High-Frequency DC-DC Conversion: Why Efficiency and Low Ripple Matter in ESS Cabinets
Sep 21, 2026
P
Peter Lu
Energy Storage Product Manager, GreenMore · Energy storage systems
Inside every modern cabinet battery converter, transistors switch thousands of times per second, chopping one DC voltage and reshaping it into another. The GreenMore GM-LDC30 and GM-LDC60 use industrial-grade high-frequency conversion to reach peak efficiencies of 98.0% and 98.6%. That sounds like a small gap. Across a 60 kW power stage running day and night, it is not.
This is the second article in our four-part DC power guide. Part 1 covered how to choose a DC power module for an energy storage cabinet. Here we look under the hood at the two numbers engineers argue about most: conversion efficiency and output ripple.
How Does High-Frequency DC-DC Conversion Work, and Why Is It Efficient?
A switch-mode DC-DC converter does not burn off excess voltage the way a linear regulator does. It stores energy briefly in an inductor and transfers it in packets timed by the switch. The ratio of on-time to off-time sets the output voltage. Because the transistor is either fully on or fully off, the average dissipation in the switch stays low.
Raise the switching frequency and the passive components shrink. Texas Instruments puts it directly: output inductance and output capacitance are both inversely proportional to switching frequency. Double the frequency and you need roughly half the inductance, so the magnetic core gets smaller and lighter. That is the core reason a 60 kW converter fits in a rack-mount enclosure today while needing a cabinet of its own a generation ago.
Efficiency is never free, though. At high frequency the loss budget shifts between three main components:
Loss type
What it is
How it changes
Conduction loss
I²R heating in switch on-resistance, inductor windings and busbars
Independent of frequency; grows with load
Switching loss
Energy lost while voltage and current overlap during each turn-on and turn-off
Increases directly with frequency
Core and drive loss
Magnetic hysteresis in the core, plus the energy needed to charge the gate each cycle
Increases with frequency
Resonant and phase-shifted control schemes soften the voltage-current overlap by timing the switch transition to a zero crossing, cutting the switching term. The designer's job is to pick the frequency, topology and silicon that balance these three. At heavy load, conduction loss dominates, so low on-resistance devices win. At light load, switching and core losses dominate, which is why the module draws under 20 W in standby rather than idling at full magnetizing current.
What Do 98.0% and 98.6% Efficiency Actually Buy You?
Efficiency converts directly into heat, which is what makes a high-efficiency DC power supply module worth its price. Take each module at its rated output:
Module
Rated output
Peak efficiency
Power that becomes heat
GM-LDC30
30 kW
98.0%
about 600 W
GM-LDC60
60 kW
98.6%
about 840 W
Both figures are stated at the favourable operating point; real efficiency varies with voltage ratio and load. Even so, the relationship is fixed. Every 0.1% of efficiency lost at 60 kW adds 60 W of heat, and that heat has to leave the module or it raises component temperatures. Intelligent air cooling handles it across the full −30 to +55 °C operating range, but hotter capacitors and semiconductors age faster. TDK's capacitor handbook states it plainly: thermal stress has a decisive effect on life expectancy, and the heat generated by ripple current is an important factor in useful life.
The energy meter notices too. At a common 60 kW output, the gap between 98.0% and 98.6% efficiency is about 360 W of dissipation, or roughly 3,150 kWh per module per year if it ran continuously. Few cabinets run flat out all year, but sites with long cycling duty see the number on their electricity bill.
That is also why the standby figure matters. Under 20 W keeps parasitic losses negligible when the cabinet is parked, instead of paying for a converter that wastes more in a day idle than it should.
Why Low Ripple Matters for Batteries and Downstream Equipment
Ripple is the periodic AC component riding on the DC output, at the switching frequency and its harmonics. It is distinct from random high-frequency noise, though the two get measured together on a scope. The output filter exists to deliver a low-ripple DC output that every load on the bus can trust.
What ripple does to a battery
Batteries are the first victim people cite. A Solis white paper drawing on 16 kW energy storage inverter testing reports that 6 A peak-to-peak ripple at around 300 Hz accelerated capacity fade by up to 15% compared with smooth DC cycling, driven by localized heating and uneven lithium intercalation.
The effect is not universally settled. Long-term academic studies reach different conclusions, and the SiCWell aging study published in Batteries found the impact scaled with ripple frequency and amplitude rather than following a single rule. The consistent finding across the literature is that low-frequency, high-amplitude ripple does the damage, while high-frequency ripple is largely absorbed by the cell's double-layer capacitance. No cabinet designer should gamble a battery warranty on that distinction when the conversion stage can simply be clean.
What ripple does to everything else on the bus
Ideal output against a rippled one, and why the filter between them decides what your loads actually see.
Sensitive loads share the same bus as the battery:
Load
How ripple hurts it
Telecom rectifiers and radios
Raises the noise floor and can push transmit signal quality out of spec
Sensors and analog front ends
Small measurement offsets and jitter at the switching frequency
Automation controllers and PLCs
Logic brownouts and communication errors when dips cross thresholds
DC-link capacitors
Extra I²R heating shortens service life, the classic wear-out path
Ripple is a steady-state power-quality problem, not a fault event. Protection such as OVP, OCP, SCP and OTP handles abnormal conditions, which is a separate subject.
How Should Output Ripple Be Measured?
A sloppy probe setup measures the probe's own antenna loop instead of the converter. Industry practice converges on a few rules. Intel's ATX 3.0 design guide defines ripple and noise as periodic or random signals over a band of 10 Hz to 20 MHz, and requires the oscilloscope bandwidth to be set at 20 MHz, which keeps wideband switching spikes out of the periodic ripple number.
On the bench:
Put the scope in AC coupling and engage the 20 MHz bandwidth limit for the ripple reading.
Use a 1:1 probe or a dedicated power-rail probe for millivolt sensitivity. A 10:1 probe throws away the vertical resolution you need.
Connect the probe tip directly across the output capacitor and use the short spring ground clip, not the dangling ground lead. A long ground lead is an antenna.
Measure both peak-to-peak and RMS, at no load and at full load.
When you compare module datasheets, check the conditions first. A ripple number without a bandwidth limit and load point tells you nothing.
Where the GM-LDC Modules Fit
The GM-LDC30 and GM-LDC60 are rack-mount DC-DC modules rated 30 kW and 60 kW. The low-voltage side operates from 200 to 900 V and the high-voltage side from 300 to 1,000 V, with maximum output current of 75 A on the 30 kW unit and 150 A on the 60 kW unit. Intelligent air cooling, IP20 ingress protection, RS485 communication with MODBUS-RTU, reverse-polarity and surge protection, insulation monitoring and a three-year warranty come standard, and modules support parallel operation for larger power stages.
They sit inside air-cooled and liquid-cooled cabinet systems as well as utility-scale containerized ESS. When your cabinet needs more than 60 kW, our next article covers parallel expansion and current sharing. The GreenMore team has worked in solar and power electronics since 2017.
Frequently Asked Questions
Does a higher switching frequency always mean better efficiency?
No. Higher frequency shrinks the inductor and capacitor, but switching loss rises roughly in proportion to frequency. There is an optimum for a given topology and load profile, and that optimum shifts with load. A converter tuned for full load is not the same converter you want idling at 5% output.
What ripple level is acceptable for a lithium battery?
There is no universal number, because it depends on frequency, cell chemistry and how the pack is cycled. The published evidence points one way: low-frequency ripple at high amplitude is the damaging case, and the Solis testing cited above saw up to 15% faster fade at 6 A peak-to-peak near 300 Hz. Check the battery supplier's datasheet for a stated limit and treat anything above it as a design constraint, not a rounding error.
Why does 0.6% of efficiency translate into hundreds of watts?
Because the base is large. At 60 kW, one percentage point is 600 W, so 0.6% is 360 W. That is a soldering iron's worth of heat running continuously inside a sealed enclosure, and it has to be moved out by the cooling system or absorbed by the components.
Can I compare ripple figures across datasheets directly?
Only if the measurement conditions match. A ripple number is meaningless without the bandwidth limit, the load point and the input voltage at which it was taken. A 200 mV figure measured at 20 MHz bandwidth and full load is not comparable to a 50 mV figure measured at 1 MHz and half load.
Talk to Us About Your DC Stage
If you are sizing the DC stage for a cabinet or container project and want help matching efficiency and ripple targets to your battery and load, talk to our engineering team. Tell us the battery voltage window, the load profile and the ambient conditions, and we will come back with a module count and a cooling check rather than a catalogue page.
P
Peter Lu
Energy Storage Product Manager, GreenMore
Peter helps integrators set realistic efficiency and ripple targets for cabinet and container projects. He reviews the operating point, thermal design and battery ripple sensitivity before a module is specified, and works with the engineering team to confirm parallel and protection configuration so the DC stage stays clean and predictable through years of cycling duty.
Updated on November 12, 2026