一个恒定跨导的cmos运放

qerqing

我设置的是不扣钱，可能管理员还得设置吧

qerqing

m31的漏极电流时怎么被，m17,m18控制的，不太懂

qerqing

有哪位大侠能吧虚线部分的电路给分析一下，谢谢了

ulsi123

Good question, qerqing. The more I reviewed this paper, the more doubts I have.

Basically, the gm-control circuit is to generate Ip that can track in+/- common mode voltage (Vicom). M22/23/24/25/26 is an amplifier to form the negative feedback loop. However, in my opinion, the node between M17 and M20 can fluctuate dramatically and the replica Ip may be decided by the difference between in+ and in- rather than Vicom. I am not sure if we should short the node between M17 and M20 with the node between M18 and M21. Maybe I'd better run some simulations to find out the answer. But my initial feeling is something might not be right with this gm-control circuit. However, I could be wrong as well.

qerqing

楼主说的有道理，不过那个负反馈网络是怎么工作的

ulsi123

   

楼主说的有道理，不过那个负反馈网络是怎么工作的
qerqing 发表于 2009-10-11 12:47

                               
登录/注册后可看大图

如何工作？

在M22-26形成的OPAMP中，M23的栅是OPAMP的正极，M22的栅则是负极，M22/M24的接点则是输出。这样一来，OPAMP会通过调节M27的电流使得M30和M20的漏极电压相等。如果M20和M30是匹配的话，流经M20的电流就会跟流经M30的电流一样，再通过电流镜，由M31流出的电流便是可控的Ip了。

qerqing

16# ulsi123
对呀，我怎么没看出来，哈哈，楼上真是厉害

refugee

1# qerqing

恒定跨导我们一般采用两种方法：一种是3倍电流的方法，一种是恒定电流的方法。前种是考虑输入差分对工作在饱和区时，使只有一个pmos差分对（或nmos差分对）工作时，获得额外的3倍尾电流，这样它的跨导就和pmos差分对和nmos差分对同时工作时的总跨导相同；第二种方法也是业界采用比较多的方法，工业设计中重要的输入差分对尺寸往往很大，使其工作在亚阈值工作条件下，从亚阈值跨导公式：gm=I/nkT可以看出，无论是nmos还是pmos，gm是和I成正比的，不考虑n的差异，只要我们保持I不变，gm也将不变。所以对于轨轨的差分输入，我们保持尾电流之和不变，那么总的跨导也将不变。

作者提出的是流过差分对的电流之和不变而非尾电流之和不变，虽然这样做的精度可能更高一点，但我个人觉得应该不会有太大的改进，只不过是新瓶老酒而已。

refugee

14# ulsi123

Hi, ulsi123. I think you have forgot that this op amp should work in feedback. So, in a stable system, the in+ and in- should be equal. and the current passed through M17 and M18 should be equal in dc value. So, I think the architecture is OK

But there might be a tricky problem. If the author bias M20 and M30 in this way, M20 and M30 should be worked in linear region. To keep the drain of M20 and M30 equal, the dc gain of  amp M22 - M27 should be large. And I'm a little bit worried about the stability of this feedback amp. Cause in case when Ip is small the phase margin would be worse, and this part of circuit might oscillate

novaming

LZ可以参考下SANSEN大叔的模拟精粹，第11章专门讲了这个恒定的轨到轨~~~