[HN Gopher] Reverse-engineering the LM185 voltage reference chip...
___________________________________________________________________
Reverse-engineering the LM185 voltage reference chip and its
bandgap reference
Author : picture
Score : 58 points
Date : 2022-04-09 18:24 UTC (4 hours ago)
(HTM) web link (www.righto.com)
(TXT) w3m dump (www.righto.com)
| antattack wrote:
| Great write-up and drawings. I've heard of fuses in processors,
| MCUs, etc but not of anti-fuses. Are anti-fuses a feature of
| older designs, based on large process size?
|
| _" The second type of fuse is an "antifuse", which has the
| opposite behavior: it does not conduct until a high current is
| applied"_
| kens wrote:
| Antifuses are used in modern chips, e.g. FPGAs:
| https://www.microsemi.com/product-directory/fpga-soc/1641-an...
| kens wrote:
| Author here for all your voltage reference questions :-)
| colejohnson66 wrote:
| How do you tell if a chip's substrate is N or P?
| monocasa wrote:
| Is there any chance you can expand on the reasoning behind the
| circular/coaxial designed transistors? The normal sorta logic
| style transistors are straight forward, and the serpentine
| nature of the high current transistors makes sense for the
| surface area, but I never understood the reasoning for the
| round transistors.
| adrian_b wrote:
| For this case of lateral PNP transistors, the reason is as
| Ken has said.
|
| Because both the emitter and the collector are on the
| surface, making them both circular ensures that the distance
| between them, which is the same as the width of the base, is
| constant.
|
| The properties of the bipolar transistors vary very strongly
| with the width of the base. If the width is not constant,
| then the current becomes crowded in only a part of the base
| and many characteristics become worse.
|
| Unlike in lateral transistors, in vertical transistors the
| width of the base is not determined by geometry, but by
| doping doses and diffusion times, so the form of the emitter
| is less important.
|
| Nonetheless, in early planar transistors the emitter was also
| circular. The reason is that in bipolar transistors with very
| narrow bases, the resistance of the narrow base layer becomes
| large and in the center of the base under the emitter, the
| base-emitter voltage drops to a lower value than at the
| terminals of the transistor, which makes the central part of
| the emitter and base non-functional (i.e. only a very small
| fraction of the current passes through there).
|
| So in vertical transistors, only the periphery of the emitter
| matters. When it is circular, the symmetry guarantees that
| the current is uniformly distributed on the periphery, for
| maximum current capability.
|
| Unfortunately, increasing the density of current per
| peripheral length of the emitter over a threshold triggers a
| positive feedback that will destroy the transistor if the
| current is not limited externally. This is usually the main
| factor that determines the specification of a maximum current
| for a bipolar transistor. If the current is non-uniform over
| the periphery, the threshold will be reached at a much lower
| current than computed by multiplying the threshold density
| with emitter perimeter.
|
| Because there is a limit for amperes per millimeter of
| emitter periphery, to increase the maximum current in a given
| area, the form of the emitter must be changed from a circle
| to a form with a longer perimeter, without increasing the
| occupied area.
|
| Early power transistors had various fancy forms for the
| emitters, e.g. Christmas tree, snow flake and so on. However,
| it was quite difficult to ensure that the current is
| distributed uniformly on the periphery of such complex forms.
|
| Later, after the photolithography had improved and smaller
| dimensions were no longer problem, instead of having an
| emitter with a complex sinuous boundary, 2 simpler solutions
| have been adopted to increase the perimeter of the emitter.
| Either the transistor had a large number of small emitters
| connected in parallel, or it had one large emitter, but with
| a large number of small holes in the emitter (mesh emitter).
| kens wrote:
| Theoretically you could make a PNP transistor by reversing
| the doping of an NPN transistor. The main problem is that
| boron diffuses rapidly, making it hard to fabricate a buried
| P-layer. Boron also has less solubility than phosphorus,
| making it hard to dope the emitter. Also, holes have only 1/3
| the mobility of electrons, so PNP and NPN aren't symmetrical.
| To deal with these issues, PNP transistors are usually built
| with lateral construction (i.e. horizontally). The ring
| structure ensures that almost all of the carriers injected by
| the emitter are intercepted by the collector.
|
| (This is based on The Art of Analog Layout, p280. I don't
| know all this doping stuff myself.)
| cushychicken wrote:
| Got any good references for how to identify silicon structures
| as certain components?
| kens wrote:
| I got an old, cheap copy of "The Art of Analog Layout"
| (Hastings), which describes these structures in detail. For
| the most part, the structures are fairly easy to recognize
| after you've seen them once or twice. But then there are the
| bizarre mystery circuits that require some puzzle-solving.
| For instance, where they combine a couple of transistors to
| save a bit of space.
| dasudasu wrote:
| Seeing that the latest edition of The Art of Analog Layout
| came out of 2005, is there any other book you'd recommend
| on analog layouts specifically?
| DiabloD3 wrote:
| I'd hazard a guess that the 20 to 50 year old tech
| described in that book hasn't changed much; anything that
| is novel is probably still under patent and will someday
| make it into a newer edition.
| kens wrote:
| I mostly look at chips from the 1970s, since modern chips
| have features that are too small for my microscope. So I
| don't have any recommendations for a "modern" book.
| londons_explore wrote:
| Lots of microcontrollers have a bandgap reference built in... And
| typically they have really rather terrible voltage tolerances -
| eg. on atmel devices the 1.1v reference doesn't even have
| guaranteed minimum and maximum voltages across the whole range of
| supply voltage and temperature, but you can only expect it to be
| somewhere between 0.9 volts and 1.3 volts....
|
| So why are these circuits so bad? Do they use a different design?
| adrian_b wrote:
| The transistors available in modern digital CMOS processes are
| worse and worse for analog functions, the more recent that
| process is.
|
| The very poor device characteristics may be mitigated only
| using very complex schematics for the analog circuit, together
| with various auto-calibration methods.
|
| The additional cost may be deemed too much for a voltage
| reference in a cheap microcontroller.
|
| One can always use a good external voltage reference, but that
| may cost as much as a microcontroller.
___________________________________________________________________
(page generated 2022-04-09 23:00 UTC)