doubt on simulating the startup circuit

[sunjiao3]

Hi, friends, I doubt some questions on simulating the startup circuits. First, why should we use both DC and transisent simulation to test it? Second, why shoudl we simulate with different slope of VDD ramp up in transisent analysis? Third, is it a real case in circuits or just a poential danger that the circuit will not start without a startup circuit?
Thank you all in advance.

 

[jiaming_lee]

without startup, the selfbiasing circuit in big possibility can not start.
There are not many situations for the different ramping up speed. Test startup circuit with different speeds which are according to the practical situations.

 

[electronrancher]

Transient is the only way to do startup sims. You should ramp Vdd very slowly (1 second for 0-3.3v for example) and watch to make sure your circuit starts. Cold is usually worse if your circuit uses PTAT currents since they are smaller at cold.

Slower ramps are always worst case, because some circuits get "knocked into startup" by capacitive coupling when you ramp quickly. This is false startup - remember, some large power supplies do take almost 1 second to ramp up to their true voltage, so if your circuit is used with a case like this, you cannot rely on a fast ramp-up of Vdd to start your circuit.

 

[sunjiao3]

Thank you, electronrancher, but, as for the transisent analysis with VDD ramp up, I have another question. Will the step of analysis, like 1ns or 1us or 1ms, affect the result of simulation? And why?

And another question, once, I simulated the self-referenced circuits (without startup circuit) with VDD ramping up at different slope. I found that, though the circuit seems to remain in zero state at first, it would "jump to start" if you increase the Tstop of transisent analysis.

 

[leichen_gy]

A starup is always required for the actual circuit to start properly everytime you power on.

I had a very bad experience, my bandgap couldn't start. Really! It couldn't start. The problem with your simulation is, it never takes care of the offset. I later figured out it's due to offset issue, I can actually start up the circuit by injecting some noise, by touching the output pin with a twizzer.

So never trust your simulation too much!

 

[sunjiao3]

Thank you, leichen_gy, thank you for your experience. But, I feel even more puzzled now than before. If your bandgap may not start with startup circuit, then, what I can do to know if the circuits, like self-biased current source will startup by simulation? And what's the meaning of startup simulation? Is it a useless and untrustable work since all the results may differ greatly from the true tested results? I am really puzzled now.

 

[leichen_gy]

I am sorry sunjiao3, I forgot to mention that I had no startup circuit with my problematic bandgap circuit.

To study the effect of the startup circuit, I have had the opportunity to build two bandgap circuits, one without startup, the other one with startup, the rest are the same. For both, an offset is deliberately created at the opamp input.

The result shows the circuit without startup doesn't start, or i need to inject a noise to start it up. Whereas the one with startup circuit always has no problem to start whenever I turned on my power supply.

So always put a startup circuit for your bandgap or your self-biased current circuits.

You startup simulations are still correct and valid, the only problem is your simulation makes some assumptions which may not be true in actual case. For my case, the offset was not taken care of.

 

[lladnar23]

Bob Pease wrote a nice article on bandgaps, and he talks a bit about how to test them, the startup circuits, pitfalls, etc...
Check it out at: **broken link removed**

 

[sunjiao3]

Thank you all for your clear answers. But the startup circuit for bandgap differs a lot from its conterpart in self-biased current source, right? And, I've found in other related topics in this forum that DC analysis is also needed. Why DC analysis is needed here? I try to sweep the VDD from 0 to 3.3, and found that the state seems right without startup circuit. So, what's the meaning of doing DC analysis?

 

[jordan76]

... Why DC analysis is needed here? I try to sweep the VDD from 0 to 3.3, and found that the state seems right without startup circuit. So, what's the meaning of doing DC analysis?

Hi sunjiao3,

The purpose of doing DC analysis is to give you an idea of what the proper operating point is supposed to be and provide your a reference for further check with transient simulations under different conditions as mentioned above.

Hope it helps you

regards,
jordan76

 

[sunjiao3]

Thank you all again. With all your answers, I went on investigate into this issue, and got a new question.

As electronranher said, slower ramps are always worst case, but I found that, without the startup circuit, if I increase the end point of transisent simulation, the circuit would "jump" from "zero" state to correct working point; this case is even more clear if I use the gear algorithem rather than the default algorithem of Hspice---trapezoidal algorithem. So, which results can I believe?

Or, I should simulate with different slope of VDD ramp up, with large enough end point of transisent analysis, and even different cases(like ff, ss corner, temps, VDD)----is it necessary? And if there is a case in which the circuit doesn't startup, I would be safe to say the circuit would not startup, right?