low noise Op Amp Design, need help

[tia_design]

low noise cmos op amp design

You guys,

I'm designing a low noise CMOS Op Amp, now I can reduce thermal noise (input referred) down to 1.5nV/rhHz, this is schematic simulation result. I don't know how much additional noise will be added after layout. Does anybody do the simular research so we can discuss it?

 

[alvays]

the additional is the disturb, not noise.

 

[mitgrace]

Excuse me . How do you run the input referred noise ? Which tools do u use it ? I am looikign for this problems , Thanks

 

[xuel]

I think your chip with a carefully designed layout will be a litter better than the worst-case postlayout simulation results .

 

[snoop835]

Hi mitgrace!

The simulate for input referred noise, you will need to run the simulation in parallel with ac analysis. The syntax are (for HSPICE);

.AC DEC 100 1 10G *****Sweep frequency from 1Hz to 10GHz*********
.NOISE (VOUT) VIN
.PRINT NOISE INOISE ONOISE
***INOISE=Input Referred Noise*****ONOISE=Output Referred Noise*********

see HSPICE manual for details.

The output waveform should be X-axis = Frequency
Y-axis = nV/sqrt(Hz)

GOOD LUCK!!

 

[jake]

Hi mitgrace!

The simulate for input referred noise, you will need to run the simulation in parallel with ac analysis. The syntax are (for HSPICE);

.AC DEC 100 1 10G *****Sweep frequency from 1Hz to 10GHz*********
.NOISE (VOUT) VIN
.PRINT NOISE INOISE ONOISE
***INOISE=Input Referred Noise*****ONOISE=Output Referred Noise*********

see HSPICE manual for details.

The output waveform should be X-axis = Frequency
Y-axis = nV/sqrt(Hz)

GOOD LUCK!!

a bandgap, for example, how to add the ac source?

 

[Vamsi Mocherla]

Dear Jake,

You can put an ac source on the supplies.

Dear tia_design,

One way to reduce your thermal noise is to increase the bias current in the circuit.

 

[jake]

Dear Jake,

You can put an ac source on the supplies.

Dear tia_design,

One way to reduce your thermal noise is to increase the bias current in the circuit.

Thanks, I do so.
I encounter a big problem about the bandgap noise. Attachment is my circuit. Who can help me to analysie how to reduce the noise(the size of the diff transistor is already big )?

 

[Vamsi Mocherla]

Dear jake,

I was looking into your circuit and i am noticing that you have very less current in the output branch. It is about 0.1u A. Is it right? I mean, that will cause a lot of thermal noise at the output. And another thing. The W/L aspect ratio of the input differential pair on the output amp is comparitively low. This will affect the relative noise in terms of gm. I would suggest that the gm of the differential pair be higher than the current mirrors for better noise performance. Finally, the lengths of the devices are a bit high, why dont you reduce them a little bit. It will give rise to a lot of unwanted parasitics...... I hope that I am clear on this.

 

[jake]

Dear jake,

I was looking into your circuit and i am noticing that you have very less current in the output branch. It is about 0.1u A. Is it right? I mean, that will cause a lot of thermal noise at the output. And another thing. The W/L aspect ratio of the input differential pair on the output amp is comparitively low. This will affect the relative noise in terms of gm. I would suggest that the gm of the differential pair be higher than the current mirrors for better noise performance. Finally, the lengths of the devices are a bit high, why dont you reduce them a little bit. It will give rise to a lot of unwanted parasitics...... I hope that I am clear on this.

THank you very much, I am sorry that I have mislabelled the output branch current, and I have improved it into 10uA now. I think that reducing the lengths of the devices is a good suggestion. I will try to do it now, and I will tall you the results.

 

[jake]

Dear jake,

I was looking into your circuit and i am noticing that you have very less current in the output branch. It is about 0.1u A. Is it right? I mean, that will cause a lot of thermal noise at the output. And another thing. The W/L aspect ratio of the input differential pair on the output amp is comparitively low. This will affect the relative noise in terms of gm. I would suggest that the gm of the differential pair be higher than the current mirrors for better noise performance. Finally, the lengths of the devices are a bit high, why dont you reduce them a little bit. It will give rise to a lot of unwanted parasitics...... I hope that I am clear on this.

Dear Vamsi Mocherla,
I have tried to reduce the lengths of the devices, they contribute little for the noise. Some one tell me that the op amplify the noise, however I reduce the gain of the OP, it can't reduce the noise. in fact, it increase the noise. Do you have another method to reduce the noise? Maybe, changing a new structure can reduce the noise. Attachment is my new structure, however, the output voltage vary with temperature seriously, why?