Choosing a programmable DC power supply for battery and PV testing
Constant voltage against constant current, what sinking and regenerating actually change, curve emulation for inverter testing, rack and paralleling decisions, interfaces, and the supply-side requirements the building has to meet.
Two numbers go into most power supply tenders: maximum voltage and maximum current. Both are usually met, and the supply still turns out to be the wrong one. It cannot absorb current from the battery it was bought to discharge, or its interface will not do what the test script needs, or it turns several kilowatts into room heat in a laboratory whose air conditioning was sized for people and computers.
Constant voltage and constant current are one supply in two states
A programmable supply holds whichever of its two limits it reaches first. Set 30 V and 2 A into a load drawing half an amp and it holds the voltage. Lower the load resistance and at some point it reaches the current limit, drops the voltage and holds current instead. The crossover is not a fault, it is the design, and using it deliberately is the difference between a bench instrument and a box that makes volts.
When powering anything new, set the current limit first, at a value the prototype cannot survive exceeding, then raise the voltage. A deliberately set current limit is the cheapest protection circuit in the laboratory and it has saved more student projects than any amount of careful soldering.
Past the two headline numbers, these are the specifications that decide whether the supply suits the work.
Ripple and noise, quoted separately as rms and peak to peak. For driving a laser diode, a detector preamplifier or any sensitive analogue circuit, this matters more than the current rating.
Programming resolution against readback accuracy. A supply can be settable in millivolts and still read back with an error hundreds of times larger. If you intend to use the supply as a measuring instrument, readback accuracy is the specification that counts.
Remote sense. A four wire connection at the load removes the cable drop, which on a high current bench is not a rounding error. Ask how the model behaves if the sense leads are left disconnected, because on some designs that is a runaway condition.
Slew rate and settling time. How fast the output moves between setpoints decides how quickly a test sequence runs, and whether a pulsed battery test is possible at all.
Output isolation, and behaviour with the output switched off. A floating output matters the moment you put two supplies in series or connect to an earthed device under test.
Stability into reactive loads. A supply that is well behaved into a resistor can oscillate into a long cable and a large capacitor. If your load is a capacitor bank or a long run to another room, say so before you buy.
Sourcing, sinking and giving the energy back
A conventional supply can only push current out. The moment a test needs to take current in, the equipment class changes and so does the price. There are three ways to arrange it.
A unidirectional supply plus a separate electronic load. Two boxes, two interfaces, a hardware changeover in the middle of the test, and every joule of discharge energy converted to heat.
A bidirectional, two quadrant supply. One box, one interface, one continuous setpoint that passes through zero current. This is what makes a clean, uninterrupted charge and discharge profile possible.
A regenerative bidirectional supply. The absorbed energy is returned to the mains rather than burnt in a resistor bank.
Regeneration is normally sold as an energy saving, and it is, but in Dhaka the stronger argument is thermal. Every kilowatt a non-regenerative load absorbs becomes a kilowatt of room heat, and the air conditioning then spends more electricity taking it out again. A bench cycling battery packs for eight hours a day changes the cooling load of the room it stands in. Size the room cooling and the electrical supply together with the test equipment, not as a variation order afterwards.
The Delta Elektronika range runs from benchtop units up to rack mounted multi kilowatt systems, with bidirectional models in the higher power series. Ask your supplier to confirm in writing which specific model sinks current and which does not, because the family name does not tell you and a single series page often covers several behaviours.
Emulating a source rather than simply powering a load
PV emulation. The supply reproduces the current voltage curve of an array, including the knee and the effect of irradiance and temperature, so that an inverter or a charge controller sees a realistic source. This is how you test a maximum power point tracker repeatably. It is a different job from a solar simulator, which illuminates a real module in order to measure the module.
Battery emulation. The supply behaves like a cell or a pack at a chosen state of charge and internal resistance, so a battery management system or a motor controller can be exercised without waiting hours for a real pack to reach the right condition.
Sequencing and list mode. A profile stored in the supply and run from its own clock is more repeatable than one sent step by step over the network, and it keeps running when the laboratory PC decides to install updates.
Programmable output impedance and slew limiting, for reproducing a weak supply, a long cable run or a deliberately poor source.
Racks, paralleling and what the building has to provide
Master and slave operation. Several units can act as one larger supply with current sharing. Check that this is a designed function with its own interface module rather than a wiring arrangement, because the difference shows up in how the group behaves during a fault.
Parallel for current, series for voltage, and check the permitted output to earth isolation voltage before you series anything.
Rack adapters and cabinet systems keep cabling short and airflow predictable. Airflow direction is a real design item: front to back cooling in a full cabinet pushed against a wall in a hot room will throttle itself by the afternoon.
The input side is where laboratory installations get caught. A multi kilowatt bench needs a dedicated three phase feed, correctly rated protection and an earth that has been measured rather than assumed. Ask for the inrush current and the power factor before the electrical contractor sizes the circuit, not after.
Decide now whether the bench must ride through a generator changeover. Some tests can be resumed from where they stopped and some cannot. A cycling test that halts halfway is a wasted week of bench time, and that requirement changes what you buy on the supply side as well.
Interfaces, and deciding who writes the test script
The interface question is really a staffing question. Settle before the purchase order who is going to automate this bench, and in what.
Isolated analogue programming, for control from another instrument or from a safety system that has to work whether or not any software is running.
Serial and Ethernet interfaces. Ethernet is normally the right default in a shared laboratory because it survives the PC being replaced, and because it lets a long run be monitored from an office.
Driver support in the environment your group actually uses. If the work is done in Python, confirm there is a driver or a documented command set before purchase. Finding out after delivery that the only supported environment needs a licence nobody in the department holds is an expensive discovery.
Interlock and inhibit inputs, so that an emergency stop or a chamber door cuts the output in hardware rather than through a software poll.
Isolated measurement amplifiers, where you need to measure a voltage that floats with respect to earth.
What actually destroys these supplies
A charged battery connected with reversed polarity. Fit a fuse and a blocking diode or a polarity protected connector on the battery side and treat it as mandatory rather than optional.
Inductive loads switched off without a freewheel path. A motor or a solenoid returns its stored energy to the output.
Sense leads reversed or left floating when the supply is configured for remote sense.
Ambient temperature above the derating point, or an air filter nobody has cleaned since commissioning. In a dusty laboratory this is the most common cause of early failure in any forced air instrument, and it is entirely preventable with a named person and a date.
Transients on the mains input during generator changeover, which show up months later as a failed control board rather than as an obvious event on the day.
Application
The specification that decides the purchase
Battery cell and pack cycling
Ability to sink current, transition speed between setpoints, and safe behaviour when a charged pack is connected
PV inverter and MPPT testing
Curve emulation with a realistic knee, dynamic response, and enough power headroom for the inverter under test
Laser diode and detector work
Ripple and noise, precision of the current limit, soft start behaviour at switch on
Electrochemistry, electrolysis and plating
Stability in constant current over long unattended runs, and corrosion resistant connections
Motor drive and regenerative load testing
Bidirectional operation with energy return, cabinet cooling, three phase supply capacity
General teaching bench
Isolated floating output, legible front panel, and over voltage protection students cannot defeat