9
\$\begingroup\$

Please, provide a summary of advantages and disadvantages of a transistor layout with multiple fingers (MF) vs single finger?

When laying out a MOSFET with a particular width and length, in an EDA tool, one has two options with regards to the shape of the gate:

1) Single stripe (classical case) (one finger);
2) Several stripes (several fingers).

Hypotheses (based on various Internet forums):

1) MF provide more flexibility in layout planning for transistor with high W/L or L/W. In other words, allows making a layout more square-like.

2) MF allow better matching of transistors, when needed. For example, if using common-centroid techniques.

3) MF layout reduces gate resistance (for AC). If so, can you explain why?

4) MF reduce current density in the gate if there are technology limitations on this.

Can someone with knowledge do a comparison of the two approaches?

Figure 1: One finger

Figure 1: One finger.

enter image description here

Figure 2: Two fingers. Two transistors in parallel (widths are summed).

enter image description here

Figure 3: Two fingers. Two transistors in series (lenghts are summed).

\$\endgroup\$
  • \$\begingroup\$ Thanks, folks. As for the decrease in gate resistance, I've read it at a forum. They say, this is due to multiple fingers' resistances in parallel. But I think it's wrong as resistance of each gate goes up (length gets smaller), the total value should stay the same. I guess, for high frequencies, this may be due to skin effect (total surface area is down), but not sure. \$\endgroup\$ – Sergei Gorbikov Jul 17 '16 at 17:21
  • \$\begingroup\$ Also read in the Internet, that the number of fingers is determined by the maximum current density through poly gate. But again, not sure. \$\endgroup\$ – Sergei Gorbikov Jul 17 '16 at 18:05
6
\$\begingroup\$

Long transistors are split into smaller ones for various reasons. One reason could be to fit them better into the overall layout of a block or simply to have a better aspect ratio. Another reason is to reduce the gate resistance of the device. The problem is that the gate-channel capacitance forms a lowpass with the gate resistance and switching speed is reduced. In addition the drain area is reduced because two fingers share a common drain and therefore parasitics are reduces as well.

Update: Splitting a transistor with some W/L into a transistor with multiple fingers is done such, that the width (W) stays the same and the length of finger is L/k, where k is the total number of fingers. The effective length and width stay the same, but the gates are now in parallel and therefore the resistance is reduced. In addition, gates are often contacted at both ends to reduce the effective resistance.

Splitting transistors can either be done by using multiple transistors with a single gate or with transistors that have multiple gate fingers.

Transistors with multiple fingers have the disadvantage that the current direction is different for two neighbouring fingers. E.g. if for the first finger the source is to the left then the source for the next finger will be to the right. The properties of transistor can change depending on the current direction. Therefore extra care has to be taken when trying to achieve good matching.

Using multiple fingers to obtained scaled current sources for example in a current mirror is also considered inferior to having multiple single gate transistors because of slightly different properties.

An analog designer therefore has to decide which option is the best for a given problem.

\$\endgroup\$
  • 1
    \$\begingroup\$ Thanks, buddy. Good answer, a lot of insights. Would appreciate if u could give more color on 1) Why is total gate resistance reduced? Remember the formula R=r*L/W. When one splits a gate into thinner fingers, resistance of each finger increases. So, net result of N parallel fingers (each with higher resistance) is the same resistance, isn't it? 2) Why are currents direction different in each finger? What properties change depending on current direction in the gate finger? 10x. Sorry for scrutinizing your answer. \$\endgroup\$ – Sergei Gorbikov Jul 19 '16 at 17:42
  • \$\begingroup\$ @Sergei Gorbikov -- See my update. \$\endgroup\$ – Mario Jul 19 '16 at 18:08
  • 1
    \$\begingroup\$ Very good update. Re 1) I thought the AC current in the gate flew along the vertical axis (wrong :)), but indeed it flows in the horizontal direction (between S and D), thus indeed, the resistance of one finger is not reduced after splitting. Re 2) now clear, 10x. Will mark as solved in a day if no one else provides more answers. Thanks, again. \$\endgroup\$ – Sergei Gorbikov Jul 19 '16 at 18:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.