Recommendations to convert SLA battery to multiple voltages

[krby]

Please let me know if this belongs in another sub-forum.

In a previous post I asked about designing a DC-DC battery charger. I haven't made progress there as life has gotten in the way. Lately, I've been thinking about the other end of the DIY UPS, powering the load. I'm currently using a separate inverter. But all the gear I'm running  runs on DC internally. So, I'm thinking about taking the SLA battery bank (~9-15VDC), doing some DC-DC conversion, making some cables and just get rid of the inverter.

I've looked at the things I need to power:
* Four devices use wall warts that output 12VDC. I have appropriate connectors for all of them. Adding up the current on the wall warts says I'll need a max of 12A at 12V. But I measured the average in use power for these devices together they total about 2.5A

* My switch has an AC/DC PSU internally, it outputs 54VDC and is rated for 1.1A/60W. There's a single connector from the PSU to the mainboard of the switch, and it's a Molex Micro-Fit Jr (I asked the PSU maker for dimensions)

* My consumer NAS box has PC-like PSU inside, I need to open it up to check, but I think it's going to be an PC ATX connector. The easiest thing is to buy a  DC-DC ATX PSU module. These typically take 12V or 24V in. I haven't checked this PSU specs, but this is probably 300W max, about 100-150W average.


I'm looking for advice on how to architect the conversion. It comes down to:
Power source is 9-15VDC.
I need to provide a peak 150W at 12V, but only about 30W continuously.
I need to provide 60W at 54V
I need to provide peak 300W, continuous 100-150W at something like 12V or 24V depending on what DC ATX PSU I find.

A simple approach is to find variable input DC converters that all use the battery as input and output the needed voltage (12V, 54V, etc). Doing some reading, I've come across the idea of using an intermediate voltage IBA. So I could step up the battery voltage to something like 60V, then step down for each output I need to provide. The advantage here I think is that each point-of-load step-down (I think I'm using that term correctly) only has to deal with a fixed input voltage and fixed output voltage, so they are simpler devices. Also, current between the initial step-up and the various step-downs in pretty low.

Is one approach better than the other? If I use the intermediate voltage, what guides the decision for that voltage? 48V seems like a common intermediate voltage, but it seems simpler to me to have a single step up that is higher than all the needed outputs so I'm only using step downs for each output. Not sure if this matters.

When I build this, I'm going to wire it up as separate modules in an enclosure. But eventually, I'd like to design a PCB that replaces this. Not sure if that matters when giving advice, but thought I would mention it.

Thanks for any help!

[thm_w]

Your switch must be POE, is that required? If you get rid of that, most standard switches would run on 12 or 5V. Then you'd just be left only needing something like a 12V buck boost converter, to give you a solid 12V output.

Consumer or custom NAS box? Most consumer NAS box run from a 12V DC supply.

If you can't get rid of the POE, then it might make sense to do battery -> 48V -> 12V.

https://www.aliexpress.us/item/3256804493402097.html
https://www.aliexpress.us/item/2255800294580515.html

[krby]

First off, thanks for the reply!

Your switch must be POE, is that required?

Yes, it is. I power some wifi APs and a couple of other things from it. I've bought it because it had PoE.

Consumer or custom NAS box?

Consumer, well I guess Prosumer? It's a QNAP cube form factor. I had it open a few years go, it felt very similar to a small PC or laptop motherboard, an AMD embedded SoC CPU, two PCIe slots, SODIMM ram slots, etc.  I'll open this weekend to check what the internal PSU actually outputs, but on a previous QNAP I had, the internal PSU had a standard ATX connections. I did find out it's only 250W max. Since it's an embedded SoC CPU, maybe I'll be lucky and the internal PSU feeds a single voltage to the main board.

If you can't get rid of the POE, then it might make sense to do battery -> 48V -> 12V.

I'm not quite following. The POE switch needs 54V to run, where does it get that?
Assuming you meant: batt -> 54V -> 12V, can you tell me why this way vs? batt -> 48V and batt -> 12v? That's essentially my question, which architecture and why.


Thanks for the links. I'm not yet looking for gear, more trying to understand how to architect it.

[Solder_Junkie]

For several years I have used an APC UPS to avoid problems with my CCTV system, more recently I have been operating amateur radio by remote and needed to keep the control desktop PC running despite occasional "glitches" in the mains supply. In the past week I recorded 4 interruptions with another APC UPS bought to supply the PC and radio equipment.

I am very impressed with the UPS, which runs on a single 12V 7Ah gel cell, has a USB connection to the desktop PC and Windows software to both monitor the mains and battery Voltage. On mains failure the PC sends me an email and if the mains remains off for 6 minutes (configurable), the PC shuts down in a controlled manner, followed by the UPS shutting the output off to conserve the battery.

While I appreciate that most of us like to build our own version of something, the features and performance of these units is something that would be hard to replicate in a home built version.

Bottom line... why not simply use a commercial UPS with various mains power supplies?

SJ

[Phil1977]

There are many advantages of using an intermediate voltage for that setup. Let me list some of them:

- Boost converters can have conveniently high efficiencies. >95% is possible even with cheap stuff. So the primary conversion does not waste much power.
- Buck converters can be chosen for adequate efficiency. Even mediocre ones can have better efficiency than most buck/boost architectures
- It´s easy to direct the power flow in such a system. If you have more than one battery or occasionally a mains power supply then it´s easy to connect so that you can control who is sending and who is receiving power.
- It´s easier to protect the attached devices if you can set a current limit for each consumer

The next can be advantage and disadvantage the same time
- You have no buffer battery in the intermediate voltage. Voltage drops and overcurrents are easier to detect than with a buffer battery. But they may occur and you may have "brownouts".

[krby]

While I appreciate that most of us like to build our own version of something, the features and performance of these units is something that would be hard to replicate in a home built version.

Bottom line... why not simply use a commercial UPS with various mains power supplies?

I used an APC UPS for this network rack for many years. Its runtime was ok, but I wanted something longer, I knew enough about batteries, charging, and inverters to build a system from discrete products, so about a year ago, I built my DIY UPS from a charger, some batteries, and an inverter. I had much of what I needed already. I wired up and Arduino and some sensors so that it pretends to be a USB port on a commercial UPS and notifies my NAS to shut down and email me.

I agree, sometimes buying is the right choice. For me, I explicitly wanted a project. It's working now, but I want to improve it. That's why I'm posting here.

[krby]

Thanks for this, so it sounds like the overall direction is boost, then several bucks. But, I lose the battery as a buffer on the intermediate -> load conversion. I think I'm ok with that.

Also, is there a different sub-forum I should be using for this? Maybe Projects, Designs, and Technical Stuff

[thm_w]

Projects or Beginners is fine.

I said 48V as its a common voltage and the POE switch might still work with that. If it actually needs 54V then yeah, can use an adjustable boost converter to get 54V.

If you want a longer run time look into an external battery, if you want long term reliability look at LifePO4 12V batteries.

[krby]

I said 48V as its a common voltage and the POE switch might still work with that.

48V is definitely more common in the world. How would I figure out if the switch would be ok with this? The reason I think it runs on 54V is I opened it up last week and found a PSU labelled nw-60a54 by APD, which says it has a 54VDC 60W output.  When the PSU powered on, I measured ~54.4V on the output. I would love to be able to run this on a more common voltage, but unsure of how to figure out if that's safe for the switch.

If you want a longer run time look into an external battery, if you want long term reliability look at LifePO4 12V batteries.

What I have now is an "DIY" UPS I have put together with external batteries, an RV charger/converter, and an inverter. This thread is about me trying to get rid of the inverter. I chose SLA instead of LiFePO4 because they're lower cost and a better fit for this exact use case. They're heavier, but they'll sit on a shelf their entire life. They'll spend nearly all of their life fully charged, which is fine for SLA but not as good for LFE. Also, since I'm not going to cycle them very often, I'm willing to discharge the SLAs to 75% DoD so the usable capacity advantage that LFE typically has over SLA isn't as big a deal compared to the higher cost.

[Phil1977]

There are good reasons to use 48V as an intermediate DC - not to hazardous, easy single step conversion to low semiconductor voltages, easy buffering with batteries etc.

Without knowing the details I suppose your PoE-Switch will also run with 48V. Usually PoE is specified for quite heavy resistive losses in the cable, so if it´s 54V nominal then 48V at the input of the router port could even be in-spec.

[MrAl]

Please let me know if this belongs in another sub-forum.

In a previous post I asked about designing a DC-DC battery charger. I haven't made progress there as life has gotten in the way. Lately, I've been thinking about the other end of the DIY UPS, powering the load. I'm currently using a separate inverter. But all the gear I'm running  runs on DC internally. So, I'm thinking about taking the SLA battery bank (~9-15VDC), doing some DC-DC conversion, making some cables and just get rid of the inverter.

I've looked at the things I need to power:
* Four devices use wall warts that output 12VDC. I have appropriate connectors for all of them. Adding up the current on the wall warts says I'll need a max of 12A at 12V. But I measured the average in use power for these devices together they total about 2.5A

* My switch has an AC/DC PSU internally, it outputs 54VDC and is rated for 1.1A/60W. There's a single connector from the PSU to the mainboard of the switch, and it's a Molex Micro-Fit Jr (I asked the PSU maker for dimensions)

* My consumer NAS box has PC-like PSU inside, I need to open it up to check, but I think it's going to be an PC ATX connector. The easiest thing is to buy a  DC-DC ATX PSU module. These typically take 12V or 24V in. I haven't checked this PSU specs, but this is probably 300W max, about 100-150W average.


I'm looking for advice on how to architect the conversion. It comes down to:
Power source is 9-15VDC.
I need to provide a peak 150W at 12V, but only about 30W continuously.
I need to provide 60W at 54V
I need to provide peak 300W, continuous 100-150W at something like 12V or 24V depending on what DC ATX PSU I find.

A simple approach is to find variable input DC converters that all use the battery as input and output the needed voltage (12V, 54V, etc). Doing some reading, I've come across the idea of using an intermediate voltage IBA. So I could step up the battery voltage to something like 60V, then step down for each output I need to provide. The advantage here I think is that each point-of-load step-down (I think I'm using that term correctly) only has to deal with a fixed input voltage and fixed output voltage, so they are simpler devices. Also, current between the initial step-up and the various step-downs in pretty low.

Is one approach better than the other? If I use the intermediate voltage, what guides the decision for that voltage? 48V seems like a common intermediate voltage, but it seems simpler to me to have a single step up that is higher than all the needed outputs so I'm only using step downs for each output. Not sure if this matters.

When I build this, I'm going to wire it up as separate modules in an enclosure. But eventually, I'd like to design a PCB that replaces this. Not sure if that matters when giving advice, but thought I would mention it.

Thanks for any help!

Hello there,

If I understand you right, you are using a UPS system that of course has battery backup so you can power many things that run purely on DC not AC.  The UPS system is an intermediate device that of course eats up some of the power as it converts the internal battery power to higher voltage AC, and that AC is then used with several power supplies (like wall warts) that put out DC that powers many other devices.
You want to be able to eliminate the UPS system because it's not really needed because there is nothing that requires an AC voltage to run when the mains power goes down.
The main idea to get rid of the UPS system is to just convert the backup battery power to several DC voltage sources.  That is a reasonable thing to do.

The problems you will have to think about are as follows...
1.  You would need a separate DC to DC converter for each device, ideally.  One that puts out several DC voltage probably is not a good idea.
2.  You have to think about isolation between the devices if any of them interconnect.

#1 is probably the minimum requirement which I think you are aware of.

 #2 though is something that has to be dealt with because it could cause a catastrophic failure if any of the devices interconnect or have metal cases that can touch accidentally and have those cases connected to one side of the DC input source.  This is because it is likely that most of the DC devices expect to have a DC input that is galvanically isolated from any other DC input source for any other device, and so if they are interconnected there could be a ground voltage difference issue, which could cause high current to flow and blow out the two (or more) DC sources involve as well as possibly the devices themselves.
When we connect two audio devices using a patch cord, the grounds of the two devices come into direct contact in most cases.  If the DC to DC converters all have common grounds that may not be a problem, but because of the way grounds work it is best to have them all isolated until that last point where they come into contact.  That's the best case however, as the grounds could actually already be at different potentials, or it could be that one or more devices does not like a ground that is connected to another ground for another device.

This just means that you have to pay attention to the grounds of all the DC to DC converters.  Now all DC to DC converters will have the same ground voltage relative to each other.  If they do, that could work unless there is significant current flow, but if they do not, then that could be a big problem.
The easiest way to handle this is to make each and every DC to DC converter have a completely isolated output voltage.  That means some sort of transformer output that isolates the input DC from the output DC completely.  That's a more complicated design, so you may not want to do that, but if not, just check the ground routes and requirements of each device and see if it will work out.

The most extreme case of this kind of problem happens when people use two phases of the mains supply coming into their house, or there is a ground isolation problem with the DC source such as a wall wart.  In that case someone could get electrocuted if they touch a water pipe.  I remember a girl got killed listening to her CD player with headphones on and sitting in the bathtub.  There was a problem with the wall wart and the headphones had metal parts connected to audio ground.

[krby]

Thanks to everyone for the replies!

Re 48V and the PoE switch. I just realized I wasn't clear. The switch is not powered via PoE. It provides PoE on several of its ports. The switch's only power input is AC via an IEC C14 inlet, which goes right to an internal AC/DC PSU that outputs ~54.4V.

Re ground isolation. Thanks for pointing this out, I had not thought of that. The devices do connect to each other in various ways, mostly ethernet cables. Router, switch, ISP "modem", NAS box. But also a couple of USB cables. I see your point: In normal use, the DC side of things is all isolated by the individual switching power supplies (whether a wall wart or an internal PSU).

Hmm. I was planning to have the 12V devices all share a single converter, so even if it was an isolated converter, they would all still share a ground. How can I test for a problem here to decide if things are ok with a shared ground?

Also, a related question: Should I be connecting any of this to my home's PE (aka earth, or ground) ?

[Phil1977]

Ah, okay, then of course feeding the switch directly with 48V could be too little.

Regarding ground isolation: It´s like said, usually the wallwarts and regular DC supplies are galvanically isolated. But most of the time it´s not as bad because there is one important exception: PC-Power supplies. GND of PCs is mostly connected to PE, so all USB-Ports are not galvanically isolated to each other and so all USB-powered devices must tolerate a common ground GND.

That also answers the question where to connect PE - connect it to GND.

The only thing that would get really nasty is if you use DC-DC-converters that have no common GND in their input and output. Especially some of the cheap DC-DC-Lab-supplies have no direct connection there. But most of the small and cheap converter boards you can buy on ebay and co. have their (-)pole just straightly connected.

Of course you can also use galvanically isolated DC-DC converters like these small metal canned devices from Hi-Link or Murata. They´re much more expensive but definitely reliable for pretentious consumers.

[MrAl]

Thanks to everyone for the replies!

Re 48V and the PoE switch. I just realized I wasn't clear. The switch is not powered via PoE. It provides PoE on several of its ports. The switch's only power input is AC via an IEC C14 inlet, which goes right to an internal AC/DC PSU that outputs ~54.4V.

Re ground isolation. Thanks for pointing this out, I had not thought of that. The devices do connect to each other in various ways, mostly ethernet cables. Router, switch, ISP "modem", NAS box. But also a couple of USB cables. I see your point: In normal use, the DC side of things is all isolated by the individual switching power supplies (whether a wall wart or an internal PSU).

Hmm. I was planning to have the 12V devices all share a single converter, so even if it was an isolated converter, they would all still share a ground. How can I test for a problem here to decide if things are ok with a shared ground?

Also, a related question: Should I be connecting any of this to my home's PE (aka earth, or ground) ?

Hi,

The way I had tested in the past was to first measure the voltage between the grounds BEFORE they are connected, maybe with a high input impedance voltmeter.  It's probably best to test for both AC and DC.
If you measure any voltage, then connect a 10k resistor from one ground to the ground to be tested.  If you still see that voltage, then reduce to 1k.  If you still see that voltage, that could indicate a problem because that means significant current could flow if you connect the two grounds directly.

If order for current to flow there has to be a voltage difference.  If that voltage difference can cause a current to flow that could mean that if you connect the two some significant current could flow and that could be very bad.

Some things will show a significant voltage difference yet cannot cause sufficient current to flow once shorted.  The voltage can be high, 120vac, 240vac, etc., but once you test with even a relatively high value resistor like 100k, 10k, etc., that voltage goes away because it was getting through a very high impedance, which could be from a normal leakage current.  Once you short this out (connect the two grounds together) that voltage potential goes away and perhaps only a tiny insignificant current flows that does not bother anything.

I measured a high potential from a bathroom shower light measuring from the neutral directly to the water pipe, which would normally indicate a dangerous situation.  However, the impedance was so high that even a 100k resistor had dropped the voltage to nearly zero.
Same situation with a regular but older light switch down in the vestibule.  With the bulb unscrewed, I could measure the full potential of the mains voltage across the bulb contacts of the socket, even with the light switch in the OFF position.  That was due to some somewhat normal leakage current in that older switch.  Connecting a 100k resistor where the bulb normally goes, the voltage reduced to nearly zero.

This is using a modern voltmeter with high input impedance like 10Megohms.

[krby]

Thanks for the detailed reply!

I understand the general problem of two different devices being at different ground potentials and that can cause data cables between them (Ethernet in my example, but could be USB, or HDMI or whatever) to carry significant current, which is dangerous. However, I'm having trouble understanding the problem with using non-isolated DC converters. I've drawn up a quick diagram showing some of the equipment and connections:


The white boxes are equipment I want to power, the yellow-ish boxes would be the DC power connections. Since the non-isolated converters have continuity on DC- from input to output, my (beginner) instinct tells me that DC- in my entire diagram is all at the same potential. So, no unintentional current flow will happen.

I understand some of these devices are isolated from each other now via AC/DC wall warts. But I don't understand how that is better than a shared DC-/Gnd reference.

Also, if it matters, the network switch (and another device not shown in the drawing) currently have internal AC/DC PSUs and those PSUs tie their AC PE (earth/ground) input to the DC- output inside the device. Also, the metal cases are also tied to this DC-. So, my current situation (powering via AC) seems to be that several devices are isolated, but two make have DC- and tied together via the PE prong the AC input?

[Phil1977]

I think you have understood the issue and are on a good way to get a stable supply.

Devices that do not work with a common, PE-grounded (-)input are rare. Especially since many devices have USB and USB ground is very commonly connected to the housing and with PE.

Anyhow, if any device needs a floating (-)input then high currents could flow and magic smoke might quickly escape. That´s (assumably) why it was proposed to first measure the devices for their respective potentials. From my point of view I´d feel safe enough if there is continuity between the USB-GND and the negative pole of the input jack.

Just one more comment to your block diagram: It seems you want to supply all 12V-devices with one converter. I'd use the following setup:

12V-Bat-->Step up 24V -->Step up 54V for Switch
                      -->Step down 12V mini PC
                      -->Step down Modem
                      -->Step down router


The reason is cost and stability. Well performing buck/boost-converters are quite expensive and often still not more efficient than 2 conversion steps.
The reason to use one DC-DC per consumer is protection against inrush current voltage drops. If you connect e.g. your router to the same 12V-bus as your modem, chances are good that the charging electrolytics in your router pull down the voltage so that your modem restarts - or vice versa.

But of course that´s only how I would set it up - as long as you know about the issues there will be many possible setups.

[MrAl]

Thanks for the detailed reply!

I understand the general problem of two different devices being at different ground potentials and that can cause data cables between them (Ethernet in my example, but could be USB, or HDMI or whatever) to carry significant current, which is dangerous. However, I'm having trouble understanding the problem with using non-isolated DC converters. I've drawn up a quick diagram showing some of the equipment and connections:
(Attachment Link)

The white boxes are equipment I want to power, the yellow-ish boxes would be the DC power connections. Since the non-isolated converters have continuity on DC- from input to output, my (beginner) instinct tells me that DC- in my entire diagram is all at the same potential. So, no unintentional current flow will happen.

I understand some of these devices are isolated from each other now via AC/DC wall warts. But I don't understand how that is better than a shared DC-/Gnd reference.

Also, if it matters, the network switch (and another device not shown in the drawing) currently have internal AC/DC PSUs and those PSUs tie their AC PE (earth/ground) input to the DC- output inside the device. Also, the metal cases are also tied to this DC-. So, my current situation (powering via AC) seems to be that several devices are isolated, but two make have DC- and tied together via the PE prong the AC input?

Hi,

Non-detailed block diagrams are not the best way to understand this situation.  For example, the green lines (nice color codiing BTW) do not show how the devices interconnect through those lines in enough detail to see if there would be a problem, but because they are at least there at all, they do show there could be a problem.  To see the problem in more detail you'd have to know what those green lines are and where they connect, and I have a feeling there is more than one wire for each green wire.

Now if the ground lead of all the green lines are at the same potential, there may not be a problem.  But each device will draw current from the power sources so there will be voltage drops in each line which means it's completely impossible for the output grounds to be at the same potential for each device, even if they connect to the grounds of the input sources.  The difference however may be small and may not cause a problem.

A really good example, but maybe not exactly what you have, is where you have two full wave bridge rectifiers powered from the line. Each output DC goes to a different device, and the minus lead is common to the output lead of each device.  Since each device has it's own rectifier though, that means that you can't connect the two output grounds together.  You can connect the two input ground leads together (and after all you do that so you can power both devices) but you cannot connect the two output lead grounds together.  If you use two isolated input supplies or in this example even one isolated and one non isolated, you can then connect the two output ground leads together.

In your diagram if you could provide more detail about the green lines that could help explain this better.  Remember though this is only a precaution right now, there is a chance it will work anyway.

Another really good example, also not exactly what you are doing though, is when someone goes to measure the AC mains power line with a non isolated scope, where the ground of the scope is connected to one side of the AC mains power line.  If they don't know that, they could connect the ground of the scope right to the hot side of the line, thus putting 120vac right across a short circuit and blowing who knows what out with a nice size spark.  A friend of mine actually did this and blew out the USB scope and the USB port he was using, and that was lucky nothing else blew out.

For your example, you'd have to, as a minimum, be sure that all the output lines take their ground from their input ground.  That's so when you connect the (green line) grounds together there is no conflict.  If by chance even ONE of the outputs is FLOATING, something is gotta give.  An example here is an audio amplifier.  IF the amplifier has a floating speaker ground (typical in high power automobile audio amplifiers) then you can never connect either side of the output to ground which would be the minus lead of the car battery.  Doing so would blow out the amplifier most likely.  Floating outputs like this are created when the design uses two internal amplifiers to generate the output audio, which is done to get more power output using just a 12vdc source.  Each output lead comes from a separate audio amplifier output, so there is NO common ground between the outputs and the inputs.  That's one of the worse situations.  The situation where they do have a common ground with the input would only be a problem if the currents being drawn from the input sources or currents sourced from the outputs create significant voltage drops or some other problem like THD.  For logic level signals the voltage levels could be shifted which MAY not damage anything but could prevent the signals from being interpreted correctly by the logic gate inputs and thus the signal would not be able to get through reliably.

Again, all this is a precaution, and measurements usually confirm or verify that a given hookup will work or not.  The alternative is to connect everything together and hope nothing blows

[Phil1977]

I really don't want to contradict to MrAIs caution, but the ethernets physical interface is quite well defined and is galvanically isolated:


(Picture from TI)

If you have any power outputs, like speaker amplifiers, motor drivers - or worst of all: programmable DC outputs e.g. for charging batteries - then caution is really advised.

With ethernet and DSL ports there´s nothing bad to be expected. Of course this statement is without warranty, it´s definitely justified to check the potentials carefully.

[krby]

Thank you both for the responses.

To be sure I understand, I'm repeat my understanding of  what I think MrAl is saying:
If my power input grounds are common, the output or interconnect ground (USB, Ethernet, whatever) may not be at the same potential and this can lead to current flow. In order to be safe, I need to test voltage across the output/interconnect grounds of devices before connecting them. I have an high impedance DMM, 10M\$\Omega\$.

If my power inputs are isolated, then I can avoid this problem.

My interconnections are all Ethernet and USB. In my exact case, Phill1977 points out that Ethernet is galvanically isolated, so I'm safe there.  I also have USB between a few devices, I'll need to read how USB is handled and then test myself.

Do I have it right?

Thanks again for taking the time and explaining this all to me.

[Phil1977]

Yes...

To be safe, you can connect your device to 12V and measure the voltage between your power input GND and USB GND - it should be close to zero. As I mentioned earlier you could also use the continuity test of your DMM and test for continuity between these two contacts in the power off state. If you read <5Ohm in resistance mode or <0.1V in diode test you can assume that these contacts are on the same potential and nothing goes wrong if they are connected by one more path.

[MrAl]

Thank you both for the responses.

To be sure I understand, I'm repeat my understanding of  what I think MrAl is saying:
If my power input grounds are common, the output or interconnect ground (USB, Ethernet, whatever) may not be at the same potential and this can lead to current flow. In order to be safe, I need to test voltage across the output/interconnect grounds of devices before connecting them. I have an high impedance DMM, 10M\$\Omega\$.

If my power inputs are isolated, then I can avoid this problem.

My interconnections are all Ethernet and USB. In my exact case, Phill1977 points out that Ethernet is galvanically isolated, so I'm safe there.  I also have USB between a few devices, I'll need to read how USB is handled and then test myself.

Do I have it right?

Thanks again for taking the time and explaining this all to me.

Hi again,

Yes there's no contradiction, my suggestion is for the general case where you want to be absolutely sure the connection will not cause a problem.  For example, I have had Ethernet ports blow out several times and I doubt that's in the spec

So anyway, my suggestion is an absolute test where no matter what you are connecting you should be able to determine the go/no-go status before anything blows out.

Now with the USB i doubt that is isolated, but I remember seeing USB isolators on the web if you run into any problem like that.  That would be a good solution just hope it's not too expensive.
When I designed a four channel voltmeter, I used an opto coupler for galvanic isolation between the RS232 port and the measuring circuit.  That handled the isolation so I did not need an isolator.  In that kind of situation you really do want absolute isolation because you might have to measure something that has no common ground of any kind.  For example, if you want to measure the AC mains voltage (120vac here) with a USB connection you don't want the USB ground connected directly to either side of the mains, not even the neutral.

They had a saying in carpentry: measure twice, cut once.
For this electrical connection issue: measure once, then either connect or find a way to isolate.

And again I can't stress enough, your intended connections may work out just fine (as was pointed out with the comment on Ethernet isolation by phil1977) but if there is going to be a problem the measurements should show it before anything goes wrong.  Remember the resistor test too.

[MrAl]

I really don't want to contradict to MrAIs caution, but the ethernets physical interface is quite well defined and is galvanically isolated:

(Attachment Link)
(Picture from TI)

If you have any power outputs, like speaker amplifiers, motor drivers - or worst of all: programmable DC outputs e.g. for charging batteries - then caution is really advised.

With ethernet and DSL ports there´s nothing bad to be expected. Of course this statement is without warranty, it´s definitely justified to check the potentials carefully.

Hello there,

Not really any contradiction, no problem, my suggestions are for the general case where you don't know if anything is isolated, or you think it's isolated and you want to make damn sure before you actually connect anything.

I actually invite contradictions, not condemn them, as long as there is a good reason for it (as you have shown).  For this reason, I thank you for the comments and look forward to future discussions with you.

[krby]

And again I can't stress enough, your intended connections may work out just fine (as was pointed out with the comment on Ethernet isolation by phil1977) but if there is going to be a problem the measurements should show it before anything goes wrong.  Remember the resistor test too.

Yup, that was clear in your replies. I didn't mention the resistor test, but I intend to do that if I run into a problem. I'm going to move forward with non-isolated supplies I already have, and rig up Ethernet and USB test cables that will let me probe their ground connections from each side.


Thanks again for teaching me the concepts!

[krby]

It seems you want to supply all 12V-devices with one converter. I'd use the following setup:

12V-Bat-->Step up 24V -->Step up 54V for Switch
                      -->Step down 12V mini PC
                      -->Step down Modem
                      -->Step down router


The reason is cost and stability. Well performing buck/boost-converters are quite expensive and often still not more efficient than 2 conversion steps.

I skipped over this, but wanted to ask a couple of follow-up question:
Is your choice of 24V as the intermediate voltage because it is common? Or is there some technical reason to always do two steps between batt and device?

I was thinking I would step-up batt voltage to 54V for the switch, then step-down from 54V for the 12V (I didn't miss your point about individual step downs and voltage sag for those). But I bet stepdown from 54V is more rare. I guess a step-down to 12V that takes variable input like 24-60V would do it.

[Phil1977]

Yes, 24V would be a good compromise between cost and efficiency.

- <30V input converters are omnipresent and cheap (some are as cheap that they are fraud, be cautious at AliXXX&co.)
- The efficiency of the converters somehow depends on the conversion ratio. A convertion ratio of around 2 often is in the region of high efficiency. In that aspect it would be bad to once go up from 12V to 48V and then back.
- 24V gives the regulators quite a lot of margin. If the intermediate voltage has some drop down, then especially individual step down modules have a very good chance to compensate
- 24V seems quite a good compromise of amperage and safety.

One more word about the AliXXX & ebay conversion modules: The cheapest kind of converter there, often sold for less than 1€ per converter, rarely reaches its advertised power - not even for short time. They are often equiped with fake semiconductors and sometimes are noisy as hell.
The modules with higher power can be okay. I´d anyhow use a "safety factor" of 2 for the required power and check output voltage with a scope.
The industrial modules you get at serious b2b-platforms are much more expensive but you get what you pay for. These are the units you should use if it needs to be reliable, e.g. if you build something that can not easily be repaired.

You can also buy galvanically isolated modules if you should find any ground loop problem. If it needs to be cheap I have good experiences with the following design:

https://de.aliexpress.com/item/1005004463474248.html?

If it needs to be isolated and you want something better you should look for the canned converters. The chinese originals from "Hi-Link" are very affordable and work flawlessly in always-on-applications. If you want even more reputation you can take the japanese "Murata". But then you are giving your DC-DC-converter a MUCH better quality than any usual wall wart ever had.

One really last remark: Have you thought about fuses?