The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Mixed-Signal Design >> reference cicuit
https://designers-guide.org/forum/YaBB.pl?num=1251484358

Message started by smarty on Aug 28^th, 2009, 11:32am

Title: reference cicuit
Post by smarty on Aug 28^th, 2009, 11:32am

Hi,
I might sound strange, but it is possible to generate a supply/temperature independent reference without using band-gap. This is for an ADC.

Thanks,
SBR

Title: Re: reference cicuit
Post by wave on Aug 28^th, 2009, 1:23pm

Typical ways to make a ref are with a BG over an external resistor,
OR
to use a crystal clock and generate something useful from there.
Perhaps a Freq. Locked Loop with the oscillating element generating a bias which is then filtered and pass onward.

Title: Re: reference cicuit
Post by RobG on Aug 30^th, 2009, 10:59am

You can do it from any number of sources, for example the Vt of a mosfet added to a PTAT source, but usually these references are not PROCESS independent. There is only one process independent reference available: the PTAT voltage generated by differing current densities in a bipolar transistor. The next best thing is the bipolar transistor base-emitter voltage, which is well behaved over process. When you add the two together you can always trim the result to a "magic voltage" of about 1.2V that is temperature stable. There is nothing out there that compares to that robustness over process.

Other methods: putting a voltage onto a floating gate has been published a few times in the last few years. I'm skeptical of the lifetime of these things, but what do I know... Buried Zener references were an old standby (better than bandgap references), but supply voltages are probably too low in your process. And using MOSFET Vts, but I don't think that gives a "magic voltage" to trim to. A long time ago there was an article in JSSC about using the difference between a "native" MOS and one that had the threshold adjust, but it isn't used much.

That said, you don't need to create a bandgap circuit to generate the reference. E.G. if you want a current, just sum up a VBE and PTAT current. Also, in the charge domain, you can get the same effect of sampling a 1.2V constant "bandgap" reference by sampling a properly weighted combination of PTAT and VBE voltages. You can even do this with one bipolar transistor and a two-level current source if you are clever.

In the end though, it is really hard to beat using the combination of PTAT and VBE to create a reference.

rg

Title: Re: reference cicuit
Post by smarty on Sep 4^th, 2009, 10:11am

Hi,
Thanks for the inputs, but for me it seems band gap is not possible since I am working in sub 1V supply ( min supply can be as low as 0.8) and the vbe can be as high as 0.76.

Best Regards,
SBR

Title: Re: reference cicuit
Post by ACWWong on Sep 4^th, 2009, 5:58pm

"not possible" is not the way I would put it... more likely performance is degraded when the temperature is cold and the voltage supply is 0.8V. For supplies of 0.85V and 0^oC bandgap in the traditional sense (delta Vbe & Vbe) is easily achieved.

Does that the ADC itself start suffering performance degradation with 0.8V irrespective of the bias ? If the ADC needs extremely accurate bias to operate under the specified conditions, then maybe its the ADC which needs modification ?

Title: Re: reference cicuit
Post by loose-electron on Sep 5^th, 2009, 11:24am

There are some IEEE JSSC articles on sub 1V bandgaps out there - try IEEE Explore or just Google it.

Also, a walk around - Use a charge pump boosting structure to generate a larger voltage and run the BG off that, then proportionally divide the voltage down to something smaller, and go back to working on 1V power.

Should be able to get a BG working on just a few uA of juice and that means the charge pump boost is not too problematic to create.