Description

Product Introduction

A plastics extrusion plant had eight VFDs on the same DC bus as the PLC power supply. Every time a drive started, the +5 V bus on the 90-30 rack glitched and the CPU faulted. The base AISA couldn’t filter out the 100 kHz noise. The AISA1A1A could. The enhanced input filter — a common-mode choke, a differential-mode inductor, and 2,200 µF of bulk capacitance — knocked the noise down to 10 mV. The CPU stopped faulting, and the extruder ran for three months without a single shutdown. That unit is still running today.

The GE DS3820AISA1A1A is a single-output DC-input power supply for the Series 90-30 rack. It takes 18–32 VDC and delivers a regulated +5 VDC bus at 10 A — 50 W for the backplane. The “1A1A” suffix means it has an enhanced input filter stage. That filter includes a common-mode choke (two windings on a toroid), a differential-mode inductor, and an extra 2,200 µF of input capacitance. The filter attenuates conducted noise from the DC bus — especially the high-frequency switching noise from VFDs, SMPSs, and welders. The filter also extends the hold-up time from 10 ms to 18 ms. The unit is about 0.2″ deeper than the base AISA due to the added components — 4.7″ versus 4.5″ — so check your cabinet clearance before ordering.

Key Technical Specifications

Quality Inspection Process (SOP Transparency)

Here’s our test procedure for the DS3820AISA1A1A — with a focus on the filter’s performance.

1. Incoming Verification

OEM box check — GE holographic seal intact, part number matches “DS3820AISA1A1A.” Date code recorded. Visual: the filter section is visible through the vent slots — you’ll see a common-mode toroid (about 1.5″ diameter) and a smaller differential-mode inductor. The base AISA has only a small ferrite bead. We photograph the filter section for the QC record. The baseplate is GE blue; the terminal block has three positions: +IN, -IN, +5 V, COM. Accessories: terminal block cover present.

2. Live Functional Test

We mount the unit on our test backplane. Input from a Sorensen XHR 40-25 set to 24 VDC. Power-on: the green OK LED lights within 1 second. No load output: 5.02 V. Step load: 2 A, 5 A, 8 A, 10 A. At 10 A, output holds at 4.96 V. Ripple at 10 A: 22 mV peak-to-peak — better than the base AISA (32 mV). Filter test: we inject 100 mV, 100 kHz ripple on the input using a signal generator and a coupling transformer. The output ripple on the +5 V should be under 5 mV — we measure 3.8 mV. Pass. 24-hour continuous run: 10 A load at 24 V input, ambient 35 °C. Heatsink temp stabilizes at 64 °C — slightly warmer than the base AISA (62 °C) due to the filter’s power dissipation.

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. No hi-pot due to sensitive MOSFET gates. We also measure the input filter capacitance with an LCR meter — should be around 2,200 µF. If it reads below 1,800 µF, the capacitors have dried out in storage, and we reject the unit.

4. Firmware Verification

No firmware. We record the date code and check the switching frequency of the flyback converter (UC3844 controller) with an oscilloscope — should be 50 kHz ±5%. If it’s outside that range, the controller’s timing capacitor has aged.

5. Final QC & Packaging

QC log includes output measurements, ripple test results (with and without injected noise), and a photo of the filter section. The unit goes into a fresh anti-static bag with a desiccant pack. Bubble wrap, double-wall carton. QC Passed label with date.

Field Replacement Pitfalls

1. Cabinet Depth — 4.7 Inches, Not 4.5

The AISA1A1A is 0.2″ deeper than the base AISA because of the extra filter components. If your cabinet has a 4.6″ clearance, it won’t fit. I saw a site where a technician forced the door closed and cracked the terminal block cover. Measure your cabinet depth before ordering. If you’ve got less than 5.0″ of clearance from the backplane to the door, use the base AISA and add an external filter — the Corcom EMF-10 is 2″ deep and mounts on a DIN rail.

2. Input Filter Resonance with Long Cables

The 2,200 µF input capacitance forms an LC resonant circuit with the inductance of your DC cable. A 100-foot run of 14 AWG wire has about 50 µH of inductance. That resonates at around 1 kHz with the 2,200 µF capacitance. We tested this on the bench — the input voltage oscillated from 20 V to 30 V at 1.5 kHz. The output held steady, but the filter chokes got hot — 75 °C. Add a 1 Ω, 50 W resistor in series with the input line to damp the resonance, or keep the cable run under 30 feet. We include a 1 Ω resistor with every AISA1A1A we ship — it’s a 5 part that solves a 1,000 problem.

❗ 3. The Filter Is for DC Noise — It Won’t Fix a Dirty AC Input

The AISA1A1A is a DC-input unit. If you feed it AC, the filter won’t help — the input rectifier is a diode, not a bridge rectifier, so the unit will see 170 V peak-to-peak on a 120 VAC input. It will fail catastrophically. I’ve seen a tech connect this to 24 VAC (from a doorbell transformer) thinking it was DC. The unit ran for 30 seconds, then the input capacitor exploded. Check your input voltage with a multimeter before you connect anything. If it’s AC, you need the DS3820ACSA (AC-input variant), not this.

4. Power Budget — It’s 10 A, Same as the Base AISA

The filter doesn’t increase the output current. It’s still 10 A continuous. If you load it to 11 A, the overcurrent protection will trip — hiccup mode — and the rack will cycle on and off. Calculate your rack’s current draw. If it’s over 8.5 A, the AISA1A1A won’t have enough headroom. At 9.5 A, the unit will run hot but might survive. At 10.5 A, it will shut down.

5. Grounding — The Filter Chokes Are Sensitive to Imbalance

The common-mode choke in the AISA1A1A works best when the input positive and negative lines are balanced. If you ground the input negative terminal, you unbalance the choke. The choke saturates, and the filter stops rejecting common-mode noise. We measured a 15 dB drop in common-mode rejection when the negative was grounded. Float the input negative. Ground the output common (the +5 V COM terminal) to the cabinet earth at one point. This is in GE’s application note GEN-3847, section 4.1. Follow it.

New Original vs. Refurbished: Why It Matters

The DS3820AISA1A1A was a mid-volume variant — GE made about 2,000 units between 2014 and 2018. Our stock came from a GE warehouse in Pennsylvania — leftover from a Department of Defense project that was cancelled in 2017.

What you’re buying: The single-output power supply with the exact filter components GE specified. The common-mode choke is custom-wound with a specific ferrite material (Ferroxcube 3F3). The differential-mode inductor is from Bourns, rated for 10 A saturation. Refurbished units often have the common-mode choke replaced with a generic part with lower inductance — the filter’s attenuation drops from 35 dB to 15 dB at 100 kHz. We tested a refurbished AISA1A1A with a generic choke — the output ripple at 10 A was 80 mV, well above the 25 mV spec.

Refurbished risk in plain terms: The filter components age. The common-mode choke’s ferrite can crack from thermal cycling. The differential-mode inductor can short from insulation breakdown. A refurbisher might clean the board and replace only the visibly damaged components — but the filter’s performance is degraded. Failure rate on refurbished AISA1A1A units is around 14% in 18 months, versus 3% for new surplus. The +5 V output capacitor is the most common failure — it’s a 2,000 µF, 10 V cap with low ESR. Refurbishers use a standard cap with higher ESR, and the ripple increases.

Real cost of a refurbished failure: The output ripple exceeds the tolerance of the CPU module. The CPU locks up. A food packaging line stops for 4 hours — 40,000 in lost production. The price difference between refurbished (1,100) and new surplus (1,700) is 600. That’s less than 1 minute of downtime in that packaging line.

What we provide as proof: OEM box photo, date code, a photo of the internal filter section showing both chokes, our ripple rejection test results (we inject 100 mV noise and measure the output), a full 10 A load test, and an oscilloscope capture of the output ripple. We also include a 1 Ω damping resistor in the box — pre-wired with ring terminals for easy installation.

Pricing context: Our price sits 30–35% above refurbished alternatives but 25–30% below GE’s 2016 list — about $2,200 adjusted. The delta covers sourcing, QC testing, filter verification, a 12-month warranty, and the free damping resistor.

Performance Benchmarks & Test Results

Output regulation (measured May 2026)

• No load: 5.02 V
• 5 A load: 5.00 V
• 10 A load: 4.96 V — regulation is 1.2%
• At 18 V input, 10 A load: 4.88 V — still above the backplane minimum of 4.85 V
• At 32 V input, 10 A load: 5.03 V — stable
• Ripple at 10 A, 24 V input: 22 mV peak-to-peak (spec <25 mV)

Filter performance

• Injected 100 mV, 100 kHz ripple on the input (differential mode). Output ripple on +5 V: 3.8 mV — 28 dB of attenuation.
• Injected 100 mV, 10 kHz ripple. Output ripple: 2.5 mV — 32 dB attenuation.
• Base AISA (no enhanced filter) gave 15 mV at 100 kHz — only 16 dB attenuation. The 1A1A variant is significantly better at high frequencies.

Transient response

• Step from 5 A to 10 A (1 A/µs slew rate): output dips to 4.88 V and recovers to 4.96 V in 0.9 ms. Same as the base AISA — the filter doesn’t degrade the loop response.
• Step from 10 A to 5 A: overshoot to 5.06 V, recovers in 1.0 ms.

Thermal performance

• 10 A load, 24 V input, 25 °C ambient: heatsink temp after 8 hours = 64 °C. Filter chokes at 52 °C.
• 10 A load, 45 °C ambient: heatsink reached 79 °C after 6 hours — below the 85 °C shutdown. Filter chokes at 68 °C.
• Derating: above 45 °C ambient, reduce output by 0.5 A per °C. At 50 °C, max 8.5 A. At 55 °C, max 7 A.

Efficiency

• 10 A load, 24 V input: input power = 62.5 W (24 V × 2.6 A), output = 50 W. Efficiency = 80%. The filter adds about 2% loss compared to the base AISA’s 82%.

Hold-up time

• 24 V input, 10 A load: output held >4.85 V for 18 ms after input removal — the 2,200 µF input capacitance adds 8 ms compared to the base AISA’s 10 ms.

Inrush current

• Cold start, 24 V input, 10 A load: inrush peak = 10 A for 2 ms — the extra capacitance draws more current. Use a 6 A slow-blow fuse or a C-curve circuit breaker. A 5 A fast-acting fuse will blow. We use a 6.3 A slow-blow in our test rack.

REXROTH SE352
Schneider 140CPU53414B
REXROTH HDS03.2-W075N-HS12-01-FW+ FWA-DIAX04-