The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> Start-up circuits
https://designers-guide.org/forum/YaBB.pl?num=1340780454

Message started by arupkg on Jun 27th, 2012, 12:00am

Title: Start-up circuits
Post by arupkg on Jun 27th, 2012, 12:00am

Hi all,

      How do we know if a start-up circuitry is essential for a circuit topology? When we do the simulation for a self-biased (e.g. beta-multiplier current reference, the results are the same irrespective of whether the start-up circuit is present or not. I am just using ADE for the simulation. Kindly share your thoughts, as to why this is happening.

Many thanks

Title: Re: Start-up circuits
Post by raja.cedt on Jun 27th, 2012, 1:48am

When you have multiple operating points, and not to let the circuit operate at undesired point we use start-up circuit. This doesn't change result but in some cases it forces the result to desired point.

In case of beta multiplier, you have 0 current and some other currents are the final solutions and if you don't use start up ckt it may settle at any point.

Thanks,
Raj.

Title: Re: Start-up circuits
Post by aaron_do on Jun 27th, 2012, 6:09am

I think the question was more pertaining to the simulation itself.

I have had cases where a circuit was unable to start up in simulation without a start-up circuit. I think it depends on how the circuit is solved by the simulator. For some circuits, the solving algorithm may lead to the desired operating point, and for other circuits, it may lead to the undesired operating point.

I'm not sure if there's any simulation to see if a circuit has start-up issues, but if there is one, its probably a transient analysis with step input on the power supply (start from zero initial conditions).


regards,
Aaron


Title: Re: Start-up circuits
Post by RobG on Jun 27th, 2012, 7:17am


aaron_do wrote on Jun 27th, 2012, 6:09am:
I'm not sure if there's any simulation to see if a circuit has start-up issues, but if there is one, its probably a transient analysis with step input on the power supply (start from zero initial conditions).


Cap coupling can provide startup current so ramping supply can miss issues. However, you do want to make sure that a glitch on the power supply won't shut it off -- a friend had that happen to him. This is the reason I don't use por-type startup circuits. DC sims also  don't tell you that it might start up in real life from parasitic leakage, but it could take an extremely long time because the inherent startup currents from leakage are very small - and may even be overcome by leakages not modeled.

The best sim I've found is a power up sim without ramping the supply. Make sure all mosfets are off. Even if you have to add extra circuitry to power it down this is the sim you want to look at.

The second extremely important simulation is mismatch (with the power up sim). You need to overcome the offsets in the opamp (or mos pair, etc) to activate the positive feedback part of the loop.

So power-up with mismatch is what you want to look at.

Title: Re: Start-up circuits
Post by aaron_do on Jun 27th, 2012, 8:09am

Hi RobG,



Quote:
The best sim I've found is a power up sim without ramping the supply. Make sure all mosfets are off. Even if you have to add extra circuitry to power it down this is the sim you want to look at.


I didn't really understand what kind of simulation this is. How do you power up something without ramping? Do you mean have the power supply already up, but set initial conditions so that the transistors are initially OFF? Actually I doubt this will guarantee catching any undesired operating points but maybe it is better than ramping the supply...not really sure.

I think if there were a way to do a DC analysis while asking the simulator to use a particular node voltage and check for a solution. You could then sweep said node voltage and see if there is more than one solution. Not sure if this is possible...of course this wouldn't tell you how long it takes to reach the solution as you pointed out earlier...


regards,
Aaron



Title: Re: Start-up circuits
Post by RobG on Jun 27th, 2012, 8:36am


aaron_do wrote on Jun 27th, 2012, 8:09am:
I didn't really understand what kind of simulation this is. How do you power up something without ramping? Do you mean have the power supply already up, but set initial conditions so that the transistors are initially OFF? Actually I doubt this will guarantee catching any undesired operating points but maybe it is better than ramping the supply...not really sure.  

Close but not quite. Yes, you power it up with the supply already up. This will require a power up/power down input that disables the circuit and make sure all devices are off -- just your basic power down circuitry. Peronsally, I'd add the circuitry even if the specs don't call for it so you can do a good test of the startup.

I suppose setting the initial conditions so that the circuit was off at time=0, but that seems very tedious and if I was to do it I know I'd spend two days getting it right  ;D

Title: Re: Start-up circuits
Post by RobG on Jun 27th, 2012, 10:13am

I should should add that it isn't good enough to just verify it starts up, you need to verify that the startup circuit is operating properly. Make sure it injects a reasonable amount of current into the startup node (I prefer a current on the order of the normal operating current, too much and you can get into an oscillating condition). Above all, make sure it turns all the way off!

Most of the problems I have had with startup circuits are that they don't turn on at cold and don't turn off at hot, so make sure you check over temperature.

Title: Re: Start-up circuits
Post by Dan Clement on Jul 11th, 2012, 7:52pm

I've had the same problems with my startup circuits Rob...

I shipped a bg once that didn't start up...  From then on I have been very careful. I like to transient sims that ramp the power supply at every conceivable ramp rate from 1ns to seconds depending on the situation. The hope is to rely on the real startup circuit and not be fooled by parasitic caps and leakages.

I also do dc sweeps of the power supply and compare to the transient in terms of the different voltage threshold as things power up and the startup circuit powers down.

Sometimes you can also do a dc voltage sweep on the startup circuit and see where the current goes to zero. This can find stable states that maybe you didn't know about.

I also agree with Rob to watch out for the mismatch...

Good luck!

Title: Re: Start-up circuits
Post by Alberto on Oct 16th, 2013, 6:20pm

I was trying to simulate a low voltage band gap voltage reference without start up circuit only with node set conditions. Then I weept the Vbias voltage (VDD) from 0 to the MAX VDD. I did not have the curve Vref("VT") vs VDD some one could help me please? I read the Ken Kundert's book about SPICE and SPECTRE simulation and I tried to adjust the convergence items in Virtuoso ADE and the circuit did not converge. The result is the shape in the bottom of the post. The envelope of this curve is the desired curve.

Title: Re: Start-up circuits
Post by AnilReddy on Oct 17th, 2013, 12:13am

Hi Alberto,

Did you check the stability of the system?

Bye

Title: Re: Start-up circuits
Post by harpoon on Oct 17th, 2013, 2:40am


Alberto wrote on Oct 16th, 2013, 6:20pm:
I was trying to simulate a low voltage band gap voltage reference without start up circuit only with node set conditions. Then I weept the Vbias voltage (VDD) from 0 to the MAX VDD. I did not have the curve Vref("VT") vs VDD some one could help me please? I read the Ken Kundert's book about SPICE and SPECTRE simulation and I tried to adjust the convergence items in Virtuoso ADE and the circuit did not converge. The result is the shape in the bottom of the post. The envelope of this curve is the desired curve.


Try simulating in transient ... looks like you have a startup issue (esp since you have no startup circuit) !

Title: Re: Start-up circuits
Post by harpoon on Oct 17th, 2013, 2:46am


arupkg wrote on Jun 27th, 2012, 12:00am:
Hi all,
      How do we know if a start-up circuitry is essential for a circuit topology?


If it can settle at 2 or more unwanted steady states (i.e. it is stable with 0 current), then you definitely need one.

If you see strange behaviour (like Alberto above), you need one.

If you are leaving parasitics to get the circuit to a steady state, you need one. (better to be safe than sorry !).

it will come with time and experience ... all the best !


Title: Re: Start-up circuits
Post by Alberto on Oct 17th, 2013, 3:35am


A Kumar R wrote on Oct 17th, 2013, 12:13am:
Hi Alberto,

Did you check the stability of the system?

Bye


Hi Anill thanks for your help. The stability is ok, the loops in the bandgap are working, One more thing,  the opamp is only a VCVS (voltage controlled voltage source) then the stability problem of the OPAMP is not an issue in this case. The TC ( Temperature coeffitient) curve, in other works the Vref vs Temp of the bandgap apear with my node set condition, but when I try to simulate the line regulation (Vref vs VDD(bias)) I had the last curve that I presented upside. And I need verity the VDD maximum and minimum where the bandgap circuit works, remember that my bandgap is a low voltage bandgap and this circuit will most work in the VDD range from 900mV to 1.2 V.

Thanks


Title: Re: Start-up circuits
Post by RobG on Oct 17th, 2013, 8:36am

I think this is a DC sweep so stability isn't any issue. Note that using a vcvs can cause problems (even AC stability since it might not have a dominant pole). Set the min and max voltages of the vcvs to your power supply, otherwise it might find an operating point out there at 500 volts or something ridiculous.

I'm not sure what your goal is, but you can't get reliable results without some sort of startup. If you are just roughing out the design you can bias both bipolars with a very small current using ideal current sources - say 0.1% of their normal current. This should ensure that the circuit starts up so you can test the rest of the circuit.

Title: Re: Start-up circuits
Post by Alberto on Oct 17th, 2013, 9:11am


RobG wrote on Oct 17th, 2013, 8:36am:
I think this is a DC sweep so stability isn't any issue. Note that using a vcvs can cause problems (even AC stability since it might not have a dominant pole). Set the min and max voltages of the vcvs to your power supply, otherwise it might find an operating point out there at 500 volts or something ridiculous.

I'm not sure what your goal is, but you can't get reliable results without some sort of startup. If you are just roughing out the design you can bias both bipolars with a very small current using ideal current sources - say 0.1% of their normal current. This should ensure that the circuit starts up so you can test the rest of the circuit.


HI Rob I do that, I put the limits of the VCVS. And it did not works.

Title: Re: Start-up circuits
Post by RobG on Oct 17th, 2013, 10:29pm

If you put up the schematic maybe we could offer suggestions - hard to help otherwise.

Title: Re: Start-up circuits
Post by Alberto on Oct 19th, 2013, 2:23pm


RobG wrote on Oct 17th, 2013, 10:29pm:
If you put up the schematic maybe we could offer suggestions - hard to help otherwise.


Hi RobG thanks for the suggestion, the schematic is the next picture

Title: Re: Start-up circuits
Post by RobG on Oct 19th, 2013, 9:40pm

I haven't seen that one before and it isn't immediately obvious that R1 won't keep it from starting up reliably unless you force the operating point quite a ways from the 0 state. Basically you have to get the diodes turned on so that the delta-Vbe is greater than the I*R1 drop or the circuit won't wind up....  The problem is that the R2s steal all the current from the diodes.

You might be able to make it work with a nodeset for the PMOS gates = 0V. It might also work if you inject a small current into node "A" to make sure it isn't in the 0 state and increase Av from 1e3 to 1e5. But I think you'll need to figure out a startup circuit before you'll be able to simulate -- assuming one exists.

Title: Re: Start-up circuits
Post by Alberto on Oct 20th, 2013, 2:46am


RobG wrote on Oct 19th, 2013, 9:40pm:
I haven't seen that one before and it isn't immediately obvious that R1 won't keep it from starting up reliably unless you force the operating point quite a ways from the 0 state. Basically you have to get the diodes turned on so that the delta-Vbe is greater than the I*R1 drop or the circuit won't wind up....  The problem is that the R2s steal all the current from the diodes.

You might be able to make it work with a nodeset for the PMOS gates = 0V. It might also work if you inject a small current into node "A" to make sure it isn't in the 0 state and increase Av from 1e3 to 1e5. But I think you'll need to figure out a startup circuit before you'll be able to simulate -- assuming one exists.


Thanks RobG i will test that solution, the BG reference exists, in a master thesis. I will be working in this issue and I will comment the results.

Title: Re: Start-up circuits
Post by Alberto on Oct 21st, 2013, 3:36am


RobG wrote on Oct 19th, 2013, 9:40pm:
I haven't seen that one before and it isn't immediately obvious that R1 won't keep it from starting up reliably unless you force the operating point quite a ways from the 0 state. Basically you have to get the diodes turned on so that the delta-Vbe is greater than the I*R1 drop or the circuit won't wind up....  The problem is that the R2s steal all the current from the diodes.

You might be able to make it work with a nodeset for the PMOS gates = 0V. It might also work if you inject a small current into node "A" to make sure it isn't in the 0 state and increase Av from 1e3 to 1e5. But I think you'll need to figure out a startup circuit before you'll be able to simulate -- assuming one exists.


Hi RobG you were right the nodes to apply node set are: node A and the gate of the PMOS. That's all, That band gap circuit does not have the problem of the "I" drop in the parallel resistors of the diodes because they are big, more or less 1MOhm each.

Thank so much for your help. I'm attaching the result.

Title: Re: Start-up circuits
Post by RobG on Oct 21st, 2013, 7:32am

Glad I could help.

Is that circuit published?

Title: Re: Start-up circuits
Post by Alberto on Oct 21st, 2013, 8:55am


RobG wrote on Oct 21st, 2013, 7:32am:
Glad I could help.

Is that circuit published?


Yes It is, It was published by Gupta student of Phillip Allen

Title: Re: Start-up circuits
Post by RobG on Oct 21st, 2013, 9:29am

I couldn't find it. Do you have the year and publication?

I assume the resistor divider + diode circuit is a Thevenin equivalent of a Vbe/2 diode with a series R of R2/2.

Title: Re: Start-up circuits
Post by Alberto on Oct 21st, 2013, 9:51am


RobG wrote on Oct 21st, 2013, 9:29am:
I couldn't find it. Do you have the year and publication?

I assume the resistor divider + diode circuit is a Thevenin equivalent of a Vbe/2 diode with a series R of R2/2.



I will send you and email please check it

The Designer's Guide Community Forum » Powered by YaBB 2.2.2!
YaBB © 2000-2008. All Rights Reserved.