GENERAL QUESTIONS about Bandgap Reference

[aredhel_vlsi]

Hi! I'd like some general info about the design of Bandgap Reference Voltage.

1) What kind of effects can somebody meet at a bandgap reference?

The offset at the input of the opamp, is one, as far as I know. Also there are the second order effects, which are met in mos transistors, such as channel length modulation (is that the same with Early Voltage effect?),Body effect, subthreshold region, mobility variation, impact ionization. anything more specific?

2)a)what kind of temperature analysis error about the temp coefficient can sombody meet? 0.3 % relative variation is good enough?

2)b)temperature analysis is a systematic error? so that meas it is because designer's mistakes?

3)what other software tools can be used except cadence? I've heard something about magic. which is better for layout and simulations?

4)what kind of a noise margin do typical bandgaps have got? I mean, how much?

5)what is the upper limit about consumption in Amperes? 0.5 mA is considered to be very high about 1.8 V supply voltage? do I have to check the whole consumption of the circuit, or should I check the nets separately, and ignore the startup consumption (which will turn off later)?

6) what kind of gain have got opamps in a bandgap?

7)Is it better to use diodes or BJTs at the bandgap core?

8)what is the lowest in um design that has been published until now?

Too many questions? I think I know some answers merely, but I want to have a general idea, especially in numbers. ie, 30dB gain, 1mA consumption etc.
Any ideas welcomed!

 

[dick_freebird]

BJTs are better because their internal current gain tends
to linearize the "diode" curve (log-linear). Diode curvature
is a second (or higher) order effect which you can't do much
about, with circuit methods. Other than lumping it in with
the other curvature terms and trying to comp it. But the
diode curvature is a special problem because you are
looking to run diodes at widely spaced current densities
while expecting a linear relation in Vf.

The "op amp" can be fairly poor and/or simple; I have
made (not good, but functional) bandgaps with just a PTAT
core, the NPN mirror is ratioed / resistored and you sit the
whole thing on top of the tempco resistor at the tail. Its
virtues include pretty good / easy stability, its accuracy
is not great. But good op amps cost area. If you want
load regulation, though, you need one for sure.

You'd be well off to set your operating current at the core,
by what gives you the best TC flatness (minimizing bowing
particularly) - staying away from both low current and
high current nonidealities. Real data is wanted, models
that foundries put out for "ESD diodes" are "for amusement
only".

 

[aredhel_vlsi]

thank you. one question about the opamp gain :

From the dc analysis I checked the Vy- Vx = -0.00494 V at the opamp input, and the output is 1.18 Volts. Thus, the gain of the amplifier is

A=Vout/(Vy-Vx) = 238.86 Volts -> 47.56 dB. Is it good or bad? what could be the ideal gain?

My amplifier output is 1.18 Volts and the Vref is 1 Volt, that means I should have more gain, to increase the output of the bandgap, right? but it is difficult to increase the Vref at 1.15 V (with a good temp curve) , so I needn't mention 1.2 Volts..
I think this is a problem of the topology..Maybe I should add some cascode amplifier which has better gain?

 

[dick_freebird]

The "magic voltage" varies with process, particularly with
the tempco of the resistor you use and somewhat with
the transistor / diode doping which drives Vbe.

If you are dead set on having a higher voltage, do it with
a resistor divider between the op amp output and the
diode/resistor rack. Contrariwise if you need lower, attach
the op amp directly to the rack and divide for the output.

 

[harry_the_elecwhiz]

I am trying to build a bandgap for 1.5V
my design for a 1.2 vol bandgap works fine but the 1.5 volts design has a positive ptat of 1.3mV/deg C. I guess we cant really call it a bandgap

the process limits my voltages to +/- 1 volt for the supplies and i am using a charge pump to provide a VDD of 3V for the bandgap....

any suggestions to get the bandgap working fine at 1.5 volts output ??

 

[tzg6sa]

The bandgap voltage will have zero tempco around 1.2V. It is due to the temperature dependence of the pn junction. The derivation of this statement can be found in almost every analog textbooks.
If you need different output voltage then you have to amplify/divide the bandgap voltage. (I heard of some sophisticated architecture where the output lays (sligthly) above 1.2V but I don't know them.)
But do you REALLY need a temperature independent reference voltage at 1.5V?

 

[harry_the_elecwhiz]

yes I really need a 1.5 temp independent supply. I hav built bandgap's which are temp independentt for 1.25 reference and am aware of the derivation you are talkin about.

in need of a scheme to push reference output to 1.5 MAINTAINING the temp independence.....

 

[harry_the_elecwhiz]

I have tried looking for journal articles on a 1.5volt bandgap reference generator but didn find anything substantial.

can any1 guide me on this one