Transcript New York 2011 Balanced Links Masterclass

AES NEW YORK 2011
Product Design Session PD1
The Quietest Link
Noise and CMRR in balanced interconnections
Douglas Self
1
The advantages of a balanced link
• Noise
due to current flowing through the link ground is
cancelled out.
• Noise
coupled to the link conductors (eg magnetically) is
cancelled out.
2
The basic balanced link
3
The basic balanced input
In all cases the input impedance at the +ve input is 20 KΩ
With balanced input drive input impedance at the –ve input is only 6.66 KΩ
The noise output (using 5532) is -104.8 dBu
4
The basic unbalanced input
Much quieter than 10K balanced input:
Noise output = -119 dBu
by 14 dB
5
A practical balanced input
EMC filters added (R5,C3 and R6,C4)
DC blocking added (C5,C6 non-polar)
Bandwidth limitation added (C1,C2)
6
A high-end balanced input
Common-mode choke added for greater EMC immunity
R4 will usually be made trimmable for optimum CMRR
7
NOISE
8
Noise advantage of a balanced output
Noise out of inverter only = -113.5 dBu
Bal input amp only noise out = -104.8 dBu
Noise out of complete link =
-104.5 dBu
For a 1 Vrms input (+2.2 dBu)
The link S/N ratio is 2.2 + 6 +104.5 = 112.7 dB.
Balanced advantage = 5.4 dB
If R2, R4 of input amp reduced to 5KΩ, for 1 Vrms internal:
advantage = 5.9 dB
9
The signal/noise ratio of a balanced
link is determined by the noise
performance of the balanced input
amplifier
10
High impedance balanced input amplifier
Noise out with R = 10K is -105.1 dBu
(22 – 22 kHz)
For minimum input impedance of 47 KΩ via IN-, R = 68 KΩ
Noise out = -98.6 dBu:
(68 KΩ x 2/3 = 47 KΩ)
6.2 dB worse than with R = 10 KΩ
11
High impedance balanced input amplifiers
Using 5532
R Ohms
Noise out dBu
Relative to 10K
10K
-105.1
0.0 dB
15K*
-103.4
1.7 dB
22K
-101.8
3.3 dB
47K
-98.6
6.5 dB
68K
-96.8
8.3 dB
100K
-94.7
10.4 dB
* R = 15 K gives a minimum input impedance of 10 K
12
Low impedance balanced input amplifiers
Too low for direct connection to outside world
Using 5532
R Ohms
Noise out dBu
Relative to 10K
10K
-105.1
0.0 dB
4K7
-107.8
-2.7 dB
2K
-110.2
-5.1 dB
1K2
-111.2
-6.1 dB
820 *
-111.8
-6.7 dB
560
-112.2
-7.1 dB
* Gives the minimum input impedance for loading a preceding 5532 stage
13
5532 opamp noise consistency
Test circuit is the R = 10K balanced input amplifier.
Noise output in dBu. (22 Hz - 22 kHz)
Manuf:
Philips
Fairchild
Texas
JRC
Signetics
#1
-105.1
-105.3
-105.0
-105.1
-104.8
#2
-104.9
-105.2
-105.0
-105.1
-104.9
#3
-104.9
-105.3
-104.9
-105.0
-104.9
#4
-105.1
-105.4
-104.7
-105.1
-105.0
#5
-105.1
-105.3
-105.0
-105.2
-105.0
All but two within a 0.5 dB range
14
Improving the noise performance of the
balanced input amplifier
1)Add input buffers so lower resistor values
can be used in the balanced amplifier stage
2) Use multiple amplifiers
15
Improving the noise performance: 1
Added input buffers and lower balanced amp resistors
Noise output = -110.2 dBu
5.1 dB better than the standard 10K balanced amp (-105.1 dBu)
16
Improving the noise performance: 2
Adding a second balanced amplifier and averaging the outputs
Noise output = -112.5 dBu
7.4 dB better than the standard 10K balanced amp (-105.1 dBu)
2.3 dB better than the previous amplifier
17
There is also an analogous improvement in CMRR
Improving the noise performance: 3
With four balanced amps, noise output = -114.0 dBu
8.9 dB better than the standard 10K balanced amp (-105.1 dBu)
1.5 dB better than the previous amplifier
18
Improving the noise performance: 4
Doubling the input buffers
Noise output = -116.2 dBu
11.1 dB better than standard 10K amp
2.2 dB better than previous version
Looks unwieldy but can be very
compact with SM technology.
Only 4 opamp packages.
19
Improving noise performance: summary
•
•
•
•
•
Standard 10K balanced amp
Buffers + 820Ω balanced amp
Buffers + dual 820Ω balanced amps
Buffers + quad 820Ω balanced amps
Dual buffers + quad 820Ω bal amps
-105.1 dBu
0 dB
-110.2 dBu -5.1 dB
-112.5 dBu -7.4 dB
-114.0 dBu -8.9 dB
-116.2 dBu -11.1 dB
20
Noise improvement:
Balanced input stage noise out = -106.7 dBu
5.8 dB better
Bandwidth defn filter noise out = -105.7 dBu
4.7 dB better
21
Noise improvement in real life
Cambridge Audio 840W
Balanced input 0.9 dB quieter than
unbalanced input
CES Innovation Award in January 2008
22
CMRR
23
Optimising balanced amp CMRR
What affects it?
•
•
•
•
•
Source impedance imbalances
Input amplifier resistance imbalances
Finite op-amp gain
Finite op-amp bandwidth
Finite CMRR of op-amp itself
24
The basic balanced link again
25
The effects on balanced amp CMRR
Source impedance imbalance
Balanced input amplifier R = 10 KΩ
Rout +
Rout-
Error
CMRR
100 Ω
100 Ω
0
Infinity
100 Ω
101 Ω
1%
-80.2 dB
100 Ω
110 Ω
10%
-60.2 dB
26
The effects on balanced amp CMRR
Source impedance imbalance
Balanced input amplifier R = 100 KΩ
Rout +
Rout-
Error
CMRR
100 Ω
100 Ω
0
Infinity
100 Ω
101 Ω
1%
-100.1 dB
100 Ω
110 Ω
10%
-80.8 dB
27
The effects on balanced amp CMRR
Source impedance imbalance
Balanced input amplifier R = 1 MΩ
Rout +
Rout-
Error
CMRR
100 Ω
100 Ω
0
Infinity
100 Ω
101 Ω
1%
-120.1 dB
100 Ω
110 Ω
10%
-100.8 dB
28
The impedance-balanced output
The XLR connector is likely to be much more expensive
than an extra 5532 section that would make it a true
balanced output with an almost 6 dB noise advantage.
29
The effects on balanced amp CMRR
Finite opamp gain alone: opamp CMRR infinite
Open-loop gain x
LF CMRR
10,000
-74.0 dB
30,000
-83.6 dB
100,000
-94.0 dB
300,000
-103.6 dB
1,000,000
-114.1 dB
30
The effect on balanced amp CMRR
Finite opamp gain alone: opamp CMRR infinite
Opamp
Open-loop gain x
LF CMRR
NE5532
100,000
-94 dB
TL072
200,000
-100 dB
OPA2134
1,000,000
-114 dB
OPA627
1,000,000
-114 dB
OP27
1,800,000
-119 dB
LM4562
10,000,000
-134 dB
LT1028
20,000,000
-140 dB
31
The effects on balanced amp CMRR
Finite opamp bandwidth
LF open-loop gain of 100,000x gives an LF CMRR of -94 dB
32
The effects on balanced amp CMRR
Finite CMRR of the opamp itself
If all resistors are exactly correct and opamp gain
infinite the CMRR of the link is that of the balanced
input amplifier opamp CMRR.
Opamp
Opamp LF CMRR
LM4562
-120 dB
OP27
-120dB
LT1028
-120 dB
OPA627
-116 dB
NE5532
-100 dB
OPA2134
-100 dB
TL072
-86 dB
33
The effects on balanced amp CMRR
Input amplifier resistance imbalances
Worst-case CMRR with all resistors at limit of their
tolerance in the most unfavourable direction
T is tolerance in %
34
The effects on balanced amp CMRR
Input amplifier resistance imbalances
Worst-case CMRR with all resistors at limit of their
tolerance in the most unfavourable direction
Resistor tolerance
Worst-case CMRR
5%
1%
0.5%
0.1%
20 dB
34 dB
40 dB
54 dB
This will never happen.
In real life 1% resistors reliably give CMRR better than -40 dB
35
The effects on balanced amp CMRR
Using 5532
• Source impedance imbalance
• Finite opamp open-loop gain
• Op-amp CMRR
• Resistor imbalances
-80 dB
-94 dB
-100 dB
-40 dB
Resistor imbalances have the greatest effect on LF CMRR
36
Optimising balanced amp CMRR
As used in top-grade mixing consoles by the thousand.
EMC & DC-blocking components not shown.
37
Optimising balanced amp CMRR
Trim improves CMRR from -50 dB to better than -80 dB up to 500 Hz
5532 opamp
38
Other balanced input
amplifier configurations
39
Other balanced input amplifier configurations
•
•
•
•
•
High-impedance 2-opamp balanced amplifier
The “Superbal” balanced input amplifier
Gain-switched balanced input amplifier
Variable-gain balanced input amplifier
The instrumentation amplifier (see later)
40
A high impedance balanced input amplifier
41
A high impedance balanced input amplifier
Built with AD8397 for operation from a +5V supply rail
With Rg = 62Ω, R1A etc = 1 KΩ, Gain = +30.6 dB
Noise out = -86.0 dBu
Equiv Input Noise = -116.6 dBu
Noise Figure = +13.8 dBu
(ref 150 Ω source res; Johnson = -130.4 dBu)
42
A high impedance balanced input amplifier
CMRR plot
Effect of capacitor to ground at end of R1A for single-rail operation
43
A high impedance balanced input amplifier
Snag: if gain is set to less than 2 times, headroom is
impaired because the first opamp stage now has a gain
of greater than one, while second stage has a gain of
44
less than one.
The “Superbal” balanced input amplifier
Equal input impedances on the two inputs
Attributed to Ted Fletcher of Alice. (Studio Sound Dec 1981)
Reported by David Birt of the BBC 1990
45
Gain-switched balanced input amplifier
Gain is intermediate with switch in mid-position
46
Variable-gain balanced input amplifier
Gain range is -20 to +10 dB
47
The instrumentation
amplifier
Noise and CMRR
48
Application for balanced input with gain of 4 times
Active crossover unit with elevated internal
levels to increase signal/noise ratio by reducing
effect of filter noise
Assumption that there is only a limited-range
level trim between crossover and power
amplifier, so that power amp will always clip first
49
Application for balanced input with gain of 4 times
The active crossover is followed by a ±10 dB gain trim
network that is normally expected to work at the 0 dB
setting, ie giving 10 dB of attenuation and so reducing
the noise from the input stage.
50
Active crossover input circuit
Gain = 3.92 times
Noise out = -100.9 dBu
Combined HP and LP filters: complete input system with
only two opamps
51
Instrumentation amplifier: 4x1
Stage 1 gain = 3.93 times
Total gain = 3.93 times
Stage 2 gain = 1 times
Noise out = -105.4 dBu
52
Instrumentation amplifier: 4x1
Measured CMRR of Stage 2 alone = 56 dB
With both stages measured CMRR = 69 dB
CMRR improvement over simple amp is 13 dB
Stage 1 gain = 7.67 times
Stage 2 gain = 0.51 times
Total gain = 3.93 times
Noise out = -106.2 dBu
0.8 dB quieter than previous version for no cost
53
Practical instrumentation amplifier: 4x1
EMC filtering, DC blocking and HF response definition added
A lot of extra components
Gain = 3.93 times
Noise out = -105.4 dBu
54
Instrumentation amplifier: 8x0.5
Stage 1 gain = 7.67 times
Stage 2 gain = 0.51 times
Total gain = 3.93 times
Noise out = -106.2 dBu
0.8 dB quieter than previous 4x1 version for no cost
55
Instrumentation amplifier: 8x0.5
Measured CMRR of Stage 2 alone = 53 dB
With both stages measured CMRR = 70 dB
CMRR improvement over simple amp is 17 dB
Previous 4x1 version improvement was 13 dB
Stage 1 gain = 7.67 times
Stage 2 gain = 0.51 times
Total gain = 3.93 times
Noise out = -106.2 dBu
56
0.8 dB quieter than previous version for no cost
Instrumentation amplifier: 8x0.5
What about headroom with unbalanced drive?
If the input is driven unbalanced, we will have 0.775 Vrms on one input
pin, and nothing on the other. You might fear we would get 6 Vrms on
one output of Stage 1. In fact things are rather better than that. The
cross-coupling of the feedback network R3, R4, R5 greatly reduces the
difference between the two outputs of stage 1 with unbalanced inputs.
With 0.775 Vrms on one input pin, and nothing on the other, the Stage
1 outputs are 3.39 Vrms and 2.61 Vrms, summing to 6 Vrms, reduced
to 3 Vrms out by Stage 2. The higher voltage is at the output of the
amplifier associated with the driven input.
Maximum output with unbalanced input is slightly reduced at 9.4 Vrms.
This slight restriction does not apply to the earlier 4x1 version.
Stage 1 gain = 7.67 times
Stage 2 gain = 0.51 times
Total gain = 3.93 times
Noise out = -106.2 dBu
57
0.8 dB quieter than previous version for no cost
Instrumentation amplifier: 21x0.19
Gain = 3.8 times
Note R3,R4,R5 scaled
Noise out = -106.2 dBu once more
No further noise improvement
Still 0.8 dB quieter than the 4x1 version for no cost
58
Instrumentation amplifier: 21x0.19
Measured CMRR of Stage 2 alone = 61 dB
With both stages measured CMRR = 86 dB
CMRR improvement over simple amp is 25 dB
The 4x1 version improvement was 13 dB
The 8x0.5 version improvement was 17 dB
Stage 1 gain = 21 times
Stage 2 gain = 0.19 times
Total gain = 3.8 times
Noise out = -10 dBu
0.8 dB quieter than previous version for no cost
59
Instrumentation amplifier: 21x0.19
What about headroom?
It is impaired with unbalanced drive at same level
Assumptions are:
1) Input drive will always be balanced
2) Maximum output need not be more than 3 Vrms
60
Unity-gain instrumentation amplifier
Instrumentation-amplifier balanced input stage with unity gain
Gain of 2 times in Stage 1 and 0.5 times in Stage 2 for overall unity gain
Measured CMRR of Stage 2 alone = 56 dB
With both stages measured CMRR = 69 dB
61
Unity-gain instrumentation amplifier
Will pass the full 10 Vrms maximum opamp level, but only in balanced mode.
In unbalanced mode Stage 1 will clip at the opamp connected to the driven
input because gain to this point from input is 1.5 times.
The maximum signal that can be handled in unbalanced mode is therefore
about 7 Vrms.
Stage 2 alone had a measured CMRR of -57 dB
Adding Stage 1 brought this up to -63 dB.
You can have 6 dB more CMRR, if you can tolerate a loss of 3.5 dB in headroom
when the stage is run unbalanced but with the same input voltage.
The noise output is -111.6 dBu
62
Other output amplifier
configurations
63
The zero-impedance output
Output impedance reduced from 68 Ω to 0.24 Ω at 1 kHz (with 5532)
Non-inverting: inverting version also possible
64
Quasi-balanced output
(Simulates floating transformer)
Distortion not of the best
Requires preset
Do not use thick-film SM quad R’s
Specialised studio use only
65
Transformer-balanced output
R1 must be as low as possible to minimise transformer LF distortion
66
Better transformer-balanced output
Zero-impedance output stage reduces transformer LF distortion
Zobel network R4,C2 eliminates peak in frequency response
67
Frequency response and Zobel networks
Sowter 3292 line out transformer
68
Output transformer LF distortion
Effect of series resistance on transformer LF distortion
Sowter 3292 line out transformer
69
Ground cancellation
(Ground-compensated outputs)
70
The ground-cancelling link
● A balanced link renders noise on the ground connection
harmless by subtraction in the input amplifier
● A ground-cancelling link renders noise on the ground
connection harmless by addition in the output amplifier
71
The ground-cancelling link
Same 3-way cable is used
72
The ground-cancelling link
Stage noise = -107.1 dBu
Stage noise = -119.0 dBu
Practical ground-cancelling link circuitry
73
The signal/noise ratio of a groundcancelling link is determined by the
noise performance of the output
amplifier, not the input amplifier
74
Ground-cancel output: inverting
For R5,R6 = 10 KΩ
Noise out = -107.2 dBu
75
Ground-cancel output: inverting
Reduce R5, R6 to reduce noise output
R5,R6 = 10 KΩ
R5,R6 = 2K2
R5,R6 = 1 KΩ
R5,R6 = 560 Ω
Noise out = -107.2 dBu
Noise out = -111.3 dBu
Noise out = -112.2 dBu
Noise out = -112.6 dBu
0.0 dB
-4.3 dB
-5.0 dB
-5.4 dB
76
Ground-cancel output: non-inverting
77
Ground-cancel output: inverting
Measurement of ground-cancel link CMRR
78
Plot of ground-cancel link CMRR
79
Comparison of balanced and ground-cancel links
● Balanced links have potentially twice the signal voltage
Noise can be lower if you can define both ends of link
For a 1 Vrms signal the bal link S/N ratio is 2.2 + 6 +104.5 = 112.7 dB
● Ground-cancel
links have only capacity for usual
signal voltage limited by supply rails
If output stage R5,R6 = 1 KΩ
Complete ground-cancel link
Noise out = -112.2 dBu
Noise out = -112.0 dBu
For a 1 Vrms signal the GC link S/N ratio is 2.2 +112.0 = 114.2 dB
So here the GC link is 1.5 dB quieter
80
AES NEW YORK 2011
The Quietest Link
Noise and CMRR in balanced interconnections
[email protected]
douglas-self.com
81