high precision bandgap without importance of temperature sensitivity

Im designing a bandgap for an application were temperature sensitivity is not important.
Instead the bandgap reference voltage should be quite insensitive to vdd changes and process variation. (5-7%)

I tried several subthreshold solutions but the process variations had been too large.

The traditional bandgap solutions with lateral pnp- devices need immense power (compared with the other solution) and I cannot use one of the main advantages: the ease of the temperature compensation

Is there any posibility either to compensate processvariations at one small temperature range of 3-5° or use a traditional bandgap with less power?

the high precision references I´ve found are always improved for temperature stability but that is of no avail.

Any Idea?

Hello,

As far as I've seen (I've done some bandgap design with genetic algorithms), the bandgap based on threshold voltage behaves quite well in respect to supply voltage changes.

Following paper is a very good collection of bandgap circuits. You might find there the circuit you need.

Christian Jesus B. Fayomi et. al. "Sub 1 V CMOS bandgap reference design techniques: a survey", Analog Integr Circ Sig Process (2010) 62:141–157

Paul

thanks Paul.

Do you mean devices with different thresholdvoltages or devices in subthreshold when you´re writing "bandgaps based on threshold voltage"?

Well, I read through the paper but it tells nearly nothing about the processdependence of the single designs.
It´s quite difficult to find reliable information about process dependence anyway. It seams that I have to implement each design in my technology to get comparable result. :-?

I implemented some designs with only single devices in subthreshold as well as designs which are completely implemented in subthreshold.

In both cases I was able to achieve a quite good vdd-independence (with cascoding or error-amplifiers).
But because of the huge dependence of the reference voltage to process parameters my monte carlo simulation in matching and/or process varied the hell.

My technology provides bipolar transistors so I could easily use them. The main advantage I see for me is to reduce the problem of process dependence by a huge amount.
I need a 500 times larger supply current and the area increases as well (because of the use of resistors in traditional implementations).
If I want to achieve the same low vdd dependence I got with the (partly or fully) subthreshold designs I need 2 current paths more to implement an amplifier (for cascoding I do not have enough voltage headroom in the saturation design). So the power consumption will increase even more.

Well, the question I was refering above is that the traditional implementations with bipolar devices (Banda et al. or similar ones) show always also a good behaviour in temperature independence. But for my application I nearly do not care about temperature. So I was thinking about a posibility to save energy, area... at this point.
But up to now I could not find a design that offers anything in this direction.

zitty said:

thanks Paul.

Do you mean devices with different thresholdvoltages or devices in subthreshold when you´re writing "bandgaps based on threshold voltage"?

...

Click to expand...

What i meant was a bandgap circuit that trims out the temperature variation of a MOS transistor threshold voltage in order to achieve temperature invariance. If you have the paper, it is figure 18.