Transcript Presentation Title Here

Characterizing Bias Current Spikes
Device Under Test
Input
Current
Wideband TIA
A wideband transimpedance amplifier was used to directly view the input
current of several chopper amplifiers
TI Information – Selective Disclosure
1
Characterizing Bias Current Spikes
VOUT   RF I B (U 1)  I B (U 2)  I N (U 1)  I N (U 2)  I Spike 
Bias Current (IB)
Current Noise (IN)
OPA657
OPA2188
2 pA
160 pA
1.3 fA/rtHz
7 fA/rtHz
Bias current spikes should be apparent above other noise sources
TI Information – Selective Disclosure
2
Input Bias Spike Measurement Equipment
• A shielded enclosure was used
to mitigate extrinsic noise
• Direct connection to the
oscilloscope via coax cable
Spikes were viewed on a 500
MHz oscilloscope
– 50 Ohm input impedance
– DC coupling (maximize
bandwidth)
– Averaging used to remove
random noise
TI Information – Selective Disclosure
3
OPA2188
• Spikes repeat at 2x the
chopper clock frequency
• Larger spike is due to the
input commutation
• Smaller spike is from the
synchronous notch filter on
the output
• Largest spike peaks at 850nA
• Total duration is ~24nS
TI Information – Selective Disclosure
4
OPA2333
Commutation frequency is
much lower
– Input clock spikes are now
of similar magnitude to
notch filter spikes
Different input topology
– Transmission gates reduce
input current spikes
– 70 nA peak
– 216 nS duration
TI Information – Selective Disclosure
5
AD8639
1.7 uA!
This is not unique to TI auto-zero topologies!
TI Information – Selective Disclosure
6
Noise Feed-Through
Without proper design considerations noise from the input current spikes
can appear in the output
TI Information – Selective Disclosure
7
Equivalent Schematic
Input current spikes can
be viewed as current
sources on the inputs
Input current spikes are
outside the opamp
bandwidth
– The opamp can be
removed to simplify
analysis
– Current spikes on the
non-inverting input are
not amplified
TI Information – Selective Disclosure
8
Contribution to Output Noise
Current spikes on the inverting input are coupled to the load by the feedback network
V1  IG 2 ( RG || RF  RLOAD )
VOUT  V1
RLOAD
RF  RLOAD
Output noise is dependant upon:
– Input current spike magnitude
– Feedback network impedance (RF and RG)
– Load Impedance (RLOAD)
TI Information – Selective Disclosure
9
Noise Measurement
• Output noise was amplified and
viewed on a spectrum analyzer
– OPA2211 with a gain of 11
– HP3588 (10Hz to 150MHz)
• Data collected using LabView
– Shielded enclosure and cables
Noise(V /
Hz )

AV
VRMS
K N  RBW
VRMS: Output voltage
AV: Gain of secondary amplifier
KN: Brickwall correction factor (1.056)
RBW: Resolution Bandwidth
TI Information – Selective Disclosure
10
Gain Effects on Total Noise
At high gains chopper noise is not a dominant noise contributor
TI Information – Selective Disclosure
11
Load Impedance Effects
High load impedances can exacerbate chopper noise from input current spikes
TI Information – Selective Disclosure
12
Feedback Network Impedance
Large feedback resistor values will also worsen the output noise
TI Information – Selective Disclosure
13
Output Filtering
fC 
1
2  ROUT  COUT
Adding an RC output filter can mitigate noise seen by high impedance
loads
– COUT chosen to have an impedance much less than RLOAD at 2x chopping
frequency
– ROUT chosen to maintain opamp stability with the chosen COUT
TI Information – Selective Disclosure
14
Output Filtering
f C ( SPIKE ) 
2  ( ROUT
1
 RF )  COUT
• The corner frequency for the input current spike is actually much lower
– The filter now includes the feedback resistance RF
– Filter corner frequency can be chosen to remove noise without affecting
desired signal
TI Information – Selective Disclosure
15
Output Filtering
fC 
fC ( SPIKE ) 
1
2  ROUT  COUT
2  ( ROUT
TI Information – Selective Disclosure

1
 159kHz
2 *100 *10nF
1
1

 1.576kHz
 RF )  COUT 2  (100  10k) 10nF
16
Output Filtering
• Harmonics due to input current spikes are completely eliminated
• Autozero noise at chopping frequency is within the opamp bandwidth
TI Information – Selective Disclosure
17
Output Filtering
OPA2188 Without Filtering
– Gain: 101, RF:10k, RG:100 Ohm
– Oscilloscope 1MOhm input
impedance is the load
– Input current spikes are visible
above other noise sources
TI Information – Selective Disclosure
OPA2188 With Filtering
– Gain: 101 RF: 10k, RG: 100 Ohm
– Oscilloscope 1MOhm input
impedance is the load
– Triggering oscilloscope becomes
difficult due to low noise
18
Comparison to Non-Autozero Amplifiers
The noise level of a filtered chopper amplifier is on-par with non-chopper topologies
TI Information – Selective Disclosure
19
Minimizing Chopper Noise Effects
• Input current spikes are not amplified by the part
– Spikes on the inverting input will be coupled to the load by the feedback
network
• Minimize feedback resistance values
– Reduces the voltage produced by current spikes
– Standard design practice for low-noise, low-drift circuits
• Load impedance directly contributes to the magnitude of voltage
produced by the spike
• An RC filter is an extremely effective way to reduce output noise
– Corner frequency can be placed outside of the signal bandwidth
– Noise through the feedback network experiences a much greater
attenuation
TI Information – Selective Disclosure
20