The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> Constant Current Bias Circuit
https://designers-guide.org/forum/YaBB.pl?num=1249023960

Message started by cmos.analogvala on Jul 31^st, 2009, 12:06am

Title: Constant Current Bias Circuit
Post by cmos.analogvala on Jul 31^st, 2009, 12:06am

I am looking for a constant-I bias circuit to generation constant current across corner. I understand that generally constant-gm bias circuit are employed as we want constant gain over process variations. Are there any constant-I bias circuit ? I need constant-I bias circuit for LVDS circuits. What kind of bias circuits are used in industry for such applications ?

-CA

Title: Re: Constant Current Bias Circuit
Post by aaron_do on Jul 31^st, 2009, 8:21am

Hi,

as far as i know, you can only create a really constant voltage (bandgap). In order to generate a constant current, you can use a constant on-chip voltage and an off-chip resistor. If the current doesn't have to be that accurate, you could use a switched capacitor to emulate a resistance since I believe on-chip capacitors vary less over process as compared to on-chip resistors. I can't say how they normally do it in industry however...

cheers,
Aaron

Title: Re: Constant Current Bias Circuit
Post by RobG on Jul 31^st, 2009, 8:43am

You can do this many ways w/o generating it from a preexisting bandgap. Take a look at the sub-1V reference from the other post:

Note that the current in R1 is PTAT (proportional to absolute temp) and that the current in R2 is CTAT (complementary to absolute temp). The temp co of the combined current is determined by the ratio of R1 and R2. For a small value of R2, the current will be mostly CTAT. For a small R1, the current is mostly PTAT.

So, you can get almost any temp-co you want.

rg

Title: Re: Constant Current Bias Circuit
Post by aaron_do on Aug 2^nd, 2009, 6:40pm

Hi,

just wanted to point out that the figure above only generates a reference voltage. The current can only be used as a reference with the help of an off-chip resistor.

cheers,
Aaron

Title: Re: Constant Current Bias Circuit
Post by Berti on Aug 3^rd, 2009, 12:04am

I would like to add that although a bandgap generates a temperature-stable voltage, it is absolute accuracy is about only ±5%. Unless you add some kind of trimming...

Cheers

Title: Re: Constant Current Bias Circuit
Post by raja.cedt on Aug 3^rd, 2009, 1:40am

hi guys,
band gap will give constant voltage across temp variation but what about process and Voltage (i assumed that according to berthi we need to use curvature compensation for better result )

Thanks,
rajasekhar.

Title: Re: Constant Current Bias Circuit
Post by RobG on Aug 3^rd, 2009, 8:38am

aaron_do wrote on Aug 2^nd, 2009, 6:40pm:

just wanted to point out that the figure above only generates a reference voltage. The current can only be used as a reference with the help of an off-chip resistor.

Sorry I was not clear, but I thought everyone could pick up on the concept I described without re-drawing the circuit. As shown, the MOSFET on the far right is biasing a resistor with a current. You can obviously bias something other than the resistor using this far right resistor.

(As an aside, you can't generate both the bandgap voltage reference AND the 0 ppm/C current reference at the same time unless the resistors have zero temp-co.)

I think you would agree that there are currents in the MOSFETs. The temp-co of the MOSFET currents can be changed from roughly +3000ppm/C to -3000ppm/C by varying the relative size of R1 and R2 (R2 being the sum of R2A and R2B). Accordingly, you can generate a 0 ppm/C current using only internal resistors.

Regarding tolerances, you can generate an accurate voltage reference using internal components, but current references will be on the order of +/- 30% since that is the tolerance of the sheet resistance.

Title: Re: Constant Current Bias Circuit
Post by cmos.analogvala on Aug 3^rd, 2009, 9:58pm

I could design constant-I bias circuit that gives Iref with +/- 15 % variation with Vdd and +/- 9% with process and mismatch at 90nm technology.

How good / bad is it ?

-CA

Title: Re: Constant Current Bias Circuit
Post by RobG on Aug 3^rd, 2009, 10:38pm

I do not understand your question. +/-15% over Vdd is not a very good design. You should be able to get less than 1% using cascode devices if your design needs it.

Your variation over process corners will be dominated by the variation of your resistors so +/- 9% seems optimistic unless your resistors are tightly controlled.

rg

Title: Re: Constant Current Bias Circuit
Post by cmos.analogvala on Aug 5^th, 2009, 9:29pm

Thanks for the reply. There was a basic flaw in what I had designed.
I have two more questions.

1. Does classical beta multiplier based reference current generation circuit become unstable in the absence of Rs ? I was just trying to understand evolution of this circuit. It was written in one of the text book that this circuit becomes unstable without Rs. I was just wondering in today's technology ro of transistors are finite. It can be easily proved quantitatively that if ros are finite, closed loop gain of beta multipier circuit is < 1 at low frequency. In this case is instability of beta multiplier also a concern ?

2. By definition this Beta multiplier based circuit with Rs is a constant-gm biasing circuit. This means it keeps gm of the transistor constant as long as Rs is constant. Does it mean to keep gm constant with process this circuit changes I such that sqrt(2 Mu CoxW/L I) remains constant. This means this circuit increases I in SS corner so that gm remains constant. Am I correct ?

-CA

Title: Re: Constant Current Bias Circuit
Post by raja.cedt on Aug 5^th, 2009, 10:00pm

hi
1.Even though you consider gds of transistors...still you have good range of stable R (i tried till 90nm tech)

2.yaa based on the corner, -ve feedback of the circuit keeps gm constant by changing current.

thanks,
rajasekhar.

Title: Re: Constant Current Bias Circuit
Post by cmos.analogvala on Aug 8^th, 2009, 3:19am

Thanks Rajashekhar,

I didn't get your answer to the first question clearly. You mean beta multiplier without Rs can also be stable at sub 100nm technology.

-CA

Title: Re: Constant Current Bias Circuit
Post by loose-electron on Aug 9^th, 2009, 6:18pm

One overlooked detail in the dialog - the application is for biasing a LVDS system.

If that's the case, you really dont care about accurate current, you care about "enough current" over process corners, and maintaining an appropriately biased set of operating points.

I do agree with the earlier posting where you can only get an accurate current (for the sake of discussion, better than 1% accuracy) with an external reference resistor, or some method of trimming.