Silicon Pr0n

User Tools

Site Tools

Trace:
fet

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision		Previous revision
fet [2016/01/03 06:49] – mcmasterfet [2016/01/03 08:02] (current) – [Xilinx XC2018] mcmaster
Line 1: Line 1:
 This page focuses on MOSFET technology.  Pre-MOSFET is mostly only of historical interest This page focuses on MOSFET technology.  Pre-MOSFET is mostly only of historical interest
- 
  
 ====== MOSFET theory 101 ====== ====== MOSFET theory 101 ======
Line 8: Line 7:
 MOSFETs build on diodes by adding a control terminal to adjust charge between two similarly doped areas.  For example, when positive voltage is applied to the gate of a N MOSFET it conducts.  Similarly, when a negative voltage is applied to the gate of a P MOSFET it conducts.  While real devices have many other critical details, this is generally sufficient detail to reverse engineer digital logic: while a logic designer has to make the system work reliably, a reverse engineer generally assumes the system works reliably. MOSFETs build on diodes by adding a control terminal to adjust charge between two similarly doped areas.  For example, when positive voltage is applied to the gate of a N MOSFET it conducts.  Similarly, when a negative voltage is applied to the gate of a P MOSFET it conducts.  While real devices have many other critical details, this is generally sufficient detail to reverse engineer digital logic: while a logic designer has to make the system work reliably, a reverse engineer generally assumes the system works reliably.
  
-{{:wiki:mosfets.png|}}+{{:wiki:mosfets.png}}
  
 This ([[http://en.wikipedia.org/wiki/MOSFET|Wiki]]) image shows schematic symbols for different types of MOSFETs.  When analyzing basic CMOS circuits only N MOSFETs (NFETs) and P MOSFETs (PFETs) are relevant.  However, NMOS and PMOS also use depletion mode transistors.  For most practical purposes, they can be inferred as resistors and the device physics can be ignored. This ([[http://en.wikipedia.org/wiki/MOSFET|Wiki]]) image shows schematic symbols for different types of MOSFETs.  When analyzing basic CMOS circuits only N MOSFETs (NFETs) and P MOSFETs (PFETs) are relevant.  However, NMOS and PMOS also use depletion mode transistors.  For most practical purposes, they can be inferred as resistors and the device physics can be ignored.
  
 A note of caution: chip designers know this is how people think when they try to copy chips.  They occasionally exploit real device physics even in digital logic to confuse reverse engineers ("traps").  For the meantime assume that circuits work as described above but be mindful that there are exceptions. A note of caution: chip designers know this is how people think when they try to copy chips.  They occasionally exploit real device physics even in digital logic to confuse reverse engineers ("traps").  For the meantime assume that circuits work as described above but be mindful that there are exceptions.
- 
  
 ====== MOSFET layout 101 ====== ====== MOSFET layout 101 ======
  
-But what are those funny triangles next to some of them?  Lets go a little deeper.  This is a typical cross section ([[http://en.wikipedia.org/wiki/File:MOSFET_functioning.svg|derivative work]]) of an N MOSFET:+{{:wiki:mosfet_functioning_modified.png}}
  
-{{:wiki:mosfet_functioning_modified.png|}}+With some high level theory, lets now look at how real world MOSFETs are built.  Above shows a typical cross section ([[http://en.wikipedia.org/wiki/File:MOSFET_functioning.svg|derivative work]]).  This is basically the theoretical discussion with the addition of a body terminal to isolate transistors from their neighbors.
  
-Specifically it represents this schematic symbol ([[http://en.wikipedia.org/wiki/File:IGFET_N-Ch_Enh_Labelled.svg|source]]):+{{:wiki:igfet_n-ch_enh_labelled.png}}
  
-{{:wiki:igfet_n-ch_enh_labelled.png|}}+That transistor is represented by this schematic symbol ([[http://en.wikipedia.org/wiki/File:IGFET_N-Ch_Enh_Labelled.svg|source]]).  Recall that a diode is formed when a P and N region meet.  And that's why the schematic symbol has a diode triangle: it represents the body diode.  For most reverse engineering purposes this can be ignored.
  
-Now we can understand that that triangle is a diode notation.  In an N MOSFET, the source is N doped sitting in a P well.  Thus the body forms a PN junction with the source and drain, connected in the middle of the channel.  That is, applying a positive voltage to the body will conduct out either the source or drain.  If the body is left unconnected it can generate some stray currents.  By connecting the body to ground it reverse biases the PN junctions (particularly the body-drain) and helps prevent leakage as well as potential latchup conditions.+There are many ways to make a real device.  To illustrate, here is a gallery of different real transistor constructions.
  
-To make the final point, the "no bulk" version above is a discrete MOSFET.  Since it isn't made in a well process there's no reason to have a P substrate.  For example, it might simply be an N wafer with P doping just in the channel region  (TODO: get some discrete MOSFET pictures.  Maybe bad example due to self aligned gate practices).+====== Gate ======
  
-{{gallery>:mcmaster:n_fet_no_metal.png}}+In general, there are three types of FET technology (in chronological order):
  
-Typical poly gate transistor from a typical standard cell based IC.  Some of the contact metal can be seen on the bottom but the gate itself is poly.+    * Early aluminum metal gate 
 +    * Polysilicon gate 
 +    * Metal gate w/ high-k dielectric
  
 +Within each of these, especially polysilicon gates, there are many variations.
  
-====== Gate ======+====== Examples ======
  
-In general, there are three types of FET technology (in chronological order): +===== Gate: poly =====
-  * Early aluminum metal gate +
-  * Polysilicon gate +
-  * Metal gate w/ high-k dielectric+
  
-Within each of these, especially polysilicon gates, there are many variations.+==== RSA SecurID 1C ====
  
 +{{:mcmaster:n_fet_no_metal.png?100}}
  
-====== Examples ====== +Some of contact metal can be seen on the bottom but the gate itself is poly.  The [[:cmos|CMOS page]] details its use in an inverter
  
-===== Gate: early metal ===== 
  
-==== MOS MPS7083 ====+==== Xilinx XC2018 ====
  
-{{:mcmaster:mos:mps7083:q_mz.jpg?300}}+{{:mcmaster:xilinx:xc2018:q.jpg?150|}}
  
-{{:mcmaster:mos:mps7083:q_dlyr1.jpg?300}}+Similar process to the SecurID.  Left middle has a via to poly.  The poly extends right over an active area to form a transistor.  The active area has contacts at top and bottom.
  
-Above top: top metal.  Bottom: Al and some SiO2 removed 
  
-Typical metal gate CMOS transistor as used in non-trivial chips (ie CPU).  The control signal+==== MOS 6522 ====
  
 +{{:mcmaster:mos:6522:q.jpg?150|}}
  
-==== Fairchild CD4011 ====+Above: delayered showing poly (textured orange), active (orange), and buried contact (shadow)
  
-{{:image:fairchild_4011_transistor.jpg?300}}+The top polysilicon connects is driven by a via at top left.  It crosses the active area to the right to form a transistor.
  
-{{:image:cmos_metal_gate_transistor_no_metal.jpg?300}}+The bottom transistor side connects directly from active to poly via a buried contact.  This poly goes off to bottom left to drive another transistor.
  
-Above  textbook style metal gate transistors 
  
 +==== MOS 6526 ====
  
 +{{:mcmaster:mos:6526:q.jpg|}}
 +
 +Above: top metal image
 +
 +Active area enters lower left and meets poly at a buried contact.  The poly sweeps right to up to form a transistor separating the via from the active area at the right of the image.
 ==== CMOS quiz ==== ==== CMOS quiz ====
  
 {{:quiz:metal_gate_cmos:q.jpg?300}} {{:quiz:metal_gate_cmos:q.jpg?300}}
  
-See quiz for analysis+[[:quiz:3h7e0f4|See quiz]] for analysis 
 + 
 + 
 +===== Gate: early metal ===== 
 + 
 +==== MOS MPS7083 ==== 
 + 
 +{{:mcmaster:mos:mps7083:q_mz.jpg?150}} 
 + 
 +{{:mcmaster:mos:mps7083:q_dlyr1.jpg?150}} 
 + 
 +Above top: top metal.  Bottom: Al and some SiO2 removed w/ SiO2 thin film interference showing oxide thickness 
 + 
 +Typical metal gate CMOS transistor as used in non-trivial chips (ie CPU).  The gate driver comes in on the right through an active area.  The SiO2 has a cutout to allow placing a (Al?) via up to M1. 
 + 
 +M1 goes left where it meets an active area.  The teal SiO2 M1 area has thick oxide that prevents it from coupling to the Si below.  However, the dark blue cutout brings it much closer (but not contacting, ie an "isolated gate" FET) such that its able to induce charge into the substrate below.  Therefore, while metal crosses several active regions only one transistor is formed. 
 + 
 +While the metal does spill over to some adjacent areas there are several things that prevent them from becoming an effective transistor: 
 + 
 +    * They do not fully cut off active areas 
 +    * These areas are likely biased such that adding additional charge wouldn't do anything 
 + 
 +==== Fairchild CD4011 ==== 
 + 
 +{{:image:fairchild_4011_transistor.jpg?150}} 
 + 
 +{{:image:cmos_metal_gate_transistor_no_metal.jpg?150}} 
 + 
 +Above top: original transistor.  Bottom: very siimlar delayered transistor 
 + 
 +These textbook style metal gate transistors are rare in real devices.  They are similar to the preceeding MOS transistor with cutouts for both the gate and vias.  However, the doping is very different.  The areas immediately to the left and right of the transistor must be opposite doping to the center and surrounging area (which apper to be doped the same per the delayer image).  The via features are due to ohmic contact / windowing.  I'm guessing the green layer is the opposite doping ie N if the center is P) and the layers on either side of it  are the same doping based on thin film interference color.  However, I'm still unclera why there are three different main regions. 
 + 
 +Consider staining these to get more info.
  
  
fet.1451803749.txt.gz · Last modified: 2016/01/03 06:49 by mcmaster