MC1403 Schematic:What is going on?!

[MrAureliusR]

So I've slowly (very slowly) been learning to reverse-engineer and understand chip layouts. I still miss obvious things and am a looooong way from even being able to recognise most structures, but when browsing the datasheet for the MC1403 bandgap voltage reference, I just stared in amazement for a little bit. I understand maybe a few tiny parts of it, but the part that had me really scratching my head is this part in the bottom left in the red square:


What is that resistor doing with 6 taps? Is this just a convenience thing when creating the schematic instead of drawing out what are presumably 7 ~143 ohm resistors? Or is it actually laid out like this on the chip, as multiple taps on a single stretch of resistive silicon? Also a question: is the P-FET with the NPN transistor directly across it the actual silicon bandgap that is being used as a reference? This stuck out to me as the only FET in an otherwise entirely bipolar layout. I still have a lot to learn, and I know explaining a chip like this can be difficult, but if someone can give me the broad strokes at least, that would be super helpful

Here's the link to the datasheet I took the screenshot from.

[T3sl4co1l]

FET is for startup, probably a few uA or something.  Then the NPN (minus the diode-strapped NPN) sets normal operating bias.  The top PNPs are a current mirror: consider what happens at nominal Vout, the PFET-NPN base is at Vout, so emitter is a diode less, then another diode less just because, then the 10k.  The 10k sets quiescent current.  (Probably the two Vbe's gives a desirable tempco that helps compensate things.)

The circled, erm rectangled, part, is characteristic.  Notice the two bases at equal voltage, and the emitters at dissimilar loads: an extra some kohms in one leg.  The two transistors likely have a large ratio of emitter areas, too.  This is the basis of the bandgap reference.

Or, maybe the emitter ratios are in the current mirrors (the lower row of three PNPs, well four if you count the dual-collector as separate), that's fine too.  Uh, I suppose they can't, not with that much emitter degeneration (if they're actually 2k, 2k and 1k as shown). 

In any case, the ratios should be obvious since you're looking at chip layouts...I guess?  Do...do you actually have a die shot of this part?  Or did you mean just traditional (bipolar?) IC schematics?  'Cuz there are a lot of things you'll be missing without that; IC designs are a bit richer than the schematic would suggest, and they never show dimensions (junction width/length/area) on these diagrams, which are often equivalent diagrams at that (this one, it seems, does show all its transistors, given the "contains" claim).  There are indeed clever mysteries to uncover if you're puzzling over just the diagram...

Oh and the tapped resistor is probably laser or fuse trimmed.  Yeah, it's... badly drawn.  Or, weird that they show it at all really.  Most times I think that would be drawn as fixed resistors, hiding the fact that it's trimmed.  Anyway, give or take schematic errors (lack of junction dots, or a way to indicate that some of those connections may be cut), that would indeed be how it's laid out on chip, a bunch of resistors with metal links between them, connected to test pads to run a fusing current, or left out in space where a laser can blast 'em.  I suppose the discrete option is more likely (pads for microprobing) since a laser process would most likely carve an 'L' track into a resistive patch and it would be silly to draw it this way.

Tim

[magic]

This looks awfully similar to AD1403. See here and a few subsequent posts.

The FET was N-ch in the original part, but I suppose they could change it if they wanted to. Standard bipolar process can produce dodgy P-JFETs; with special processing, better P-JFETs are available (LF156, TL072, etc.). Another possibility is that the schematic is simply wrong; certainly wouldn't be the first time.

This resistor is weird, nobody normally draws such things and I'm not even sure how to interpret that symbol. Maybe it represents different taps which were created on the resistor to enable changing its value by adjusting the metal layer only, when the design was being tuned for final production.

Individual fabricated units may have been further tuned by zenner zapping or similar means, I don't remember if this IC was trimmed. I'm pathologically lazy, am I not? I looked it up, the Analog Devices part was supposedly laser trimmed.