A constant gm circuit's startup

[twonsr]

constant gm bias

Dear all :

A constant gm bias circuit need a startup circuit to set the correct operation point.
But in fact , there are some leakage current in MOS transistor.

I have a idea , if don't use a startup circuit , but use these leakage current to let the circuit shift to the correct opeartion point.

Is this idea ok ?

 

[srivatsan]

constant gm biasing

zero current is a stable one. i fabricated just current reference without startup. it usually did not start. even increasing Vdd from 0 to the actual value, does not excite it as often as required. so leakage current cannot be "trusted".

srivats

 

[twonsr]

constant gm circuit

Excuse me , how do you make sure that the fail is from the issue you said.
Did you measure the signal with probe station ?

 

[Humungus]

start up circuit

What srivatsan says is the pure truth. Never rely on selfstartup. Leakage currents can set several stable points faw away from your expected behavior.

 

[lastdance]

startup circuit

i guess it's alright to ask.
I have not done anything that's going to start up using leakage current, nor would i use one because of the risk involved.
However, I would like to know what do u mean by "Leakage currents can set several stable points faw away from your expected behavior. "
Also, if the circuit has not started up, have you verified that it's due to leakage current not large enough or not present? how do u verify that? have u tried injecting a short pulse of few uA externally and see it start?
Theoretically, everyone knows what's happening. But to verify experimentally is another thing.

 

[raymond33]

constant gm bias start up

As srivastsan told you , if you are into a stable point , as is usual with gm constant circuit ,you can do what do you want but if is point is stable (and current equal to zero is a stable point), your circuit won't work properly in silicon.
Start up circuit is mandotory when you have two stable points .

 

[srivatsan]

start-up circuit

The question is: "How do I know that 'Leakage Current' is NOT the answer?"

Well, Everytime I get a chip fabricated, I look at them under the microscope and just "feel happy" at the way they are fabricated... I am 100% sure that the circut was fabricated the way I wanted to do so. Now you do need to understand "stable". The circuit seems to be happy k: in that value and "injecting 10uA" is never clear for me. So the only way I see, is to increase (or toggle) the VDD and Hopefully the circuit gets into Ibias value. I really never got it working except at situations where I dont understand how it starts. So instead of pursuing the complete reason (which might be of not much use ... atleast right now), I went ahead and used start-up circuit. Then it Definitely works. I am okay with it.

Also I take equations as "rule of thumb" rather than 100% prefect. The only way to understand the working is sometimes proving other conditions cannot occur and Hence this is the "only way" it works...

I hope I am clear. I am always open to other view points. I hope to get different opinion on this topic.

Regards,
Srivats

 

[lastdance]

constant gm startup circuit

In other words, u have not pursued to find out the reason because of the limited use u thought the work could bring. Well, I agree on this.
However, I do not quite understand what u meant by "looking", u should say "probing", right? How could u "look" under microscpoe and be sure of what's going on in there?
As for the few uA, u see, if ur circuit, the one without the start up, has no current in it after u jack up the supply Vcc. I assume it's in one of the stable state with zero current. Now, I go ahead and inject a pulse into the high impedance node to bring the whole thing into disequilibrium again. Theoretically, it should have started because the initial positive feedback should have multiplied the current and stabilised it thru negative feedback.
I guess u have fabricated this chip during ur college years, my boss has specifically banned one from omitting the startup, even if it takes up space.

 

[srivatsan]

gm constant biasing

Dear lastdance,
Of course, you cannot "see"voltages or current; you probe them. I just (for the fun of it, ) see the fabricated devices using the microscope.

I truly attest your boss.... Startup is very important.

Regards,
Srivats

P.S.: well, i am still a college student.. will be graudated this spring...