I used Claude to pull power supply specifications from numerous sources, the JCAT INITO 3 looks the best based on the data the manufacturer supplies. Since Acopian is an industrial supplier, their noise and ripple would be the most accurate. The other manufacturer's info may be cherry picking low numbers or marketing, there is no way to tell.
The LHY Audio LPS-15 has excellent published specifications, you would need a JCAT to get lower ripple numbers.
I use an Acopian 5EB100 5v power supply with a JCAT Femto USB card. I haven't tried other power supplies with this card. Using the 5EB100 lps with a Paul Pang USB card very slightly increased the bass and made the sound cleaner, but the change was slight. Upgrading the LHY lps to a JCAT lps is entering the land of diminishing returns. You will definitely hear a difference after spending $1090.00.
JUMPINGThe largest jump in quality was buying a Hapa Audio AerØ USB cable, it was much bigger than upgrading to the JCAT Femto USB card.
The second largest jump in quality was the HoloAudio May KTE DAC and
HQPlayer combo. I prefer the sound of DSD so all music files are upsampled to DSD256, PCM sounds sharper and doesn't sound as good in my system. Other DACs like the ANK 5.1 Signature sound better if the source is strictly Redbook PCM. Not all DACs work with HQPlayer, the website has an approved list, the main requirement is a NOS DAC. Almost all DACs oversample so their internal filters work. PS Audio DirectStream DAC automatically upsamples all inputs to DSD256 so not a good choice with HQPlayer. Music playing software has it own built in sound, JRiver has a mild mid-bass bump that makes everything sound warmer. I haven't tried Jplay. HQPlayer's sound can be customized using a dizzying array of filters (only a handful are used so not as complicated as it initially looks).
One reason I'm not thinking of a USB power supply upgrade is the HoloAudio has a lot of filtering:
The May uses a discrete multi-stage voltage regulation circuit (not off-the-shelf regulator ICs) built specifically for each internal rail, paired with a separate outboard power supply chassis so the transformer and rectification stay physically isolated from the sensitive DAC/analog circuitry.
Holo Audio's own marketing claims the regulation achieves better than 0.2μV output noise — for context, that would put it below even the JCAT/LT3045-class parts, though as with most of these figures, it's a manufacturer claim without a published third-party measurement methodology. The KTE tier goes further with an O-core transformer using flat-wire windings, which produces less flux leakage than a toroidal and is claimed to improve regulation performance further. The May's jitter/latency is also buffered and re-clocked.
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Here is the info Claude put together from published data sheets:
Acopian 5EB100The Acopian 5EB100's actual datasheet ripple is 1000μV (1mV RMS) — a solid, industrial-grade, conservatively-rated number that holds across the whole EB family regardless of voltage/current. Against the LHY LPS-15's claimed 17.28μV, that's a ~58x gap on paper. But remember the asymmetry: Acopian's number is a guaranteed worst-case spec from a formal datasheet; LHY's is an unqualified marketing figure with no stated load or bandwidth — so in practice the real-world gap is likely smaller than the raw numbers suggest, though the Acopian's rated ripple is still probably a more trustworthy floor to build expectations around.
LHY LPS-15On paper, the LHY LPS-15's advertised ~17μV ripple figure is roughly 60x lower than the Acopian's rated 1mV (1000μV) ripple spec. A few caveats worth flagging before you lean on that number for a real decision:
The Acopian spec is an industrial datasheet number — conservative, tested to a defined standard, and something you can trust as a worst-case guarantee across the full family.
The LHY figure is a consumer marketing spec with no stated test conditions (bandwidth, load, or measurement method aren't published), so it's likely a best-case or no-load number rather than a worst-case guarantee.
Different measurement bandwidths (e.g., 20MHz vs. 100kHz) can shift reported ripple numbers by an order of magnitude on their own, so a like-for-like comparison really needs matching test conditions — which the LHY listing doesn't give you.
Acopian L5MC500 Gold Box Infinity - $40 eBay
Ripple: The L5MC500 is meaningfully better — 0.25mV RMS vs. the EB-series' 1mV RMS, a 4x improvement. And critically, Acopian actually states the test bandwidth (25MHz) for the Infinity series, which the older EB-series datasheet doesn't do — so the L5MC500 number is more rigorously documented, not just lower.
Regulation: Interesting reversal here — the Infinity series specs regulation in absolute millivolts (±2mV for both load and line) rather than percentage, which at 5V output is actually tighter than the EB-series' percentage-based spec once you do the math (±0.25% load reg on 5V is ±12.5mV, over 6x looser than the Infinity's flat ±2mV).
Bottom line for your comparison purposes: If you're sizing this against the LHY LPS-15's claimed 17.28μV, the L5MC500's 250μV (0.25mV) is still roughly 14x higher than LHY's marketing number — closer than the EB-series' 58x gap, but still a real difference, especially since (again) Acopian states its ripple over a real measurement bandwidth and the LHY figure doesn't.
JCAT INITIO 3JCAT doesn't publish a hard ripple/noise spec for the INITIO 3 itself, but there's enough here to build a solid comparison. Key facts: it uses LT3045 ultra-low-noise linear regulators (same architecture as JCAT's flagship OPTIMO 3 DUO, which JCAT explicitly rates at <2μV noise), a dual-output design (5V or 12V, 3A total), and a shielded/epoxy-potted custom transformer. The LT3045 IC itself (Analog Devices/Linear Tech) is datasheet-rated at 0.8μV RMS typical output noise (10Hz–100kHz) — one of the lowest-noise linear regulators commercially available, which is why it's become a favorite in audiophile PSU designs.
TeraDakThe TeraDak is the weakest entry in this whole set purely on the basis of documentation — not necessarily performance. It's a well-regarded product in the DIY/streaming-audio community (repeatedly mentioned favorably alongside things like the UpTone LPS-1 and HDPlex units in forum discussions), but TeraDak as a company simply doesn't publish engineering specs the way Acopian does, or even make aspirational marketing claims the way LHY and JCAT do. One relevant data point from a forum thread on a related TeraDak product (the ATX-series linear supply): TeraDak told a buyer they couldn't share ripple numbers because they "don't have the numbers on hand" — which suggests this isn't a case of them hiding a number, but of the company not routinely measuring or publishing this spec at all.
Bottom line: if a documented, comparable ripple number is what you actually need to make this decision, the TeraDak DC30 gives you nothing to go on — you'd be relying entirely on anecdotal listening reports rather than anything you could put in a spec sheet next to the Acopian units.
How to think about this rankingIf you trust the manufacturer claims at face value, the rough order from noisiest to quietest is:
Acopian 5EB100 (1mV) > Acopian L5MC500 (250μV) > LHY LPS-15 (~17μV) > JCAT INITIO 3 (likely low single-digit μV, based on LT3045/OPTIMO precedent)But the caveats matter a lot here, and they cut in opposite directions depending on the product:
Acopian's numbers are the most trustworthy — both are from formal engineering datasheets with defined worst-case guarantees (and the L5MC500 even states its measurement bandwidth). These are industrial parts built to a spec, not marketed for subjective "sound quality."
The LHY figure is unverified marketing copy with no test conditions disclosed — it could be a best-case, no-load number that's not representative of real operating conditions.
JCAT doesn't publish a ripple number for the INITIO 3 at all. The <2μV figure belongs to its more expensive sibling (OPTIMO 3 DUO, ~$2,000), and while INITIO 3 shares the same regulator IC, JCAT explicitly markets INITIO 3 as an "entrance" tier product built to a lower price point — so it likely doesn't hit the same number even if the underlying LT3045 topology is capable of it. The LT3045's own datasheet spec (0.8μV RMS) is the theoretical floor the IC can achieve under ideal conditions, not what any implementation delivers.
