lateral and vertical bipolar transistor for designing BGR circuit

[Junus2012]

Hi

Kindly, I designed a BGR currebt source based on the lateral PNP I have in my technology (P-Well). is there any advantage if i redesign the circuit using the vertical NPN transistors (in this case I think I must connect the collector to VDD)) ??

second, if the parasatic bipolar is lateral, is it mean that I am free to connect the collector to VDD or to ground as I need or it should be fixed??

please if you can assist your answer with a pictures it will be very healpful

 

[jamaleddin mollasalmani]

Hi Junus,

As I know one of issues which can be discussed is fabrication.
Generally making a PNP BJT in CMOS technology is more easier.
you can find this in this picture.

 

[Junus2012]

Thank you

but actually I dont care for the difficulty of the technology as I use it directly. i have both the lateral and the vertical available in the technology I am using

Hi Junus,

As I know one of issues which can be discussed is fabrication.
Generally making a PNP BJT in CMOS technology is more easier.
you can find this in this picture.

 

[erikl]

... based on the lateral PNP I have in my technology (P-Well). is there any advantage if i redesign the circuit using the vertical NPN transistors (in this case I think I must connect the collector to VDD)) ??

Use that one which has larger B (current gain). Collector (and base) to VDD.

second, if the parasatic bipolar is lateral, is it mean that I am free to connect the collector to VDD or to ground ...??

... to VDD.

please if you can assist your answer with a pictures it will be very healpful

 

[Junus2012]

Dear Erikl

Nice to hear from you after the new year holiday

if the thing is up to the current gain, then the NPN bipolar I have in my technology has more B. but if the comparision is limited to B, why 99.9% of the current reference circuit is implemented by the PNP transistor ??

 

[erikl]

... if the comparision is limited to B, why 99.9% of the current reference circuit is implemented by the PNP transistor ??

Because isolated NMOS and vertical NPN BJTs need the costly triple well option (2 more masks and implants, and less process yield), which may be 99.9% of all bosses won't allow for if not absolutely necessary ;-)

 

[Junus2012]

Hello Erikl

As I understood from you that even even if the technology provide the vertical NPN (like the one I am using), but still it more expensive to implement it and Hence the people prefer to use the vertical PNP.

Now please the last thing I would ask, when you have a lateral and vertical PNP, which one you would like to use and why?

Thank you very much

Because isolated NMOS and vertical NPN BJTs need the costly triple well option (2 more masks and implants, and less process yield), which may be 99.9% of all bosses won't allow for if not absolutely necessary ;-)

 

[erikl]

... when you have a lateral and vertical PNP, which one you would like to use and why?

That one with the larger B .

 

[Junus2012]

Thank you Erikl

Now I know which one to use

if you allow me to ask what the effect of B on the diode connected transistor so you prefer the higher one??

 

[erikl]

what the effect of B on the diode connected transistor so you prefer the higher one??

I explained that in a former forum thread which I now don't remember; you might want to search for it. Shortly, the by the factor B smaller base current and its smaller variation in relation to the total current's (Ic + Ib) variation is responsible for higher accuracy.

 

[dick_freebird]

I prefer vertical transistors for the work I do because their
conduction is subsurface and a small portion of the base is
exposed to field oxide and its lousy interface qualities. A
lateral is likely to have an inferior low-current beta even if
the peak betas are close; there is a lot of base surface to
see recombination.

The BJT does (as transdiode) "idealize" the E-B diode and
can make a better bandgap. But in any case you have to
tweak to what you've got, and a more ideal diode may or
may not give you the best tempco and linearity of it.

If you want a ground referred forward biased PN diode in
a Psub single well technology, you will be using a PNP -
like it or not. Because you need the N-well (base) to be
the cathode, and it sits in the Psub (collector). So the
structure has only one right answer and you had best
model it as what it is, not pretend it's "just a diode" and
parasitics don't matter.

Now if you have "real", three-free-terminals (not one of
them pinned to substrate) BJTs you can do simpler PTAT /
bandgap designs, not requiring a whole op amp to equal-
bias the reference diodes but making those part of the
amplifier front end and operating current-mode; like a
6-8 transistor bandgap.