fig.(b). Transistor design rules (n-MOS, p-MOS and c-MOS) Contact Cuts : While making contacts between poly-silicon and diffusion in nMOS circuits it should be remembered that there are three possible approaches--poly. to metal then metal to diff., or a
9

Dr.Y.Narasimha Murthy Ph.D [email protected]

buried contact poly. to diff. , or a butting contact (poly. to diff. using metal). Among the three the latter two, the buried contact is the most widely used, because of advantage in space and a reliable contact. At one time butting contacts were widely used , but now a days they are superseded by buried contacts. In CMOS designs, poly. to diff. contacts are always made via metal. A simple process is followed for making connections between metal and either of the other two layers (as in Fig.a), The 2. x 2. contact cut indicates an area in which the oxide is to be removed down to the underlying polysilicon or diffusion surface. When deposition of the metal layer takes place the metal is deposited through the contact cut areas onto the underlying area so that contact is made between the layers. The process is more complex for connecting diffusion to poly-silicon using the butting contact approach (Fig.b), In effect, a 2. x 2 contact cut is made down to each of the layers to be joined. The layers are butted together in such a way that these two contact cuts become contiguous. Since the poly-silicon and diffusion outlines overlap and thin oxide under polysilicon acts as a mask in the diffusion process, the poly-silicon and diffusion layers are also butted together. The contact between the two butting layers is then made by a metal overlay as shown in the Fig.

Fig.(a) . n-MOS & C-MOS Contacts

10

Dr.Y.Narasimha Murthy Ph.D [email protected]

Fig.(b). Contacts poly-silicon to diffusion In buried contact basically, layers are joined over a 2. x 2. area with the buried contact cut extending by 1, in all directions around the contact area except that the contact cut extension is increased to 2. in diffusion paths leaving the contact area. This helps to avoid the formation of unwanted transistors .So, this buried contact approach is simpler when compared to others. The, poly-silicon is deposited directly on the underlying crystalline wafer. When diffusion takes place, impurities will diffuse into the poly-silicon as well as into the diffusion region within the contact area. Thus a satisfactory connection between poly-silicon and diffusion is ensured. Buried contacts can be smaller in area than their butting contact counterparts and, since they use no metal layer, they are subject to fewer design rule restrictions in a layout. Double metal MOS process rules : In the MOS design rules a powerful design process is achieved by adding a second metal layer. This gives a much greater degree of freedom, in distributing global VDD and Vss(GND) rails in a system. From the overall chip inter-connection aspect, the second metal layer in particular is important and, although the use of such a layer is readily envisaged, its disposition relative to its connection. to other layers using metal1 to metal 2 contacts, called vias ,can be readily established .
11

Dr.Y.Narasimha Murthy Ph.D [email protected]

Usually, second level metal layers are coarser than the first (conventional) layer and the isolation layer between the layers may also be of relatively greater thickness. To distinguish contacts between first and second metal layers, they are known as vias rather than contact cuts. The second metal layer representation is color coded dark blue (or purple). The important process steps for a two-metal layer process are given below. The oxide below the first metal layer is deposited by atmospheric chemical vapor deposition (CVD) and the oxide layer between the metal layers is applied in a similar manner. Depending on the process, removal of selected areas of the oxide is accomplished by plasma etching, which is designed to have a high level of vertical ion bombardment to allow for high and uniform etch rates. Similarly, the bulk of the process steps for a double polysilicon layer process are similar in nature to those already described, except that a second thin oxide layer is grown after depositing and patterning the first polysilicon layer (Poly.1) to isolate it from the now to be deposited second poly. layer (Poly.2). The presence of a second poly. layer gives greater flexibility in interconnections and also allows Poly.2 transistors to be formed by intersecting Poly. 2 and diffusion. The important features of double metal process are summarized as follows : Use the second level metal for the global distribution of power buses, that is, VDD and GND ( Vss), and for clock lines. Use the first level metal for local distribution of power and for signal lines. Lay out the two metal layers so that the conductors are mutually orthogonal wherever possible.

CMOS Lambda-based Design Rules: The CMOS fabrication process is more complex than nMOS fabrication . In a CMOS process, there are nearly 100 actual set of industrial design rules . The additional rules are concerned with those features unique to p-well CMOS, such as the p-well and p+ mask and the special 'substrate' contacts. The p-well rules are shown in the diagram below.

12

Dr.Y.Narasimha Murthy Ph.D [email protected]

In the diagram above each of the arrangements can be merged into single split contacts.

From the above diagram it is also clear that split contacts may also be made with separate cuts.

13

Dr.Y.Narasimha Murthy Ph.D [email protected]

Fig. Particular rules for p-well CMOS Process.

The CMOS rules are designed based on the extensions of the Mead and Conway concepts and also by excluding the butting and buried contacts the new rules for CMOS design are formed. These rules for CMOS design are implemented in the above diagrams. General Observations on the Design Rules : The microscopic dimensions of Silicon circuits always cause some problems in the design process.The major problem is presented by possible deviation in line widths and in interlayer registration. If the line widths are too small, it is possible for lines to be discontinuous in places. If separate paths in a layer are placed too close together, it is possible that they will merge in places or interfere with each other. For the lambda-based rules , the design rules are formulated in terms of a length unit which is related to the resolution of the process may be viewed as a limit on the width deviation of a feature from its ideal 'as drawn' size and also as a bound on the maximum misalignment of any one mask. In the worst case, these effects may combine to cause the relative position of feature edges on different mask levels to deviate by as much as 2 in their interrelationship. Inevitably, a consequence of using the lambda-based concept is that every dimension must be rounded up to whole values and this leads to layouts which do not fully exploit the capabilities of the process.

14

Dr.Y.Narasimha Murthy Ph.D [email protected]

Similar concepts underlie the establishment of 'micron-based' rule sets, but actual dimensions are given so that full advantage can be taken of the fabrication line capabilities and tighter layouts result. Layout rules, therefore, provide strict guidelines for preparing the geometric layouts which will be used to configure the actual masks used during fabrication and can be regarded as the main communication link between circuit/systems designers and the process engineers engaged in manufacture. The goal of any set of design rules should give optimize yield while keeping the geometry as small as possible without compromising the reliability of the finished circuit. On the questions of yield and reliability, even the conservative nature of the lambda based rules can stand reevaluation when these two factors are of paramount importance. In particular, the rules associated with contacts can be improved upon in the light of experience. Fig.(a) sets out aspects that may be observed for high yield and in high reliability situations. In our proposed scheme of events in creating stick layouts for CMOS, it is assumed that poly. and metal can both freely cross well boundaries and this is indeed the case, but we should be careful to try to exclude poly. from areas which lie within p+ mask areas where possible. The reason for this is that the resistance of the poly. layer is reduced in current processes by n- type doping. Clearly the p+ doping which takes place inside the p+ mask will also dope the poly. which is already in place when the p+ doping step takes place. This results in an increase in the n- doping poly. resistance which may be significant in certain parts of a system.

15

Dr.Y.Narasimha Murthy Ph.D [email protected]

The 3. metal width rule is a conservative one but is implemented to allow for the fact that the metal layer is deposited after the others and on top of them and several layers of silicon dioxide, so that the surface on which it sits is quite 'mountainous' . The metal layer is also light-reflective and these factors combine to result in poor edge definition. In double metal the second layer of metal has an even more uneven terrain on which to be deposited and patterned. Hence metal 2 is often wider than metal 1. Metal to metal separation is also large and is brought about mainly by difficulties in defining metal edges accurately during masking operations on the highly reflective metal. All diffusion processes are such that lateral diffusion occurs as well as impurity penetration from the surface. Hence the separation rules for diffusion allow for this and relatively large separations are specified. This is particularly the case for the p-well diffusions which are deep diffusions and thus have considerable lateral spread. Transitions from thin gate oxide to thick field oxide in the oxidation process also use up space and this is another reason why the lambda-based rules require a minimum separation between thinox regions of 3. In effect, this implies that the

16

Dr.Y.Narasimha Murthy Ph.D [email protected]

minimum feature size for thick oxide is 3.The simplicity of the lambda-based rules makes this approach to design an appropriate one for the novice chip designer and also, perhaps, for those applications in which we are not trying to achieve the absolute minimum area and the absolute maximum performance. Because lambda-based rules try 'to be all things to all people', they do suffer from least common denominator effects and from the upward rounding of all process line dimension parameters into integer values of lambda. The performance of any fabrication line in this respect clearly comes down to a matter of tolerances and definitions in terms of microns (or some other suitable unit of length).Thus, expanded sets of rules often referred to as micron-based rules are available to the more experienced designer to allow for the use of the full capability of any process. Also, many processes offer additional layers, which again adds to the possibilities presented to the designer. In order to properly represent these important aspects, the next section introduces Orbit Semiconductor's 2m feature size double metal, double poly. n-well CMOS rules which also offer a BiCMOS capability.

Acknowledgment: The author is thankful to Prof.Pucknel and Prof .Kamran Eshraghian with out whose book this class notes would not have been prepared. This class notes is based on the book Basic VLSI Design By Douglas A.Pucknel and Kamran Eshraghian and other internet resources. It is purely meant for academic purpose not for any commercial use.

17