Anyone has a circuit as a sub-threshold example?

i just learned that there might be some applications in low-voltage or low-power design. anyone can give me an example? thx

I've successfully designed sub-threshold opamps, OTAs, voltage regulators, references (current and voltage), and ADCs.

The main problems with subthreshold region are speed (very low current = serious speed issues), and the exponential current gain can throw power higher than desired at points (eg. one of my voltage references used 30nW of power at 20°C, but 100nW of power at 60°C, 170nW at 100°C).

Basically, the first order approximation of the weak inversion region is identical to that of a BJT.

Does that help at all?

I have some questions as well. I tried to design a current mode bandgap reference to provide a nanoamperes of current that needs resistors in range of Mohms. Is there a way to reduce the resistor value or we just have to live with it? And how about the noise performance of circuits biased in weak inversion as opposed to circuits biased in strong inversion? Thanks

I am currently designing sum-product blocks using gilbert multipliers. All the transistors (except current mirrors) operate in subthreshold region.

pseudockb said:

I have some questions as well. I tried to design a current mode bandgap reference to provide a nanoamperes of current that needs resistors in range of Mohms. Is there a way to reduce the resistor value or we just have to live with it? And how about the noise performance of circuits biased in weak inversion as opposed to circuits biased in strong inversion? Thanks

Click to expand...

My voltage references aren't designed as the standard opamp/current mirror/sub-threshold fets/resistors style - so I can't say offhand.

How many nano-amps are you trying to limit your current to? If it is around 200nA, I'd suggest using multiple divider stages to get a reference voltage of around 40mV and use a resistance of around 200kOhms. This will get you out of the MOhms and should still be easy to implement. I'd also suggest, perhaps, using a lower supply voltage to acheive this. I have working bandgap references at supplies as low as 300mV. If you are using a 1V+ supply and are trying to use the standard bandgap structure, may I also suggest cascode current mirrors to drop the voltages on the active components so that the current is naturally lowered, therefore lowering the resistance needed.

Does any of this help. Do you have any other questions?

Hi |IAngel|, thanks for the suggestion. You mentioned that you use divider stages to get a 40mV of Vref, I am not sure how you generate the voltage reference (V1) to be used in the divider stages. Did you use pure resistive strings as the divider stages?
Moreover, I am not able to find other power supply invariant and temperature invariant voltage reference circuits apart from typical bandgap reference circuit. If you are able to show me some references, kindly enlighten me. I am also curious as to how did manage do design a bandgap reference working at 300mV. Did you special techniques/process to achieve that? Thanks for your guidance.

You can generate that low a reference with purely resistive stages, yes. Much in the same way that it has been done for a few years - divide the output by a resistive network to get a stable voltage at a much lower voltage. I can post a couple of papers on this if you need them.

As for my work, it is awaiting patenting, so I can't mention too much about it right now, but once the patent goes through I can

I didn't use any special methods. I used a standard 0.13um process and designed everything in subthreshold (including current mirrors, as the mirror didn't need to be accurate). That's why I'm becoming a major fan of weak inversion - it works very well.

For your application, I'd suggest a resistive step-down to a lower voltage and then a resistance in the couple-hundred k range. Also - as I mentioned - I'd suggest using a cascode current mirror, as that will drop the current naturally, therefore dropping the needed resistance and improving the PSRR to boot...

Well, this is my first time trying to design circuits in weak inversion for low power low voltage implementation. Could you please post some papers about the divider stages to get a lower reference voltage? I am just wondering, don't the resistor strings draw some currents as well? I might be wrong about the concept. Anyway, thanks for your help. Oh yeah, hope your get your patent.

They would take up extra power (and current), but it isn't as bad as it seems. I'll post a MASc paper when I find the pdf for it. It's a great brief on sub-1V bandgaps. It has an adjustable reference voltage due to a voltage divider at the output (ie. scalable to any desired voltage). That being said - it uses BJTs for PTAT/CTAT generation, but those can be replaced with FETS. It's the idea behind the resistive division that counts.

I am currently designing sum-product blocks using gilbert multipliers. All the transistors (except current mirrors) operate in subthreshold region.

many IEEE papers