Biomedical

Instrumentation

Noise Reduction
Noise & artifact in biomedical
signals
Noise in Biomedical
Measurement Systems
 Biomedical measurements involve low-level signals which are
susceptible to noise interference.

 We have 3 common sources of noise in biomedical

 Improper grounding
 Electrostatic Coupling to AC signals
 Electromagnetic coupling to AC signals
Noise from Electronics
Components
 Every electronic component has noise
• thermal noise
• shot noise
• distribution noise (or partition noise)
Thermal Noise
 Thermal noise due to agitation of
electrons
 Present in all electronic devices and
transmission media
 Cannot be completely eliminated
 Function of temperature
Thermal noise
 thermal noise is caused by the thermal
motion of the charge carriers; as a result
the random electromotive force appears
between the ends of resistor;
Johnson Noise, or Thermal Noise, or
Thermal Agitation Noise
 Also referred to as white noise because of
gaussian spectral density.

2
 Vn  4kTRB where
• Vn = noise Voltage (V)
• k = Boltzman’s constant
• Boltzman’s constant = 1.38 x 10 -23Joules/Kelvin

• T = temperature in Kelvin
• R = resistance in ohms (Ώ)
• B = Bandwidth in Hertz (Hz)
Eg. of Thermal Noise
• Given R = 1Kohm
• Given B = 2 KHz to 3 KHz = 1 KHz
• Assume: T = 290K (room Temperature)
• Vn2 = 4KTRB units V2
• Vn2= (4) (1.38 x 10 –23J/K) (290K) (1 Kohm)
(1KHz)
• = 1.6 x 10-14 V2
•Vn = 1.26 x10
–7 V = 0.126 uV
Shot noise
 Shot noise appears because the current
through the electron tube (diode, triode
etc.) consists of the separate pulses
caused by the discontinuous electrons;
Shot Noise
 Shot Noise: noise from DC current flowing
in any conductor

2
 In  2qIB where I n  2qIB
• In = noise current (amps)
• q = elementary electric charge
• = 1.6 x 10-19 Coulombs
• I = Current (amp)
• B = Bandwidth in Hertz (Hz)
Eg: Shot Noise
 Given I = 10 mA
 Given B = 100 Hz to 1200 Hz = 1100 Hz
 In2= 2q I B =
= 2 (1.6 x 10 –19Coulomb) ( 10 X10 –3A)(1100 Hz)
= 3.52 x10 –18 A2
In = (3.52 x10–18 A2) ½ = 1.88 nA
 Noise cont
 Flicker Noise also known as Pink Noise or 1/f noise
is the lower frequency < 1000Hz phenomenon and
is due to manufacturing defects
• A wide class of electronic devices demonstrate so
called flicker effect, its intensity depends on
frequency as 1/f, ~1, in the wide band of
frequencies;
• For example, flicker effect in the electron tubes is
caused by the electron emission from some
separate spots of the cathode surface, these spots
slowly vary in time; at the frequencies of about 1
kHz the level of this noise can be some orders
higher then thermal noise.
Distribution noise
 Distribution noise (or partition noise)
appears in the multi-electrode devices
because the distribution of the charge
carriers between the electrodes bear
the statistical features;
Signal to Noise Ratio = SNR

 SNR = Signal/ Noise

 Noise Power Pn
• Pn = kTB, where
•Pn =noise power in watts
•k = Boltzman’s constant
• Boltzman’s constant = 1.38 x 10 -23Joules/Kelvin
• T = temperature in Kelvin
• B = Bandwidth in Hertz (Hz)
 Internal and External Noise

 Internal Noise
 External Noise
 Total Noise Calculation
Internal Noise
 Internal Noise: Caused by thermal
currents in semiconductor material
resistances and is the difference between
output noise level and input noise level
External Noise
 External Noise: Noise produced by
signal sources also called source noise;
cause by thermal agitation currents in
signal source
External Noise
 Total Noise Calculation = square root of
sum of squares Vne = (Vn2+(InRs)2) ½
necessary because otherwise positive
and negative noise would cancel and
mathematically show less noise that what
is actually present
Noise Factor
 Noise Factor = ratio of noise from real
resistance to thermal noise of an ideal
resistor
 Noise Factor
 Fn = Pno/Pni evaluated at T = 290oK (room
temperature) where
• Pno = noise power output and
• Pni = noise power input
 Noise Factor
 Pni =kTBG where
• G = Gain;
• T = Standard Room temperature = 290oK
• K = Boltzmann’s Constant = 1.38 x10-23J/oK
• B = Bandwidth (Hz)
 Noise Factor
 Pno = kTBG + ΔN where
• ΔN = noise added to system by network or
amplifier

Fn 
kTBG  N 

N
kTBG kTBG
Noise Figure
 Noise Figure : Measure of how close
is an amplifier to an ideal amplifier
 NF = 10 log (Fn) where
• NF = Noise Figure (dB)
• Fn = noise factor (previous slide)
Noise Figure
 Friis Noise Equation: Use when you have a
cascade of amplifiers where the signal and
noise are amplified at each stage and each
component introduces its own noise.
• Use Friis Noise Equation to calculated total Noise
F2  1 F3  1 Fn  1
FN  F1    ... 
G1 G1G2 G1G2 ...Gn 1
• Where FN = total noise
• Fn = noise factor at stage n ;
• G(n-1) = Gain at stage n-1
 Example: Given a 2 stage amplifier where A1
has a gain of 10 and a noise factor of 12 and
A2 has a gain of 5 and a noise factor of 6.

FN  12 
6  1
 12.5
10
Noise Reduction Strategies
1. Keep source resistance and amplifier input
resistance optimum (High resistance with
increase thermal noise)
2. Keep Bandwidth at a minimum but make
sure you satisfy Nyquist’s Sampling Theory
3. Prevent external noise with proper ground,
shielding, filtering
4. Use low noise at input stage (Friis Equation)
5. For some semiconductor circuits use the
lowest DC power supply
 Feedback Control Derivation
Vo  G1  E
Vin + E Vo E  Vin  Vo
Σ G1
+
Vo  G1Vin  Vo 
β
Vo  G1Vin  G1Vo
Vo  G1Vo  G1Vin
Vo 1  G1   G1Vin
Vo G1

Vin 1  G1 
 Use of Feedback to reduce Noise

Vn = Noise
Vin + V1 V1G1 + V2 V2G2 Vo
Σ G1 Σ G2
+
B Vo
Β
V 1  Vin  Vo
V 2  V 1G1  Vn
V 2  Vin  Vo G1  Vn
Vo  V 2G 2
 Use of Feedback to reduce Noise

Vn = Noise
Vin + V1 V1G1 + V2 V2G2 Vo
Σ G1 Σ G2
+
B Vo
Β

Vo  Vin  Vo G1  Vn G 2

Vn = Noise
Vin + V1 V1G1 + V2 Vo
Σ V2G2
G1 Σ G2
+
B Vo
Β
Vo 
G1G 2Vin  G 2Vn 
Thus Vn is reduced by Gain G1
1  G1G 2  
G1G 2Vin G1  G 2Vn 
Vo  
1  G1G 2   G1  1  G1G 2  
G1G 2  Vn 
Vo  Vin 
1  G1G 2    G1 
 Noise Reduction by Signal Averaging
 Un processed SNR Sn =20 log (Vin/Vn)
 Processed SNR Ave Sn = 20 log (Vin/Vn/ N1/2)
• Where
• SNR Sn = unprocessed SNR
• SNR Ave Sn = time averaged SNR
• N = # repetitions of signals
• Vin = Voltage of Signal
• Vn = Voltage of Noise
 Processing Gain = Ave Sn – Sn in dB
 Noise Reduction by Signal Averaging
 Ex: EEG signal of 5 uV with 100 uV of
random noise
• Find the unprocessed SNR, processed SNR
with 1000 repetitions and the processing Gain
 Noise Reduction by Signal Averaging
 Unprocessed SNR
• Sn = 20 log (Vin/Vn) = 20 log (5uV/100uV) = -26dB
 Processing SNR
• Ave Sn = 20 log (Vin/Vn/N1/2)
= 20 log (5u/100u / (1000)1/2) = 4 dB
 Processing gain = 4 – (- 26) = 30 dB