A linear regulated power supply drops excess voltage as heat to achieve ultra-low output noise (typically <10 µV RMS), making it the mandatory choice for precision analog circuits, Battery Management System (BMS) calibration, and sensitive RF loads where switching ripple is unacceptable. While switching power supplies dominate high-efficiency applications, a linear regulated power supply remains the undisputed king of signal integrity on the bench.
Why Choose a Linear Regulated Power Supply Over Switching?
When designing power for a 48V battery emulator or a high-precision coulomb counter, the choice between a linear regulated power supply and a Switch-Mode Power Supply (SMPS) dictates your noise floor. An SMPS chops DC at high frequencies (100 kHz to 2 MHz), creating switching spikes that can alias into your ADC readings. A linear regulator operates in the active region, acting as a smart, variable resistor to drop voltage without high-frequency chopping.
Here is how the two topologies compare across the metrics that matter on the bench:
| Metric | Linear Regulated Power Supply | Switch-Mode Power Supply (SMPS) |
|---|---|---|
| Efficiency | Low ($V_{out}/V_{in}$). A 5V output from 12V input is ~41%. | High. Typically 85% to 95% across a wide load range. |
| Heat Dissipation | High. Excess power is burned as heat ($P_d = (V_{in}-V_{out}) \times I$). | Low. Requires only minimal heatsinking for high-current designs. |
| Output Noise | Ultra-low. Typically <40 µV RMS broadband, no switching spikes. | High. 10 mV to 50 mV peak-to-peak ripple, high-frequency EMI. |
| Cost & BOM | Low component count, but heavy/expensive low-frequency transformers. | Higher component count (inductors, MOSFETs), but cheap lightweight magnetics. |
| Transient Response | Excellent. Reacts instantly to load steps (limited only by output cap). | Slower. Requires multiple switching cycles to correct output voltage. |
The Verdict: Choose a linear regulated power supply when your load draws under 2A and requires a pristine voltage rail for analog-to-digital conversion or precision sensing. Choose an SMPS when efficiency, size, or high current (>3A) are the primary constraints.
Design Example: 5V/1A Low-Noise Supply for BMS Calibration
Let’s design a 5.0V, 1.0A linear regulated power supply to power the analog front-end of a BMS calibration jig. We need ultra-low noise to prevent jitter in the cell voltage measurements. For this, we will use the Analog Devices LT3080, a modern 1.1A low-dropout (LDO) regulator capable of parallel operation and programmable output down to 0V.
Input and Output Specifications
- AC Input: 120VAC nominal, 60Hz
- Transformer: 9VAC, 1.5A secondary (e.g., Triad Magnetics F-281P, ~$12)
- Unregulated DC: ~11.5VDC after bridge rectifier and bulk capacitor
- Target Output: 5.0VDC at 1.0A maximum
Dropout and Headroom Math
A common mistake in linear design is ignoring the dropout voltage ($V_{do}$). The LT3080 has a maximum dropout of 0.5V at 1A over temperature. Therefore, the minimum input voltage at the regulator pin must be:
$V_{in(min)} = V_{out} + V_{do} = 5.0V + 0.5V = 5.5V$
Our rectified 11.5VDC provides 6.0V of headroom, keeping the regulator safely out of dropout even if the AC mains sags by 10%. However, this large headroom comes at a thermal cost, which we must calculate next.
Setting the Output Voltage
The LT3080 uses a precision internal current source of 100 µA that flows through a single external resistor ($R_{set}$) to ground. The output voltage is set by:
$R_{set} = V_{out} / 100\mu A$
For 5.0V, $R_{set} = 5.0 / 0.0001 = 50,000\Omega$. Use a 49.9 kΩ, 1% tolerance metal film resistor for exact calibration. Bypass the SET pin with a 0.01 µF ceramic capacitor to ground to reduce resistor thermal noise.
Thermal Management and Derating Realities
Linear regulators burn excess voltage as heat. You cannot skip the thermal math, or the part will hit its internal thermal shutdown (typically 165°C) under full load.
Power Dissipation Calculation:
$P_d = (V_{in} - V_{out}) \times I_{out} = (11.5V - 5.0V) \times 1.0A = 6.5W$
The LT3080 in a TO-220 package has a junction-to-ambient thermal resistance ($\theta_{JA}$) of roughly 50°C/W without a heatsink. A 6.5W dissipation would cause a 325°C temperature rise, instantly destroying the silicon. We must add a heatsink.
Heatsink Sizing
Assume a maximum ambient temperature ($T_a$) of 35°C (inside an enclosed test chassis) and a target maximum junction temperature ($T_j$) of 110°C for reliability.
Max allowable $\theta_{JA} = (T_j - T_a) / P_d = (110 - 35) / 6.5 = 11.5°C/W$
We subtract the junction-to-case ($\theta_{JC} \approx 3.0°C/W$) and case-to-sink ($\theta_{CS} \approx 1.0°C/W$ with thermal paste) resistances:
Required Heatsink $\theta_{SA} = 11.5 - 3.0 - 1.0 = 7.5°C/W$
Select an extruded aluminum heatsink rated for ≤ 7.5°C/W, such as the Aavid Thermalloy 530704B02500G (approx. $6.50). If your enclosure lacks ventilation, derate the heatsink by 20% to account for localized ambient heating from the transformer.
Input Protection and Ripple Expectations
A properly designed linear power supply must survive mains transients and load faults. The unregulated DC bus is vulnerable to voltage spikes, and the regulator output needs protection against reverse currents from battery loads.
Input Protection Network
- Primary Fuse: 0.5A slow-blo fuse on the 120VAC primary to handle transformer inrush current.
- Transient Suppression: An SMBJ15A TVS diode across the unregulated DC bus. If a mains surge pushes the DC bus above 15V, the TVS clamps it, protecting the LT3080 (which has an absolute maximum input rating of 36V).
- Reverse Battery Protection: When powering a battery emulator or charging a cell, the load can push voltage back into the regulator. Place a 1N5819 Schottky diode in reverse bias from the LT3080 output to its input. If the output is forced higher than the input, the diode conducts, preventing internal junction breakdown.
Ripple and Noise Expectations
After a full-wave bridge rectifier and a 4,700 µF bulk capacitor, the 11.5VDC unregulated rail will have approximately 1.0V peak-to-peak of 120Hz ripple.
The LT3080 boasts a Power Supply Rejection Ratio (PSRR) of 75 dB at 120 Hz. A 75 dB attenuation means the input ripple is reduced by a factor of 5,623.
Output Ripple = $1.0V_{pp} / 5623 \approx 177 \mu V_{pp}$ (0.18 mV)
Combined with the broadband noise floor of ~40 µV RMS, this linear regulated power supply delivers a pristine 5.0V rail, easily meeting the <1 mV ripple requirement for 16-bit ADCs used in precision battery testing.
Frequently Asked Questions
Can I use a linear regulated power supply to charge lithium batteries directly?
No. A linear regulator only provides Constant Voltage (CV). Lithium-ion and LiFePO4 cells require a Constant Current / Constant Voltage (CC/CV) charge profile. If you connect a deeply discharged battery to a 4.2V linear regulator, the regulator will attempt to source massive current to bring the battery voltage up instantly, likely exceeding the regulator's current limit or triggering thermal shutdown. You must use a dedicated charge controller IC (like the MCP73831) or a buck converter with CC/CV feedback loops downstream of your linear supply.
What is the maximum input voltage for a standard linear regulator?
It depends on the silicon. The classic LM317 has an absolute maximum input-to-output differential of 40V, but its maximum operating input is often limited by power dissipation. The LT3080 has an absolute maximum input rating of 36V. Always consult the specific datasheet. More importantly, design for transients: a 24VAC transformer can produce over 40VDC unregulated under no-load conditions, which will instantly punch through the oxide layer of a 36V-rated regulator. Always use a TVS diode or bleeder resistor to manage no-load voltage rise.
How do I reduce the heat dissipation in a high-dropout linear design?
If you need to drop 24V down to 5V at 2A, a linear regulator will dissipate 38W—requiring a massive, fan-cooled heatsink. To solve this while maintaining low noise, use a switching pre-regulator. A circuit like the LM3485 (a hysteretic PFET buck controller) can step the 24V down to 6.5V efficiently. You then feed that 6.5V into your linear regulator to drop to 5.0V. The linear regulator only dissipates $(6.5V - 5.0V) \times 2A = 3W$, and it filters out the switching noise from the buck stage, giving you the best of both topologies.
Why is my linear regulator oscillating or ringing on the output?
Oscillation in linear regulators is almost always an output capacitor Equivalent Series Resistance (ESR) issue. Older regulators like the LM317 or L200 were designed when electrolytic capacitors were standard; they require a specific ESR range (typically 1Ω to 5Ω) to maintain loop stability. If you use modern, ultra-low-ESR ceramic capacitors with these older parts, the control loop will ring or oscillate. Modern LDOs like the LT3080 or TI's LP5907 are internally compensated and specifically designed to be stable with low-ESR ceramic capacitors (e.g., 10 µF X7R). Always check the "Output Capacitor Requirements" table in the datasheet before selecting your dielectric.
