Close

Low voltage power distribution?

A project log for today's assorted project ramble "grab-bag"

Assorted project-ideas/brainstorms/achievements, etc. Likely to contain thoughts that'd be better-organized into other project-pages

eric-hertzEric Hertz 10/22/2019 at 23:008 Comments

...is the best search term I've come up with, and still no relevant info...

So here's the deal: say I've got a multi-voltage DC power-supply ~10ft away from my equipment... the equipment consists of a [single-board] computer and various peripherals, all powered from that same supply.

Now, say the computer runs off 5V, it's connected to a hard disk which runs off 5V and 12V, and it's connected to the computer via USB.

Forget "ground loops", here we're talking *several* factors which could affect data, or even inject weird/negative voltages where they don't belong.

There's the obvious loop on the ground, around 25ft [10ft from the power supply to the computer, 10ft from the supply to the drive, 5ft between the drive and computer]. But, again, who cares about the audio-nuts' much-loathed "ground-loop" "hum" in a digital system?

Forget that for a second, let's look at the devices. The computer uses 3A max, but varies dramatically depending on CPU load. The hard disk's 5V is used only for digital stuff [I'm presuming], whereas the 12V is used for motors. Let's look only at the 5V system for now, and also presume that the disk's 5V load stays relatively small and somewhat constant.

Now, again, looking at the "ground" [or maybe more appropriately "common"] signal from *just* the perspective of the 5V loads, there will be different voltage-drops on those 10ft cables to the compy and drive. Say the drive uses half an amp at 5V, and say there's 1ohm in its 10ft connection to "common"... that leaves the drive's "common" [at the drive] at +0.5V relative to the power supply.

Now say the compy is asleep, and drawing only 0.1A, its "common" rests at 0.1V WRT the supply. Compy and disk are attached via 5ft USB cable; their "commons" also. 

What happens, now? A USB cable, with its comparatively-thin wires [unless both sides' shields are connected to common, which is often the case], say 1ohm, has 0.4V across its common wire... it *should* not carry current [except in the case of a USB-powered device, but that's another matter], but V=IR, I=V/R=0.4/1=0.4A. 

That's nearly *all* of the disk's circuitry current, travelling not through the disk's power wiring, but through the USB cable and back to the power-supply through compy's power cable.

Now, obviously this can't be the case... because those calcs rely on compy's common carrying 0.1A, and our result is that it's carrying 0.5A, etc. So, some balance must be achieved. 

Let's assume, for a moment, somehow that balances to 0 current through the USB's common. [Note that I can't quite visualize how, in consideration of such scenario].

Now, consider the transients; a load changes suddenly, maybe compy's awakened to draw 2A, it will take time for the system to balance out, again. There will be a moment, however brief, where current [plausibly a lot] will flow through the USB cable's common. A surge of current [nevermind the voltage difference] through a long wire parallel to several others which are coupled inductively... a transformer, sending a surge of current through data lines.

Now, sure, maybe CRC-errors will result in data retransmission... but this is looking kinda ugly already, and we haven't even considered the 12V system of noisy power-hungry motors, etc. sharing that same "common," nor other large/varying loads such as a USB-powered DVD drive deriving its power through compy's power cables. Further, we haven't considered nearly identical effects on the 5V "loop," nor external interference/"hum" commonly-associated with "ground" loops [that could just-as-well occur on a 5V loop].

So far we have reasonable potential for data-voltages that, at least, aren't within high/low voltage tolerances for data transmission. It's also plausible such signals may *exceed* input-specs. I.E. If the drive's 5V circuitry's "common" is at 0.4V higher than compy's, and it outputs a 5V data signal, compy will see a 5.4V input. And the reverse, compy might output a 0V data signal, which will be -0.4V into the drive's circuitry. 

And all this is only in consideration of *inter*-system data communication. What about the effects of a 5V digital circuit's *power rails* fluctuating so dramatically?

And, of course, there's plenty of room for our 5V source wires to also drop those same voltages due to the same length and wire-resistance... 0.5V on the drive's circuitry's "common" *and* 0.5V on its 5V wires, resulting in only 4V for the circuitry. [This is where remote-voltage-sense is appropriate, but can't do that with multiple endpoints!]

These values may seem extreme, but we are talking *amps* of current, non-negligible load fluctuations, and long wires.

It's almost a wonder a typical full-tower PC can even keep things straight with various drives and multiple endpoints, floppy and IDE cables with interconnected commons looping all over the place, motor-currents [however briefly] surging through ribbon cables with alternating data wires.

I'm a bit baffled as to how to best go about powering these devices. Almost thinking it best to just send 12V to each and have a local 5V regulator. But even still the long "common" wires, USB cables, voltage offsets, etc.

Oh, and I meant to go further into thoughts on the overall "balance" of the system. I'm almost certain it involves current through the USB cable's power wires, maybe even a lot, even in steady-state. Am somewhat convinced it may wind-up that the two long power cables both carry pretty much the same current, regardless of the endpoints' different needs. Then the USB cable must handle that balancing, so compy running full-tilt at 3A while the disk is sleeping means 1.5A through a measly usb-cable? And, further, reversed in direction... 1.5A *leaving* compy's "common," through the USB cable, *to* the device... I'm not liking this.

Discussions

Ken Yap wrote 10/22/2019 at 23:56 point

>It's almost a wonder a typical full-tower PC can even keep things straight with various drives and multiple endpoints, floppy and IDE cables with interconnected commons looping all over the place, motor-currents [however briefly] surging through ribbon cables with alternating data wires.

This is a different situation. Here the power cables are thick and relatively short. Moreover in the case of SATA and USB interfaces, differential signalling instead of single ended signalling is used. Also the motors use the 12V supply, not the 5V one and this ground has a separate return line. Even when single ended signalling is used for PATA, the bus interface chip at the disk end of the cable is referenced to signal ground, not power ground.

As for locating a PSU 10ft away from the mobo, just don't.

  Are you sure? yes | no

Eric Hertz wrote 10/23/2019 at 00:48 point

"this ground has a separate return line." I always thought the same, but I've taken apart probably three-dozen drives of various sorts and vintages, and I don't recall even one where the two ground pins weren't both connected by copper fill at the PCB. [Key factor: at some point I began *looking* for it]. 

Also, your point about signal vs power-ground at controller chips... makes sense, except that, again, usually those two are connected *somewhere* on-board. Most'd be pretty unhappy with a 0.5V difference between 'em... oh, I think I've seen it done with inductors, hmm... 

And a full tower PC can have power cables nearly 3ft long with three drives attached... 6A @ 3ft is nothing to scoff at, they're not *so* heavy...

USB is a weird differential-signalling bastard-child, I dunno if it counts... but yeah, other differential systems make sense for systems like these.

  Are you sure? yes | no

Ken Yap wrote 10/23/2019 at 00:53 point

Not on the PCB, in the connector. Obviously they have to be connected on the PCB for the motor controller circuitry.

I have a full tower with 3 drives in RAID 5 plus two other drives. Power cables are nowhere near 1m long, more like 0.3m. SATA of course, all my PATA drives died long ago.

  Are you sure? yes | no

Eric Hertz wrote 10/23/2019 at 01:33 point

"Obviously they have to be connected on the PCB for the motor controller circuitry."

That would make sense, unless, say, they had isolated grounds [via optocoupler?] But they don't, usually, as far as i know...

So, again, say your 5V circuitry is 0.5A and your motor circuitry is 1.5A. They use the same length and AWG wires. If the PCB didn't connect the two ground pins at the connector, one ground-wire would carry 0.5A, the other 1.5A, the voltage drop on the second would be 3x that of the first, say 0.1V and 0.3V. Now your motor-controller chip, presumably running off motor-ground, has to work with -0.2V control-signal input. Prb not a great idea, unless planned-for. So, you're saying they connect the two ground wires both at the PSU and at the drive. So, now, a voltage-drop of 0.2V across them both. Now the digital/5V circuitry interfaces with the mobo, whose cable is shorter, and has *many* grounds... say 0.01V drop. Now your mobo communicates with your PATA [for decades] with 0.01V, while the chip on the drive is running at 0.2V, so the drive sees -0.19V input signals.

But, no, it doesn't, because also on the drive's PCB [and mobo] are signal grounds, which *somewhere* on both boards is connected to the power ground, if even through a 6mil trace. Because those are connected at both ends, the voltage at your drive won't be 0.2V, nor the mobo 0.01V. Instead, the drive will draw current *through* the signal grounds, to use for power. Maybe not much, but some. And now your signal and power return paths are kinda moot.

  Are you sure? yes | no

Ken Yap wrote 10/23/2019 at 01:54 point

The motor will draw most of the current through the power ground not through signal ground. So if there is a voltage drop across the power ground cable the difference can be dealt with at the motor drive chip. It's not hard to design controller chips that are insensitive a fraction of a volt. Secondly the data signals are digital not analog so also not sensitive to a fraction of a volt. If you've got more than a fraction of a volt drop then you shouldn't be using such long power cables.

  Are you sure? yes | no

Eric Hertz wrote 10/23/2019 at 02:15 point

again, the point is moot. Yes, chips can be designed with tolerances to such differences, but motor current *is* going through signal ground, maybe not much, but some, because those systems' grounds are *not* isolated. That same motor noise is now affecting the motherboard's circuitry, much of which literally runs on little more than fractions of volts.

"If you've got more than a fraction of a volt drop then you shouldn't be using such long power cables."

People designing case-mods and such seldom think to even check these things. Further, people running Ardiuno servo-motor circuits off ATX->Bench supplies with undersized 6ft "wallwart" wire, similarly. My point remains: it's a wonder these systems function as well as they do.

  Are you sure? yes | no

Ken Yap wrote 10/23/2019 at 02:24 point

>That same motor noise is now affecting the motherboard's circuitry, much of which literally runs on little more than fractions of volts.

That difference is manifest in the PATA interface drive outputs, not the other signals on the mobo which are referenced to local signal ground.

One reason SATA went to differential signalling is better resistance to these artefacts. I'm sure a lot of glitches in home builds can be explained by people going past sensible limits. None of their designs would pass scrutiny in enterprise environments where subsystems are designed to be electrically isolated, e.g. Ethernet, Fibre Channel, etc.

  Are you sure? yes | no

Eric Hertz wrote 10/23/2019 at 02:37 point

indeed. Oddly, looking up a completely different topic just the other day, this same issue arose. E.G. usb-attached hard disk enclosures, dvd drives, etc. Many so cheaply-designed, and-or so poorly-implemented by the user, that e.g. chip-manufacturers get blamed for things like under-rated power supplies, poor cables, etc. Again, looking at these things, it's a wonder they work at all, nevermind as well-enough as they must to've become commonplace. USB 2.0 cables with nary a twist, and no shielding, tiny wires driving 2A phone chargers. So-on. Motherboard manufacturers who don't even supply 500mA on their ports, DVD drives needing 700mA... the list is endless.

  Are you sure? yes | no