# Center tap transformer pairing with ATX PSU

I'd like to use both a beefy center tap transformer (220v to 24-0-24, 1000VA) and an ATX PSU togheter, to power some lab equipment. So I isolated from case the ATX GND on DC side as the case is earth ground referenced, to make the ATX GND floating; and I'd like to connect the ATX ground to the transformer's center tap in order to have +-24v rails from the transformer and being able to use that rail togheter with the ATX rails.

Pic. 1, Simplified circuit diagram

Pic. 2, Noise filter 1

Pic. 3, Rectifier and Noise filter 2

Things I'm investigating before powering the circuit:

1) Given I don't know the ATX circuit, how can I be sure that some AC (example: 77v, a few uArms) from the center tap won't burn the ATX? Are Filter2+Filter3 enough? Or I need some diodes (example: "Moral of this story: include diode D1 even if you think that it isn't necessary." )?

2) protections: I placed glass fuses both on the primary AC input and the DC output connectors in order to have a basic overcurrent/shortcircuit protection. How can I improve that? Example: a bulb could be both a limiting resistor and a visual signal proportional to voltage/current output but I don't know how to dimension it (ie: do I need a 12v, 24v, ..., 220v bulb? Can I place it on the center tap?). Are there 10-20A resettable fuses (or PTCs) similar to the ones used on smaller circuits?

After the first powering up I'd like to further refine the filtering. So I need a way to measure and do some math for proper capacitor placing and dimensioning. I've a cheap oscilloscope but I have no idea how to track ripple and choose the right caps. Any advice?

EDIT2: modified question to include EMI issues raised in the comments.

EDIT3: circuit diagram fixed, with filters and higher resolution.

• What is the reason it is necessary to modify ATX supply not to be ground referenced? The 24VAC transformer is already isolated. – Justme Aug 24 '19 at 23:15
• No leakage or EMI line or CM on DC noise specs, BAD idea to do this with line filter Y caps inside. – Tony Stewart EE75 Aug 25 '19 at 0:00
• @Justme Can be useful when using an oscilloscope and other circumstances. In any case I kept the case to ground so that a bad psu failure can blow the mains fuse. And I added the earth ground connector close to the gnd one, so that I can easily earth ground the whole power supply in case I need it to be earth grounded. – Anichang Aug 25 '19 at 0:07
• @SunnyskyguyEE75 can you elaborate please? What's "CM"? – Anichang Aug 25 '19 at 0:12
• @SunnyskyguyEE75 Please consider 2 things: 1) the atx metal case is intact, I just place it as is inside a bigger plastic case, and the case is earth grounded. 2) I isolated the DC side of the PCB only, there's a 3rd pcb-to-case connection on the AC side of the PCB and I didn't isolate it. I tought that could be a security feature to leave it as is and there's no need to remove it as there's no galvanic continuity between that one and the DC ones. – Anichang Aug 25 '19 at 0:29

If you really need a “floating ground” then use 1nF to 4.7nF plastic cap across earth disconnect to block DC and 1nF = 1M5 ohms @ 100 Hz.

• don’t modify ATX box unless you have to, for a good reason.

• Certs. becomes void with mods.
• It is safe to bypass open gnd with small plastic cap. Yet connect RF to Earth.
• ground short fault into 4n7 cap= 220V/1M5=150 uA safe. Low V, low I so would not trip fuse but might possibly permit filter current to be sensed as small tingle thru body to earth or humid air. No worries just a natural consequence of SMPS CM filter leakage current < 0.5mA in America or maybe < 2.5mA in EU. Safety limit.

• Yet at 100 MHz, 4n7 cap is 1.5 Ohms which shunts stray noise (high Z) very well.

• But case is no longer earth bonded, so use ATX PSU supply as is with earth-bonded ground for best RF results.
• make custom cable harness using surplus harnesses from old PC’s to make pretty. But I just used wire nuts to connect wires.
• treat ATX supply the same as you would in a PC
• add Pwr ON jumper to MoBo plug as required.
• Then use it for any peripheral as you wish and enhance design with variable or fixed V,I limits
• not as good as a battery + charger for heavy motor loads , where start current=V/DCR = 800 to 1200% rated current, but ok for ramped acceleration servo motors or ECU’s. (> 250us)
• Beware of surge switch effects from your in or out connection or switch.
• consider enhanced protection to load or ATX PSU;

• These can be ganged in a range of sizes or applied in series or parallel with predictable, expected results and switched in or bypassed or out as you desire/design in your adapter or generally use interface.

Certs - In our country one cannot modify inside a cUL product even with a switch or regulator and keep safety certs on product. But outside the box, you can do whatever.

    - Not even for 1st generation shipments to a corporation. Some have lost their P Eng status in Canada doing so.  So  our company paid \$50/ power supply for a cUL waver to circumvent cUL requirement and add our mods to a Canadian Tire battery charger in early 80’s for an SLA backup. The waiver is obtained from local  authorities (province of Mb , in our case) so any potential incident would not create a legal liability.

• Or you may want a fancy PFET with 50mV shunt amplifier as a fancy current limiter with low gain feedback or high gain , you may choose to use this before external (10uF) cap near or in load or after an added bulk cap to improve transient response or reduce unwanted transient effects depending on load and application and how it is switched ON.
• some ancient or poor PSU designs needed a <10% or 10W dummy load on 5V only to function well, that was a normal practise at one time.

Anecdotal

• web buys are pot luck
• I recently got a nice variable 3~24V 3A supply incl LED DVM , cheap, to power external ~ 16V custom fence & garden LEDs
• it worked really well until it didn’t....
• 3 months operating ~75% current rating in dry cool environment, no lightning storms. It was on a digital timer possibly Triac controlled, which may have caused more stress if async abrupt daily cycle to input current surge. Not verified but fault on input AC side.
• We don't have certs problems here in (most of) EU. Between the public net and the house there are (mandatory) multiple technologies (magnetic, thermal) switches for safety purposes. All we have to do is checking the earth pole in the garden every few years. And then the switches will save our asses ... – Anichang Aug 26 '19 at 15:56
• Please check the diagram. I didn't isolate the ATX AC side from ground. And I'm not sure to understand when you wrote "use 1nF to 4.7nF plastic cap across earth disconnect to block DC": did you mean to place that cap instead of some insulating tape? – Anichang Aug 26 '19 at 15:59
• Your block diagram has no Earth Ground. Then why do you need a transformer if not to isolate earth ground. I meant if you MUST isolate AC from Earth ground any least bypass switch or open connection with an RF plastic cap. I suggest you scrap your design. – Tony Stewart EE75 Aug 26 '19 at 17:16
• Regarding safety CERTS in EU AC powered units or adapters. This is mandatory globally. The Cert becomes void if unit is modified by 3rd party.. But no AC Safety requirements for DC powered units, just the AC adapters. Regarding RF Certs, these are controlled by local CE, and FCC. Yet in a lab, you can break any rule you want, take measures and suffer incident consequences. – Tony Stewart EE75 Aug 26 '19 at 17:18
• there are 2 earth ground connections: 1 in "Noise filter 1" section (not explicit, but if you look at the noise filter picture, you'll see it), and 1 in "Noise filter 3" (probably the wrong symbol, as I couldn't find any better, but I named it "Earth"). And despite those 2 connections, the DC side stays isolated from AC and earth ground, as the first connection is on AC side, and the second is via in line caps (ie: DC have no connection, but caps remove AC noise on GND). – Anichang Aug 27 '19 at 2:30