在线咨询
eetop公众号 创芯大讲堂 创芯人才网
切换到宽版

EETOP 创芯网论坛 (原名:电子顶级开发网)

手机号码,快捷登录

手机号码,快捷登录

找回密码

  登录   注册  

快捷导航
搜帖子
楼主: austxuhui

[讨论] 基本电流镜仿真结果结果不对,需要注意什么地方

[复制链接]
发表于 2012-8-9 21:36:04 | 显示全部楼层
回复 20# austxuhui

While trying to copy current it is necessary to set v(A)=v(B). Otherwise you will have VDS modulation and so error. For this process Vth maybe at 0.4V level so your first simulation with 0.9V is more appropriate (3V is too high and so you will have much bigger error due to VDS modulation and other second order effects). If I were you I would not sweep the current but set a value like 0.8uA and VDD=0.9V and sweep R value to find out where to branch currents are equal. When currents are equal I expect V(A)=V(B) would be equal. Once you get equal branch currents than you can sweep input current around that nominal current and observe V(A), v(B) and current mismatch. Current mirrors does not copy every current value so there is a range for current values they work with an acceptable error. Good luck. 而试图复制当前有必要设定V(一)= V(B)。否则,你将有VDS的调制等错误。也许在这个过程VTH为0.4V水平与0.9V的模拟是比较合适的(3V太高,所以你将有更大的错误,由于VDS的调制及其他二阶效应)。如果我是你我不会扫的电流,但像0.8uA和VDD=0.9V和扫描R值设置的值,找出分支电流是相等的。当电流是相等的,我希望V(A)= V(B)将是平等的。一旦你得到平等的分支电流比你可以打扫,额定电流输入电流和观察五(一),V(B)和电流不匹配。电流镜不复制每一个电流值,所以是他们与一个可以接受的错误工作电流值的范围。祝你好运。
发表于 2012-8-10 10:12:08 | 显示全部楼层
1. mos current mirror 得在工作區 ..
     constant current 會和 load 有關的
2. mos Vds 會影響到 mirror . 一般要 mirror 夠好就是 L => large
   但因此 L large =>  W & M 都拉大..
一般 0.5um  多選 L >3u  5u  ..要  Vds  能 1~3v都還能穩定流 ,  L=15u W=20u M=6 ..

32nm process analog design VCC 會多低阿??
 楼主| 发表于 2012-8-10 14:57:16 | 显示全部楼层
回复 21# AcoAco


     以下是我对英文部分的理解:
    需要设置v(A)=v(B)来拷贝电流。否则会有VDS调制错误等问题。在这个工艺下,Vth可能是在0.4v左右,因此你原来用0.9v的仿真是恰当的(3V太高了,由于VDS调制和其他二阶效应你可能会有更大的错误)。如果我是你,我不会来变动(sweep)电流而是设置大约0.8uA和VDD=0.9V,然后去变动电阻R,这样去找分支电流相等的地方。当电流相等以后,我想v(A)=V(B)大致就实现了。一旦你获得了相等的分支电流,你可以在这个正常的电流值附近变动电流值去观察V(A),V(B)以及电流不匹配的情况。电流镜只能在可以接受的失配范围内拷贝一定范围内的电流而不能拷贝任意电流。祝你好运。

   非常感谢您的回复,我试了该方法很好。只是您为什么要确定电流为0.8uA呢?
 楼主| 发表于 2012-8-10 15:02:27 | 显示全部楼层
回复 22# peterlin2010

感谢回复。

    2. mos Vds 會影響到 mirror . 一般要 mirror 夠好就是 L => large
   但因此 L large =>  W & M 都拉大..
一般 0.5um  多選 L >3u  5u  ..要  Vds  能 1~3v都還能穩定流 ,  L=15u W=20u M=6 ..

这里,“L large =>  W & M 都拉大..”是什么意思啊?L和W是宽长比中的L和W吗?M是什么呢?

32nm process analog design VCC 會多低阿??
32nm制程的标称VCC为0.9V。
发表于 2012-8-10 17:19:48 | 显示全部楼层
回复 23# austxuhui


    Since I do not know your M1, M2, R sizes and models I wanted to speed you up to a good enough point in your simulations. There is no magic about 8.3uA current, I got it from your simulation result curves shown at your post number 14. On your plot with your existing M1, M2 and R values, 8.3uA  is giving you aproximately v(A)=v(B) (Eye-balling from your plot I got 8.3uA, so it is aproximate). Now if you like the results while still watching v(A)=v(B) you can change current value provided that you keep current density through M1 and M2 the same i.e. change current and device sizes together so you have same current density on CMOS transistors. If you satisfied with your new nominal current value, device sizes and results than check VDsat=Vgs-Vt of transistors and make sure VDsat is about 200mV for better matching results (If you go closer to weak inversion i.e. 70-100mv of VDsat your mirror error will be worse since Vgs mismatch dominates). I realize Google translation is not as clear as it should be,  I do not have any other way, but you understood so it worked. Good luck.
既然我不知道你的M1,M2,研究尺寸和型号,我想加快你在你的模拟足够的好点。有没有约8.3uA电流的魔力,我从你的仿真结果曲线显示在您的后14号。在您与您现有的M1,M2和R值的情节,8.3uA给你aproximately至五 V(A)= V(B)(从你的阴谋眼球我8.3uA,因此它是aproximate)。现在,如果你喜欢的结果,而仍然看至五 V(A)= V(B)你可以改变电流值提供您保留通过M1和M2相同的,即电流变化和器件尺寸在一起让你有相同的电流密度的CMOS电流密度晶体管。如果您满意您的新的额定电流值,器件尺寸和比支票VDSAT= VGS-VT晶体管的结果,使确保VDSAT是为更好的匹配结果约200mV的(如果你去接近弱反转,即70-100mv的VDSAT你的镜子错误会更糟,因为VGS不匹配占主导地位)。我知道谷歌翻译是因为它应该是不清晰的,我没有任何其他方式,但你了解它的工作。祝你好运。
 楼主| 发表于 2012-8-10 21:38:26 | 显示全部楼层
回复 25# AcoAco


       谢谢回复。
   1.    您说您根据14楼的仿真图得到8.3uA,不过21楼您说的是0.8uA,不过我大致理解了,这个电流值不是最重要的,重要的是要根据电流值保证N1和N2都工作在饱和区,不知道我的理解是否正确?
   2.    还有一个问题我很茫然,因为电流镜工作在饱和区意味着要满足“Vds≥Vgs-Vth”,由于N1的栅极和漏极接在一起,所以我想N1一定工作在饱和区的,那么是不是只要保证N2工作在饱和区就能保证电流镜的电流拷贝作用?
   3.    这个地方考虑电流密度的目的是什么?N2对N1能够按比例拷贝电流不就行了么?
发表于 2012-8-11 16:09:27 | 显示全部楼层
回复 26# austxuhui
Answer to A: It is a typo.  Yes you understand it right, For a desired nominal current (and for the desired current range around it) we want current mirror transistors are operating in saturation region. N1 and N2 have a relatively large ~200mV VDsat and V(A)=V(B). Since transistors Gate and sources are common making drain voltages equal sets the both transistors operating conditions being exactly equal so current will copy exactly (within the practical limits due to mismatches etc). Now making V(A) = V(B) is hard in practice since V(B) need to be forced by extra circuits. Goal is making these voltages as close as practical so VDS modulation error will be small.

Answer to 2: You are right. N2 needs to be in saturation. N1 is forced in saturation as you said sice gate and drain shorted but N2's operation mode is forced through N1. If N1 is not sized right to begin with N2 will not be in saturation. If N1 and N2 are the same size, N2 is in saturation and V(A)=V(B) than N2 copies the current of N1.

Answer to 3: This is for any individual transistor  I =K(W/L)(Vgs-Vt)*(Vgs-Vt)= K(W/L)*VDsat*VDsat  So chosing Vdsat~200mV (I/(W/L))= current density =K(VDsat*VDsat)=constant. Therefore for a desired current there is a corresponding W/L value for a given process. For a mirror with equal size transistors both N1 and N2 will have the same current density. Good luck.


回答:这是一个错字。是的,你明白它的权利,对于所需的额定电流(以及围绕它所需电流范围),我们希望电流镜晶体管在饱和区。 N1和N2有一个比较大的~200mV的VDSAT和 V(A)= V(B)。由于晶体管栅极和源是常见的漏电压等于设置的两个晶体管的工作条件是完全平等的,所以电流将完全复制(内由于不匹配等实际限制)。现在使 V(A)= v(B)是在实践中很难的,因为V(b)需要额外的电路被强迫。目标是使这些电压,如实际接近VDS的调制误差会小。

回答2:你说得对。 N2的需要是处于饱和状态。 N1是被迫处于饱和状态,如你所说SICE门和漏极短路,但氮气的运作模式是通过N1的被迫。如果大小的权利开始与N2 N1是不是会不饱和。如果N1和N2是相同的大小,N2饱和度和 V(A)= V(B)比氮气复制N1的电流。

回答3:这是任何单个的晶体管是我= K(下的W / L)(VGS-VT)*(VGS-VT)= K(下的W / L)* VDSAT* VDSAT所以VDSAT艇员选拔~200mV的(我/(宽/ L),)=电流密度= K(下VDSAT* VDSAT)=常数。因此,所需的电流有一个相应的W / L值给定的进程。对于与同等大小的晶体管的一面镜子,N1和N2都将有相同的电流密度。祝你好运。
发表于 2012-8-13 11:44:43 | 显示全部楼层
如果 mos current mirror 要準,  L 變大下
你要達到同電流大小下 W 就是得變大 .
 楼主| 发表于 2012-8-15 08:39:37 | 显示全部楼层
回复 27# AcoAco


    有事最近几天没上论坛,很抱歉,非常感谢您的回复。此贴的问题您给的帮助非常有用,再次感谢您。
 楼主| 发表于 2012-8-15 08:42:10 | 显示全部楼层
回复 28# peterlin2010


       感谢回复。   
       L和W都要大点,这是电流镜的一般要求,不知道我理解您的意思对不对?
您需要登录后才可以回帖 登录 | 注册

本版积分规则

关闭

站长推荐 上一条 /2 下一条

×

小黑屋| 关于我们| 联系我们| 在线咨询| 隐私声明| EETOP 创芯网
( 京ICP备:10050787号 京公网安备:11010502037710 )

GMT+8, 2024-5-27 18:13 , Processed in 0.027956 second(s), 6 queries , Gzip On, Redis On.

eetop公众号 创芯大讲堂 创芯人才网
快速回复 返回顶部 返回列表