How to Pick a Bench DC Power Supply: Envelope, CV/CC and Ripple

The two biggest numbers on a bench power supply listing are the two easiest to misread.

A title that says "60V 10A" leads many people to assume the unit will give you 60 V and 10 A at the same time. I checked the official spec pages and official manuals from KORAD, Hanmatek, Siglent, RIDEN and OWON (fetched 2026-10-22) and found that assumption holds for only about half of them — one of these units is rated for half of volts-times-amps, and one is not a complete supply at all.

Two disclosures up front: every performance number below is a manufacturer figure. I do not own these units and took no measurements of my own. Prices and stock change constantly, so check the listing page when you order. And the Amazon links below are affiliate links — I earn a commission if you buy through them, which does not change how models were selected here.

TL;DR: Where Each of the Five Stops

• Lowest ripple, power only, no automation: KORAD KD3005P — 30 V / 5 A / 150 W linear, rated ripple ≤2 mVrms (20 Hz–20 MHz). See the KORAD KD3005P listing.
• High current, switching noise accepted: Hanmatek HM310T — 30 V / 10 A / 300 W, rated ripple 10 mVrms. See the HANMATEK HM310T listing.
• Three channels, series and parallel: Siglent SPD3303X-E — two 32 V / 3.2 A channels plus a fixed logic rail, 220 W total. See the Siglent SPD3303X-E listing.
• 60 V, if you already have a DC bus: RIDEN RD6006W — 60 V / 6 A / 360 W, but it is a DC-DC module that needs a separate 6–70 V DC input. See the RIDEN RD6006W listing.
• Wide voltage range, constant-power design: OWON SPE6103 — 60 V / 10 A / 300 W, where 60 V and 10 A cannot both be delivered. See the OWON SPE6103 listing.

Check 1: Rated Power Is an Envelope, Not V x A

Start with a counterexample the manufacturer publishes itself. OWON's official SPE series spec page lists six models; put the voltage ceiling, the current ceiling and the rated power side by side and the pattern is unmistakable.

Model	Max V	Max A	V x A	Rated	V x A ÷ W
SPE3051	30 V	5 A	150 W	150 W	1.00
SPE3102	30 V	10 A	300 W	200 W	1.50
SPE6102	60 V	10 A	600 W	200 W	3.00
SPE6053	60 V	5 A	300 W	300 W	1.00
SPE3103	30 V	10 A	300 W	300 W	1.00
SPE6103	60 V	10 A	600 W	300 W	2.00

On the SPE3051, SPE6053 and SPE3103, V x A equals rated power exactly — a full-rectangle envelope where the whole voltage/current grid is usable. The SPE6102 sits at 3x and the SPE6103 at 2x, and both are constant-power designs: at 60 V you do not get 10 A, and at 10 A you do not get 60 V. The usable region is the area under a hyperbola, not a rectangle.

So the correct way to compare power is to divide max volts times max amps by rated watts. The closer the ratio to 1.00, the closer to a full envelope. Run it on the five here: KORAD KD3005P is 30 x 5 / 150 = 1.00, Hanmatek HM310T is 30 x 10 / 300 = 1.00, RIDEN RD6006 is 60 x 6 / 360 = 1.00, OWON SPE6103 is 60 x 10 / 300 = 2.00. Siglent needs its own arithmetic: one channel is 32 V x 3.2 A = 102.4 W, two channels 204.8 W, against a 220 W total — so a single channel can only ever reach 46.5% of the total rating.

This matters because it decides whether you are paying for half a supply you will never use. If you live at 12 V, a 60 V constant-power unit will not give you more current than a 30 V full-envelope one.

Check 2: A Ripple Number Means Nothing Without Its Bandwidth

Line up the five official ripple figures and the spread reaches 50x — but these numbers are not directly comparable, because the test bandwidths differ completely.

Unit	Official ripple	Bandwidth the vendor states
Siglent SPD3303X-E	300 µVrms / 2 mVpp	5 Hz–1 MHz
KORAD KD3005P	≤2 mVrms	20 Hz–20 MHz
OWON SPE6103	≤5 mVrms / ≤50 mVp-p	ripple 1 MHz, noise 20 MHz
Hanmatek HM310	10 mVrms	not stated in the manual
RIDEN RD6006	100 mV VPP typical	not stated

KORAD publishes 20 Hz–20 MHz, Siglent publishes 5 Hz–1 MHz, and OWON goes further: its footnote reads "Noise bandwidth 20MHz, ripple bandwidth 1MHz" and requires a 10 µF electrolytic in parallel with a 0.1 µF ceramic across the output terminals during the test. Measure the same unit over a different bandwidth and the number moves. Comparing two brands' numbers mostly compares their test conditions.

One more layer: RIDEN quotes peak-to-peak (100 mV VPP typical) while Siglent quotes both peak-to-peak and RMS, and Hanmatek and KORAD quote RMS. Put Siglent's 2 mVpp next to RIDEN's 100 mVpp and you get 50x — and that gap is real. It comes from linear versus buck switching architecture, not from test conditions.

One unit trap worth flagging: the Hanmatek HM310 manual lists current ripple as "8 mVrms". Current ripple should be in mA. It is a unit typo in the official table; read it as mArms.

Check 3: Resolution Is the Step Size, Accuracy Is the Confidence Interval

Nearly every listing advertises "10 mV / 1 mA resolution". Resolution tells you how finely you can turn the knob. It does not tell you how close you land.

KORAD's official KA3005P spec page is unusually complete: setting resolution 10 mV / 1 mA, readback resolution also 10 mV / 1 mA, and setting accuracy (at 25 °C ± 5 °C) of ≤0.5% + 20 mV for voltage and ≤0.5% + 10 mA for current. Work it out at 5 V: 0.5% of 5 V is 25 mV, plus the 20 mV fixed term, so the error ceiling is 45 mV — 4.5x the 10 mV step. Set the front panel to "5.00 V" and anything from 4.955 to 5.045 V is within the published spec.

Run the same arithmetic at the 30 V full-scale point: 0.5% of 30 V is 150 mV, plus 20 mV, for a 170 mV ceiling. The last digit stops meaning "I can set this" and starts meaning "I can set this, but I should not trust it."

Siglent's own lineup makes the point cleanly. Per the official spec table compiled at siglent.eu, the SPD3303C, SPD3303X-E and SPD3303X have identical three-channel output; they differ in display, resolution and interface. The X-E is 10 mV / 10 mA; the X is 1 mV / 1 mA. On accuracy, the X-E is ±(0.5% + 2 digits) for voltage while the X is ±(0.03% of reading + 10 mV). At 30 V that is roughly 170 mV versus 19 mV — about 9x apart. So "the SPD3303 series" is not one machine; the suffix marks two different accuracy grades.

Check 4: CV and CC Are Two Regions on One Curve

Keysight states this directly in the N6700C user's guide and in the
application note, and it is worth quoting: constant voltage (CV) is the mode where the source holds output at the programmed voltage despite changes in load, AC line, or temperature; constant current (CC) is the mode where it holds output at the programmed current limit despite those same changes.

The key point is that you do not pick the mode — the load does. Keysight's criterion: for R_LOAD greater than V_SET/I_SET the supply runs in CV; for R_LOAD less than V_SET/I_SET it runs in CC; when they are equal the supply sits at mode crossover. Most single-quadrant supplies are optimized for voltage priority, and the mode at turn-on is determined jointly by the voltage setting, the current setting and the load resistance. It cannot be programmed.

That explains a very common confusion: you set 12 V, yet the CC indicator lights and the output sits at 8 V. That is not a fault. The load resistance is below 12 V divided by the current limit, so by definition the supply has moved into the constant-current region.

Keysight also documents a third state, UNR (unregulated), where neither voltage nor current is held and the settled values are unpredictable. Five causes are listed, and two of them are easy to hit on a hobby bench: another source connected across the output (two supplies feeding one board), and the moment of transition between CV and CC.

One official wiring note is worth carrying over. In the remote-sense section, Keysight points out that lead resistance grows with length and gauge — every three-gauge increase roughly doubles copper resistance — and requires that sense leads not be twisted or bundled with load leads, and that the sense terminals never be left open, since they are part of the output's feedback path. None of the five units here list remote-sense terminals in their official specifications; if you run high current over long leads, confirm with the seller before ordering.

The Five, One by One: Official Specs and Real Flaws

KORAD KD3005P (30 V / 5 A / 150 W, linear)

KORAD's official KA3005P product page gives: 0–30 V / 0–5 A adjustable, setting and readback resolution both 10 mV / 1 mA, ripple ≤2 mVrms (20 Hz–20 MHz), load regulation ≤0.01% + 2 mV, line regulation ≤0.01% + 3 mV, RS232 and USB interfaces, 4.3 kg, 265 x 110 x 165 mm.

Real flaws. First, a 10 mV step paired with ±(0.5% + 20 mV) accuracy means the last two digits are not trustworthy (see Check 3). Second, the current ripple figure is inconsistent across four sources: the official KA3005P page says ≤5 mArms, the official KA3005PS page says ≤3 mArms, the Reichelt dealer page says ≤3 mA rms, and the Amazon A+ spec table prints current ripple as "≤2 mVrms" — wrong unit entirely. Third, it is single channel only, has no LAN, and cannot be series- or parallel-combined for more power.

The listing is KORAD KD3005P programmable linear supply. That listing is titled KD3005P while the official spec page covers the KA3005P; both are 30 V / 5 A class, so confirm the exact variant on the listing page.

Hanmatek HM310T (30 V / 10 A / 300 W, switching)

The Hanmatek official manual gives: output range 0–32 V / 0–10 A, 300 W rated, setting and readback resolution 10 mV / 1 mA, voltage ripple 10 mVrms, load regulation ≤0.1% + 5 mV, line regulation ≤0.01% + 5 mV, setting accuracy ≤0.1% ± 1 digit, 1.95 kg net, 80 x 150 x 230 mm. Hanmatek's official blog describes the HM310T as 0–30 V / 0–10 A with six memory groups (M1–M6), dual front USB outputs at 5 V ≤1.5 A, and rear USB for PC programming.

Real flaws. First, a switching architecture puts ripple at 10 mVrms, 5x a linear unit like the KORAD — you will see it on analog or audio front ends. Second, official weight disagrees between two sources: the official blog says about 2 kg, the official manual says about 2.7 kg. Third, the input voltage figures also disagree: the blog's spec table says "Input Voltage 110 V (AC)" while the HM310 manual says 110 V / 220 V selectable. Confirm the variant that matches your mains before ordering.

The listing is HANMATEK HM310T programmable DC supply.

Siglent SPD3303X-E (three channels, 220 W, linear)

Per Siglent's official product pages and the official spec table compiled at siglent.eu: CH1 and CH2 are 0–32 V / 0–3.2 A, CH3 is a fixed 2.5 / 3.3 / 5 V at 3.2 A (±8%), 220 W total; series mode reaches 60 V and parallel mode 6.4 A; CV load regulation ≤0.01% + 2 mV, transient recovery <50 µs (50% load change, minimum load 0.5 A); the X-E has USB Device plus LAN while the SPD3303C has USB only; roughly 8 kg at 225 x 143 x 278 mm.

Real flaws. First, the X-E is only 10 mV / 10 mA — you need the SPD3303X for 1 mV / 1 mA, and the two are about 9x apart on accuracy. Second, ripple is published as three different values on three official pages: the Japan site says 300 µVrms / 2 mVpp (5 Hz–1 MHz), the Chinese product page says ≤350 µVrms (5 Hz–1 MHz), and siglent.eu's official datasheet says ≤1 mVrms (5 Hz–1 MHz). Cite whichever you use. Third, channel 3 is a fixed rail, not a second adjustable output. Fourth, at about 8 kg this is a stay-on-the-bench instrument, not something you carry around.

On price, Siglent's own comparison table lists "￥3,080 起" for the SPD3303X/X-E/X-C against "from $1,949" for the Keysight E3631A series and "from $1,240" for the Tektronix 2230G-30-1, with the table dated 2019-07-19 — useful for order of magnitude only.

The listing is Siglent SPD3303X-E triple output supply.

RIDEN RD6006W (60 V / 6 A / 360 W, DC-DC module)

Hangzhou Ruideng Technology's official manual (version matched to RD6006 V1.42, dated 2025.7.5) gives: input voltage range 6–70.00 V DC, output 0–60.00 V / 0–6.000 A / 0–360 W; resolution 0.01 V and 0.001 A; output ripple 100 mV VPP typical; output accuracy ±(0.3% + 3 digits) for voltage and ±(0.5% + 5 digits) for current; CV response time 2 ms (0.1 A–5 A load); buck mode requires a drop of more than 1 V and more than 10%; fan starts above 40 V output, above 4 A output, or above 45 °C system temperature; about 0.58 kg with packaging, 167 x 81 x 65 mm.

Real flaws. First, it is not a complete power supply. The input range is 6–70 V DC, so you need a separate AC-DC supply or a kit that bundles a case and PSU. The official manual's PSU selection section explicitly warns against e-bike chargers, second-hand supplies and modified supplies, because they can output excessive voltage and damage your load, and it requires the no-load output to stay below 65 V with reliable grounding. Second, buck topology means output is always below input, with a required drop of more than 1 V and more than 10% — so the advertised 60 V output needs more than 60 V in, which in practice means a 48 V or 65 V bus. Third, ripple at 100 mV VPP typical is not in the same league as a linear supply.

The listing is RIDEN RD6006W digital control step-down module.

OWON SPE6103 (60 V / 10 A / 300 W, constant power)

OWON's official SPE series spec page gives: 0–60 V / 0–10 A / 300 W; setting and readback resolution 10 mV / 1 mA; accuracy (within 12 months, 25 °C ± 5 °C) of ≤0.1% ± 30 mV for voltage and ≤0.05% ± 10 mA for current; load regulation ≤30 mV and ≤20 mA; response time ≤1.0 ms (50%–100% rated load); voltage temperature coefficient 100 ppm/°C.

Real flaws. First, it is a constant-power unit: 60 V x 10 A = 600 W against a 300 W rating, so high voltage and high current are mutually exclusive (see Check 1). Second, response time is ≤1.0 ms, about 20x slower than Siglent's published <50 µs, which matters for dynamic load testing. Third, the official ripple footnote requires the 10 µF electrolytic plus 0.1 µF ceramic across the output — outside that condition the number will not reproduce. Fourth, dealer eVision's SPE6103 page lists "Resolution 1 mV / 1 mA", 10x off from OWON's own 10 mV / 1 mA; go with the manufacturer.

The listing is OWON SPE6103 programmable DC supply.

A Decision Tree: Work Backwards From Your Load

Four steps beat staring at spec tables.

1. Measure your load before choosing a supply. Write down three numbers: the voltage you normally use, the peak current, and whether you need to switch between two voltages often. Multiply normal voltage by peak current to get the power you actually need, then add 30% headroom.

2. Check whether that power falls inside the envelope. Say your load is 12 V / 8 A = 96 W: the KORAD is out immediately at a 5 A ceiling; the Hanmatek's full 30 V / 10 A envelope covers it; the OWON SPE6103 also covers it, limited by its own 10 A ceiling — though at 60 V it drops to 5 A.

3. Then decide whether you can live with the ripple. For digital circuits, motors and LEDs, switching noise in the 10 mVrms range rarely matters. For analog front ends, audio, or a precision ADC reference, move toward the linear end.

4. Only then look at channels and interfaces. Need bipolar (±12 V) or a logic rail? Take a three-channel unit. Need automation? Confirm USB or LAN with SCPI support. Need to carry it? Note that the KORAD is 4.3 kg, the Siglent about 8 kg and the Hanmatek about 2 kg.

If you are still undecided after four steps, here is a fallback from the data above: for a quiet, low-ripple supply that feeds microcontrollers and modules, take the KORAD KD3005P; for more than 10 A where switching noise is acceptable, take the Hanmatek HM310T.

Limits of This Guide

A few things I did not do, stated plainly:

• No measurements. Every number comes from official vendor spec pages, official manuals and Keysight documentation, fetched 2026-10-22. I do not own these units and did not measure load regulation or ripple myself.
• Prices are not usable. This machine's egress IP geolocates to China, so Amazon's price and stock data cannot be trusted as evidence. No current prices are listed here; check the listing when you order.
• Remote sense was not verified per model. I only cite Keysight's official guidance on sense. Whether these five have sense terminals should be confirmed with the seller.
• Not covered: multi-channel synchronization, SCPI command-set details, and long-term drift. All three affect a buying decision, but the published material does not support a fair comparison.

FAQ

Q: Why can't a 60V 10A supply deliver 60 V and 10 A at once?

A: Because it is a constant-power design. The OWON SPE6103 is rated 300 W while 60 V x 10 A = 600 W, twice the rating. The spec page lists the voltage ceiling and the current ceiling separately, which does not mean the whole rectangle is usable. Before buying, compute max volts times max amps divided by rated watts: 1.00 is a full envelope, 2.00 means high voltage or high current, not both.

Q: The CC indicator stays lit — is the supply broken?

A: Per Keysight's definition, when load resistance is below V_SET/I_SET the supply is supposed to run in constant current; that is normal behavior. Raise the current limit or reduce the load before concluding anything. If it is neither a stable CV nor a stable CC, Keysight also documents a third state, UNR (unregulated), whose listed causes include another source connected across the output and the instant of transition between CV and CC.

Q: Does 10 mV / 1 mA resolution mean the setting is accurate to 10 mV?

A: No. Resolution is the smallest step; accuracy is the confidence interval. KORAD publishes 10 mV / 1 mA resolution alongside ±(0.5% + 20 mV) accuracy: about 45 mV of error at a 5 V setting and about 170 mV at 30 V, both far larger than 10 mV. Read the accuracy row, not the resolution row.

Q: Linear or switching — how do I choose?

A: In this group the KORAD and Siglent are linear, with published ripple in the hundreds of µVrms to 2 mVrms range; the Hanmatek, OWON and RIDEN are switching, published at 5–10 mVrms and 100 mV VPP. Choose linear for analog, audio and precision measurement. For motors, LEDs and digital circuits, switching wins on size, weight and price.

Q: I want to series two supplies for more voltage. What should I watch for?

A: Keysight's application note lists series and paralleling outputs for more power as a fundamental technique, but also lists "another source connected across the output" among the causes of the UNR state. Confirm per model that the vendor documents the mode you plan to use — Siglent, for example, publishes 60 V maximum in series and 6.4 A in parallel for the SPD3303X-E — and never tie two independent supplies directly in parallel.

Wrapping Up

None of these five is stronger across the board. The differences come down to three numbers you can compute: the envelope ratio, the bandwidth behind the ripple figure, and the multiple between resolution and accuracy. Pull those three out of the official spec sheet and there is very little left to guess at.

If you are rounding out the measurement and supply side of your bench, two related reads: How to Read Multimeter CAT Ratings covers the static-level side, and Oscilloscope Buying Guide: Where 100 MHz Stops covers the time domain. For the supply side of power protection, UPS Capacity Math uses the same trick of deriving efficiency from published runtime figures, and that method transfers directly.

👉 Join Xiaomi MiMo Platform: Leading AI model platform with cost-effective inference

👉 Join Aliyun AI: Top AI products with exclusive coupons for business innovation

📌 This article was AI-assisted generated and human-reviewed | TechPassive — An AI-driven content testing site focused on real tool reviews