1. Noise Experiments
Ray
Oct 13th, 2015

2. Preamp Analysis
frequency domain
• Procedure:
1. Set its gain to 500V/V, input to gnd
2. Change the cutoff frequency of low-pass filter in preamp to
see its output spectrum
• Observations:
1. Independent of the cutoff frequency, the high frequency thermal
noise is always around -81dBVrms
2. When the cutoff frequency is between 1KHz to 10KHz, there are
two plateaus with the second one as -81dBVrms

3. Preamp Analysis
frequency domain cont.
• Conclusions:
1. Since the thermal noise in the first plateau is from the preamp input while the
second is from preamp itself, we need to adjust the cutoff frequency to make the
first plateau appear.
2. Also, we need the first plateau to be as long as possible to get rid of the roll-off
effect from either flicker noise and preamp cutoff frequency.
1/f noise
thermal noise
due to preamp input
thermal noise
due to preamp
fL

4. Noise Analysis
frequency domain
• Measurement Setup:
A
To Spectrum
Analyzer
Preamp
1Mohm
VGS

5. Noise Analysis
frequency domain
RD
VGS
Weak Inversion
Condition

6. Noise Analysis
frequency domain
• Setup detail:
1. No pole introduced in preamp
2. Device works in room temperature 73F
3. VGS=0.4V
Pole (detector
induced*)

7. Noise Analysis
Integration Method
Pole in preamp:
10Hz
1/f noise dominant

8. Noise Analysis
Integration Method

9. Noise Analysis
Integration Method

10. Noise Analysis
Integration Method
Pole in preamp:
10KHz
thermal noise dominant
Rload=1Kohm

11. Noise Analysis
Integration Method

12. Noise Analysis
Integration Method

13. Noise Analysis
Integration Method
Region 1
Region 2

14. Noise Analysis
Integration Method
• Region 1:
1. noise highly depends on the temperature
2. This temperature dependence is from the channel noise
• Region 2: noise is almost independent of the
temperature
This happens because RD sits outside oven.
• No matter what is the reason, measuring temperature from 73F to
83F can tell us the channel noise dependence on the temperature

15. Noise Analysis vs Temperature
frequency domain T=73F(296K)
• After averaging using two frequencies and five
resetting, -55.023 dBVrms noise power
spectrum density is derived and it equals to
3.146uV/sqrt(Hz).
• Since preamp gain is 500V/V, the equivalent
noise spectrum density from the detector is
3.146u/sqrt(500), which is 0.1407uV/sqrt(Hz)
fL = 56.5KHz

16. Noise Analysis vs Temperature
frequency domain T=83F(301.5K)
• After averaging using two frequencies and five
resetting, -53.872 dBVrms noise power
spectrum density is derived and it equals to
4.1uV/sqrt(Hz).
• Since preamp gain is 500V/V, the equivalent
noise spectrum density from the detector is
4.1u/sqrt(500), which is 0.18336uV/sqrt(Hz)
fL = 47.616KHz

17. Noise Analysis vs Temperature
frequency domain T=93F(307K)
• After averaging using two frequencies and five
resetting, -53.792 dBVrms noise power
spectrum density is derived and it equals to
4.1764uV/sqrt(Hz).
• Since preamp gain is 500V/V, the equivalent
noise spectrum density from the detector is
4.1u/sqrt(500), which is 0.18677uV/sqrt(Hz)
fL = 45.4KHz
Noise actually drops in this temperature

18. Noise Analysis vs Temperature
frequency domain theoretic result
Channel Noise
Loading resistance noise
Frequency domain experiment result:

19. Noise Analysis vs Temperature
frequency domain theoretic result
Channel Noise Weak inversion approximation
Loading resistance noise
Flicker Noise
Channel Noise strong inversion approximation
Frequency domain experiment result: