A start-up circuit for a voltage reference

[vadim888]

Dear All,

sorry, a new question...

I just need a small hit, because I can't fully understand how that (fig bellow) start-up circuit is working (three transistors on the left side):



Best regards,
Vadim

 

[erikl]

At start, Vref=0, so M3 is closed, long-length weak PFET M1 is open, and M2 is opened, too. The drain of M2 is at GND potential, opening the 3 PFET current sources. Then the Vref circuit goes starting to work, the OpAmp - with its low output impedance easily winning against the weaker M2 - reaches its correct operation point, so does Vref, which opens M3, which then wins against the weak M1 and draws M2's gate potential down to GND. M2 is open now, and the 3 PFETs' bias is fully determined by the opAmp.

No current through M2 any more now, but always cross current through M1-M3. That's why M1 should be rather weak (very long, i.e. W/L≪1).

 

[vadim888]

I should buy you a beer someday, thank you!

 

[vadim888]

So, I have an issue with that start-up circuit... it's oscillates. However, if I increase a width of the M1, hence and the current, it works good. But the current consumption is around 1mA...

Does anyone has a clue why I have oscillations in my start-up circuit?

 

[erikl]

Instead of increasing the width of M1 (it should be a long transistor anyway, s. below), try reducing the width of M3.

Oscillation originates from not enough gain in the (positive) feedback loop during the switch-off period. Also try increasing the width of M2 - so increasing this gain. M2 is totally off after start, so no current through it, then. Only M1 - M3 leads current after startUp, which is determined by the W/L ratio of M1. That's why M1 usually is a very long transistor (W/L « 1).

 

[dick_freebird]

The trick with DC startup circuits is to get them to be always
effective, but also always "shut up" once the output is up at
its proper value. This depends to some extent on the loop amp's
gain at lower than normal output voltage (amp must "snap in"
nicely). By the time the output reaches designed setpoint, the
output voltage must be enough to make the startup circuit
quit pushing current at the P mirror rack altogether. If amp
gain or common mode range is inadequate then you may see
oscillation (if weak @ corner) or tempco issues (if strong @
corner) from startup current continuing to bleed into the
loop.

 

[vadim888]

That's why M1 usually is a very long transistor (W/L « 1).

but it will increase the static power consumption of the circuit. SO maybe I should try another start-up circuit?

- - - Updated - - -

If amp
gain or common mode range is inadequate then you may see
oscillation (if weak @ corner) or tempco issues (if strong @
corner) from startup current continuing to bleed into the
loop.

you were right, the opamp's gain was low. Thank you for your advise.

 

[erikl]

That's why M1 usually is a very long transistor (W/L « 1).

but it will increase the static power consumption of the circuit. SO maybe I should try another start-up circuit?

In contrary: Long transistors (W/L « 1) provide low currents (low static power consumption).