De-Integrating Integrated Circuit Preamps

Transcription

De-Integrating
Integrated Circuit Preamps
Less Delivers More?
Les Tyler, President, THAT Corporation
1
De-integrating Integrated Circuit Preamps
Rev 1.4
Copyright © 2014, THAT Corporation
Today’s Topics
• Where we’ve been, and where we’re
going
– The old to the “new” topology
• Optimizing this new topology
– Noise
– Bandwidth
– Gain control, taper, & dc offset
– Output configuration
• Wrap-up
2
Rev 1.4
Out with the Old
… In with the New
Some historical perspective…
3
Rev 1.4
In the Beginning (for IC Preamps)
• Solid State Music
– Ron Dow
– Dan Parks
RA
• SSM 2011 (~1982)
– Integrated preamp
– External feedback
R9
R6
IN+
IN-
RG
IN1
RG1
OUT1
RG2 OUT2
IN2
OUT
+
R7
–
R8
• SSM2015 (~1983)
RB
– Integrated preamp
– One external
resistor controls
gain: RG
IC Preamp
4
Rev 1.4
“Standard” IC Preamp Configuration
• Differential input
• Single-ended
output
• Current feedback
• Single resistor
controls gain
RA
R9
R6
IN+
IN-
RG
IN1
RG1
OUT1
RG2 OUT2
IN2
OUT
+
R7
–
R8
RB
– RG
IC Preamp
• Minimum gain: 0dB
– Requires infinite RG
5
Rev 1.4
Many Followed In SSM’s Footsteps
• SSM2016 (1986)
– Derek Bowers’ design
RA
• SSM2017 (1989)
– By this time, SSM was part of
Analog Devices
• SSM2019 (2003)
– Derek Bowers (at ADI)
• TI INA163 (~2000)
– INA217 (2002)
R9
R6
IN+
IN-
• THAT 1510 (2004)
• THAT1512 (2004)
RG
IN1
RG1
OUT1
RG2 OUT2
IN2
OUT
+
R7
–
R8
RB
– External RG
– Diff-amp gain: -6dB
– Credit: Cal Perkins (at Mackie)
IC Preamp
• Not considered: AD524 (1982)
– Also included internal RA & RB, as
well as choice of (internal) RG
– Scott Wurcer’s design
– Noise too high for a mic preamp
(~5nV/√Hz)
6
Rev 1.4
Benefits of “Standard” Topology
• Wide bandwidth at high gain due to
current, not voltage feedback
• Can be very quiet at high gains
– Many reach 1nV/√Hz voltage noise
• Easy to control gain with single RG
• Integrated approach allows wide
input dynamic range
– See G. Hebert’s presentation at the 2010
US AES convention
7
Rev 1.4
Detriments of “Standard”
Topology
• Feedback network adds noise at low gains
– Resistor self-noise
– Current noise in RG pins drawn across the
feedback network’s impedance
• Maximum gain is affected by pot’s
effective end-resistance
• Smooth taper is hard to achieve
– Depends on RG vs. RA & RB
• Must convert output from single-ended to
differential to drive A/D converters
8
Rev 1.4
Let’s De-Integrate the Topology
• Start with the
“standard” IC
RA
R9
R6
IN+
IN-
RG
IN1
RG1
OUT1
RG2 OUT2
IN2
OUT
+
R7
–
R8
RB
IC Preamp
• Remove the
output diff amp
RA
IN+
IN-
RG
OUT+
IN1
RG1
OUT1
RG2 OUT2
IN2
OUT-
RB
IC Preamp
9
Rev 1.4
Let’s De-Integrate the Topology
• Remove
internal RA & RB
IN+
IN-
RG
OUT+
IN1
RG1
OUT1
RG2 OUT2
IN2
OUT-
IC Preamp
• Take away the
external RG
IN+
OUT+
IN1
RG1
IN-
OUT1
RG2 OUT2
IN2
OUT-
IC Preamp
10
Rev 1.4
What Does That Leave?
• Uncommitted
– Completely configurable
• Differential In
• Differential Out
– 0dB common-mode
gain
IN+
OUT+
IN1
RG1
IN-
• Current Feedback
• Low Voltage Noise
• THAT’s the 1570 &1583
11
OUT1
RG2 OUT2
IN2
OUT-
IC Preamp
Rev 1.4
The 1570/1583 “Uncommitted”
Topology Makes Optimization Easy
• Noise vs. gain
• Bandwidth vs. gain
• Pot end resistance
• Gain vs. pot rotation
• Output amp performance
– Noise
– CMRR
12
Rev 1.4
Optimization Details
Noise
Bandwidth
Gain control, taper & DC offset
Output & Common-Mode Rejection
13
Rev 1.4
Noise Model for
1570/1583-Type Preamp
• Voltage (en) & Current (in)
noise flows in each input
pin
• en is amplified by gain
RA
IN+
– 0dB at minimum gain
– +60dB at 60dB gain
i n IN+
e n IN+
i n RG1
e n RG1
RG1
RG
RG2
• in creates a noise voltage
based on the impedance
it flows through
OUT+
IN1
i n RG2
e n RG2
IN2
i n IN-
e n IN-
RB
OUT1
OUT2
OUT-
IN-
– Then amplified by gain
• Sources are uncorrelated
– Add in RMS fashion (root
of the sum of the squares)
14
Rev 1.4
Noise Model for
1570/1583-Type Preamp
• enRG can be lumped into the
enIN+ and enIN- sources
• Contributions of inIN+ & inINdepend on source impedance
at In+ & In= (R3+R4+RS)||(R1+R2)
RA
IN+
i n IN+
R3
10
R1
1k0
e n IN+
i n RG1
RG1
RS
150
RG
RG2
• Contribution of inRG depends
on feedback network
impedance
OUT2
OUT-
R4
10
= (RA + RB)||RG
– Current times impedance
generates the voltage
OUT1
IN2
i n RG2
IN-
OUT+
IN1
R2
1k0
i n IN-
e n IN-
RB
• Because gain varies with RA, RB
& RG , relative contribution of
each source depends on gain
15
Rev 1.4
1570/1583 Noise At High Gains
(60dB shown)
• RG is small, so inRG
contribution is small
• RS is small, so the inIN+
& inIN- contributions
are small
RA
IN+
i n IN+
R3
10
R1
1k0
e n IN+
i n RG1
RG
4
RG2
OUT1
OUT2
OUT-
IN2
i n RG2
IN-
OUT+
IN1
RG1
RS
150
– But, if RS is large (e.g.,
open inputs), inIN
contributions
can be significant
2k
R4
10
R2
1k0
i n IN-
e n IN-
RB
2k
• enIN+ & enIN- dominate,
along with the selfnoise of RS
High-gain noise depends more
on IC characteristics and source
impedance than anything else
16
Rev 1.4
1570/1583 Noise At Low Gains
(0dB shown)
• enIN+ & enIN- are small,
along with the self-noise
of RS
• RS is small, so the iIN+ & iINcontributions are small
• RG is open, so the two
inRG currents flow
through RA & RB
RA
IN+
i n IN+
R3
10
R1
1k0
e n IN+
i n RG1
2k
RG1
RS
150
RG
RG2
OUT2
OUT-
R4
10
– inRG currents dominate
the noise floor
OUT1
IN2
i n RG2
IN-
OUT+
IN1
R2
1k0
i n IN-
e n IN-
RB
2k
• To reduce noise at low
gains, reduce RA & RB
Low-gain noise depends more on
IC characteristics and feedback
(RA & RB) impedance than
anything else
17
Rev 1.4
1510 (Front End) Noise At High Gains
(60dB Shown)
• RG is small, so inRG
contribution is small
• RS is small, so the inIN+
& inIN- contributions are
small
RA
IN+
i n IN+
R3
10
R1
1k0
e n IN+
i n RG1
RG
10
RG2
OUT1
OUT2
OUT-
IN2
i n RG2
IN-
OUT+
IN1
RG1
RS
150
• As with 1570, if RS is large
(e.g., open inputs), inIN
contributions
can be significant
5k
R4
10
R2
1k0
i n IN-
e n IN-
RB
5k
• enIN+ & enIN- dominate,
along with the selfnoise of RS
High-gain noise depends more
on IC characteristics and source
impedance than anything else
18
Rev 1.4
1510 (Front End) Noise At Low Gains
(0dB Shown)
• enIN+ & enIN- are small, along
with the self-noise of RS
• RS is small, so the inIN+ & inINcontributions are small
• RG is open, so inRG1 flows
through RA, & inRG2 flows
through RB
RA
IN+
i n IN+
R3
10
R1
1k0
e n IN+
i n RG1
5k
RG1
RS
150
RG
RG2
IN-
OUT2
OUT-
R4
10
• Since RA & RB are fixed
(internal), designers don’t
have freedom to affect this
noise source
OUT1
IN2
i n RG2
– inRG currents dominate the
noise floor
OUT+
IN1
R2
1k0
i n IN-
e n IN-
RB
5k
Noise at low gains depends only
on IC characteristics: no flexibility
with the “standard” topology
19
Rev 1.4
Controlling Noise:
“Standard” vs. “New” Topology
• High-gain noise is
same (1nV/√Hz)
across 1510, 1512,
& 1570
• Low-gain noise can
be controlled in
new topology, but
not in standard
topology
Preamp Noise vs Gain
-95
-100
Source Impedance = 0 Ohm
-105
Bias resistors = 1k Ohms
1570 R A = R B = 2k Ohms
E IN (d B u )
-110
-115
-120
-125
-130
1570
1512
1510
-135
-140
0
– Requires the
designer to supply
RA & RB
10
20
30
Gain (dB)
40
50
60
1570’s 0dB gain noise is
~9dB lower compared to 1510,
~5dB lower compared to 1512
20
Rev 1.4
Controlling Noise:
“Standard” vs. “New” Topology
• 1583 high-gain noise
(1.9 nV/√Hz, or
5.6dB) is higher than
1510, 1512, & 1570
• But, low-gain noise
for the 1583 is almost
as low as the 1512
• New topology
preserves better lowgain noise, even with
a higher-noise part
Preamp Noise vs Gain
-95
-100
Source Impedance = 0 Ohm
-105
Bias resistors = 1k Ohms
1570, 1583 R A = R B = 2k Ohms
E IN (d B u )
-110
-115
-120
-125
1583
1570
1512
1510
-130
-135
-140
0
10
20
30
Gain (dB)
40
50
60
1583’s 0dB gain noise is
~3dB lower than 1510
~1.4dB higher than 1512
21
Rev 1.4
Noise
Bandwidth
22
Rev 1.4
Optimizing Bandwidth Vs. Gain
• Bandwidth is determined by amplifier design and feedback
resistor (RA & RB)
• In 1510/12, you’re limited to RA = RB =5k
• In 1570 & 1583, you can vary RA & RB
– Lower values => higher bandwidth
– Minimum value is 2k
Gain (dB)
0
6
10
20
30
40
50
60
1510
10.39
10.22
10.14
9.95
9.48
8.11
5.25
2.28
MHz
1512
11.95
11.84
11.65
11.15
9.76
6.66
3.04
1.07
MHz
1570 (2k )
16.78
16.78
15.65
12.71
7.83
3.65
1.38
0.46
MHz
1570 (5k )
4.19
4.19
4.19
3.91
3.41
2.41
1.12
0.43
MHz
1570 (10k )
1.93
1.91
1.91
1.86
1.72
1.39
0.87
0.40
MHz
1583 (2k )
13.97
13.00
12.01
8.73
4.33
1.69
0.59
0.19
MHz
1583 (5k )
3.92
3.60
3.40
2.95
2.20
1.24
0.52
0.19
MHz
1583 (10k )
1.56
1.50
1.47
1.38
1.19
0.84
0.44
0.17
MHz
Part
23
Rev 1.4
Noise
Bandwidth
24
Rev 1.4
Practical Considerations for RA & RB
• RG is determined by
maximum gain and
RA, RB values
• To minimize output
(differential) offset
with gain, use CG
• CG works against RG
to determine LF
cutoff (f0)
RA
IN+
5k
OUT+
IN1
RG1
RG2
RG
10
CG
3300u
OUT1
OUT2
IN2
OUT-
INRB
5k
– Small RG and low f0
means big CG
25
Rev 1.4
Varying Gain in the “Standard” Circuit
• RG varies to set gain
• Max gain when RGV is
maximum
• Min gain when RGV is
minimum
• Highest lowfrequency cutoff
occurs at max gain
• CG depends on
desired LF cutoff and
RGF value
26
RA
IN+
5k
RG
RG1
RG2
R GV
10k
OUT+
IN1
R GF
10
CG
3300u
OUT1
OUT2
IN2
OUT-
INRB
5k
CG=3300 f, RG=10 , fo= 4.43Hz
Rev 1.4
Practical Considerations for Varying
Gain in the “Standard Circuit”
• Effective end resistance in
RGV limits max gain
– Reduce RGF to offset
– Variation in effective end
resistance will alter max
gain
• Pots used by APB
Dynasonics have 2~3
(measured) end resistance
• Thanks to John Petrucelli
for samples!
RA
IN+
5k
RG
RG1
RG2
R GV
10k
OUT+
IN1
R GF
10
CG
3300u
OUT1
OUT2
IN2
OUT-
INRB
5k
– Check your spec sheet for
your tolerances …
27
Rev 1.4
Practical Considerations for Varying
Gain in the “New” Circuit
• The conventional circuit is shown
at right
• To minimize low-gain noise,
select RA & RB as low as possible
RA
IN+
2k
– For 1570, that’s 2k
– For 1583, that’s ??
RG
• For 60dB gain,
RGF = 4
• End resistance may be a
significant fraction of RGF
• To maintain < 5Hz cutoff,
CG > 7300 f
RG1
RG2
R GV
10k
OUT+
IN1
R GF
4
CG
6800u
OUT1
OUT2
IN2
OUT-
INRB
2k
– That’s a big, expensive cap
– Is < 5Hz cutoff a good idea at
60dB gain?
28
Rev 1.4
Consider a Dual-Element Pot to vary
RG, RA & RB Simultaneously
• Using a dual-element
pot allows RA & RB to
vary in addition to RG
RA
IN+
– Lowers RA & RB at min
gain, but raises them at
max gain
• This allows smaller CG
for the same cutoff
2k
R GV1
4k
RG1
R GF
12
CG
2200u
OUT+
IN1
CW
RG2
CW
OUT1
OUT2
IN2
R GV2
OUT-
4k
IN-
– In circuit shown
fo ≈ 5.5Hz
– 60dB gain noise is still
very low: 1.18nV/√Hz
(-129.1dBu with 150
source)
RB
2k
29
Rev 1.4
Pot Taper
• Measured taper of
“5% reverse log” taper
pots (thanks to APB)
• Actually two linear
sections with
different resistances
in each section
• What curve of gain
vs. rotation will this
produce?
Resistance vs Pot Rotation
5000
4500
R e s is ta n c e in O h m s
4000
3500
3000
2500
2000
1500
1000
500
CW
0
0
50
100
150
200
250
300
Rotation in degrees
30
Rev 1.4
Gain vs. Rotation Compared:
Single and Dual-Element Circuits
• Theoretical “ideal”
curve shown in black
• Single-element pot
(circuit of slide 26)
results in red curve
• Dual-element pot
(circuit of slide 29)
results in blue curve
• Dual-element pot
trajectory is much
closer to “ideal”
Gain vs Pot Rotation
60
50
1 5 7 0 G a in in d B
Ideal Response
40
Dual Pot
Configuration
30
20
Single Pot Configuration
10
CW
0
0
50
100
150
200
250
300
Rotation in degrees
31
Rev 1.4
Noise
Bandwidth
32
Rev 1.4
Single-Ended Output Stages
• The 1570/1583 topology is differential in,
differential out
– Common-mode gain is always unity (0dB)
– Differential gain varies with RG, RA, & RB
(0~>60dB)
– CMRR is equal to differential gain
– Output has a (negative) common-mode DC
offset of 1 diode drop (~-0.6V)
IN-
OUT2
12k
33
6k
VOUT
From 1570
or 1583
OUT1
IN+
12k
6k
OUT
REF
1246/1256
• To provide CMR at low gains, add a
differential amplifier after the
1570/1583
• Choose carefully to avoid adding noise
and limiting bandwidth
– 1510/1512 includes a pretty quiet, wide-band
amp
– 1570/1583 allows designers flexibility in
choosing this amplifier for even greater
performance
– Circuit at right (w/ 2114) compromises low-gain
noise floor by only ~2.6 dB (for the 1570) and
~0.14 dB (for the 1583)
SENSE
C5
22p
R15
1k13
R14
OUT2
From 1570
or 1583
OUT1
2k15
R16
–
2114 or
5532
OUT
+
2k15
R17
1k13
C6
22p
Rev 1.4
Differential Output Stages
• In many cases (e.g., driving
ADC), differential outputs are
needed
• But, many designers want to
remove common-mode signals
• Simple circuit (top, using 1286)
has great common-mode
rejection, fair noise
performance
• More complex circuit (bottom,
after Birt) maintains very low
differential noise, but commonmode rejection depends on
resistor matching
IN-
OUT2
From 1570
or 1583
OUT1
IN+
SENSE
REF
SENSE
IN-
OUT-
1286/
1296
REF
IN+
OPTIONAL
Vcm
INPUT
J1
C5
R14 180p
OUT2
From 1570
or 1583
OUT1
R15
2k15
–
2k15
R21
OUT+
2114
R16
49R9
+
2k15
C6
180p
R18
U1
R17
2k15
C7
33p
R22
OUT–
49R9
R19
4k99
–
OPTIONAL
Vcm
INPUT
4k99
2114
+
R20
2k49
34
OUT+
1286/
1296
U2
Rev 1.4
Differential Output Stages
• Variation of Birt circuit
suggested for driving A/D
converters
– Includes attenuation
suitable for ~2VRMS
differential drive
From 1570
or 1583
R9
IN-
2k49
R10
IN+
R11
C17
_ 2n2
NJM
2114
+
2k49
C18
• 1570/1583 differential
output allows designers to
spend on highperformance circuits &
opamps when necessary,
or save money when cost
is more important than
performance
301
R12
R13
47R
301
C20
2n7
2n2 R14
VCM
IN
R17
2k49
4k99
_
NJM
2114
+
C19
33p
OUT+
R15
OUT-
47R
R16
4k99
35
Rev 1.4
Summing Up
Why choose the “new” topology?
36
Rev 1.4
Summary
• “New” 1570 & 1583 topology is a subset of the
“standard” one
• Removing feedback resistors gives designers
freedom to change their values
– Optimize noise
– Optimize bandwidth
– Minimize blocking capacitor value while
maintaining optimum noise performance
– Optimize gain vs. rotation for analog control
• Removing output amp offers more flexibility
– Naturally provides a differential output
– Allows designer to set common-mode rejection
– Optimize noise performance
• Less really is more!
37
Rev 1.4
Thank You to …
• Joe Lemanski (THAT’s Applications Engineering
Manager), for many of the facts & figures contained
herein
• Gary Hebert (THAT’s Chief Technical Officer) & Fred
Floru (THAT’s Principal IC Design Engineer) for general
review and much tutoring in RA/RB‘s influence on noise
• Dave Lail (THAT’s Art Director) for providing the
drawings, and putting up with endless revisions
• Steve Green (THAT’s Technical Marketing Manager) for
many suggestions
• Dan Parks (now President of Cruz Tools, was the
marketing guy at SSM), for help with early preamp
chronology and credits.
• All of you in the audience for attending and supporting
us!
38
Rev 1.4

De-Integrating Integrated Circuit Preamps

Transcription

Similar documents

SERVICE MANUAL K4

MSR400 - Cjoint

Pre-Coronation Invite - Give to Dillard University