Description

Product Introduction

The paper mill’s wet end had twenty-four analog inputs and twelve analog outputs on a single 90-30 rack. That’s 15.6 A of current draw just from the analog cards, plus the CPU, plus the comms module. The standard DC power supply — the AIPA with 10 A — didn’t stand a chance. The DS3820AIQA was the answer. It’s a 14 A DC-input unit, designed for exactly this scenario. We installed one in 2016, and it’s still running today, though the rack is now down to eighteen inputs — they moved six to a remote I/O drop.

The GE DS3820AIQA is a DC-input power supply that delivers 14 A at +5 VDC. That’s 70 W. It’s physically identical to the AIPA models — same blue baseplate, same three-slot width, same terminal block — but the internal converter is a different topology. The AIPA uses a buck-boost converter. The AIQA uses a full-bridge converter with synchronous rectification. That’s why it can push 14 A without overheating. Efficiency is 82% at full load, compared to 76% on the AIPA. The downside? It has a higher minimum input voltage — 18 V versus 16 V on the AIPA. The full-bridge needs more headroom to regulate. And it’s harder to find — GE made fewer of these because most plants used the AC-input version (18 A) if they had AC available.

Key Technical Specifications

Quality Inspection Process (SOP Transparency)

Here’s our SOP for the DS3820AIQA — and it’s slightly different because of the higher power.

1. Incoming Verification

OEM box check — GE holographic seal intact, part number matches “DS3820AIQA.” Date code recorded. Visual: the baseplate is blue, but the heatsink is visibly larger than the AIPA — it extends about 0.3″ higher off the board. That’s a quick visual differentiator. We also look for the full-bridge converter’s MOSFETs — four of them, visible through the vent slots in a line. If there are only two (a half-bridge), it’s not an AIQA. Accessories: terminal block cover present. No other items.

2. Live Functional Test

We mount it on our test backplane. Input from a Sorensen XHR 40-25 — capable of 25 A output, so it handles the 5.2 A input draw. We set input to 24 VDC. Power-on: OK LED lights within 1 second. Output at no load: 5.02 V. Then we load it in steps: 2 A, 6 A, 10 A, 14 A. At 14 A, the output holds at 4.97 V — that’s 0.6% regulation, actually better than spec. Ripple at 14 A: 32 mV peak-to-peak. We sweep the input from 20 V to 32 V at 14 A — output stays 4.95–5.01 V. At 18 V input, the output drops to 4.90 V — still within the 4.85 V minimum for the backplane, but GE’s spec warns of degraded regulation below 20 V. We note that. 24-hour continuous run: 14 A load, 24 V input, ambient 35 °C. We measure the heatsink temp with a thermocouple — it stabilizes at 72 °C, which is acceptable for a 85 °C maximum.

3. Electrical Parameters

Insulation resistance: Fluke 1587 megger at 500 V between input and output — >10 MΩ. Between input and chassis ground — >10 MΩ. Ground continuity: <0.1 Ω. We don’t hi-pot this unit — the full-bridge converter has sensitive MOSFET gates that can be damaged by high-voltage stress.

4. Firmware Verification

No firmware. We record the date code and the revision of the main controller IC — a UC3875 phase-shift PWM controller. Some early revisions had jitter issues at light load — we check for that by measuring the switching frequency at no load. If it oscillates, we reject the unit.

5. Final QC & Packaging

QC log includes all data, photos, and a note on the input voltage regulation test. The unit goes into a fresh anti-static bag, desiccant pack, bubble wrap, double-wall carton. QC Passed label with date. Test photos available.

Field Replacement Pitfalls

1. Input Cable Gauge — 5.2 A Needs 14 AWG Minimum

The AIQA draws up to 5.2 A at 18 V input. If you’re using 18 AWG cable over a 50-foot run, the voltage drop is about 2 V. At the power supply’s input terminals, you’ve got 22 V instead of 24 V. That’s still within the operating range, but if the battery bank sags to 22 V under load, the power supply sees 20 V — and regulation starts to degrade. Use 14 AWG for runs under 50 feet, 12 AWG for runs up to 100 feet. We’ve seen sites use 16 AWG and wonder why the rack shuts down during motor starts. Voltage drop is the culprit.

2. Heatsink Clearance — It Runs Hot

The AIQA’s heatsink is larger than the AIPA’s, but it still runs hotter — 72 °C at full load in a 35 °C ambient. That’s normal. But if you mount this in a cabinet with no airflow, the heatsink temp can hit 85 °C and the unit will shut down. The AIQA’s thermal shutdown is at 90 °C — tighter than the AIPA’s 95 °C. Leave at least 2 inches of clearance above and below the unit for convection airflow. I saw a site where they packed modules on both sides — the AIQA was sandwiched. It ran for 3 months, then shut down every afternoon when the sun hit the cabinet. We moved it to the top slot, with open space above it. Problem solved.

❗ 3. Input Fuse — 8 A Slow-Blow, Not Fast-Acting

The AIQA’s inrush current is about 10 A for 2 ms at cold start. A fast-acting 6 A fuse will blow — we’ve seen it. Use an 8 A slow-blow (time-delay) fuse on the input line. GE’s spec says 7.5 A minimum, but 8 A gives you margin. If you’re using a circuit breaker, use a C-curve (motor-start) breaker, not a B-curve (general-purpose) breaker. B-curve will nuisance-trip.

4. Output Capacitance — Don’t Overdo It

The AIQA has a 1,500 µF output capacitor internally — more than the AIPA’s 1,000 µF. If you add external capacitance beyond 2,200 µF, the inrush at startup can trip the overcurrent protection. I’ve seen techs add a 4,700 µF cap to smooth out a noisy analog bus. The power supply hiccuped on startup for 5 seconds before it finally came up. Keep external capacitance under 2,000 µF on the +5 V bus. If you need more filtering, use a separate DC-DC converter for the analog cards.

5. Grounding — The AIQA Has Less Isolation Margin

The AIQA’s full-bridge converter has lower common-mode isolation capacitance than the AIPA’s buck-boost. That means it’s more susceptible to ground loops. If you ground the input negative to the rack ground, you might get a 100 mV AC noise on the output. We tested this on the bench. The output ripple went from 32 mV to 120 mV when we grounded the input negative. Float the input negative. Ground the output common (the +5 V return) at the rack ground. That’s the recommendation in GE’s application note GEN-3847, section 3.2. Follow it.

New Original vs. Refurbished: Why It Matters

The DS3820AIQA was a specialty variant — GE made fewer than 5,000 units total, and production ended in 2018. Surplus is hard to find. Our stock came from a single OEM warehouse in Illinois — leftover units from a cancelled pipeline automation project. They’ve never been installed.

What you’re buying: The full-bridge converter with synchronous rectification — a design that uses MOSFETs instead of diodes for the output rectification. That’s what gives the AIQA its 82% efficiency. Refurbished units often have the synchronous MOSFETs replaced with Schottky diodes because the MOSFETs are harder to source. Efficiency drops to 72%, and the unit runs hotter. We’ve seen refurbished AIQAs with diode replacements — they don’t have the same output rating.

Refurbished risk in plain terms: The output MOSFETs are the critical components. They run at 200 kHz and switch 14 A. After years of thermal cycling, their on-resistance (Rds(on)) increases. A refurbisher might not test the full 14 A load — they’ll test at 5 A and call it good. At 14 A, the MOSFETs overheat and the unit shuts down. Failure rate on refurbished AIQAs is about 17% in the first year — versus 3% for new surplus.

Real cost of a refurbished failure: A high-current DC power supply fails. The rack goes down. A plastic extrusion plant loses 4 hours of production — 30,000 in scrap and downtime. The price difference between refurbished (1,700) and new surplus (2,500) is 800. That’s less than 2 minutes of downtime in that plant.

What we provide as proof: OEM box photo, date code, a photo of the internal board showing the four synchronous MOSFETs (we have a special QC camera for this), our full load test at 14 A, and a thermal image of the heatsink at full load. The unit goes out with a test report and a QC label.

Pricing context: Our price sits 30–35% above refurbished alternatives but 25–30% below GE’s 2016 list — about $3,200 adjusted. The delta covers sourcing, testing, and a 12-month warranty.

Performance Benchmarks & Test Results

Load regulation (measured April 2026)

• No load: 5.02 V at 24 V input
• 7 A load: 5.00 V
• 14 A load: 4.96 V — regulation is 1.2%
• At 20 V input, 14 A load: 4.92 V. At 18 V input: 4.88 V. Regulation is about 2.8% at 18 V — GE’s spec says ±2% only between 20–32 V.
• Ripple at 14 A, 24 V input: 32 mV peak-to-peak.

Input range and transient testing

• 32 V input, 14 A load: output at 5.03 V. Stable.
• 40 V transient (1 ms pulse) at 24 V input: output glitched to 5.10 V, recovered in 300 µs. No trip.
• 45 V transient: the overvoltage protection on the input tripped — the unit shut down and latched. We had to cycle power to reset it.
• Reverse polarity: diode clamps — unit survived a 2-second reverse at 1 A.

Thermal performance

• 14 A load, 24 V input, 25 °C ambient: heatsink temp after 8 hours = 65 °C.
• 14 A load, 45 °C ambient: heatsink reached 82 °C after 6 hours. Derating rule: above 45 °C ambient, derate 0.5 A per °C. So at 50 °C ambient, max load is 11 A. At 55 °C, 9 A. GE’s datasheet says “derate above 45 °C” — they don’t give a curve, so we derived this from our testing.
• No load at 25 °C ambient: heatsink temp is 38 °C — the full-bridge dissipates about 5 W even at no load.

Efficiency

• 24 V input, 14 A load: input power = 85.4 W (24 V × 3.56 A), output = 70 W. Efficiency = 82%.
• 7 A load: 78%.
• 24 V input, 14 A load, 40 °C ambient: efficiency dropped to 80% because the MOSFET on-resistance increases with temperature.

Hold-up time

• 24 V input, 14 A load: output held >4.85 V for 8 ms. GE spec says 8 ms minimum — exactly on the line. Don’t rely on this for hot-swap or brownout conditions.

Switching frequency

• 200 kHz nominal — we measured 198 kHz at 14 A load. The UC3875 controller has a ±5% tolerance, so this is normal.
• At no load, we saw the controller drop into skip mode (pulsing at 20 kHz) to save power. Some PLCs don’t like the acoustic noise — you can hear a faint whistle at no load. If that’s a problem, add a 1 A dummy load to keep it in continuous mode.

A-B 1753-L32BBBM-8A
ABB PPC322BE
B&R 5PP320.1505-39
FLUKE FLUKE190-202