Any interest in a DCDC UPS build report?

To be frank, it’s actually easier to island the entire house from the grid and run off decent size battery stacks then a small, purpose built UPS. At least in the UK.

A simple earth spike, a hybrid inverter (or two in parallel) and a couple of LiFePo4 batteries and everything just works. It doesn’t have to be that expensive but it can scale pretty high, up to apocalypse style doomer scale, especially with solar panels.

This DC-DC backup just looks incredibly complicated in comparison, but it looks like an interesting project and definitely cheaper.

FWIW, my solar electricians actively dissuaded me from same. At least over here the islanding feature works so well that several folk had it ripped out. I was told to wait for next generation islanding Intertie from Enphase.

All I want is a solar backup for fridge and heat. Have solar but it’s load following so if the grid dies, it must also. House, fridge would be fine if they only ran 5 hrs a day in the depths of winter.

I’ll put together a system diagram. In some ways it’s actually simpler than the traditional DCACDC approach since you skip all the PSUs.

Stickied this thread. Looks like a fun project with great goals.

I’m a bit of a horder of computers and parts. So, in many ways, I have the opposite problems with my off contract old enterprise gear…between…space and power…

Semi-ish related story:
The previous guy(s) who owned my house was a…real “Tim the Tool Man Taylor” kind of guy. I’ve been slowly replacing alot of his “work”…but one of the unfortunate design decisions I am stuck with is the layout of the mechanicals which have been changed far too many times for a house of this age.

I have to share my homelab space with a sewage injection pump :grimacing: (for the shower/toilet in my basement). The only other place it could go would be near a hot water hearter…and thats always pressurized water instead of intermittenly pressurized water, I’m not sure which is worse.

Either way, I have to fit in a 4x8 closet. I chose the one with the sewage injection pump because it has the electrical panel/telco demarcs. It made it quick and easy for my to get more power - that room has been upgraded to have 2x20 Amp circuits in it so I actually have proper-ish A/B power with two Eaton 5PX2200s. But like…trying to do things the “right way” with the space constraints in a home environment is easier said than done…it kinda sucks but I love it.

I do still have the same problems you do in a small build but at a larger scale.
My “MDF” consumes about 600-800W continous on both 20A circuits as reported by the Eaton UPSs front screen.

So like, I don’t have the heavy duty “Enterprise Grade” “Double Online Conversion” style UPSs which I wanted…the ones I have are “line interactive”…https://www.eaton.com/us/en-us/skuPage.5PX2200RTUS.pdf

Spec sheet says its “EFFICIENCY AT MAX LOAD is 97%”. I’ll call it 95% because I never believe the spec sheet. All of the power supplies in all of my servers and switches are all Gold or better. 80 Plus - Wikipedia
We’ll call that 85% that silver is rated for just to have nice round numbers.

So from the wall, I am consuming ~630W on the low end to like ~840W during “normal load” for each circuit so 1260W to 1680W ish.

I’ll call it 2000W for simplicities sake and adding some margin of error on the safe side. I’m assuming electrical load can actually increase this much for my equipment under heavy load.

So now if we assume the 80 Plus silver again (which is also conservative) I have generated about 300 Watts of waste heat. Theres also AC fans and some other things here so this is an oversimplification.

In any case, no matter how you measure it. My cooling requirements have increased pretty substantially just by virtue of plain old waste heat. 115% ish is really not bad, its amazing computer equipment is as efficient as it is. But it’s still signficant enough to matter in my wallet.

AC-DC conversion sucks. But its fine :slight_smile:

2 Likes

Good news… I think it is all working out as expected. So, when the battery level gets very low (43V), the VIGN signal will go low. That in turn makes the DCDC-NUC send the shutdown command by emulating a power key press (USB). No special software and the NUC can do it hardwired too, if you prefer that approach.

So, I still have some cleanup to do (i.e. gather all those PSUs, make sure each is labeled, put in a sealed plastic bag) and the APC UPS that used to feed that circuit has been retired for now. There are far fewer wires flying around and overall neatness is up a notch.

With a tip of the hat to @ericloewe for the sticker inspiration.

4 Likes

You would make a killing on etsy or eBay selling these as a kit all fancy schmancy and safe ish. But open source. You should look into it :slight_smile:

This guys business model is super cool so you can see what I mean:

I would buy one.

Well, at the rate at which my client is getting decimated by executive branch cuts, that may have to become my backup plan… my team is down over 50% and the client losses are similar. Just wait until the RIFs go through… :frowning:

On a side note: everything here is fused. But I’m having a lot of trouble finding 58VDC rated blade fuses. It’s one of the limitations of this design - most ATO / ATC fuses are only rated up to 32VDC… the ATR and like series from Optifuse go higher. But they are as rare as hen’s teeth.

Never underestimate the ability of DC to maintain an arc. I really started to appreciate that when I built a whole house humidifier. I kept welding relays shut until I saw this video and went for a set of high-voltage DC rated solid state relays instead. It’s been quiet since.

I am not sure, whether you are interested in some improvement ideas?
If, yes, just let me know!

1 Like

Absolutely, would love some feedback!

OK, so some ideas then:

  • all components could be mounted on a DIN rail, either by getting that version of them (the main PSU) or you can buy or 3D print special holders to do that.
  • You can buy a small, plastic (or metal) box (I am not sure of the proper english term for them I think they are called electrical switchboxes?), that is specially desinged for this DIN rail system. It is usually really cheap, it has professional look, and it makes your mounting tidy and easy. Also, it has prepared, but originally closed openings, tat are even water tight in most of the cases. So, if you want to route any wires in- or out of that box, you yout break/cut out the opening and pull the wire accross.
    (however, as I see that DCW is really a deep unit)
  • Such a cabinet makes your makeshift wooden spacing blocks unnecessary, since you can just simply route the wires through from the top to tha bottom.
  • Maybe, you could replace those automotive fuse panel with a couple circuit breakers too, since those can be just simply swithed back, while these fuses, needs to be replaced, if they blow. (However, it is always good to have a different mechanism protection in place as redundancy.) Those are also nicely fit on the DIN rail.
  • Such a box prevents it from dust and also unwanted touch.

Otherwise, this is already a great and practical build!

1 Like

I don’t remember ever having said that, but it is so what I might have said.

1 Like

You made me look it up! You are right, you didn’t say that, it was @Stux and my wetware misremembered. Apologies! :slight_smile:

1 Like

Great suggestion except I wasn’t able to find a 94% efficient PSU @ 48VDC in a DIN rail format. Even Nextys doesn’t make them even though they are part of TDK-Lamda.

Great suggestion for a commercial product though you have to take adequate care about waste heat. Let’s start with the Meanwell PSU - at 94%, you are getting 6% conversion loss or 9W of heat at about 150W. That has to be dissipated and the smaller the box the less surface you can use as a passive radiator. Then there is the DCW20 and the DCDC NUC which probably add another 6W or so of heat, i.e. 15W total. Put a 15W lightbulb in a small, well sealed box and see what happens… you’ll potentially be frying the electronics in their own juices.

The meanwell manual is pretty specific about the size of the radiator that they want you to screw the PSU to. I used 12x12x1/8" aluminum with a standoff to allow convection on the front and the rear of the Aluminum plate, resulting in a surface that is only very slightly warmer than the surrounding air. I really wanted something as passive as possible.

Another great suggestion. My main reason with going to fuses is that fuses are relatively inexpensive and circuit breakers that can handle 50+VDC are not. Three DIN-RAIL-mountable 1A circuit breakers cost as much as the entire meanwell PSU. That becomes kind of hard to justify, especially if the main reason to have them is to prevent unexpected catastrophic damage.

Given the difficulty and expense of ATR fuses ($3 ea vs. $15 for a circuit breaker) a more mature design likely would use 5x20mm fuses as a baseline and circuit breakers on a DIN-Rail as a upgrade option.

The most important thing is that there is protection that can interrupt the very high currents that large batteries can develop. Your circuit breaker better have the interrupting capacity given the arcs that DC systems can develop.

One of my final projects around this assembly is better taming / protecting the wires. I may use some of the that fun Aluminum strut stuff to mount a 12x12 plexi sheet in front and on the sides, while leaving the top and bottom open so the structure can continue to vent as nicely as it does.

Thank you again.

“Great suggestion except I wasn’t able to find a 94% efficient PSU @ 48VDC in a DIN rail format. Even Nextys doesn’t make them even though they are part of TDK-Lamda.”
That is totally a valid reason.
I did not do this research.

“…you have to take adequate care about waste heat.”
That is also true.
However, I did a real life experiment for a premium German car manufacturer exactly for such concerns:

  • It was a new sedan prototype.
  • The trunk vas covered in a carpet -like material, wrapped aroud a hard foam board
  • NO forced heat exchange was provided or been possible in this setup
  • Our audio amplifier, rated for about 3-400W contious power at about 75% efficiency, is located in the about 32l compartment in the trunk aboutwhere the spare wheel used to be when is still was a thing.
  • We played white noise (the worst case scenario for Class D amplifiers efficiency) on the system with full power for about 10 hours and the temperature raised until 54°C.
  • After that we placed an actual heat gun next to the amp, but we were not brave enough to set the temp above 90°C. Another 10 bours raised the femp up to almost 70°C
    So our conclusion is that overheating such thermal systems is really a hard task.
    SO, I would say, that you are quite safe to put all 15W of that waste heat into even a plastic box.
    If you pick a Steel or aluminium box, that makes heat exchange even better!
    Of course, you are free to not to do it, if you want to stay on the safe side!

“The most important thing is that there is protection that can interrupt the very high currents that large batteries can develop.”
Yes, I am aware of all those caveats, and that is why I recommended using fuses as a “last line of defence” in my reply!
If your 3 USD fuse is blown, it will cost you 3 USD, the 15 USD circuit breaker, you just switch back. It is obvious, that you catch up with your costs with the fuses to circuit breakers. (you use 3 fuses, that is 9USD vs 45USD of the circuit breakers. That leaves you 12 fuses away from turn to a loss.)
Also, circuit breakers have different characteristic, and you have to pick the correct one (speed, overload capacity, etc) for your purposes.
It is usually more practical to build a tiered protection for electronics systems.

  • Have a fast protection for the subsystems, that reacts quickly on any irregular events in that local subsystem.
  • Then have a slover reacting (like one order of magniture slower), but much biefier protection for a group of sub systems (like all the power circuits, but not for the control device, so even if any part breaks, at least the control or monitoring system still functions )
  • Have the actual “Last line of defence”, when all goes south, this, really powerful, but slow system shuts everything down, to save the rest of the property.
    The first two levels can be implemented by circuit breakers ant the last one by fuses.

THe plexi cover is a good idea. I dont konw, how dusty or humid your enviroment is, so I cannot judge, if that is OK or not.

1 Like

All valid points. Thank you.

I reckon your amplifier also benefitted on the bottom side from conduction? That’s the issue with DIN Rails, they offer relatively little cross sectional contact area for conduction to happen. Instead, air likely has to convect most of the heat away and the larger the surface area of the box and the better its emissive, the cooler the electronics will be able to run.

Presumably, that is why Nextys suggests air spacing between each DCW20 even though they are relatively efficient? The DCW20 case is optimized for convection with lots of holes on top and bottom while the sides are solid.

The meanwell has a giant flat underbody perfect for conduction while the top is largely open.

Neither unit should be used anywhere you expect potential water damage, etc. That would likely end as unhappily as a fanless ATX PSU getting rained on.

Anyhow, a lot to chew on an consider in case I ever want to take this install to the next level. Thanks again.

Actually, all computation is a side-effect of heat generation :wink:

CPUs are just fancy silicon heaters.

Your 3000W is probably generating a lot more than 300W of “excess” heat…

3 Likes

I can honestly say I have not turned the heat on in the basement. :).

This thread has been very fascinating to read. You had me at are any of you interested in hearing about this quest.

This is an interesting prototype, can’t wait to see what you may do with it over time. Transplant it into a more professional looking case, slap the name of your new company “Ghetto but Awesome” logo on the front. Have some nice quick disconnect plugs on the rear and lots of adapters.

I think your next step, generate specifications of the device. What can it supply, voltage/current/time and so fourth. Heat generation as well.

Do you have other photos worth sharing of the stack of batteries and connections to the equipment itis powering?

I like the fact that you took your time to layout the wires nicely. Many people do not do that, they slap it together, it works, they do not complete it by organizing the cabling.

Comment, in examining the cabling, I noticed that the output of the power supply (top left), the wire connectors “look” a bit close. I can’t see the actual gap but it all looks like you know what you are doing, I’m sure it just appears close from the photo.

Second Comment, the rather large cables entering the DCUPS, are the cables terminated using crimped pins or loose wire, maybe soldered at a minimum? Same type of comment, electrical short concern.

I’ve seen some nasty electrical short, seen some bad accidents with DC power. One fella had his mouth (upper lip and cheek) burned off. He was very traumatized. He could not return to work after that, ever. So high voltage and current always gets my attention.

It is very safe to assume that all energy consumed by a computer or computer like device is radiated as heat. The exceptions are things that eventually turn into heat (for example the light coming from a display hits a surface and … is turned into heat).

I’m with Joe on this. Thanks for taking the time to detail your experience. All your projects are thought-provoking and inspirational - I can say that I bless your pointers on tools useful when updating wiring and cabling in old houses each time I use them (Magnepul and Depstech). To them I add a set of fiberglass probe rods to make up my set of “infrequently used life savers”.
One of these years I’m going to make a Halloween lawn decoration to compete with our neighbors’ inflatable awfuls.