Designer’s Guide Community

tulip

Community Member

Offline

Posts: 41

Does every bias circuit need a start-up circuit?
Aug 20^th, 2007, 1:25am

Does every bias circuit need a start-up circuit?
I simulated a constant-gm biasing circuit, by using trans and DC simulation,
it do not have start up circuit and it works properly.

My question is : does this curcuit has sescond stable condition
when all comonents are off, how can I find the second stable
condition by simulation?

Back to top

Not every morning wake up your alarm clock, but dream!

IP Logged

vivkr

Community Fellow

Offline

Posts: 780

Re: Does every bias circuit need a start-up circui
Reply #1 - Aug 20^th, 2007, 2:04am

Yes, every self-biased circuit needs a startup, and that covers pretty much all bias circuits that I know of.

You can do a DC sweep of the current in the 2 branches of your bias circuit to see where the circuit has stable
DC operating points. You will see that your circuit probably has a second operating point for the case where
no current flows into either branch.

This is actually a metastable state but even then, the circuit is unlikely to startup well without a
well-designed startup circuit. It is worth the effort to make this, and to make it well unless
you want to make it after your silicon is back

Regards
Vivek

Back to top

IP Logged

tulip

Community Member

Offline

Posts: 41

Re: Does every bias circuit need a start-up circui
Reply #2 - Aug 20^th, 2007, 2:10am

Thank you very much.

I know this circuit has a second stable state with current=0

But someone says that because of leakage current, the
circuit is sure to start, just slower than when it has a start-up
circuit. I suspect what he said, but I can not find a way to
persuade him, because the simulation all show the circuit is
working well.

best regards.

Back to top

Not every morning wake up your alarm clock, but dream!

IP Logged

RobG

Community Fellow

Offline

Posts: 570
Bozeman, MT

Re: Does every bias circuit need a start-up circui
Reply #3 - Aug 20^th, 2007, 12:26pm

tulip wrote on Aug 20^th, 2007, 2:10am:

But someone says that because of leakage current, the
circuit is sure to start, just slower than when it has a start-up
circuit. I suspect what he said, but I can not find a way to
persuade him, because the simulation all show the circuit is
working well.

best regards.

Vittoz was telling people that a few years ago, so there is probably a lot of truth to it with MOS circuits (bipolar circuits fail because of base currents). However, start-up can take several seconds, or if your leakage current isn't enough to overcome other effects (like injected noise or leakage current in the wrong direction) then it may not start up at all.... and in my career I've had to slap a few people for dumb start-up circutis that didn't work so this is an issue you should take seriously! What you don't want is it to work 99.9% of the time, meaning you have some hard-to-track-down error in a production device. Put a decent start-up circuit on it.... (Also make sure the SU circuit shuts off!)

A proper start-up simulation usually requires you to have it powered down with full supply (short a gate with an ideal switch). When you open the switch the bias should start-up. The worst case seems to be cold with mismatched devices that lower the gain of the positive feedback path.

Back to top

IP Logged

RobG

Community Fellow

Offline

Posts: 570
Bozeman, MT

Re: Does every bias circuit need a start-up circui
Reply #4 - Aug 20^th, 2007, 1:47pm

By the way, there are circutis that don't need start-up circuits. I've used variations of the following circuit with good success (and there are countless other ways to do this), but you almost always improve performance by self bias -- in this case bootstrapping the output current to the M1a/b devices to provide their current instead of relying on the power supply, which means you will need a start-up circuit.

Back to top

IP Logged

tulip

Community Member

Offline

Posts: 41

Re: Does every bias circuit need a start-up circui
Reply #5 - Aug 20^th, 2007, 8:26pm

RobG wrote on Aug 20^th, 2007, 12:26pm:

A proper start-up simulation usually requires you to have it powered down with full supply (short a gate with an ideal switch). �When you open the switch the bias should start-up. �The worst case seems to be cold with mismatched devices that lower the gain of the positive feedback path.

by using the simulation method you proposed, I can see the circuit start up in 5ms.

If delay is not concerned, can this circuit be used without start up circuit?

Thank you.

Back to top

Not every morning wake up your alarm clock, but dream!

IP Logged

RobG

Community Fellow

Offline

Posts: 570
Bozeman, MT

Re: Does every bias circuit need a start-up circui
Reply #6 - Aug 21^st, 2007, 11:18am

I'd weigh the risks versus the benefits. Your simulation may have missed something; in that case it may not start-up in real life. I would use a start-up circuit unless you REALLY have a good reason not to, and a good story to tell your boss if it don't work.

Even with low power circuits, where the SU circuit are costly, I have always errer on the side of caution and used one with the common self-biased circuits.

Back to top

IP Logged

tulip Community Member Offline Posts: 41	Re: Does every bias circuit need a start-up circui Reply #7 - Aug 21^st, 2007, 6:14pm Many thanks. I really appreciate your help.
Back to top	Not every morning wake up your alarm clock, but dream! IP Logged

joel Community Member Offline Posts: 43	Re: Does every bias circuit need a start-up circui Reply #8 - Sep 26^th, 2007, 2:30pm I've had good luck by building a digital start-up circuit constructed from a one-shot triggered by RST or some other sequential initialization event. This then yanks up one side of the bias circuit, making sure it doesn't lock into it's off-state. Just a thought...
Back to top	IP Logged