LDO的AC分析与tran分析结果不一样!!

lsh0211

目前设计一个LDO,整个芯片静态电流小于10uA, 因为电流太小,所以不论是BandGap还是op的mos管都工作在亚阈值状态. 完成补偿网络设计后, ac仿真后调了一下补偿电阻电容,得到较好的相位和增益裕度. 但是tran分析发现一直振荡. 回过头在ac分析下再调相位和增益裕度, 当在tran分析下还是振荡(ac分析的方法确定是正确的).
   最后项目要到期,实在不得已,在tran分析下去调整补偿元件, 经过长时间优化,终于在不同的输出电容esr/不同负载条件下tran分析正常, 且最差情况下过冲也较小.再回头做ac分析,发现增益/相位曲线与先前的很不一样, 同时相位裕度很差, 以为这次ac分析正常了,于是又在ac下把相位裕度调好, 再tran分析,又振荡!!!!
   反复了几次,实在没时间和耐心搞下去了,准备交差, 可是偶然见又发现大负载(300mA)和小负载(1uA)下都正常,可是在1k(2~3mA)下又是振荡的!! 我简直也晕掉了! 周末加了两天班就是这个结果!!  
    现在ac分析已经完全失去意义了, 所以也没有办法去通过ac分析看零点/极点在负载变化时的影响. 不知道是不是由于mos管工作在亚阈值区的原因, 据说bsim3 model在亚阈值区的模型精度不是很高(与饱和区相比), ekv model在此区域好像精度比较高.
   不知道哪位碰到类似的现象? 有没有什么建议?   我准备换不同的补偿方式试试, 可是电流实在太小,可选择的余地也很小,郁闷!

hyy95

画图来看看

prof3

As a matter of fact, I believe your compensation is traditional. This is why your LDO works for heavy load and light load, the cases of which are easy to compensation using pole-spliting concept. However, It may not work that well under median load conditions because now the output pole is also getting lower and this is why the worst case Phase margin at median load. So traditional miller type compensation is hard to stabilize such a large range. You can fix it using dynamic compensation scheme as many others do or you can design a compensation scehem such that it always uses internal node as dominant pole.

I also believe AC simulations can tell us the loop stability information for the LDO type circuit although AC simulations sometimes can "miss" something such as doublet etc. You have to make sure all the loading condition correctly considered when running AC such as ESR, Iload, Rload etc.

Of course, if your LDO also need low noise, high PSRR and also low current, that will be another story. Traditional way hard to achieve all.

lsh0211

楼上说得不错，高手！ 芯片对噪声倒是没有高的要求，只是PSRR要求不低。
我也在考虑使用动态补偿，可是由于静态电流太小，怕是难以挤出足够的电流来用于动态补偿，而且我发现即使同样的环路增益，如果电流太小，PSRR也会下降不少， 试试看吧。
多谢！！

dengqin

eetop

prof3

If yout try dynamic compensation, let say standby current change with Iload, so you only need to pay attention at maximal Iload(to meet standby budget)

icsb

PM will change with load Current
so when you simulate AC,what's your loading condition??
I suggest you simu PM with Loading changes

lsh0211

我现在采用了两种方案, 一是把op增益从100db降低到85db,使得瞬态仿真不同负载/温度等条件下输出最终稳定.
再就是设计动态补偿电路, 可能是op的电流太小(两级不超过2.5uA), 在大负载条件下, 零点/极点很难设计满足要求, 简单调试了一下,也只是达到和第一种方案差不多的要求. 准备按照零点极点分布要求重新设计op试试,不再是简单的修改调试.

JJmeow

OP 只有 2.5uA , 哇 那你的最重loading 是多少呢

122013137

使用动态补偿