ET
发帖数: 10701 | 1 gate voltage noise density:
vn^2=kf/(w*l*cox*f)
drain current noise density (due to gate noise) =
in^2=vn^2*gm^2
gm^2=2*Id*un*cox*(W/L)
so in^2=kf*Id/(L^2*f)
为何我看到的公式是：
i^2/(delta f) = kf*Id/(L^2*cox*f)
谁能给解释解释？ |
m********e
发帖数: 585 | 2 If there's deltaf (which is the bandwidth), it's from the total power.
Others are Power Spectral Density
【在 ET 的大作中提到】
: gate voltage noise density:
: vn^2=kf/(w*l*cox*f)
: drain current noise density (due to gate noise) =
: in^2=vn^2*gm^2
: gm^2=2*Id*un*cox*(W/L)
: so in^2=kf*Id/(L^2*f)
: 为何我看到的公式是：
: i^2/(delta f) = kf*Id/(L^2*cox*f)
: 谁能给解释解释？
|
x****g
发帖数: 2000 | 3 请给出你看到的式子的出处
【在 ET 的大作中提到】
: gate voltage noise density:
: vn^2=kf/(w*l*cox*f)
: drain current noise density (due to gate noise) =
: in^2=vn^2*gm^2
: gm^2=2*Id*un*cox*(W/L)
: so in^2=kf*Id/(L^2*f)
: 为何我看到的公式是：
: i^2/(delta f) = kf*Id/(L^2*cox*f)
: 谁能给解释解释？
|
m*****t
发帖数: 3477 | 4 I have a question that related to this.
The equation I got from K.Martin's book between input and output refereed
noise:
Vno(f)^2 = Vni(f)^2 * A|j.2.pi.f|
Which means the output noise spectre is not only amplified by the gain, but
also shaped by the transform function. In other words, the transistor’s
bandwidth also kicks in at high frequency.
As a result the noise corner is pushed to the far end of frequency axis for
the output-refereed noise (10MHz range) compared to the input-refereed noi
【在 ET 的大作中提到】
: gate voltage noise density:
: vn^2=kf/(w*l*cox*f)
: drain current noise density (due to gate noise) =
: in^2=vn^2*gm^2
: gm^2=2*Id*un*cox*(W/L)
: so in^2=kf*Id/(L^2*f)
: 为何我看到的公式是：
: i^2/(delta f) = kf*Id/(L^2*cox*f)
: 谁能给解释解释？
|
ET
发帖数: 10701 | 5 这2公式都是power spectral density..
最大的区别是我认为cox应该被消掉了。。
【在 m********e 的大作中提到】
: If there's deltaf (which is the bandwidth), it's from the total power.
: Others are Power Spectral Density
|
ET
发帖数: 10701 | 6 got it..
bsim 4 model..nnd..
【在 ET 的大作中提到】
: gate voltage noise density:
: vn^2=kf/(w*l*cox*f)
: drain current noise density (due to gate noise) =
: in^2=vn^2*gm^2
: gm^2=2*Id*un*cox*(W/L)
: so in^2=kf*Id/(L^2*f)
: 为何我看到的公式是：
: i^2/(delta f) = kf*Id/(L^2*cox*f)
: 谁能给解释解释？
|
ET
发帖数: 10701 | 7 想了半天，终于明白了你这个问题了。
你的意思是说，flicker noise的corner frequncy ，在经过一个single ended amplif
ier (2-pole filter),这个corner frequency 会会被spread out?
这个flicker noise的noise corner是基于noise density.
而经过一个single-ended amplifier后（two-pole), 算出的noise是total noise. 这时
的bandwith是这个filter的bw.
我认为不是一个概念。
but
for
noise
【在 m*****t 的大作中提到】
: I have a question that related to this.
: The equation I got from K.Martin's book between input and output refereed
: noise:
: Vno(f)^2 = Vni(f)^2 * A|j.2.pi.f|
: Which means the output noise spectre is not only amplified by the gain, but
: also shaped by the transform function. In other words, the transistor’s
: bandwidth also kicks in at high frequency.
: As a result the noise corner is pushed to the far end of frequency axis for
: the output-refereed noise (10MHz range) compared to the input-refereed noi
|
m*****t
发帖数: 3477 | 8 to simplify my question:
If you plot the output noise current power spectre, the corner frquency is
much higher than that of the equivalent input noise voltage power spectre.
why is that?
amplif
这时
【在 ET 的大作中提到】
: 想了半天，终于明白了你这个问题了。
: 你的意思是说，flicker noise的corner frequncy ，在经过一个single ended amplif
: ier (2-pole filter),这个corner frequency 会会被spread out?
: 这个flicker noise的noise corner是基于noise density.
: 而经过一个single-ended amplifier后（two-pole), 算出的noise是total noise. 这时
: 的bandwith是这个filter的bw.
: 我认为不是一个概念。
:
: but
: for
|
l****o
发帖数: 184 | 9 如果你指的是mosfet 的corner frequency,那一般来说从几十Hz到几十kHz.但是output
noise frequency是取决于op amp的bandwidth。 我也觉得你概念混淆了
有具体的电路么？
【在 m*****t 的大作中提到】
: to simplify my question:
: If you plot the output noise current power spectre, the corner frquency is
: much higher than that of the equivalent input noise voltage power spectre.
: why is that?
:
: amplif
: 这时
|
m*****t
发帖数: 3477 | 10 Just one individual mosfet.
I believe the 10Hz to 10KHz is the input noise corner.
While when you extract the noise model, you can not measure input noise, you
can only measure output noise current at Drain, and our measurement data
shows it's in the 10MHz to 100MHz range.
output
【在 l****o 的大作中提到】
: 如果你指的是mosfet 的corner frequency,那一般来说从几十Hz到几十kHz.但是output
: noise frequency是取决于op amp的bandwidth。 我也觉得你概念混淆了
: 有具体的电路么？
|
|
|
ET
发帖数: 10701 | 11 output noise current power spectre里的corner frequency是啥？
【在 m*****t 的大作中提到】
: to simplify my question:
: If you plot the output noise current power spectre, the corner frquency is
: much higher than that of the equivalent input noise voltage power spectre.
: why is that?
:
: amplif
: 这时
|
s*****o
发帖数: 22187 | 12 I dont think Flicker corner can reach 10MHz level. Can you print the noise
summary from ADE?
you
【在 m*****t 的大作中提到】
: Just one individual mosfet.
: I believe the 10Hz to 10KHz is the input noise corner.
: While when you extract the noise model, you can not measure input noise, you
: can only measure output noise current at Drain, and our measurement data
: shows it's in the 10MHz to 100MHz range.
:
: output
|
R*****o
发帖数: 204 | 13 我想你是说，在flicker noise的实际测量中，由于测量仪器的limited bandwidth，
观察到的corner frequency比理论上的要高很多。
you
【在 m*****t 的大作中提到】
: Just one individual mosfet.
: I believe the 10Hz to 10KHz is the input noise corner.
: While when you extract the noise model, you can not measure input noise, you
: can only measure output noise current at Drain, and our measurement data
: shows it's in the 10MHz to 100MHz range.
:
: output
|
s*****o
发帖数: 22187 | 14 Never mind I thought it's simulation.
【在 s*****o 的大作中提到】
: I dont think Flicker corner can reach 10MHz level. Can you print the noise
: summary from ADE?
:
: you
|
ET
发帖数: 10701 | 15 不是。。你这corner frequency是怎么定义的？
fcorner 是flicker noise density = thermal noise density那点
的frequency ..
所以，理论计算上只能是一个点。 怎么会从10hz to 100khz?
you
【在 m*****t 的大作中提到】
: Just one individual mosfet.
: I believe the 10Hz to 10KHz is the input noise corner.
: While when you extract the noise model, you can not measure input noise, you
: can only measure output noise current at Drain, and our measurement data
: shows it's in the 10MHz to 100MHz range.
:
: output
|
m*****t
发帖数: 3477 | 16 That's the point.
One student from Berkely simulate output noise using our model and complain
the "noise corner frquency"(where the slop turn to flat) is too high.
I believe he mistaked the input noise corner (KHz) with the output noise
spectrum.
He/She should simulate input refeerd noise, where you set Vd as noise source
and prob the noise at G. I want to confirm that.
【在 ET 的大作中提到】
: output noise current power spectre里的corner frequency是啥？
|
R*****o
发帖数: 204 | 17 我理解他说的应该是fc所在的频率范围吧
【在 ET 的大作中提到】
: 不是。。你这corner frequency是怎么定义的？
: fcorner 是flicker noise density = thermal noise density那点
: 的frequency ..
: 所以，理论计算上只能是一个点。 怎么会从10hz to 100khz?
:
: you
|
m*****t
发帖数: 3477 | 18 It's a range, different technologies will give you different corner frquency
.
【在 ET 的大作中提到】
: 不是。。你这corner frequency是怎么定义的？
: fcorner 是flicker noise density = thermal noise density那点
: 的frequency ..
: 所以，理论计算上只能是一个点。 怎么会从10hz to 100khz?
:
: you
|
s*****o
发帖数: 22187 | 19 Fc is same no matter you calculate input referred or output referred, isn't
it?
complain
source
【在 m*****t 的大作中提到】
: That's the point.
: One student from Berkely simulate output noise using our model and complain
: the "noise corner frquency"(where the slop turn to flat) is too high.
: I believe he mistaked the input noise corner (KHz) with the output noise
: spectrum.
: He/She should simulate input refeerd noise, where you set Vd as noise source
: and prob the noise at G. I want to confirm that.
|
m*****t
发帖数: 3477 | 20 So you believe when you plot the output noise spectrum, you will get the
frequency dependent part meet the noise floor around KHz?
t
【在 s*****o 的大作中提到】
: Fc is same no matter you calculate input referred or output referred, isn't
: it?
:
: complain
: source
|
|
|
s*****o
发帖数: 22187 | 21 I guess so. If it's only one mosfet, input and output noise transfer is the
gain. And gain is flat at the freq we talk about. I cannot see anything
cause the fc drift. Please correct me if I am wrong.
【在 m*****t 的大作中提到】
: So you believe when you plot the output noise spectrum, you will get the
: frequency dependent part meet the noise floor around KHz?
:
: t
|
ET
发帖数: 10701 | 22 啥technology node?
现在的65nm 45nm flicker noise corner 上MHz也完全现实的。
input referred noise的唯一物理意义是可比性，无法可测。
都是通过测量output noise density, 折算回去。 在simulation中，
也不要设定noise souce, 因为transistor本身就是个noise source, 要measure的就是
ouput noise density..
complain
source
【在 m*****t 的大作中提到】
: That's the point.
: One student from Berkely simulate output noise using our model and complain
: the "noise corner frquency"(where the slop turn to flat) is too high.
: I believe he mistaked the input noise corner (KHz) with the output noise
: spectrum.
: He/She should simulate input refeerd noise, where you set Vd as noise source
: and prob the noise at G. I want to confirm that.
|
ET
发帖数: 10701 | 23 是这个意思。
the
【在 s*****o 的大作中提到】
: I guess so. If it's only one mosfet, input and output noise transfer is the
: gain. And gain is flat at the freq we talk about. I cannot see anything
: cause the fc drift. Please correct me if I am wrong.
|
s*****o
发帖数: 22187 | 24 I second this.
【在 ET 的大作中提到】
: 啥technology node?
: 现在的65nm 45nm flicker noise corner 上MHz也完全现实的。
: input referred noise的唯一物理意义是可比性，无法可测。
: 都是通过测量output noise density, 折算回去。 在simulation中，
: 也不要设定noise souce, 因为transistor本身就是个noise source, 要measure的就是
: ouput noise density..
:
: complain
: source
|
ET
发帖数: 10701 | 25 昨晚simulted a 45nm nmosfet using predictive spice model.
attached here. the spectral density also included a 10K resistor's noise
performance tho.
the flicker noise corner frequency is in the range of ~Mhz..
【在 s*****o 的大作中提到】
: I second this.
|
ET
发帖数: 10701 | 26 显示不出来？
【在 ET 的大作中提到】
: 昨晚simulted a 45nm nmosfet using predictive spice model.
: attached here. the spectral density also included a 10K resistor's noise
: performance tho.
: the flicker noise corner frequency is in the range of ~Mhz..
|
s*****o
发帖数: 22187 | 27 I guess you are right. I only thought about the big size case for limiting
flicker noise.
【在 ET 的大作中提到】
: 显示不出来？
|
x****g
发帖数: 2000 | 28 就应该是这个量级没错的
【在 ET 的大作中提到】
: 昨晚simulted a 45nm nmosfet using predictive spice model.
: attached here. the spectral density also included a 10K resistor's noise
: performance tho.
: the flicker noise corner frequency is in the range of ~Mhz..
|
R*****o
发帖数: 204 | 29 schematic and results please.
你的图挂了，貌似
【在 ET 的大作中提到】
: 昨晚simulted a 45nm nmosfet using predictive spice model.
: attached here. the spectral density also included a 10K resistor's noise
: performance tho.
: the flicker noise corner frequency is in the range of ~Mhz..
|
ET
发帖数: 10701 | 30 为啥贴不上呢？
【在 R*****o 的大作中提到】
: schematic and results please.
: 你的图挂了，貌似
|
|
|
s*****o
发帖数: 22187 | 31 65nm: W/L=1u/1u, Fc=3M
【在 ET 的大作中提到】
: 为啥贴不上呢？
|
ET
发帖数: 10701 | 32 90nm: 1/90n
45nm : 1/45n
工具比我的强大呀.. 我只能眼睛看了..
【在 s*****o 的大作中提到】
: 65nm: W/L=1u/1u, Fc=3M
|
s*****o
发帖数: 22187 | 33 赞眼神:D，不用dB也能看出来？
【在 ET 的大作中提到】
: 90nm: 1/90n
: 45nm : 1/45n
: 工具比我的强大呀.. 我只能眼睛看了..
|
ET
发帖数: 10701 | 34 那不至于。。我不需要那个数字。
我用的ltspice应该也能测出来，需要点功夫吧。
【在 s*****o 的大作中提到】
: 赞眼神:D，不用dB也能看出来？
|
ET
发帖数: 10701 | 35 你这model是real？
【在 s*****o 的大作中提到】
: 赞眼神:D，不用dB也能看出来？
|
s*****o
发帖数: 22187 | 36 tsmc
【在 ET 的大作中提到】
: 你这model是real？
|
ET
发帖数: 10701 | 37 mosis?
【在 s*****o 的大作中提到】
: tsmc
|
H***F
发帖数: 2501 | 38 input noise fc仿真的和output noise fc不一样啊。你再仿真一下？
65nm: W/L=1u/1u, Fc=3M
【在 s*****o 的大作中提到】
: 65nm: W/L=1u/1u, Fc=3M
|
s*****o
发帖数: 22187 | 39 想请教一下，如何仿真input-referred noise呢？以CS amp with R load为例，谢谢。
【在 H***F 的大作中提到】
: input noise fc仿真的和output noise fc不一样啊。你再仿真一下？
:
: 65nm: W/L=1u/1u, Fc=3M
|
H***F
发帖数: 2501 | 40 你用cadence工具不？直接选input noise不就行了？
想请教一下，如何仿真input-referred noise呢？以CS amp with R load为例，谢谢。
【在 s*****o 的大作中提到】
: 想请教一下，如何仿真input-referred noise呢？以CS amp with R load为例，谢谢。
|
|
|
s*****o
发帖数: 22187 | 41 没有input source noise啊? 如果是同样的电路，你仿真的input referred noise啥样
子呢？
【在 H***F 的大作中提到】
: 你用cadence工具不？直接选input noise不就行了？
:
: 想请教一下，如何仿真input-referred noise呢？以CS amp with R load为例，谢谢。
|
ET
发帖数: 10701 | 42 你不是都说了吗?
output noise density/(intrinsic gain)
以你的setup的话，这个65nm的technology, gmR <10
所以input refered noise比output noise density曲线相似，小40db.
【在 s*****o 的大作中提到】
: 没有input source noise啊? 如果是同样的电路，你仿真的input referred noise啥样
: 子呢？
|
s*****o
发帖数: 22187 | 43 Yes, I think so. But HighF said Fc is different for input and output noise.
So I want to learn how to simulate input referred Fc from him.
【在 ET 的大作中提到】
: 你不是都说了吗?
: output noise density/(intrinsic gain)
: 以你的setup的话，这个65nm的technology, gmR <10
: 所以input refered noise比output noise density曲线相似，小40db.
|
ET
发帖数: 10701 | 44 spice simulator会有个ｉｎｎｏｉｓｅ （ｌｔｓｐｉｃｅ)
ｓｐｅｃｔｒｅ应该也有类似的吧。
从ｉｎｏｉｓｅ曲线上和ｏｕｔｐｕｔ ｎｏｉｓｅ曲线一样。
自然能看到一个flicker noise corner..
.
【在 s*****o 的大作中提到】
: Yes, I think so. But HighF said Fc is different for input and output noise.
: So I want to learn how to simulate input referred Fc from him.
|
