Noty an70 Linear Technology

Application Note 70

October 1997

A Monolithic Switching Regulator with 100µV Output Noise

“Silence is the perfectest herald of joy ...” Jim Williams

INTRODUCTION

Size, output flexibility and efficiency advantages have made switching regulators common in electronic apparatus. The continued emphasis on these attributes has resulted in circuitry with 95% efficiency that requires minimal board area. Although these advantages are welcome, they necessitate compromising other parameters.

Switching Regulator “Noise”

Something commonly referred to as “noise” is a primary concern. The switched mode power delivery that permits the aforementioned advantages also creates wideband harmonic energy. This undesirable energy appears as radiated and conducted components commonly labeled as “noise.” Actually, switching regulator output “noise” is not really noise at all, but coherent, high frequency residue directly related to the regulator’s switching.1 Figure 1 shows typical switching regulator output noise. Two distinct characteristics are present. The slow, ramping output ripple is caused by finite storage capacity of the regulator’s output filter components. The quickly rising spikes are associated with the switching transitions. Figure 2 shows another switching regulator output. In this case the ripple has been eliminated by adequate filtering and linear postregulation, but the wideband spikes remain. It is these fast spikes that cause so much difficulty in systems. Their high frequency content often corrupts associated circuitry, degrading performance or even disabling operation. Noise gets into adjacent circuitry via three paths. It is conducted out of the regulator output lead, it is conducted

back to the driving source (“reflected” noise) and it is radiated. The multiple transmission paths combine with the high frequency content to make noise suppression difficult. Unconscionable amounts of bypass capacitors, ferrite beads, shields, Mu-metal and aspirin have been expended in attempts to ameliorate noise-induced effects.

10mV/DIV

(AC COUPLED)

1µs/DIV AN70 F01

Figure 1. Typical Switching Regulator Output “Noise.” Wideband Spikes Are Difficult to Suppress, Causing System Interference Problems. Ripple Component Has Low Harmonic Content, Is Relatively Easily Filtered

20mV/DIV

(AC COUPLED)

Note 1. Noise contains no regularly occurring or coherent components. As such, switching regulator output “noise” is a misnomer. Unfortunately, undesired switching related components in the regulated output are almost always referred to as “noise.” As such, although technically incorrect, this publication treats all undesired output signals as “noise.” See Appendix B, “Specifying and Measuring Something Called Noise.”

50µs/DIV AN70 F02

Figure 2. Linear Regulator Eliminates Ripple, but Wideband Spikes Remain. Peak-to-Peak Amplitude Exceeds 30mV (Just Visible Near 2nd, 5th and 8th Vertical Graticule Divisions)

, LTC and LT are registered trademarks of Linear Technology Corporation.

AN70-1

Application Note 70

Alternate approaches involve synchronizing switching regulator operation to the host system or turning off switching during critical system operation (an “interrupt driven” power supply). Another approach places critical system operations between switch cycles, literally running between electronic rain drops.

The difficulty of debugging a noise-laden system and the compromises involved in synchronized approaches could be eliminated with a low noise switching regulator. An inherently low noise switching regulator is the most attractive approach because it eliminates noise concerns while maintaining system flexibility.

A Noiseless Switching Regulator Approach

The key to an inherently low noise regulator is to minimize harmonic content in the switching transitions. Slowing down the switching interval does this, although power dissipated during the transition causes some efficiency loss. Reducing switch repetition rate can largely offset the

losses, resulting in a reasonably efficient design with small magnetics and the desired low noise. Noise reduction by restricting harmonic generation has been employed before, although the implementations were complex and narrowly applicable.3 A monolithic approach, broadly usable over a range of magnetics and applications, is described here.

A Practical, Low Noise Monolithic Regulator

Figure 3 describes the LT®1533, a monolithic regulator designed for low noise switching supplies. Figure 4 details the pin functions. Figure 3’s functional blocks show a fairly conventional push-pull architecture with a major exception. The push-pull approach has good magnetics utilization (power transfer is always occurring in the transformer; the core does not store energy) and pulls current

Note 2: See References 2 and 3 for details and practical examples of these techniques.

Note 3: See Appendix A, “A History of Low Noise DC/DC Conversion.” See also References 4 through 10.

NFB

SYNC

SHDN V

LDO REGULATOR

NEGATIVE FEEDBACK

AMP

–

INTERNAL V

CURRENT

PGND COL A COL B

AMP

OUTPUT

DRIVERS

–

1.25V

–

ERROR

AMP

OSCILLATOR

COMP

SLEW CONTROL

VSL

CSL

Figure 3. LT1533 Simplified Block Diagram. 1A Slew-Controlled Output Stages Provide Low Noise Switching

AN70-2

GND DUTY

AN70 F03

Application Note 70

COL A, COL B: Output transistor collectors which switch out-ofphase.

DUTY: Grounding this pin forces the outputs to switch at a 50% duty cycle. This pin must float if not used.

SYNC: Used to synchronize to an external clock. Float or tie to ground if unused.

CT: Oscillator timing capacitor. RT: Oscillator timing resistor. FB: Used for positive output voltage sensing. NFB: Used for negative output voltage sensing. GND: Analog ground pin. PGND: High current ground return. Should be returned to

ground via ≈ 50nH (≅1" of PC trace or wire, or a small ferrite bead). See Appendix F or schematic (Figures 5 and 26) notes for details. In some package options, this pin may be internally connected to “GND” pin.

VC: Frequency compensation node. SHDN: Normally high. Grounding this pin shuts the part down.

= 20µA.

SHDN

: Current slew control resistor.

CSL

: Voltage slew control resistor.

VSL

VIN: Input supply pin. 2.7V to 30V range. Undervoltage lockout

at 2.55V.

Figure 4. LT1533 Short Form Pin Function Descriptions

continuously from the source. The even, continuous current drain from the source eliminates the fast, high peak currents required by flyback and other approaches. The source sees a benign load and is not corrupted. The switches also receive nonoverlapping drive, ensuring they do not conduct simultaneously. Simultaneous conduction

would cause excessive, quickly rising currents, degrading efficiency and generating noise.

The design’s most significant aspect is the output stage. Each 1A power transistor operates inside a broadband control loop. The voltage across each transistor and the current through it are sensed and the loop controls slew rate of each parameter. The voltage and current slew rates are independently settable by external programming resistors. This ability to control the switching’s rate-ofchange makes low noise switching regulation practical. Operating the switching transistors in a local loop permits predictable, wide range control over a variety of situations.4 Figure 5 is a 40kHz, 5V to 12V converter using the LT1533 in a push-pull, “forward” configuration. The feedback resistor’s ratio produces a 12V output. A two-section LC filter provides high ripple attenuation, although a single section will give good performance. It is particularly noteworthy that high frequency noise content (as opposed to the 40kHz fundamental related ripple) is unaffected by output filter characteristics. This is so simply because there is so little high frequency energy developed in this circuit. If there’s nothing there, it doesn’t need to be filtered!

L2 provides compensation for the output current control loop. In practice, L2 may be a length of PC trace, a small inductor, a coiled section of wire or a ferrite bead. See Appendix F, “Magnetics Considerations” for complete discussion.

Note 4: Patent pending.

AN70 F05

100µH

OPTIONAL

(SEE TEXT)

C3 47µF

12V

47µF

1N4148

4.7µF

SHDN

DUTY

3300pF

18k

0.01µF

SYNC

COL A

COL B

LT1533

PGND

NFB

VSL

CSL

GND

15k

2.49k 1%

L1, L3: COILTRONICS CTX100-3

L2: 22nH TRACE INDUCTANCE, FERRITE BEAD OR T1: CTX02-13665-X1 (SEE APPENDIX F FOR DETAILS)

100µH

1N4148

21.5k 1%

INDUCTOR (SEE APPENDIX F) COILCRAFT B-07T TYPICAL

Figure 5. 100µV Noise 5V-to-12V Converter. Output LC Section May Be Deleted If Low Frequency Ripple Is Acceptable

AN70-3

Application Note 70

Measuring Output Noise

Measuring the LT1533’s unprecedented low noise levels requires care.5 Figure 6 shows a test setup for taking the measurement. Good connection and signal handling technique combined with judicious instrumentation choice should yield a 100µ V noise floor in a 100MHz bandwidth. This corresponds to the noise of a 50Ω resistor in a 100MHz bandwidth.

Before measuring regulator output noise, it is good practice to verify test setup performance. This is done by running the test setup with no input. Figure 7 shows a noise base line of 100µ V in a 100MHz bandwidth, indicating the instrumentation is operating properly. Measuring Figure 5’s noise involves AC coupling the circuit’s output into the test setup’s input. Figure 8 shows this. Coaxial connections must be maintained to preserve measurement integrity.6 Figure 9’s waveforms detail circuit operation. Traces A and C are switching transistor collector voltages, B and D are the respective transistor currents. The test setup’s output, representing circuit output noise, is Trace E. Wideband spiking and ripple, just visible in the noise floor, is inside 100µ V, even in a 100MHz bandpass.

This is spectacularly good performance and is, in fact, actually better than the photo shows. Removing all probes from the breadboard leaves only Trace E’s coaxial connection. This eliminates any possible ground loop-induced error.8 Figure 10’s trace shows 40kHz ripple with about the same amplitude as in Figure 9. Switching related spikes, just faintly outline in the noise, are reduced.

Measurement bandwidth is reduced to 10MHz in Figure 11, attenuating test fixture amplifier noise. Switching and ripple residue amplitude and shape do not change, indicating no signal activity beyond this frequency. Figure 12’s

Note 5: Equipment selection and measurement techniques are detailed in Appendix B, “Specifying and Measuring Something Called Noise.” See also Appendix C, “Probing and Connection Techniques for Low Level, Wideband Signal Integrity.”

Note 6: Again, see Appendices B and C for extended treatment of these and related issues.

Note 7: It is common industry practice to specify switching regulator noise in a 20MHz bandpass. There can be only one reason for this, and it is a disservice to users. See Appendix B for tutorial on observed noise versus measurement bandwidth.

Note 8: See Appendix C for related discussion and techniques for triggering oscilloscopes without invasively probing the circuit.

OSCILLOSCOPE

0.01V/DIV VERTICAL SENSITIVITY

100µV/DIV REFERRED TO AMPLIFIER INPUT

HP461A

AMPLIFIER

X40dB

INPUT

= 50Ω

Figure 6. Test Setup Noise Baseline Is 100µV 50Ω Resistor Noise Limited. BNC Cable Connections and Terminations Provide Coaxial Environment, Ensuring Wideband, Low Noise Characteristics

BNC CABLE

50Ω TERMINATOR HP-11048C OR EQUIVALENT

in 100MHz Bandwidth. Performance Is

P-P

AN70 F06

AN70-4

Application Note 70

100µV/DIV

20µs/DIV AN70 F07

Figure 7. Oscilloscope Verifies Test Setup 100µV Noise Floor in 100MHz Bandwidth. Indicated Noise Is That of a 50Ω Resistor

HP461A

COUPLING

OUT

CAPACITOR

SWITCHING

REGULATOR

UNDER TEST

BNC CABLE

AND

CONNECTORS

HP-10240B

LOAD (AS DESIRED)

AMPLIFIER

× 40dB

= 50Ω

100µV/DIV

10µs/DIV

AN70 F07

Figure 10. Removing Probes from Figure 9’s Test Eliminates Ground Loops, Slightly Reducing Observed Noise. Switching Artifacts Are Just Discernible Above Noise Floor

OSCILLOSCOPE

0.01V/DIV VERTICAL SENSITIVITY

100µV/DIV REFERRED TO AMPLIFIER INPUT

BNC CABLE

Figure 8. Connecting Figure 5’s Circuit to the Test Setup. Coaxial Connections Must Be Maintained to Preserve Measurement Integrity

A = 10V/DIV

B = 1A/DIV

C = 10V/DIV

D = 1A/DIV

E = 100µV/DIV

20µs/DIV AN70 F09

Figure 9. Waveforms for Figure 5 at 100mA Loading. Traces A and C Are Voltage; B and D are Current, Respectively. Switching Transistion’s Noise Signature Appears in Trace E, the Circuit’s Output Noise

100µV/DIV

50Ω TERMINATOR HP-11048C OR EQUIVALENT

10µs/DIV AN70 F11

AN70 F08

Figure 11. Reducing Measurement Bandwidth to 10MHz Attenuates Amplifier Noise. Switching Residue Characteristics Remain Unchanged, Indicating No Signal Activity Beyond This Frequency

AN70-5

Application Note 70

horizontal expansion of Figure 10 returns to 100MHz bandpass. The switching spike appears in the center screen region. At 2µs/division sweep, there is no wide- band activity observable. Figure 13, a 10MHz bandpass version of Figure 12, retains all signal information, further suggesting no signal power beyond 10MHz.

Figure 14 is the noise floor of an HP4195A spectrum analyzer in a 500MHz sweep. When Figure 5’s circuit is AC coupled into the analyzer, the output (Figure 15a) is essentially identical. The analyzer is unable to detect switching-induced noise in a 500MHz bandpass. Some 40kHz fundamental-related components are detectable in Figure 15b’s 1MHz wide plot, although the rest of the sweep is analyzer noise limited. Additional filtering or a linear postregulator could eliminate the 40kHz ripplerelated residue if desired.

The preamplified oscilloscope is a more sensitive tool for these measurements because its triggered operation has the advantage of synchronous detection. This is demonstratable by free running the preamplified oscilloscope sweep; the switching-related components are indistinguishable in the noise background.

Figure 14. Noise Floor of Test Fixture and HP-4195A Spectrum Analyzer in a 500MHz Sweep

100µV/DIV

2µs/DIV AN70 F12

Figure 12. Horizontal Expansion of Figure 10 Shows No Wideband Components. Switching Originated Noise Appears in Center Screen Region

100µV/DIV

2µs/DIV AN70F13

Figure 13. A 10MHz Band Limited Version of Figure 12. As Before, Signal Information Is Retained, Although Amplifier Noise Is Reduced. Results Indicate No Signal Power Beyond 10MHz

Figure 15a. Figure 5’s Circuit Connected to the Spectrum Analyzer Produces Essentially Identical Results to Figure

14. Circuit’s Noise Is Undetectable

Figure 15b. Reducing Analyzer Sweep to 1MHz Width Reveals 40kHz Related Components. Remainder of Plot Is Analyzer Noise Floor Limited, Even in Sensitive 455kHz Band

AN70-6

Figure 16 studies ripple at the first LC filter section output. The ripple’s 40kHz fundamental is clearly seen, although no wideband spikes are visible. Figure 17 horizontally expands Figure 14’s time scale, but high frequency harmonics and spikes are not observable.

Low frequency noise is rarely a concern, although Figure 18 shows it is inside 50µV in a 10Hz to 10kHz bandpass. Input current noise is usually of more interest. Excessive “reflected” noise can corrupt the regulator’s driving source, causing system level interference. Figure 19 shows Figure 5’s input current as DC with a small, 40kHz fundamentalrelated sinusoidal component. There is no high frequency content, and the sinusoidal variations are easily handled by the driving source.

System-Based Noise “Measurement”

In the final analysis, the effect of switching regulator output noise on the system it is powering is the ultimate test. Appendix K, “System-Based Noise “Measurement,” presents results when the LT1533 is used to power a 16-bit A/D converter.

Application Note 70

5mV/DIV

10µs/DIV AN70 F16

Figure 16. Ripple at Figure 5’s First LC Output Has No Wideband Spikes

5mV/DIV

2.5µs/DIV AN70 F17

Figure 17. Time Expansion of Previous Figure. No High Frequency Content Is Visible

Transition Rate Effects on Noise and Efficiency

In theory, simply setting transition rate to low values will achieve low noise. Practically, such an approach, while workable, wastes power during transitions, lowering efficiency. A good compromise sets transition time at the fastest rate permitting desired noise performance. The LT1533’s slew adjustments allow easy determination of this point. Figure 20’s photographs dramatically demonstrate the relationship between transition time and output noise for Figure 5’s circuit. The sequence shows > 5:1 noise reduction as switch transition time slows from 100ns (20a) to 1µ s (20d). Figure 20d’s displayed noise is actually lower, as the probing-induced error caused by monitoring the switch corrupts the measurement.

Figure 21 graphically summarizes Figure 20’s information. Significant noise reduction coincides with descending transition slew time until about 1.3µ s. Little additional noise benefit occurs beyond this point. Figure 22 shows efficiency fall-off with slew time. There is a 6% penalty between 100ns and 1.3µs, the same region where noise performance improves by a factor of 5 (per previous

Note 9: See Appendix C, “Probing and Connection Techniques for Low Level, Wideband Signal Integrity” for guidance.

50µV/DIV

10ms/DIV

Figure 18. Low Frequency Noise in a 10Hz to 10kHz Bandpass

10mA/DIV

AC COUPLED

ON 200mA

DC LEVEL

10µs/DIV AN70 F19

Figure 19. Figure 5’s Small Sinusoidal Input Current Variations Contain No High Frequency Content and Are Easily Absorbed by Input Supply

AN70 F18

AN70-7

Application Note 70

A = 5V/DIV

B = 100µV/DIV

A = 5V/DIV

B = 100µV/DIV

500ns/DIV AN70 F20a

(a)

500ns/DIV AN70 F20a

(c)

A = 5V/DIV

B = 100µV/DIV

500ns/DIV AN70 F20b

(b)

A = 5V/DIV

B = 100µV/DIV

500ns/DIV AN70 F20a

(d)

Figure 20. Output Noise (Trace B) vs Different Switch Slew Rates (Trace A). Highest Slew Rate (Figure a) Causes Largest Noise. Retarding Slew Rate (Figures b and c) Decreases Noise Until Lowest Noise Performance Is Achieved (Figure d)

700

600

500

400

300

200

f = 40kHz

100

PEAK-TO-PEAK WIDEBAND NOISE (µV)

= 100mA

LOAD

400

1200 1600

800

SLEW TIME (ns)

2000

AN70 F21

Figure 21. Figure 5’s Noise vs Slew Time at 40kHz Switching Frequency. Noise Reduction Beyond 1.3µs Is Minimal

80 78 76 74 72 70 68

EFFICIENCY (%)

66 64 62 60

1.3µs SLEW TIME =

f = 40kHz

= 100mA

LOAD

400

<100µV NOISE

(SEE FIGURE 21)

800

SLEW TIME (ns)

1200 1600

2000

AN70 F22

Figure 22. Figure 5’s Efficiency Drops 6% as Slew Time Extends to 1.3µs. Operation Beyond This Point Gains Little Noise Performance (See Previous Curve) with 6% Efficiency Penalty

1.3µs SLEW TIME

EFFICIENCY (%)

f = 40kHz

= 100mA

LOAD

100 200 300 400 700

WIDEBAND PEAK-TO-PEAK NOISE (µV)

500 600

AN70 F23

Figure 23. Efficiency vs Noise for Figure

5. Data Shows Significant Efficiency FallOff for Noise Below 80µV

AN70-8

Application Note 70

figure). There is an additional 6% penalty beyond 1.3µs, although no significant noise reduction occurs (again, per Figure 21). As such, operation in this region is undesirable. Figure 23 clearly shows the inflection point in the efficiency versus noise trade-off.

Negative Output Regulator

The LT1533 has a separate feedback input that directly accepts negative inputs.11 This permits negative outputs without the usual discrete level shifting stage. Figure 24’s 5V to –12V converter is similar to Figure 5’s circuit, except that the negative output is fed back to the negative feedback input. The feedback scale factor change is necessitated by the higher effective reference voltage. In all other respects, the circuit (and its performance) is similar to Figure 5.

Floating Output Regulator

Figure 25’s isolation stage permits a fully floating, regulated output. The LT1431 shunt regulator compares a portion of the output to its internal reference and drives the optoisolator with the error signal. The optoisolator’s collector output biases the LT1431’s VC pin, closing a feedback loop to regulate circuit output. The 0.22µ F capacitor stabilizes the loop and the 240kΩ resistor biases the optoisolator into a favorable operating region. This circuit’s operation and characteristics are similar to Figure 5 with the added benefit of the isolated output.

4.7µF

SHDN

DUTY

3300pF

18k

0.01µF

SYNC

COL A

COL B

LT1533

PGND

NFB

VSL

CSL

GND

15k

2.4k 1%

Floating Bipolar Output Converter

Grounding the LT1533’s “DUTY” pin and biasing FB forces the device into its 50% duty cycle mode. Figure 26’s output is full wave rectified with respect to T1’s secondary center tap, producing bipolar outputs. The forced 50% duty cycle combined with no feedback means the outputs are unregulated, proportioning to T1’s drive voltage. An output inductor is usually not required, as in Figure 5’s “forward” converter. At the very highest output currents, some inductance may be necessary to limit inrush current. If this is not done, the circuit may not start. Typically, linear regulators provide regulation.

Figure 26’s waveforms appear in Figure 27. Collector voltage (Traces A and C) and current (Traces B and D) are shown, along with the indicated output noise (Trace E). In this case linear regulators and an output filter are in use. In Figure 28 all probes except the coaxial output connection are removed. This eliminates probing induced parasitics,13 allowing a higher fidelity signal presentation. Here, the switching residuals are barely detectable in the noise floor. Removing the optional output filter (Figure 29) allows linear regulator contributed noise and switching spikes to rise, but noise is still below 300µV

Note 10: The noise and efficiency characteristics appearing in Figures 20 to 23 were generated at the bench in about ten minutes. All you CAD modeling types out there might want to think about that.

Note 11: See Figure 3’s Block Diagram. Note 12: See Appendix E, “Selection Criteria for Linear Regulators.” Note 13: See Appendix C, “Probing and Connection Techniques for

Low Level, Wideband Signal Integrity,” for relevant discussion.

100µH

AN70 F24

OPTIONAL

(SEE TEXT)

C3 47µF

–12V

47µF

1N4148

L1: 22nH INDUCTOR. COILCRAFT B-07T TYPICAL,

L2, L3: COILTRONICS CTX100-3

T1: COILTRONICS CTX02-13665-X1 (SEE APPENDIX F)

100µH

1N4148

9.6k 1%

TRACE INDUCTANE OR BEAD (SEE APPENDIX F)

P-P

Figure 24. A Negative Output Version of Figure 5. LT1533’s Negative Feedback Input Requires Minimal Configuration Changes. Noise Performance Is Identical to Positive Output Version

AN70-9

Application Note 70

4.2k

820Ω

L1: 22nH INDUCTOR. COILCRAFT B-07T, TRACE INDUCTANCE OR BEAD. SEE APPENDIX F L2: COILTRONICS CTX100-3 T1: COILTRONICS CTX-02-13665-X1. SEE APPENDIX F

4.7µF

5961610

3300pF

VSL

GND

15k

121314

CSL

LT1533

15k

PGND V

18k

L1 22nH

COL A

COL B

4N28

240k

4.7µF

OUT

GND

1N4148

0.22µF

LT1431

OPTIONAL SECOND

LC SECTION

(SEE TEXT)

100µH

2.5V

–

9.5k 1%

2.5k 1%

Figure 25. An Optoisolated Output Variant of Figure 5. Loop Closure to VC Pin Bypasses LT1533 Error Amplifier, Enhancing Loop Stability. Noise Performance is Maintained.

TO OUTPUT

COMMON

47µF

AN70 F25

OUTPUT 12V

OUTPUT COMMON

4.7µF

14 13 12

GND FB

3300pF

43k

L1: 22nH INDUCTOR. COILCRAFT B-07T TYPICAL,

PC TRACE, OR FERRITE BEAD. SEE APPENDIX F

T1: COILTRONICS CTX-02-13666-X1. SEE APPENDIX F

10k

: 1N4148

15k

VSL

LT1533

CSL

6 5

18k

15k

DUTY

PGND

COL A

COL B

151689

4.7µF

17V

100µF

TO OPTIONAL LINEAR REGULATORS AND/OR LC FILTERS, TYPICALLY 100µH/47µ

–17V

OUTPUT COMMON

AN70 F26

Figure 26. A Bipolar, Floating Output Converter. Grounding “DUTY” Pin and Biasing FB Puts Regulator into 50% Duty Cycle Mode. Floating, Unregulated Outputs Proportion to T1’s Center Tap Voltage. Linear Regulators Are Optional

AN70-10

As in Figure 5’s case, spectrum analyzer measurements are instrument limited. Figure 30 shows the analyzer’s noise floor in a 500MHz sweep when monitoring the unpowered Figure 26’s breadboard. In Figure 31, the breadboard is powered, but analyzer output is noise limited and essentially indistinguishable from the unpowered case. Similarly, Figure 32’s 1MHz wide “poweron” plot is identical to Figure 33’s noise floor limited “power-off” sweep. Note that linear postregulation is in use and the 40kHz fundamental components are not detectable. Figure 5’s circuit did not have linear postregulation and 40kHz fundamental residue appeared in Figure 15b.

A = 10V/DIV

B = 500mA/DIV

Application Note 70

500µV/DIV

10µs/DIV

Figure 29. Removing Optional LC Filter Causes Linear Regulator-Contributed Noise and Switching Spikes to Rise. Peak-to-Peak Noise Is Still <300µV

AN72 F29

C = 10V/DIV

D = 500mA/DIV

E = 100µV/DIV

20µs/DIV

Figure 27. Waveforms for the Floating Output Converter at 100mA Loading. Linear Postregulator and Optional LC Filter Are Employed. Slew-Controlled Collector Voltage (Traces A and C) and Current (Traces B and D) Produce Output (Trace E) with Under 100µV Noise

100µV/DIV

AN70 F27

Figure 30. HP4195A Analyzer’s Noise Floor in a 500MHz Sweep When Connected to Unpowered Figure 26

10µs/DIV AN70 F28

Figure 28. Removing All Probes Except Coaxial Output Connection Reveals Figure 27’s True Noise Figure. Switching Residue Is Just Detectable in Amplifier Noise

Figure 31. Figure 26’s “Power-On” Output Noise Is Undetectable in Analyzer’s Noise Floor Limited 500MHz Sweep

AN70-11

Application Note 70

Figure 32. Linear Postregulation Eliminates 40kHz Fundamental-Related Components in 1MHz Sweep

Battery-Powered Circuits

The basic configurations may be battery-powered for use in portable apparatus. Figure 34, similar to Figure 5, runs from 2.7V

(e.g., three NiCd batteries), producing 12V

MIN

output. This design induces no noise-based error when powering a fast 16-bit A/D converter, something almost no DC/DC converter can do. Appendix K contributes compelling testimony to this somewhat boastful claim.

2.7V TO 4V

(3 NiCd BATTERIES)

4.7µF

SHDN

DUTY

3300pF

18k

0.01µF

SYNC

COL A

COL B

GND

LT1533

PGND

NFB

VSL

CSL

15k

4.99k 1%

Figure 33. Turning Circuit Power Off Verifies Figure 32’s Plot Is Analyzer Noise Floor Limited. Sweep Results Are Identical to Figure 32’s “Power-On” Data

Figure 35 also operates from three NiCd cells, producing a 9V output. This design achieves 100µV output noise, qualifying it as the electronic equivalent of a 9V battery.

Performance Augmentation

In some cases it may be desirable to augment LT1533 performance characteristics. Usually, this involves additional circuitry, and may necessitate trading off performance in one area to gain the desired benefit.

1N5817

L1, L3: COILTRONICS CTX100-3

L2: 22nH TRACE INDUCTANCE, FERRITE BEAD OR

INDUCTOR (SEE APPENDIX F) COILCRAFT B-07T TYPICAL

T1: CTX02-13665-X1 (SEE APPENDIX F FOR DETAILS)

100µH

R1 15k 1%

AN70 F34

OUT

C3 47µF

OPTIONAL FOR

()

100µH

LOWEST RIPPLE

47µF

AN70-12

Figure 34. Circuit Delivers 5V from Three NiCd Batteries, Has 100µV Wideband Output Noise. This Design Contributes No Noise-Based Error When Powering a 16-Bit A/D Converter (See Appendix K)

Application Note 70

2.7V TO 4V

(3 NiCd BATTERIES)

1N4148

4.7µF

SHDN

DUTY

18k

0.01µF

SYNC

3300pF

COL A

COL B

LT1533

PGND

NFB

VSL

CSL

GND

15k

3.48k 1%

L1, L3: COILTRONICS CTX100-3

L2: 22nH TRACE INDUCTANCE, FERRITE BEAD OR T1: CTX02-13665-X1 (SEE APPENDIX F FOR DETAILS)

100µH

1N4148

21.5k 1%

INDUCTOR (SEE APPENDIX F) COILCRAFT B-07T TYPICAL

Figure 35. Electronic Equivalent of 9V Battery Operates from Three NiCd Cells. Output Noise Is Below 100µV

OUT

AN70 F35

OPTIONAL FOR

()

100µH

LOWEST RIPPLE

C3 47µF

47µF

Low Quiescent Current Regulator

The LT1533 has a quiescent current of about 6mA. Figure 36’s circuit reduces this figure to 100µA by running an on-off control loop around the device. The control loop replaces the normal error amplifier, achieving regulation by switching the IC in and out of shutdown in accordance with loop demands.

Comparator C1 compares a scaled version of the output with its internal reference and biases the regulators shutdown pin. Loop hysteresis is obtained by utilizing the phase shift (e.g., time delay) of the output LC components. In a normal continuously closed loop this phase shift must be minimized and compensated. In this case it promotes the desired hysteretic control characteristic. Local AC positive feedback at C1 ensures clean transitions. Figure 37 shows the loop at work. When circuit output drops below the regulation point, C1’s output (Trace A) goes high. This enables the regulator and it responds with a burst of drive (Trace B) to the transformer. The output is restored and C1 goes low until the next cycle. During C1’s low time the regulator is shut down, resulting in the extremely low quiescent current noted. The loop’s on-off control characteristic causes low frequency output noise related to LC tank ring. Trace C shows 600µV peaks, although no wideband components are observable.

High Voltage Input Regulator

The LT1533’s IC process limits collector breakdown to 30V. A complicating factor is that the transformer swings to 2× supply. Thus, 15V represents the maximum allowable input supply. Many applications require higher voltage inputs and Figure 38 uses a cascoded14 output stage to achieve such high voltage capability. This 24V-to-50V converter is reminiscent of previous circuits, except that Q1 and Q2 appear. These devices, interposed between the IC and the transformer, constitute a cascoded high voltage stage. They provide voltage gain while isolating the IC from their large collector voltage savings.

Normally, high voltage cascodes are designed to simply supply voltage isolation. Cascoding the LT1533 presents special considerations because the transformers instantaneous voltage and current information must be accurately transmitted, albeit at lower amplitude, to the LT1533. If this is not done, the regulator’s slew control loops will

Note 14: The term “cascode,” derived from “cascade to cathode,” is applied to a configuration that places active devices in series. The benefit may be higher breakdown voltage, decreased input capacitance, bandwidth improvement, etc. Cascoding has been employed in op amps, power supplies, oscilloscopes and other areas to obtain performance enhancement. The origin of the term is clouded and the author will mail a magnum of champagne to the first reader correctly identifying the original author and publication.

AN70-13

Application Note 70

(3 Ni-Cd BATTERIES)

4.7µF

L1, L3: COILTRONICS CTX100-3

L2: 22nH TRACE INDUCTANCE, FERRITE BEAD OR T1: CTX02-13665-X1 (SEE APPENDIX F FOR DETAILS)

SHDN

DUTY

3300pF

SYNC

18k

0.01µF

INDUCTOR (SEE APPENDIX F) COILCRAFT B-07T TYPICAL

2.7V TO 4V

LT1533

GND

COL A

COL B

PGND

NFB

OPTIONAL

()

SEE TEXT

100µH

12V

C3 47µF

21.5k 1%

150k

2.32k 1%

47µF

1.18V V

INTERNAL

TO LTC1440

AN70 F36

1N4148

LTC1440

10pF

100µH

–

1N4148

15k

VSL

15k

CSL

Figure 36. Hysteretic “Burst ModeTM” Loop Lowers Quiescent Current to 100µA While Maintaining Low Output Noise

A = 5V/DIV

B = 10V/DIV

C = 500µV/DIV

(10kHz HIGH PASS)

1ms/DIV

AN70 F37

Figure 37. Operating Waveforms for the Low Quiescent Current Converter. Comparator Output (Trace A) Restores Output Voltage by Turning LT1533 On (Trace B). Output Noise Shows LC Ringing (Trace C), Although High Frequency Content Is Negligible

AN70-14

Burst Mode is a trademark of Linear Technology Corporation.

Application Note 70

not function, causing a dramatic output noise increase. The AC compensated resistor dividers associated with the Q1-Q2 base collector biasing serve this purpose. Q3 and associated components provide a stable DC termination for the dividers. Figure 39 shows waveforms for Q1’s operation (Q2 is identical, although of opposing phase). Trace A is Q1’s emitter, Trace B its base and Trace C the collector. T1’s ring-off obscures the fact that waveform fidelity is maintained through the cascode, although inspection reveals this to be the case. Additional testimony is given by circuit output noise (Trace D), which measures about 100µV peak.

24V-to-5V Low Noise Regulator

Figure 40 extends Figure 38’s cascoding technique in a step-down design.15 Inputs from 20V to 50V are converted to a 5V/2A capacity output. Q3 and Q4 protect the regulators VIN pin from the high input voltages. The cascode must accommodate 100V transformer swings. In this instance MOSFETs (Q1-Q2) are utilized, although the divider technique is necessarily retained. RC gate damper networks prevent transformer swings coupled via gatechannel capacitance from corrupting the cascode’s waveform transfer fidelity. Figure 41 shows that resultant cascode response is faithful, even with 100V swings. Trace A is Q1’s source, with Traces B and C its gate and drain, respectively. Under these conditions, at 2A output, noise is inside 400µV peak. Note that Q3 and Q4 protect the regulator from excessive input voltages.

10W, 5V to 12V Low Noise Regulator

Figure 42 boosts the regulator’s 1A output capability to over 5A. It does this with simple emitter followers (Q1Q2). Theoretically, the followers preserve T1’s voltage and current waveform information, permitting the LT1533’s slew control circuitry to function. In practice, the transistors must be relatively low beta types. At 3A collector current their beta of 20 sources ≈150mA via the Q1-Q2 base paths, adequate for proper slew loop operation.

The follower loss limits efficiency to about 68%. Higher input voltages minimize follower-induced loss, permitting efficiencies in the low 70% range.

Figure 43 shows noise performance. Ripple measures 4mV (Trace A) using a single LC section, with high frequency content just discernible. Adding the optional second LC section drops ripple below 100µV (Trace B), and high frequency content is seen (note ×50 vertical scale factor change) to be inside 180µV.

7500V Isolated Low Noise Supply

A final form of performance augmentation is extremely high voltage isolation. This is often required in situations where circuitry must withstand high common mode voltage effects. Figure 44 is similar to Figure 25’s isolated supply, except that it has 7500V (peak) breakdown capability. Transformer and optoisolator changes permit this. The remaining operating and performance characteristics are identical to Figure 25.

Note 15: This circuit was developed from a design by Jeff Witt of Linear Technology Corporation. Note 16: Operating the slew loops from follower base current was suggested by Bob Dobkin of Linear Technology Corporation.

AN70-15

Application Note 70

24V

10k

3300pF

18k

0.01µF

SHDN

DUTY

SYNC

1N4148

1N752

5.6V

1µF

GND

LT1533

24V

(20V TO 30V)

COL A

COL B

PGND

VSL

CSL

NFB

0.003µF

68Ω

360Ω

15k

2.49k 1%

3.3k

47µF

3.3k

68Ω

360Ω

L1: COILTRONICS CTX100-3 L2: 22nH TRACE INDUCTANCE, FERRITE BEAD OR

Q1, Q2: ZETEX ZTX-853

Q3: 2N2222A T1: CTX-02-13665-X1 (SEE APPENDIX F FOR DETAILS)

MUR-110

100µH

50V

OPTIONAL FOR

()

100µH

LOWEST RIPPLE

+ +

47µF

MUR-110

97.6k 1%

0.003µF

INDUCTOR (SEE APPENDIX F) COILCRAFT B-07T TYPICAL

AN70 F38

47µF

Figure 38. A 50V Output Low Noise Regulator. Cascoded Bipolar Transistors Accommodate 60V Transformer Swings, Permitting 24V (20VIN to 30VIN) Powered Operation

A = 5V/DIV

B = 5V/DIV

C = 20V/DIV

D = 100µV/DIV

1µs/DIV AN70 F39

Figure 39. Cascode Transmits Instantaneous Voltage and Slew Information, Permitting LT1533 to Maintain Low Noise Output. Trace A is Q1 Emitter, Trace B Is Its Base and Trace C the Collector. Transformer Ring-Off Obscures Cascode Action, but Study Reveals Faithful Transmission. Output (Trace D) Has 100µV Noise

AN70-16

Application Note 70

SYNC DUTY SHDN

PGND

NFB

VSL

CSL

LT1533

COL A

COL B

215

4 3

1500pF

0.002µF

17k

0.01µF

0.002µF

10µF

10k

100µH

OUT

7.5k 1%

2.49k 1%

220µF

MBRS140

L1, L3: COILTRONICS CTX100-3

L2: 22nH TRACE INDUCTANCE, FERRITE BEAD OR

INDUCTOR (SEE APPENDIX F) COILCRAFT B-07T TYPICAL

Q1, Q2: MTD6N15

T1: COILTRONICS VP4-0860

AN70 F40

100µF

OPTIONAL

SEE TEXT

()

10k

12k

GND

24V

(20V TO 50V)

Q3 MPSA42

Q4 2N2222

4.7µF

10k

220Ω

10k

220Ω

MBRS140

Figure 40. A Low Noise 24V-(20VIN to 50VIN)-to-5V Converter. Cascoded MOSFETs Withstand 100V Transformer Swings, Permitting LT1533 to Control 5V/2A Output

Figure 41. MOSFET-Based Cascode Permits Regulator to Control 100V Transformer

A = 20V/DIV

B = 5V/DIV

(AC COUPLED)

C = 100V/DIV

10µs/DIV

AN70 F41

Swings While Maintaining Low Noise 5V Output. Trace A Is Q1’s Source, Trace B Q1’s Gate and Trace C the Drain. Waveform Fidelity Through Cascode Permits Proper Slew Control Operation

AN70-17

Application Note 70

4.7µF 11

SHDN

DUTY

1500pF

18k

0.01µF

SYNC

LT1533

GND

COL A

COL B

PGND

NFB

10k

VSL

10k

CSL

2.49k 1%

1N4148

330Ω

0.003µF

680Ω

1N5817

300µH

AN70 F42

12V

()

33µH

LOWEST RIPPLE

+ +

100µF

OPTIONAL FOR

100µF

0.05Ω

4.7µF

21.5k 1%

Q1, Q2: MOTOROLA D45C1

0.05Ω 330Ω

1N4148

L1: COILTRONICS CTX300-4 L2: 22nH TRACE INDUCTANCE, FERRITE BEAD OR

INDUCTOR (SEE APPENDIX F) COILCRAFT B-07T TYPICAL

L3: COILTRONICS CTX33-4 T1: COILTRONICS CTX-02-13949-X1

: FERRONICS FERRITE BEAD 21-110J

Figure 42. A 10W Low Noise 5V-to-12V Converter. Q1-Q2 Provide 5A Output Capacity While Preserving LT1533’s Voltage/Current Slew Control. Efficiency Is 68%. Higher Input Voltages Minimize Follower Loss, Boosting Efficiency Above 71%

A = 5mV/DIV

B = 100µV/DIV

2µs/DIV

AN70 F41

Figure 43. Waveforms for Figure 42 at 10W Output. Trace A Shows Fundamental Ripple with Higher Frequency Residue Just Discernible. Optional LC Section Produces Trace B’s 180µV

Wideband Noise Performance

P-P

AN70-18

Application Note 70

4.7µF

GND

3300pF

4.2k

820Ω

L1: 22nH INDUCTOR. COILCRAFT B-07T, TRACE INDUCTANCE OR BEAD. SEE APPENDIX F L2: COILTRONICS CTX100-3 T1: COILTRONICS CTX-02-13950. SEE APPENDIX F

15k

13 12

VSL

LT1533

15k

CSL

18k

PGND V

Figure 44. A 7500V Isolation Version of Figure 25. Transformer and Optoisolator Are Changed to Achieve Isolation and Noise Immunity. Circuit Operation Is as Before

L1 22nH

10166

COL A

COL B

MOC-8112

4.7µF

1N4148

OUT

GND

OPTIONAL SECOND

0.22µF

– +

LT1431

100µH

2.5V FB

LC SECTION

(SEE TEXT)

9.5k 1%

2.5k 1%

TO OUTPUT

COMMON

47µF

AN70 F44

OUTPUT 12V

OUTPUT COMMON

Note: This Application Note was derived from a manuscript originally prepared for publication in EDN magazine.

AN70-19

Application Note 70

REFERENCES

1. Shakespeare, William, “Much Ado About Nothing,” II, i, 319, 1598-1600.

2. Williams, Jim, “Design DC/DC Converters to Catch Noise at the Source,” Electronic Design, October 15, 1981, page 229.

3. Williams, Jim, “Conversion Techniques Adapt Voltages to Your Needs,” EDN, November 10, 1982, page 155.

4. Tektronix, Inc. “Type 535 Operating and Service Manual,” CRT Circuit, 1954.

5. Tektronix, Inc. “Type 454 Operating and Service Manual,” CRT Circuit, 1967.

6. Tektronix, Inc. “7904 Oscilloscope Operating and Service Manual,” Converter-Rectifiers, 1972.

7. Hewlett-Packard Co. “1725A Oscilloscope Operating and Service Manual,” High Voltage Power Supply,

1980.

8. Arthur, Ken, “Power Supply Circuits,” High Voltage Power Supplies, Tektronix Concept Series, 1967.

9. Williams, Jim and Huffman, Brian, “Some Thoughts on DC/DC Converters,” Low Noise 5V to ±15V Converter and Ultralow Noise 5V to ±15V Converter, pages 1 to 5, Linear Technology Corporation Application Note 29, 1988.

10. Williams, Jim and Huffman, Brian, “Precise Converter Designs Enhance System Performance,” EDN, October 13, 1988, pages 175 to 185.

12. Williams, Jim, “High Speed Amplifier Techniques,” Linear Technology Corporation Application Note 47,

1991.

13. Witt, Jeff, “The LT1533 Heralds a New Class of Low Noise Switching Regulators,” Linear Technology, Vol. VII, No. 3, August 1997, Linear Technology Corporation.

14. Morrison, Ralph, “Noise and Other Interfering Signals,” John Wiley and Sons, 1992.

15. Morrison, Ralph, “Grounding and Shielding Techniques in Instrumentation,” Wiley-Interscience,

1986.

16. Sheehan, Dan, “Determine Noise of DC/DC Converters,” Electronic Design, September 27, 1973.

17. Hewlett-Packard Co. “HP-11941A Close Field Probe Operation Note,” 1987.

18. Terrien, Mark, “The HP-11940A Close Field Probe: Characteristics and Application to EMI Troubleshooting,” RF and Microwave Symposium, available from Hewlett-Packard Co.

19. Pressman, A.I., “Switching and Linear Power Supply, Power Converter Design,” Hayden Book Co., Hasbrouck Heights, New Jersey, 1977.

20. Chryssis, G., “High Frequency Switching Power Supplies, Theory and Design,” McGraw Hill, New York, 1984.

11. Tektronix, Inc. “Type 1A7A Differential Amplifier Instruction Manual,” Check Overall Noise Level Tangentially, pages 5-36 and 5-37, 1968.

AN70-20

APPENDIX A A HISTORY OF LOW NOISE DC/DC CONVERSION

Application Note 70

Why are batteries low noise power sources? Why do 60Hz AC power line derived linear regulators have low output noise? As with most innocent questions, thoughtful answers provide surprising insights. These sources have low output noise because they have low harmonic energy content. A 60Hz fundamental driven supply produces some harmonic activity, but power becomes very small well inside 1kHz. A battery is even better.

These conclusions set a direction towards designing low noise DC/DC converters. If the goal is low noise, the key is reduction of harmonic energy, in particular, wideband harmonics. This simple guideline is central to LT1533 operation, although refinements are necessary for a generally applicable IC.

History

The notion of minimizing harmonics in DC/DC conversion to get low output noise is not new. Oscilloscopes have used this technique to generate high voltage CRT accelerating potentials without degrading instrument operation.

Designing a 10,000V output DC/DC converter that does not disrupt a 500MHz, high sensitivity vertical amplifier is challenging.

Figure A1 shows the CRT DC/DC converter from a Tektronix 454 oscilloscope. Q1430, configured as a modified Hartley power oscillator, drives T1430. T1430’s output is multiplied by the diode-capacitor tripler, producing 12,000V. Feedback to Q1414 is summed against a 75V derived reference, closing a regulation loop around the power oscillator.

The sine wave transformer drive (see waveforms in the figure) has low harmonic content, resulting in the desired low conducted and radiated noise. This approach is not very efficient—Q1430 operates in its linear region—but the power loss is acceptable in a 125W instrument.

Tektronix 7000 series oscilloscopes used a resonant, offline converter to power the entire instrument. As before, CRT high voltage was generated separately (see Footnote

1). Figure A2, a partial schematic of a Tektronix 7904 power converter, shows a series resonant network, L1237C1237 in the Q1234-Q1241 drive path. This results in sine wave drive to output transformer T1310, despite Q1234Q1241’s rectagular waveshape. Feedback (not shown) closes a loop around this stage, stabilizing its operating point. The resonant, sine wave transformer drive provides the desired low noise characteristics with good efficiency.

A less specific example appears in LTC Application Note

29. Figure A3, a partial schematic of Application Note 29’s Figure 4, shows a sine wave oscillator (A1 based) driving a power amplifier (A3 and Q2 to Q6). L3, the output transformer, provides voltage boosted secondary drive to linear regulators (not shown). This brute force approach provides a converter with extraordinarily low noise, but is complex and inefficient. Q4 and Q5, operating in their linear regions, dissipate considerable power, and efficiency is 30%.

Figure A4’s approach, also from AN29’s Figure 1, achieves better efficiency. The partial schematic shows source followers driven from 100Ω-0.003µF edge slow-down networks. This slows down the transistor’s transitions, resulting in harmonic reduction and low noise. Unfortunately, the drive scheme is complex and somewhat inflexible, requiring bootstrapped voltages to fully switch the transistors on and off. Additionally, a transformer change would require drive rework to maintain efficiency and low noise characteristics. Finally, the dynamic voltage and current control in the transistors is passively determined and not very well controlled.

The LT1533 uses closed-loop control2 around its output stages to tightly control voltage and current slewing. This allows a variety of circuits and magnetics to be easily accommodated, resulting in a true general purpose solution. Text Figure 3 and the associated discussion provide more LT1533 operating details.

Note 1: Ancillary benefits include eliminating a complex and expensive high voltage winding in the main power transformer, avoidance of long, high voltage wire runs and space and weight savings.

Note 2: Patent pending.

AN70-21

Application Note 70

Reproduced by permission

AN70-22

Figure A1. Tektronix 454 CRT Circuit Uses Sine Wave Drive for Low Noise DC/DC Conversion.

Efficiency Is Poor, Because Q1430 Remains in Linear Region

+ 50 hidden pages

Noty an70 Linear Technology

Specifications and Main Features

Frequently Asked Questions

User Manual