Over the years, many people have asked me if it’s better to use one big power supply and wire all the devices under test (DUTs) in parallel when testing multiple DUTs or if it’s better to use one small power supply for each DUT.
Almost always, my answer is that one power supply per DUT is ultimately better than one large, common, shared power supply. I make this recommendation even if, at first inspection, it looks more expensive to buy more, smaller supplies. Why? Isolation between DUTs.
Individual Power Supplies
When each DUT has its own power supply, anything that happens within one DUT (let’s call it DUT x) doesn’t affect what happens within another DUT (let’s call it DUT y) when it comes to events such as inrush or DUT failure (Fig. 1). Now, let’s compare that to powering multiple DUTs from one large shared power supply (Fig. 2).
Should a problem on DUT x disturb the output of the large shared supply, the test process running DUT y will have no way to know that DUT x invalidated the power stability, unless every test process also is forced to monitor the supply voltage at high speed in real time, watching for disturbances.
Note that this kind of event is not limited to a failure. Even the simple asynchronous act of connecting and powering up DUT x can cause an inrush current into DUT x that momentarily disturbs the common dc power rail going to all devices.
Having one power supply per DUT simplifies safety shutdown. A failure of a DUT can be detected, its power supply shut down, and the testing can continue for the remaining DUTs.
Also, having one power supply per DUT means that you can use the built-in current measurement functions of the individual power supplies. If you have a large shared power supply, any current measurement will be for the total current consumption of all DUTs.
A Shared Supply
If you need to measure the individual currents being supplied by a large shared power supply, you will need to use a current sensor (typically a current sense resistor or current shunt) on each individual DUT line. Then you will need to use a scanner to route the current sense resistor voltages to a system digital multimeter (DMM), adding wiring complexity, calibration of the shunts, and additional programming.
One power supply per DUT means each individual power supply is smaller and therefore able to make more accurate measurements. If you decided to use the built-in current measurements of the large common power supply to detect the changes in total current due to an individual DUT, consider this: With a large power supply, you may not be able to see variations in any one DUT because the variation is smaller than the accuracy of the supply.
For example, powering eight DUTs from one large power supply means the large supply will be sourcing eight times the current of one DUT. If you rely on the measurement accuracy of the large shared power supply to achieve the same measurement performance of an individual power supply, it means you need 3 bits more resolution on the large shared supply than you do on the individual, smaller power supplies.
If you determined that a 14-bit power supply (which is commonly available and reasonably priced) was good enough to measure the current when powering an individual DUT, when you connect eight DUTs in parallel, you’ll now need a 17-bit power supply (and there are virtually none on the market over 16 bits) to achieve the same measurements. High-power, high-accuracy supplies are expensive—or possibly even unavailable.
When using one large shared power supply, if the test process involves changing the voltage across an individual DUT, including something simple like powering the DUT on or off, then this will not be possible, as all DUTs would share a common voltage rail.
The Better Choice
Should your test process require galvanic disconnect, this is more easily achieved with one power supply per DUT, because smaller power supplies are sometimes available with built-in relays. On the contrary, implementing galvanic disconnect with a large shared power supply will mean a custom designed power relay setup with a relay for each of the many DUTs that could be connected in parallel.
Using one power supply per DUT is more fault tolerant, as one power-supply failure does not stop the testing on the other DUTs. It is possible to take that one power supply offline for repair or calibration, and the remaining channels can continue to operate. Should your large shared power supply fail or need calibration, the entire test system is down.
There are many advantages to implanting parallel testing with multiple smaller power supplies. It is likely that multiple smaller supplies may be more expensive. But when you factor in the costs of engineering the “glue” functions of monitoring and disconnect per DUT, the savings may be a false economy. Furthermore, the additional complexity of developing the software to manage the system based on a large shared power supply could easily outweigh any savings in initial power-supply hardware costs.