Fresh From the Bench: RME Audio DPS-2 Power Supply and LNI-2 DC Filter and Stabilizer

When RME Audio released the new DPS-2 Power Supply and LNI-2 DC Filter and Stabilizer, the German audio company left many of its faithful customers intrigued. The two new solutions are designed  to improve the performance of its ADI-2 series of converters in critical applications. How do they deliver? One major shift in electronics over the past decade or two is the use of “wall warts” or external power supplies that plug into and power equipment. Looking at the power strip that runs my audio system, I have two wall warts for the miniDSPs I use for woofer control, a power brick for my computer, a power brick for my powered speakers, a wall wart for my RME ADI-2 Pro FS R DAC/ADC, another for my WiiM Pro streamer, another for my mic preamp/interface... and fitting them all in is often a challenge. The driver for this change in electronics is the CE and UL certifications — if, as a manufacturer, I can buy the power supplies with the certifications already done, this greatly reduces my development costs and time-to-market since it’s the power end that needs to be safety tested.

Of course, audiophiles will always be suspicious of simple solutions to simple problems, and external power supplies come under suspicion because they’re just too easy. This is especially the case for the very lightweight switch-mode power supplies (SMPS). Audiophilia nervosa impels lots of people to seek aftermarket solutions, especially linear supplies, which must be better because they are bulkier and more expensive! I have been assured by many people in online discussions that SMPS are inherently noisy and have poor regulation. Never mind that the very quietest supply I have ever measured (Jan Didden’s Silent Switcher) is of course an SMPS. Likewise, the noise components of a linear supply operate at much lower frequencies than those of SMPS making them more difficult to filter out.

Now, this is not to say that there aren’t sometimes issues with external SMPS supplies — some of them are just poorly engineered and indeed noisy. Some have poor isolation and can have leakage currents from line noise. Thankfully, for engineered equipment, these issues are the exception rather than the rule, and it’s relatively easy for a manufacturer to find an OEM provider of SMPS that do the job most of the time.

There’s been an uptick in aftermarket power supplies with the explicit promise of reducing noise and potential ground loops, and the tacit promise of quelling anxiety. Recently, RME has added two such products to its line, the DPS-2 and the LNI-2 DC. I’ve had experience over the years with RME gear and as I mentioned I own their DAC/ADC. Each product I’ve used was absolutely solidly engineered, immaculately documented (their manual writer’s English is excellent, near native quality), completely reliable, and packed with useful features. The user interfaces will sometimes take time to get used to but are consistently logical. It is one of a handful of brands where I would have 100% confidence in any purchase I made from them. They will rarely be the most economical option, but it’s easy to see what you’re paying for.

So given my positive experiences with their products, I jumped at the chance to put these new power units on my test bench.

The Products
The two products here are quite different in application and design, other than the commonality of putting out a DC voltage for audio products to use. We’ll start with the LNI-2 DC, which is a solid-feeling aluminum brick weighing in at 1-1/3 pounds (Photo 1). It is not actually a power supply, it’s a regenerator, that is, you feed it any voltage between 9V to 18V, noisy or quiet, leakage currents or no. Internally, it galvanically isolates the input voltage and uses it to feed RME’s proprietary μFilter switching regulator to generate a clean, quiet, low impedance 12V DC output. There’s also an elaborate set of protection circuits for overcurrent and thermal limiting (Figure 1).
    The US price is about $650, which admittedly isn’t cheap, but it is not outrageous for what it promises to deliver and the engineering/construction behind it. As befits a high-end device, the packaging is slick and deluxe, even having black drawstring bags to contain the brick and the special cables.

The LNI-2 has two indicator LEDs on the top surface, one showing acceptable input voltage range, the other indicating output status (normal, overload, or short). The user plugs the existing (or other) power supply into the LNI-2 DC, which uses a standard 5.5mm × 2.1mm DC connector, then a provided special cable is run from the DPS-2 to the unit it’s powering, again terminated in a standard 5.5mm × 2.1mm connector. There’s also a chassis grounding screw so that if you’re in a situation where a specific ground reference is needed to minimize noise, that option is available. Throughout my testing and use, I never found it necessary, but it’s one more bit of versatility.

The special cable to which I alluded runs from the regenerator to the device being powered, and is a four-conductor wire that is set up as a four-terminal or Kelvin connection (see Resources). There are two power and two sense lines, with the sense lines connected to the power lines at the load. This allows elimination of any effect of the cable resistance with varying load currents. It’s a common method in precision laboratory supplies but rarely seen in audio.

By contrast, the DPS-2 (Photo 2) is a complete power supply system. Physically, it’s an even heavier and larger brick than the LNI-2. It comprises a conventional linear raw supply with multiple switch-selected grounding options, elaborate differential and common mode line-side filtering, and... then the fun begins (Figure 2). There are two sets of power supply filters that each have their own output. The user can select between a conventional linear regulator/filter with current and thermal protection or a μFilter system similar to that used in the LNI-2. Only one output at a time can be used. Note that if the user selects the linear option, the four-terminal connection shorts the power and sense lines at both ends of the cable, so the four-terminal sensing is inactivated.
  It is of course active for the μFilter option, but the availability of both options gives people with fiddly fingers something to keep them occupied. RME has permanently endeared themselves to me by their answer in the FAQ section of their manual: “I can’t decide whether Linear or μFilter sounds better.” “Unfortunately, we cannot help you there.” This added functionality does carry a cost, and the US price is about $1300. Still, for the materials and engineering involved, this is not unreasonable. As before, the unit has LED indicators for electrical and thermal overload. I never had them show anything but normal through my testing phase.

The user’s guides for both these products are available online (see Resources) and go into excruciating detail on their operation and features, so I will limit my rehashing of the details. But I will say that these are unusually good manuals, refreshingly free of hype and non-technical claims, and should be a model for anyone in the industry to emulate.

I should also mention that each unit comes with two four-conductor power cables — one with a locking 5.5mm × 2.1mm plug and one with a standard non-locking plug.

Measurements
As usual, my measurement setup consists of a combination of an Audio Precision APx525 audio analyzer, a calibrated Hewlett Packard 3466A bench multimeter, and for these measurements, I used an RME ADI-2 Pro FS B as a test load and signal source. This is an important caveat—while resistive loads can be highly useful in characterizing a power supply, ultimately it will be connected to a noisy and complex active load. For example, I duplicated a few of the provided measurements for noise, an example of which is shown in Figure 3.

This was taken using the DPS-2 loaded with a 10R non-inductive power resistor. The overall noise for a 100kHz bandwidth was about 3.0µV, an outstanding figure and in good agreement with RME’s data (to no surprise). However, with an ADI-2 Pro as the load, things look quite a bit different as we will see. This is unlikely entirely a function of the power supply, since the load device certainly will be kicking noise back. But I wanted to show this up front so that if you compare my measurements with RME’s, you will understand that the difference is in the setup conditions, not a disagreement between them and me under the same conditions.
  For most of the measurements, I loaded down the ADI-2’s headphone outputs with 33R fixed resistors to simulate a headphone load. To access the supplied DC power, I rigged up male and female 5.5mm × 2.1mm connectors with pigtails for making contact. Admittedly, this adds a bit of uncompensated series resistance, but it’s well under a milliohm so of unlikely consequence. The orange duct tape is used for insulation, not connection, and it’s consistent with my rule that all reliable measurements must use duct tape somewhere! To prevent any issues with measurement ground loops, the APx525 balanced analog input was used, and the ADI-2 was addressed via optical TOSLINK (Photo 3).
  I began by baselining the stock RME supply that came with the ADI-2. Using a 192k sample rate (90kHz bandwidth), the power supply noise at the ADC/DAC power input was measured at 3.6mV RMS. The frequency spectrum of the noise is shown in Figure 4, and is dominated by a few ultrasonic components, especially a big hump at 40kHz to 46kHz as well as spikes at 33kHz and its harmonics. There’s also a rising noise shelf above 80kHz. Note that although the spectrum looks quite messy, the actual levels are moderately low. Still, there’s a lot of room for improvement. In the time domain, the waveform is very sawtooth appearing, which is likely from the charge and discharge of a capacitor (Figure 5).
 
The LNI-2 DC’s noise spectrum with the ADI-2 load is shown in Figure 6. It’s far less messy than the raw supply, and with about 10dB less overall noise. If I were to speculate, the noise here is not dominated by the power supply, but by the active load. Note that the largest noise component is still the spike at 33kHz, which is something that recurs throughout all the measurements. This is a clue to me that the supplies (which are quite different) are not the source.
 
As the Figure 6 title suggests, this is taken with the ADI-2 on, but not outputting any signal. If we have the ADI-2 output a 1V, 1kHz signal from both channels into a 16R dummy load, we get the spectrum shown in Figure 7. This looks almost identical except for the appearance of some noise spikes at 1kHz and 2kHz, and a couple smaller spikes at 4kHz and 5kHz. Note that these are at -90dBV and lower, an extremely small effect, and no other part of the spectrum is affected.
  Turning to the DPS-2, I tested it in both the Linear and μFilter modes. Starting with the μFilter mode and applying a signal as before, I obtained the spectrum shown in Figure 8. Unsurprisingly, it looks quite similar to the spectrum of the LNI-2, which uses a similar μFilter technology. And again, there are the same tiny spikes at 1kHz, 2kHz, 4kHz, and 5kHz from the applied signal.
 
The Linear option under the same conditions gives the spectrum shown in Figure 9. The similarity in line shapes and positions to the spectra from the μFilter again lends credence to the idea that the limitation is the active device, not the supplies.
 
Conclusions
Aftermarket supplies are sometimes a must-have but far more often are just a nice-to-have. If you’re experiencing audible system noise, they may be worth a try. And if you are going to use an aftermarket supply, a well-engineered one is a must.

What we have here is a pair of very, very well engineered supplies. The other part of nice-to-have is the “peace of mind” factor as well as “pride of ownership,” knowing that you have spared no expense or effort in making your system as perfect as it can be top to bottom.

Which one to choose is a slightly more complex question. The LNI-2 DC is less expensive and offers top performance but does require you to keep your stock supply to run it. The DPS-2 has an impressive array of line-side options regarding grounding and ground lifts (while retaining its safety certification).

And of course, if you enjoy changing things and listening for differences — not my bag, but you do you! — the two filtering options are one more thing with which to entertain yourself. As usual, RME’s engineering delivers the goods. aX


Resources
“DPS-2 User’s Guide,” www.rme-audio.de/downloads/dps2_e.pdf
“Four-Terminal Sensing,” https://en.wikipedia.org/wiki/Four-terminal_sensing
“LNI-2 DC User’s Guide,” www.rme-audio.de/downloads/lni2_dc_e.pdf
S. Yaniger and B. Prescott, “RME ADI-2 Pro Bi-Directional Converter with Dual Headphone Amp in Review,” audioXpress, September 2017

This article was originally published in audioXpress, December 2024