anyone know startup circuit, please come in

[sunjiao3]

Hi, friends, Is it an already-tested phenomenon that the circuits like self-referenced current sources, bandgap reference circuits will stay in “zero state” if they don’t have a so called “startup” circuit? Or the startup problem is just a potential danger we get by theoretical analysis? The attachment2 is the simulation result I got while doing startup simulation using hspice,and attachment1 is the circuit.

There is a strange phenomenon in attachment2: without the startup circuit,(MS0,MS1,MS2,MS3), and VDD rise with a slope of 17.5V/MS, simulate with .TARN 1US 30MS 1US, the output currents seem to oscillate. So I decrease the step by setting .TRAN 1US 30MS 0.01US, however, the oscillation still exists. I set .option method=gear, and got another group of graph(attachment3), in which the output currents “jump” to correct operation state. So, which result can be trustable and why?

Thank you in advance.

 

[nxing]

Not very sure your results but you can try two things:
(1). look at HSpice's manual to see the difference between the option gear and trap (they are actually just two ways to make the circuit converge.)
(2). increase your simulation time to see the result.

 

[sunjiao3]

Thank you, nxing. As for the TRAP and GEAR method, I've read the manu of Hspice. Yes, Trap is the fastest and most accurate method in most time, but it may cause some problem of oscillation like what is shown in the graph. And gear is like a filter, which will fiter the oscillation in the simulation results got by TRAP method. And I try to further decress the step as you suggested, with TRAP method, it sitll oscillation at the begining, but soon, the result stabilized to "0". The problem now is, the results got with TRAP and GEAR method is different. Without the startup circuit, the TRAP method simulate the circuit would remain in zero state, but the GEAR method simulate the circuit would start. Which one I can trust? TRAP or GEAR?

 

[wangyufn]

I think transient simulation may luckily capture startup problem. But is it sufficient? And another problem is whether the current would keep zero forever if it is not properly started when giving the voltage supply?