AT PSU Voltage Sag? and Mains Earth Isolation, for ATX and USB

[rwgast_lowlevellogicdesin]

So I just connected an old 230watt AT power supply to a 2amp load and the thing started sagging majorly, on the 5v rail it dropped down to 3.9v and on the 12v rail it fell to about 7.8v. I have seen people use ATX PSUs as bench supplys, but I had this AT around which is a lot easier to work with. Is it common for a supply to drop like this? Will ATX do the same thing?

I want to build digital desktop supply at some point using some of the techniques in daves 8 part series. The thing is you still need a way too convert and step down mains power. I have a few transformers around, but I figured that maybe using a computer PSU for the main supply would be a better way to go, as its much more energy efficient and I thought it had plenty of power? I had planned to step each rail up by 3 volts using a boost regulator and then attach the LT2085's or what ever LDO's I would use to get a nice clean regulated signal. I also wanted to parallel them to draw bookoo amps, but im not sure if a computer PSU can handle this now.

[IanB]

Where did you put the load?

Computer power supplies may not expect to see a load on the 12 V output without also seeing a load on the 5 V output. Try putting a dummy load on the 5 V rail and see what happens.

[mariush]

Since it's an old AT power supply, it relies on the 5v output to regulate the other voltages.  The 12v and 3.3v output won't be stable and will sag down, if there's nothing using the 5v output.

Get a 10 ohm (5w or better) resistor and put it on the 5v, it should use about 0.5A and make the psu a bit more stable.

An AT/ATX power supply is not always the better way to go.  There's a big difference between a linear power supply with a classic transformer and an at/atx power supply, when it comes to voltage ripple and ability to limit current and other things. The differences may matter a lot, depending on what you need the psu for.

And the AT/ATX power supply is probably not isolated (almost all of them aren't), while a linear power supply is most often fully isolated.

[rwgast_lowlevellogicdesin]

I loaded the 5v rail first as I mentioned above and it saged down quite a bit, over 1V. This old AT supply only has +12, -12, +5, -5 no 3.3 rails.  Like I said if I were to use a PSU as the base I would definately boost the voltages and step them back down with linear regulators.

As far as PSUs not being isolated, I was under the impression ALL transformers were isolated from the primary side, both PSUs I have have multiple smaller transformers.

[darko31]

Since it's old ATX PSU, have you checked inside to see if all capacitors are good? Bad output filter caps could cause voltage dropping.

[alm]

The transformer provides isolation from the phase/neutral mains wires, but the secondary is tied to ground, so the outputs are all ground-referenced. This means you can't connect multiple supplies in series for higher voltage unless you perform potentially dangerous modifications to the power supply. Lab supplies are also usually isolated to prevent ground loops.

[mariush]

The ground lead in the connectors on the output is tied to ground of the mains connector.  You can easily test this by putting your multimeter on continuity and (with your power supply unplugged from mains!)  put a probe on a black pin of a random connector and the other on the earth pin on the back of your power supply.
The neutral is also tied to earth somewhere in the house electric panel or in your building.

If you accidentally put the probes of an oscilloscope where you shouldn't, current can flow through this earth wire, through earth, and form a connection through the oscilloscope's probes... and boom, you blow up some traces or probes, whatever is thin enough to act as a fuse.

I have for example an old Delta DPS-100tb-1 pulled from an old 486 computer, made a thread about it here:  http://www.badcaps.net/forum/showthread.php?t=20875&highlight=delta+hp+vectra

You can see if you read the posts that by default the PSU has about 5.3 and 11v on the 12v rail, but when I connected a lightbulb giving about 0.25-0.3A load on 5v, the voltage went up to about 11.4v on the 12v rail.

Never finished that project, got busy with other things, but I can tell you I am able to turn that white potentiometer and tweak the voltages. Without any load, I could turn the potentiometer quite a lot and get as low as 11.8v (stopped turning there, could still turn) but when I checked the 5v rail, it was already at 5.8v!
Obviously, the pot is a rudimentary way to tweak the voltages, if I really wanted or if you want it, it's possible to read the datasheet of the controller and change some resistors to alter the outputs.

A classic transformer has the two windings fully separated and the output can be left floating, no need to ground it.  In fact, that's why a lot of adjustable power supplies ( like the Tenma I posted about here have 3 plugs for each output - the negative is floating, but can be tied to earth if needed.

[Monkeh]

The neutral is also tied to earth somewhere in the house electric panel or in your building.

Please don't make assumptions and broad statements like that.

[rwgast_lowlevellogicdesin]

So you saying if I use a PSU, to power a project, the its ground becomes referenced with the earth ground system of the house? An actual bench supply would not be referenced to earth ground correct? So would putting opto isolators on the output of each supply rail be a good idea?

Also I am trying to build a solder station with a 60W 24V transformer it only has a two wire plug. I planed to chop that off and and a grounded cord, then lugging the ground wire to the chassis and the transformers core. Now on the secondary side will I still be isolated from earth ground? Also a Hakko iron has an ESD pin to ground the tip, I was assuming this went to the secondary sides DC ground after rectification would that be correct?

[mariush]

How would current flow through optoisolators?  An optoisolator is just a led and a optical sensor, they're for isolating signals.

What I'm saying is that by default AT/ATX power supplies are not isolated, they're ground referenced. Sure, you could hack them to be floating, but it seems you're really a beginner at this so you shouldn't do it.

The AT/ATX power supplies also have other downsides.. the transformers used are for high frequency and the power supply runs at something like 40kHz-150kHz so the output sees high ripple and the voltage is super stable. A linear power supply with a classic (big, heavy) transformer gives you much more stable voltage and less ripple, as the regulation is done differently.

An adjustable linear power supply has a transformer with a primary winding and several secondary windings. Each output is powered from a secondary winding, the control mechanism (opamps, microcontroller if any) is powered from a separate secondary winding, relays (if any) are powered from a secondary winding, everything is floating. You can optionally reference each output to ground, but you don't have to.

Soldering stations don't have ground, only 2 wire plugs, because the transformer inside is the classic kind, big and heavy, and it's shielded from the user by the plastic case. User can't tough any high voltage stuff, and the 24v is from the secondary winding which is floating. If you use a grounded cable and tie the ground lead to the metal of the transformer you don't gain any benefit, the transformer was already protected inside and user can't possibly touch it, and the whole station is powered from the floating 24v like I said.

[Alana]

Typical AT/ATX ground [black wires] and earth ground are connected by connection pad near one of the mounting holes on LV side.
Insulating washer and earth connection is gone but case is still grounded so is HV side EMI filter.

[IanB]

Soldering stations don't have ground, only 2 wire plugs, because the transformer inside is the classic kind, big and heavy, and it's shielded from the user by the plastic case. User can't tough any high voltage stuff, and the 24v is from the secondary winding which is floating.

Whoa there, hold your horses!

The Hakko FX-888 has a three pin grounded plug, and the ground pin has low ohms continuity with the tip of the iron. That's how it can be "ESD safe" -- the iron tip is grounded and should never have a floating voltage on it that can damage components.

[rwgast_lowlevellogicdesin]

Im sorry ive been playing with electronics for a while, im also an electrical apprentice. So I know what I am doing for the most part, but I haven't really had the nerve to start plugging projects and transformers in to the wall until recently. The opto Isolater thing was a pretty stupid comment, and I realized it as soon as I started thinking about it when I walked away from the computer.

As far as the switching frequency of the PSU, im not really concerned about that to much, you should be cleaning up signals in your project anyway. It wont always be run with a nice linear bench supply, unless it never leave the breadboard. Most of my projects use DC to DC buck regulators that step down a switching wal wart. On the other and the isolation thing does worry me a bit, but I think that is probably because I dont really understand why having your projects ground tied to the earth reference is a bad thing? Isnt this exactly why a hakko has that ESD pin that goes to the earth ground? Im trying to read up on the whole ground isolation thing.

[mariush]

[alm]

As far as the switching frequency of the PSU, im not really concerned about that to much, you should be cleaning up signals in your project anyway.

Lab supplies are often used in early stages of development, or when troubleshooting. I don't want to be worrying about power supply noise when trying to power up a design for the first time. Or when I suspect the power supply is faulty, so I temporarily power it from a bench supply. Lab supplies are to exclude the power supply circuit, not to stress test it.

On the other and the isolation thing does worry me a bit, but I think that is probably because I dont really understand why having your projects ground tied to the earth reference is a bad thing? Isnt this exactly why a hakko has that ESD pin that goes to the earth ground? Im trying to read up on the whole ground isolation thing.

It's essentially a matter of control. You can easily ground a floating output. Most bench supplies feature a third terminal, connected to earth ground, in addition to the positive and negative output. So you can easily connect one of the outputs to ground. Making a grounded power supply float (safely) is not so easy. Some examples of applications of floating power supplies:

- You want more than the say 0-30 V your supply can deliver. So you put two in series for a 0-60 V voltage range. Connect your circuit ground (and optionally earth ground) to the middle for symmetrical power supplies.
- You want a negative supply relative to ground. Just connect the positive terminal to ground and the negative terminal is suddenly negative to ground.
- You want to power a circuit that is already ground referenced. No current will flow from the floating power supply through the ground lead, regardless of the connections within the DUT. An example is the scope probe mentioned in the above video. You can easily clip the ground lead to the +5 V rail of your circuit as long as the circuit is floating. Even if you clip the ground lead to circuit ground, then there may still be currents flowing through the scope probe due to tiny potential differences between the instruments (ground loops). These currents can induce noise into sensitive circuits, just ask the audio guys.

[rwgast_lowlevellogicdesin]

Oh wow, I completely understand after watching that video. I never really thought of the scopes probe causing a short through the mains wiring! I also know why Ive killed so many USB hubs experimenting with stuff!

What if one were to just disconnect the earth ground from the neutral inside the PSU, would that keep it from floating? Or what if you just totally disconnect the ground pin for the psu?

[alm]

If you disconnect the earth connection within the PSU, then you are defeating safety and EMC features. If you need to ask, then I would suggest that you don't have enough knowledge and experience to do this safely.

[rwgast_lowlevellogicdesin]

Well I thought I would share a little bit. First of all if I needed to disconnect the earth ground to the PSU, my plan was to put it in a plastic enclosure and then connect a breaker or GFI to it, I would assume this would have taken care of any saftey issues correct?

But alas I did not need to do this I have checked every old PSU I have laying around AT and ATX style, there earth ground only runs to the chassis that is it. Reading resistance between earth ground on the plug and, every wire on the PSU gives an OL reading. Im not sure if it is just because there old and cheap, these PSUs definitely have floating outputs though! Is this uncommon? The ATX supply was bad though, when I flipped it on some of the smaller signal diodes around back up power let out smoke, and the thing wouldn't stay on even with the POWER OK (Grey) wire connected to 3.3 or 5v. The AT supply still work fine, but other than running fans I really do not see a use for it, so I am just going to go the route of a linear supply using some old 3AMP+ 24v transformers, i have managed to salvage from typewriters, printing calculators etc.

In a semi related note... after watching Daves video about earth mains I tested my USB ports just to make sure, and sure enough there grounds are low impedance to earth ground, about 500 ohms. Considering I do tons of work in my projects using USB, and even use a USB PC scope (Don't judge it is a parallax item, not a great scope but it was only $50 instead of $200, and comes with a nice book that teaches you how to use a scope, it was great for learning and I do realize an ebay analog and rigol combo are what I need to shoot for) this was upsetting news. Ever since I have gotten in to electronics I have blown a few passive USB hubs out, and even ruined one USB port on my PC. I now know that the earth ground is the reason for this, along with drawing to much power when I short something.

I want to fix this problem! So I looked in to USB isolators, might I say WOW WAY TO MUCH MONEY. They are close to $100 for a single port! I then found this kit, https://www.circuitsathome.com/products-page/usb-interfaces/adum4160-usb-isolator-board-populated for $50. Now before I found this kit I was thinking about how one may go about putting together a USB isolator. And came up with two ideas, I am hoping someone could tell me whether I was on the right track,

#1 just build a board that plugs in to a USB port, then connect opto isolators on the signal lines, while floating the V+ and V-. On the isolated side of the opto couplers, add a battery/wal wart powered 5v LDO to power V+/V- and then do whatever is needed to bring the optocoupled signals up to 3.3v. Sorry have not used an opto-isolator I do not know exactly how the signal gets powered on the out put. Im not even sure you would need the isolators, if you just used the original signal with a stand alone LDO for power, I don't see where there would be an earth reference.

#2 some kind of elaborate system, using an isolation transformer. This would involve converting DC to AC then back to DC. Seems complicated. I'm wondering if this is how a commercial USB isolator works, if so it would explain there price.

Now if isolating USB lines is as simple as floating the V+/V-, and using a floating LDO for power well that is pretty simple and probably the best way to go....

But if it requires more than that I found something that people may be interested in! The kit I linked to above is not open source, so no schematic. But there is also, very little on the board. Looking at it I could tell right away most of the passives were for the switching regulator, and had nothing to do with USB. The only thing besides a switching regulator is a chip, ADuM4160. So I went to Analog Devices and checked this thing out

http://www.analog.com/en/interface-isolation/usb-isolators/products/index.html

turns out it is a full USB isolation solution, that requires NO external components and has a built in 3.3v regulator. Not only can you sample these chips, but DigiKey sells them for $6 in one of's. This would be perfect to easily and cheaply build a powered USB hub with a few isolated ports. Much less than $100 for a single port, prebuilt solution. I really hope that these $100 single port isolators are not built using these chips.. other wise it is a huge rip.

[mariush]

You're right, usb isolators are expensive, but that's because they can do full usb speed usually (480mbps).

I am actually planning to make a Youtube series or a long video with how to build a power supply with usb isolation in the following days (going through transformer selection, diodes/bridge rectifier, capacitors, choosing linear regulators and heatsinks etc )

Going to use Adum3160 for the usb data isolation, it's a cheaper version of Adum4160 but for home usage should do the job. 

That site HAS the schematic, but it's not much different than the one in the official adum datasheets: https://www.circuitsathome.com/wp/wp-content/uploads/2010/10/USB_Isolator_r10.pdf

These isolators can only do 12 mbps, but that's enough for probably 99% of anything we hobbyists would have.

You would normally power one side of the Adum chip from the usb port (computer) and the other from an isolated 5v source... I'm going to use a linear regulator for it, probably  ld1117v (min. 800mA , 1.2v dropout), the datasheet above uses a switching regulator but the basic ideea is to produce 5v so anything would work.