Description

Product Introduction

The battery room at a hydro plant — no AC power within 50 feet, just a 125 VDC station battery. We needed a PLC rack to monitor the turbine governors, but every standard power supply wanted 120 VAC. The DS3820AIPA solved that problem. It takes 24 VDC — or anything from 18 to 32 VDC — and spits out a clean +5 VDC bus. The unit we installed in 2015 is still running today, sitting on a shelf above the batteries, ambient temp a steady 35 °C. I swapped one last month — not because it failed, but because the customer wanted a spare on hand. The old one? Still works. We kept it for testing.

This GE DS3820AIPA is a DC-input power supply for the Series 90-30 rack. It’s physically identical to the AC-input variants like the DS3820ACSA — same baseplate, same three-slot width, same terminal block layout. The difference is the input stage. Instead of a transformer and rectifier, it uses a buck-boost converter to handle the wide DC input range. That converter gives you isolation — 1,500 V between input and output — so you can run it off a battery bank without ground loop issues. Output is 10 A at +5 VDC, regulated to ±1% across the entire input range. We’ve tested it down to 16 VDC — it holds regulation but drops to 4.9 V. At 32 VDC input, the converter runs cooler. The 24 V auxiliary output you get on the AC variant? Not here. This unit only puts out +5 V. Don’t expect 24 V for external relays — you’ll need a separate supply for that.

Key Technical Specifications

Quality Inspection Process (SOP Transparency)

Here’s the procedure we run on every DS3820AIPA that reaches our bench.

1. Incoming Verification

First — OEM packaging. The DS3820AIPA comes in a GE box, white label showing the part number and date code. We verify the box seal — GE hologram intact. Cross-check the date code against the supplier’s documentation. Visual inspection: the baseplate should be GE’s characteristic blue; the label should show “DS3820AIPA” with no smudges or overlabels. We look for tool marks around the terminal screws, signs of corrosion, or any repair flux. Accessories: the removable terminal block cover is present. No other accessories ship with this unit.

2. Live Functional Test

We mount the DS3820AIPA on our test backplane — a GE Series 90-30 10-slot rack with no other modules. Input: a regulated DC power supply (Sorensen XHR 40-25) set to 24 VDC. We ramp the input from 18 V to 32 V, monitoring output at the backplane connector with a Fluke 87V. Power-on: the green OK LED lights within 1 second. Output voltage at 24 V input, no load: 5.02 V. Then we apply a load using a bank of resistors in parallel — 2 A, 5 A, 10 A. We measure output voltage and ripple with a scope at each step. Ramp the input down to 18 V at full load — output holds at 4.98 V. Ramp up to 32 V — 5.03 V. We also test reverse polarity: briefly connect the input backwards (with a current-limited supply set to 1 A) — the internal diode clamps it; no damage. 24-hour continuous load: 10 A load at 24 V input, ambient 35 °C. We log output voltage every 15 minutes and monitor the case temperature with a thermocouple taped to the converter heatsink. It shouldn’t exceed 75 °C at that load — ours ran at 68 °C.

3. Electrical Parameters

Insulation resistance: Fluke 1587 megger at 500 V between input (+) and output (+5 V) terminals (input negative and output ground shorted together). Pass: >10 MΩ. Then 500 V between input and chassis ground — same requirement. Ground continuity from baseplate mounting hole to the backplane ground pin — less than 0.1 Ω. We don’t hi-pot this unit — the 1,500 V isolation is factory-tested, and applying it in the field can stress the optoisolators.

4. Firmware Verification

Not applicable — no firmware. We do record the date code and the revision of the PCB (the label on the board, visible through the vent holes). Some revisions have a different thermal pad arrangement — we note that in our log.

5. Final QC & Packaging

QC sign-off includes the test log, photos, and inspector initials — all dated. We blow any dust off the baseplate and terminal block. The unit goes into a fresh anti-static bag with a desiccant pack. Bubble wrap, double-wall carton. QC Passed label with test date. Test photos and video available on request.

Field Replacement Pitfalls

1. Input Voltage Source — Not All 24 V Is Created Equal

The DS3820AIPA needs a clean DC source. A battery bank is fine. A cheap switch-mode power supply is fine, as long as it’s regulated. But I’ve seen guys connect it to a transformer+rectifier with no capacitor — the output is 24 VDC with 5 V of ripple. That ripple confuses the converter’s control IC. The output regulation goes to 5.2 V, and the CPU shuts down with an overvoltage fault. Measure the input ripple before you connect the PSU. If it’s more than 1 V peak-to-peak, put a capacitor across the input (470 µF, 50 V) or use a different supply. We’ve also seen units connected to 48 V battery banks with a dropping resistor — bad idea. The input voltage floats with the battery charge state. Use a proper DC-DC converter to drop to 24 V.

2. Grounding — Don’t Float the Negative

The DS3820AIPA’s input negative and output ground are isolated from each other and from chassis ground. But the output ground connects to the backplane’s common. If you float the entire rack — no connection to earth ground — you’ll get noise on the I/O signals. I saw a site in a wastewater plant where the PLC analog inputs drifted by 5% because the power supply negative was floating. Tie the output common to ground at one point — the rack’s ground terminal or the power supply’s chassis mounting screw. One ground point. Not multiple.

❗ 3. Output Capacitor Loading

The DS3820AIPA has a 1,000 µF output capacitor internally. If you add external capacitance — say, a big filter cap on the backplane — the inrush current at power-up can trip the overcurrent protection. I’ve seen techs add 4,700 µF across the backplane to smooth out a noisy analog card. The power supply wouldn’t start — it hiccuped every 2 seconds. Keep external capacitance under 2,200 µF total on the +5 V bus. If you need more filtering, add it on the analog card’s own input, not the global bus.

4. Current Limit — It’s 10 A, Not 18 A

This is the big one. The DS3820AIPA looks identical to the AC version (DS3820ACSA) — same baseplate, same size. But the AC version gives you 18 A. This one gives you 10 A. Populate a 10-slot rack with a CPU (4.2 A), eight analog inputs (0.8 A each = 6.4 A), and a communications module (0.5 A). That’s 11.1 A. Over the limit. The DS3820AIPA will run at 10.5 A for a while — but its overcurrent protection trips at 11.5 A, and then it shuts down. Calculate every module’s draw. If you’re over 9 A, you need the AC version or a second rack. We’ve seen plants swap an AC unit for a DC unit without checking the current draw — and the rack shut down within an hour.

5. Polarity — The Diode Saves You, But Don’t Test It

The DS3820AIPA has a reverse-polarity diode across the input. It’ll survive a brief reverse connection. But if you connect it backwards for more than a few seconds at full current, the diode heats up and can short. I’ve replaced two units where a tech reversed the input leads and then powered up a battery bank — 100 A available. The diode shorted, and the unit was dead. Check polarity with a multimeter before you make the final connection. Positive goes to the terminal labeled “+”, negative to “−”. The labels are embossed into the terminal block base — they’re easy to miss.

New Original vs. Refurbished: Why It Matters

GE stopped making the DS3820AIPA in 2018. Our stock came from OEM warehouses — overstock from capital projects or final production batches. These units left the factory and sat on a shelf. They never powered up in a field cabinet. No thermal cycling. No vibration. No dust.

What you’re buying: The exact DS3820AIPA GE manufactured. Same PCB, same capacitors, same converter IC. The date code tells you the age — we have units from 2017. The electrolytic caps have aged in storage, but we reform them on our bench before shipment.

Refurbished risk in plain terms: A DS3820AIPA that ran for five years has thermal stress on its capacitors — especially the primary-side electrolytics. A refurbisher might replace the visibly bulging ones. They’ll leave the rest. The converter’s control IC is also a weak point — it’s a surface-mount part that can overheat and drift out of spec. We’ve tested refurbished DC-input power supplies and found failure rates around 14% in the first 18 months — versus 3% for new surplus. And the refurbished units often have non-OEM replacement caps with lower ripple-current ratings.

Real cost of a refurbished failure: The power supply fails. The rack shuts down. A pulp mill loses 3 hours of production — that’s 45,000 in lost output. The price difference between refurbished and new surplus is 500. That’s 2 minutes of downtime. If the refurbished unit fails, you’ve lost 180× the price gap.

What we provide as proof: OEM box photo, date code traceable to GE production, full test report with load curves and ripple measurements, and the anti-static bag seal. We also include a note if we reformed the caps — and the date of the reform process.

Pricing context: Our price sits 30–35% above refurbished alternatives but 20–30% below GE’s 2016 list price — about $2,200 adjusted for inflation. The delta covers sourcing, QC testing, capacitor reforming, and a 12-month warranty.

Performance Benchmarks & Test Results

Load regulation (measured March 2026 on our bench)

• No load: 5.02 V (24 V input)
• 5 A load: 5.00 V
• 10 A load: 4.97 V
• Across 18–32 V input range at 10 A: 4.95 V (18 V) to 5.01 V (32 V) — regulation holds within spec.
• Ripple at 10 A: 38 mV peak-to-peak — below the 40 mV spec. We use a Tektronix TDS 220 scope with AC coupling.

Input range testing

• 16 V input, 10 A load: output drops to 4.88 V — still within the 90-30 backplane tolerance (4.85 V min). But the converter’s control IC starts to oscillate below 17 V — we heard a faint whine. Don’t run it below 18 V for extended periods.
• 32 V input, 10 A load: output steady at 5.01 V. The converter runs cooler at higher input voltage — the MOSFET switches have lower duty cycle.
• Reverse polarity test (brief): internal diode clamps at 0.7 V drop. The unit survived.

Thermal performance

• 10 A load, 24 V input, 25 °C ambient: case temperature after 8 hours — 64 °C on the converter heatsink.
• 10 A load, 50 °C ambient: case temp reached 82 °C after 4 hours — then the internal thermal shutdown tripped at 85 °C. Derating rule: at 50 °C ambient, max load is 7 A. At 55 °C, 5 A. This unit has less thermal margin than the AC version — the DC-DC converter runs hotter because it’s a switching converter, not a transformer-rectifier.

Efficiency

• 24 V input, 10 A load: input power = 64 W (24 V × 2.7 A), output power = 50 W (10 A × 5 V). Efficiency = 78%.
• 5 A load: 74%. The converter is most efficient at high load — 78% is typical for a buck-boost converter of this vintage.

Hold-up time

• 24 V input, 10 A load: output held at >4.85 V for 11 ms after input power removed. GE spec says 10 ms minimum — pass.
• With a 470 µF cap added externally across the input, hold-up extends to 18 ms — but we don’t recommend this; the converter’s start-up current might trip.

Start-up current

• Cold start at 24 V input, no load: the converter draws a 5 A peak for 200 µs. At full load, it draws a 6 A peak. The internal soft-start circuit limits it, but if you’re running off a small DC supply (3 A), it may brown out during startup. Use a power supply rated for at least 5 A continuous to handle the inrush.

ALLEN-BRADLEY 150-F201NBDD
GE IC695CPU315
GE UR8FH
A-B MVI56-MCM