Car battery charging below float voltage?

A great many years ago, we had problems with a low voltage power supply not delivering the right voltage. It was eventually traced to a 5A glass enclosed fuse which had a resistance of 1 Ohm. A whole bad batch!

I used a Henley fuse, the ones they put on incoming feeds to houses in the UK. No holder, hoseclipped on, like this:

Surprised that didn't melt and therefore blow.

When I designed a radio receiver hifi amplifier i got called to the production line where all the radios suddenly developed 0.25% distortion And yet on my test bed, thy were all 0.05%.

The test line was using the same connector that was in the original radio. Designed to be plugged in once, and that was it, the silver plating was now covered in silver oxide, making a first class - well now, about a tenth rate - rectifier...

Here is the original question.

I have now deleted all of your misconceptions about electricity and wiring. As they are irrrelevant to properly running 8 computers with a total of

15 graphics cards.

*******

List the assets.

In each column, you fill with the number of units needed for the load to the right of it.

Mains CCT UPS column ATX supply --- mobo --- 2 graphics cards

Mains CCT UPS column ATX supply --- mobo --- 2 graphics cards

UPS column ATX supply --- mobo --- 2 graphics cards

ATX supply --- mobo --- 2 graphics cards

ATX supply --- mobo --- 2 graphics cards

ATX supply --- mobo --- 2 graphics cards

ATX supply --- mobo --- 2 graphics cards

ATX supply --- mobo --- 1 graphics cards

On each ATX supply, you total up the electrical loads, and buy the next largest supply. A 1000W supply can power a 600W system, and have 400W of unallocated margin left over. The real draw from the wall is 600W, not 1000W. It is the actual load that determines the billable consumption.

Each ATX supply is independent of the other ATX supplies. There are no bus bars, copper plates, dangling wire assemblies going all over the place. Each of these, is assigned room on your rack shelving. No misconceptions about electricity are required, if you do it like the gamers and scrotes do it. The money wasted on a leisure battery, could have bought you three cheap ATX supplies instead, until you have the eight independent supplies needed. There should be excess supplies on Ebay, from coin miners going out of business.

+--------------------------------------------------+ \ ATX supply --- mobo --- 2 graphics cards \___ Neat and tidy packaging / +--------------------------------------------------+

You total up the real consumption of each system, if fitting UPS boxes.

You pick UPS boxes (if you really want or need UPS) according to the real load of the PCs. The PCs can each have a 1kW supply, but the UPS is gauged by the 600W load (or whatever). You follow the rules for loading a UPS.

A Kill-O-Watt meter can make measurements on each system, and making actual measurements on each motherboard plus 2 graphics cards, allows you to do a better fit of UPS boxes to loads. The meter measures in W as well as VA, and you can compare those numbers to the W and VA limits of the prospective UPS.

You might not need eight UPS boxes (of the cheap tier). Maybe fewer boxes will be needed, for the systems that have the weak past generation graphics cards. You can't fool us into believing the fifteen graphics cards are 4090. They aren't.

You allocate the UPS boxes to mains circuits, according to the breaker ratings in the panel and associated outlets. Using the real/measured 600W value or whatever the measurement shows, in each case. If you run Prime95 and Furmark on a system or a similar kind of max_load, you can get a better idea of what the max_load really is.

Summary: If you do stuff like normal people do it, there won't be any leisure batteries or bus bars or other crap in the picture. The wiring will be neat and tidy. Your fire insurance person will be well pleased and you will continue to have fire insurance.

Paul

Your guess agrees with the results here. The first voltage/current recorded was, from memory, 12.5 volts at .25 amps. This was about 6 or 7 days ago. A couple of days ago I weighed the battery and in the process wiggled it a bit. Next day the voltage & current were noticably higher. Due to the wiggling?

Hul

So I can't see them and refute your reasoning for dismissing them.

I assume CCT means circuit? The free dictionary shows 132 possibilities. I'll take it to mean a 13A 240V plug.

Ugh, I guess I need to change to fixed width text for this....

You've made too many assumptions. The above is nothing like what I have. The GPUs are not spread evenly through the MBs. I don't use UPSs.

No no no. The computers themselves - the MB, RAM, CPU, SSD run off normal PC power supplies.

Stupid way of doing things. With my current setup, I can add a GPU anywhere on the shelving and connect it to any computer. And I get a rock solid voltage of whatever I set the PSUs to(they're variable from about 10 to 15 volts. The PCI Express specs say 11.4 to 12.6V, so I've set it to 12.6V, which means even if there's any voltage drops on wires, it has plenty of room, but it will never go over the specs when unloaded. It also means less ucrrent going into the input side of the VRMs on the GPUs.

Those things do not have a stable voltage output, and are often a fair bit below 12V.

I'm not interesting in neatness, I'm interested in stability and ease of adding parts, swapping parts, testing parts.

I use an amp clamp on the 12V buses.

Old ones use the same electrical power, they just do less work. The designers tend to get limited by the heat they can dissipate in the size of a video cassette.

Fuse.

Nah. 30A 240V going to the garage. Three 13A sockets. If something gets warm I shift a load.

I use Milkyway at Home to fully load them.

The whole point of the bus bars was to get rid of the clutter of many ATX supplies loading individual cards with cables everywhere. It actually looks about 3 times tidier than before. With bud bars tacked along the shelving, there's only a 6 inch cable from each PSU to that bar, and another 6 inch cable to each GPU.

I don't have fire insurance.

It's surprising when the battery is 12V open circuit, an extra half volt only supplies 0.25 amps. I guess we need to think about the chemistry of the battery to know why it's so low. It must be more than a simple resistance, otherwise giving it 14.5 volts would only charge it at 1.25 amps.

the voltage is too low for a 6 cell lead-acid battery, and too high for a 5 cell battery. maybe try supercapacitors instead.

Get serious. Battery terminal clamps are not expensive. Niether are CNL or ANL automotive/marine fuses.

Fuses that interrupt DC are different animals.

Your illustrated clamp has obviously not run at more than few tens of amps. At 100A you'd be able to measure the voltage drop and probably be able to smell it.

Whatever happened to 50% derating on power sources?

Distributed computing through BOINC is supposed to employ unused computing resources. It's not intended to tip the ballance of global climate, or to mine bitcoin.

I expect you find your winter heating bills are unaffected, but summer must be pretty intollerable in that room.

RL

Keep in mind that charging current is limitited both by the battery & the charger. The data I stated was from a charger with a max current due to a 12v transformer rated at 2 or 3 amps.

Hul

He's in the wilds of Scotland, it never gets that warm there in summer.

Hose clips are precisely what you see in a car to attach a battery for the purpose of a few hundred amp starter motor etc.

Bollocks. The voltage matters, and I'm giving it way less then the 240V it's rated at.

The clamp isn't carrying the current.

WTF are you talking about? The power source is rated to x, it can do x. Unless it's Chinese, then you divide by two.

Why would it matter if they're unused resources or ones bought for that purpose? They use the same power.

I'm an atheist, I realise climate change is f****ng baloney.

Who mentioned bitcoins?

The room has windows duh.

Look again. The aim of a battery connector is to provide maximum air-free contact to the round terminal's surface.

Read more. The I^2t rating of a fuse is good only in it's rated application. AC current is self-interupting, by source reversal. DC isn't and will arc, potentially with explosive results.

With such modest terminal contact, the iron clamp will be developing the same (increasing) voltage drop as the terminal contact degrades. Measure it.

<snip>

As long as the battery's no actually doing anything, as your app describes, you'll not have to worry about it.

Happy number crunching (WCG since Y2K on 3+ machines)

RL

Yip, and it does, the clip goes round the terminal, look in your own car.

12V DC will not arc across a 2 inch gap.

Half of it.

Bullshit. The fuse touches the terminal.

The drop across the fuse and the clamps in total is 6.5mV at 10.3A. So at the rating of the fuse, 80A, it would drop 0.05V and dissipate 4W. Negligible.

Well one of the power supplies failed - I opened it and found two shorted out power transistors which were not tightly attached to the heatsink and had overheated, despite the power supply never having even blown body temperature air out of it. There was a transistor smell in the room, not sure if it was those (which didn't look damaged, only the dust on them was a little brown) or the other supply which I measured as running at the full 83 amps, so was presumably limiting itself (it doesn't actually say on the specs anything about limiting, just short circuit protection and overheat protection, so I don't know if it shuts off or slows down). Some of the cards were misbehaving, which doesn't make a lot of sense.

I can't find anywhere (can you?) a graph of what voltage to expect from a 130Ah leisure battery (as in the thicker plates variety presumably) at various current draws? It would only have been drawing 7 amps and I can't believe that made a drop of 1.2 volts (which should be required to make the cards play up). Mind you I'm pretty sure some of the older cards get upset at 12.3 volts, because since I installed the bus bars two are working fine that didn't used to work at all.

Lightweight! :-P

My current annoyance is LHC, their CMS tasks. They transmit a huge amount of data while running, from inside a virtualbox. But they're very impatient. When my FTTC pitiful 6Mbit uplink is saturated with them, some of them crash or give up. I'm having to manually adjust how many run, and do some of their Atlas tasks aswell, since if you ask for "anything", you only get CMS!

<snip>

Wearing eye protection and gloves, you might wany to investigate this yourself some time. Through a suirable current-limiting impedance (~your cables are intended to do this), apply a short circuit at the far end of the circuit and then remove the short (if you can). Actually, welding goggles might be recommended.

You'll probably do this without flinching only once in your life. The flinch becomes automatic on all subsequent occasions. DC battery holders will generally be provided with a (useless) cover.

Two circular objects do not mate with half of any contact surface area, no matter how it's calculated.

That's about 10x what is intended for a fresh joint. It will degrade with time, use and temperature cycling.

Derating of power from labels is common safe practice. You should consult the manuals for limiting characteristics if you want predictable load performance near or above ratings.

The closest you'll find will be 'load test' voltages, often stated as nominal under either 15 second loading, or to 'end of discharge', from an innitial SOC or open cct terminal voltage.

A 'load test' is nominally 1/2 of the Cold Cranking Amps (CCA) or

3 times the Amp-Hour rating. A 12V lead acid battery is considered as dead <12V.

Your application load being ~ constant and self-determined, sets your loading for test at your site.

I've been avoiding 'Mapping Cancer Markers' because, unless there's a cure, there's no pojnt in identifying predicators - it just becomes another method for insurance companies to screw you.

RL

Most people have shorted a car battery at some point in their life. A spanner etc. You pull it off and there's no arc. An attempt for it to weld, yes, but once there's a gap it doesn't arc.

Since the smaller area is only present for a billionth of a mm, and soon gets a lot wider on both sides, utterly irrelevant.

The temperature does not change much. The use is occasional. Something connected stays connected, since the bits touching each other are not subject to corrosion from the air.

Why should I assume something won't do what it says on the tin?

Do you have a link to some data as per what you just said?

I've found a perfect compromise level. I'm running as many CMS (which seems to be their top priority, since they send only those out if you express no preference) as I can to make the internet connection get almost saturated, so they can all send and receive whenever they need to. Then I've set the other machines to get anything other than CMS. This is usually Atlas, sometimes Theory aswell if it asks for many at once. I've moved to 31st place worldwide in recent credit, I expect this to rise to about 11th place. 126 cores should get me somewhere....

All data is useful. What they learn can be used by others to develop cures. It's absurd in the 21st century we still can't fix cancer. My neighbour's wife died of it a year ago, despite him praying for her. Fuck knows why he still believes in god, who clearly either doesn't exist or doesn't give a shit.

You might consider Folding at Home (various biology, although it's non-Boinc so I avoid it, easier to have everything under one scheduler), or Sidock (currently on Covid stuff) or Denis (run by a guy called Jesus!, doing brain cell stuff, but currently out of work), or Rosetta (misc. biology, currently very rare work), or TN-Grid (genetics).