Clifton Laboratories 7236 Clifton Road  Clifton VA 20124 tel: (703) 830 0368 fax: (703) 830 0711

E-mail: Jack.Smith@cliftonlaboratories.com
 

Revision History
07 August 2009. Original

I recently had occasion to look at the output of several inexpensive "wall wart" power supply modules. The results were interesting and worth sharing. I won't identify the particular modules because who knows how many other names they are sold under and because it's unlikely that anyone else owns this particular combination of modules.

I'll call these devices "modules' on this page because "wall wart" seems an infelicitous name for something that serves a useful purpose, although the term has a certain applicability.

This exercise came about because some months ago I had noticed one particular 9V 650 mA switching design module seemed to radiate more noise than others I had used. At the time, I looked at the radiated noise and also the amount of switching frequency energy coupled back into the AC power mains.

I bought a substitute linear regulated module, but recently I was presented with a  replacement switching module for  the original unit. The replacement is said to be quieter and of higher quality.

I'm not really equipped to make radiated noise measurements, nor do I have the necessary line impedance stabilization  network for conducted noise measurements, so I decided to compare the old and new units by simply looking at their DC output with an HP 3562A Dynamic Signal Analyzer, in essence a 100 KHz spectrum analyzer with excellent noise figure and sensitivity.

I know that noise on the DC output is not the same as radiated noise, of course.

While I was at it, I also looked at several other modules and four laboratory variable voltage power supplies with the setup illustrated in the block diagram below.

The particular equipment powered by the 9V module required almost exactly the current drawn by a 50 ohm resistor, so I used a 100 watt 50 ohm dummy load as the test load. The 3562A has a high impedance input with switchable AC or DC input, so no further isolation or protection is  required. (I also looked at higher frequency noise using an Advantest R3463 spectrum analyzer. To protect it from DC voltage, I used a 1uF blocking capacitor.)

The devices tested are:

• Replacement "quieter" 9V digital module
• A "CE" marked 12V switched module
• A 12V analog module I bought to replace the noisy 9V switcher
• The 3562A noise floor, with a 50 ohm termination on the input.

Click on the image below for the full size plot.

That a device complies with radiated emission rules is, of course, not a guarantee that the DC output will have any particular purity. However, my experience is that in order to meet the radiated emission rules some attention has to be paid to the quality of the DC output as it is a secondary source of radiated emissions.

Linear power supplies without digital electronics are not required to have FCC certification.

The plot has several surprises:

• The linear supply, which one expects to be quiet, with perhaps some broadband noise, has serious problems. It has both grossly excessive levels of broadband noise, as well as discrete 49 KHz and 98 KHz spectral components (98 KHz being the second harmonic of the 49 KHz signal, I'm sure.) One of two possibilities strike me. First, the supply has an oscillating regulator, due to component failure or poor design, or both. Second, the so called "linear" power supply may have a step down transformer but it uses a switching regulator to stabilize the 12V output. Since the module has a sealed case, I can't disassemble it to resolve these possibilities.
• The plot shows only the new 9V switcher as there's no significant difference between it and the original module, which leads me to believe there is no design change to reduce the noise.
• The CE marked switching supply is markedly quieter in terms of broadband noise than any of the other modules and indeed comes close to the HP 3562A's noise floor.

Whether the cleanliness of the DC output of these supplies is suitable for the equipment attached to it is an interesting question. It almost certainly says that when designing equipment to be powered from any random external power module, one had best consider and deal with potentially high levels of broadband noise and discrete spectral line signals. One might argue that in the worst case switcher, the discrete spectral line is still only around 1 mV in amplitude, a point I cannot disagree with. 

I also looked at the DC output with an Advantest R3463 spectrum analyzer from 0-20 MHz with the results below. Above 8 MHz or so, the spectrum analyzer shows all supplies as equal. In fact, the signals are short wave broadcast signals picked up on the AC wiring and DC power cabling.

I have not labeled the plot as neatly, but the blue and magenta traces are the two 9V switching supplies, one original and one replacement. There's about 10 dB difference in the noise level between 5 and 7 MHz, but it's not clear to me if this is real or an artifact. The  red  trace is the 12V CE marked switching supply. The gold trace is the analog supply and the green trace is the spectrum analyzer's noise floor with no input.

Note that  the analog supply, although pumping out a great deal of trash in the lower frequency range, have very little noise in the RF spectrum, contrasted with the two 9V supplies. The CE labeled switcher has less noise than the non-marked supplies, but is somewhat worse than the linear supply.

• HP E3610A analog power supply (number 1) with digital voltage and current read out.
• HP E3610A analog power supply (number 2)
• Heath 2718 triple output (5V and two 0-20V variable, analog supply and analog meter.)
• HP 6205C dual analog supply. (Analog meters for voltage and current.)

• One of my E3610 supplies has more noise, including some quasi-discrete spurious signals, than  the other. I don't know whether these are junk from the digital voltage and current displays getting back into the output or if there are from another source. Both are within HP's specification.
• The Heath supply is not a bad performer, and in fact below 25 KHz has less noise than the two HP E3610A supplies. Above 25 KHz, however, the Heath supplies have the most noise of the bench supply units tested. (The line without a label points to the 2718's B output.)
• The hands down winner is the 6205C 100% analog power supply, with an output noise too low to be distinguished from the 3562A's noise floor.

When comparing the bench supplies with the inexpensive modules, it's interesting  to note that the CE marked switching supply is quite respectable, save for the discrete switching frequency related artifacts.