CMOS Bias current generator and Bandgap reference voltage

Hi,

I have a few questions on Bias current generators and Bandgap reference voltages:

1. To generate a bias current why is a PTAT current generator used. The current will be proportional to absolute temperature so won't that affect the circuit performance. Instead shouldn't a bandgap reference voltage be generated and then use that voltage to generate the bias current?

2. To generate a bandgap reference voltage I read a Razavi's book. The book mentions the Brokaw cell which used an opamp in negative feedback. Another method that was mentioned was bootstrapping the and then mirroring the PTAT current and then adding Vbe to it to get a bandgap voltage. The latter method seems simpler. So then why use the Brokaw cell?

3. I need to design a current source for an opamp and a few more current sources and voltage sources for additional blocks. What should I design? Should I design a master Bandgap reference voltage and derive all the currents and voltages from it or should I have a PTAT master current source and then mirror all the currents that I need from it? Again I don't understand how a PTAT current source would help because it would vary with temperature.

I'd really appreciate if someone could answer these questions and help me out.

Thanks.

Re: CMOS Bias current generator and Bandgap reference voltag

calculus_cuthbert said:

... 1. To generate a bias current why is a PTAT current generator used. The current will be proportional to absolute temperature so won't that affect the circuit performance. Instead shouldn't a bandgap reference voltage be generated and then use that voltage to generate the bias current?
...
3. ... Again I don't understand how a PTAT current source would help because it would vary with temperature.

Click to expand...

This depends very much on your circuit resp. application: You probably know that most transistors get slower when their temperature rises (w.c. conditions include the highest temperature in most cases). They generally get faster, however, if they receive higher operating currents. That is why it is possible to (sort of) compensate their speed/frequency decrease with temperature by spending them a PTAT operating current.

calculus_cuthbert said:

3. I need to design a current source for an opamp and a few more current sources and voltage sources for additional blocks. What should I design? Should I design a master Bandgap reference voltage and derive all the currents and voltages from it or should I have a PTAT master current source and then mirror all the currents that I need from it?

Click to expand...

Again this depends on your circuit needs: As long as you don't have precision measurement circuits like DAC, ADC, you usually won't need a bandGap controlled voltage or current, so you usually can live with a relatively simple PTAT current source with its mirrors and their scaling possibilities. With a DAC or ADC, however, you'll need a temperature-independent reference, whose change over the full environmental temperature range is less than ½LSB of your converter. In such case a BGR is indispensable, and the PTAT current source is included anyway (in most reference designs).

Re: CMOS Bias current generator and Bandgap reference voltag

Thanks Erikl,

Your comments were very useful. I do not have ADC or DAC in my blocks. So I guess I should be looking at PTAT current bias generator.

What about the voltage biases needed in my circuit. Should I just use diode connected transistors to generate those voltages or should it be derived from the PTAT source.

I have attached a schematic for the PTAT and bandgap source. Could you please let me know if this is the way to go or is the brokaw cell better?

Thanks again.

Re: CMOS Bias current generator and Bandgap reference voltag

calculus_cuthbert said:

I do not have ADC or DAC in my blocks. So I guess I should be looking at PTAT current bias generator.

Click to expand...

Ok.

calculus_cuthbert said:

What about the voltage biases needed in my circuit. Should I just use diode connected transistors to generate those voltages or should it be derived from the PTAT source.

Click to expand...

I think this is equivalent: You feed the PTAT current into diode connected transistor(s) and use the generated voltage as bias for matching current sources.

calculus_cuthbert said:

I have attached a schematic for the PTAT and bandgap source. Could you please let me know if this is the way to go or is the brokaw cell better?

Click to expand...

I don't know the brokaw cell. As you don't need a BGR (bandgap reference), you could use a simpler circuit for just generating a PTAT current source (you don't need pnp transistors for that). Anyway, the emitter arrows are on the wrong side of your BJTs.

Re: CMOS Bias current generator and Bandgap reference voltag

Thanks Erikl,

Yes that was an error in the schematic.

Added after 38 minutes:

Erikl,

could you also suggest circuits for current sources as you mentioned it can be done in a simpler way.

Thanks.

Re: CMOS Bias current generator and Bandgap reference voltag

calculus_cuthbert said:

could you also suggest circuits for current sources as you mentioned it can be done in a simpler way. Thanks.

Click to expand...

I've often - and successfully - used Eric Vittoz' suggestions for a (partly) weak inversion PTAT current source, s. the PDF below.

Re: CMOS Bias current generator and Bandgap reference voltag

HI Erikl,

I had a few more questions on PTAT current generation. How do I figure out the slope of the PTAT current source?

Could you please let me know if my understanding is correct?

So what I think is that first I would have to figure out the negative slope of the transconductance of the amplifier (wrt to temperature) and then based on the I need to figure out the positive slope of the PTAT.

Is this understanding correct?

Thanks.

Re: CMOS Bias current generator and Bandgap reference voltag

Hi cuthbert,

this depends very much on the method how you want to generate the PTAT current: if you use the scheme of Fig 8.8. (a) from the PDF above, the temperature dependency of transconductance doesn't play a role for the PTAT current generation: because of the exponential characteristics in weak inversion operation, the PTAT dependency results only from the thermodynamic voltage's Ut=kT/q temperature dependency. You just have to make sure that all involved transistors operate in the weak inversion region.

Fig. 8.8. (b) shows a circuit for nearly temperature independent current generation. Only M6 & M7 must operate in strong inversion. In this case, the (Ut)² dependency of the current is (almost) compensated by the 1/T² dependency of the carrier mobility.

erikl