[myself248]

Hi, automotive electrical is my job.

There's quite a bit of very thick wiring in a car, not just the starter wire, but boring stuff like audio amplifiers, rear window defrosters, power seat motors. Those things don't draw a ton of power, like maybe just a few hundred watts, but at 12 volts even modest powers require extraordinarily thick wires, especially when you account for bundle derating.

This requires large terminals, which requires larger connectors, and there's the complexity, because MOST of the wiring in the car is just signals, or low-power stuff, which can run over thin wires and small terminals. (Minimum size is limited by mechanical durability rather than electrical conductivity.) Making a "hybrid" connector that has a couple large cavities for large terminals, and a bunch of small cavities for small terminals, is a pain. Having separate connectors for heavy power and for signals introduces more assembly work and negatively impacts testability. The wires have different stiffness and bend behaviors, they exert different amounts of force on weather seals, they have to be terminated on different machines at different points in the assembly process.

By allowing power wires to be nearly as thin as signal wires, you can use simpler connectors with unified terminals. Manufacturing gets simpler, harnesses get lighter, assembly gets faster and easier.

Weight is also a huge deal, every ounce counts. There's upwards of 100 lbs of wiring harness in most cars, more in larger or premium models with a lot of accessories. If half of that weight is signals and won't change with voltage, but the other half is heavy power circuits that'll get 4x thinner at 48v, it's significant weight savings.

Furthermore, switching heavy current means massive relays or FETs and the heatsinks thereon. If you can reduce the current, those components get lighter too. Audio amplifiers get lighter, speakers get lighter (stupid heavy-wound 2-ohm speakers to get reasonable volume out of low voltage drive? Nah, use standard 8-ohm now that you have real voltage at the amplifier!), all sorts of things get lighter.

That's all in addition to the electric power steering already mentioned by others. EPS can easily move 1kw for short periods, and has stupidly huge wiring to do that at 12v. It's still chunky at 48v, but a lot less so, and can use more common terminals and connectors. Replacing a hand-assembled bolted connection with a machine-crimped and clicked-together connector improves reliability or reduces testing process overhead.

It's really significant, and it's embarrassing that the industry fell flat on its face in the late 90s last time they tried. Here's hoping this takes off.

[Tuna-Fish]

And to explain why this hadn't been done before/how we got here:

Nothing in a car actually wants 12V DC. Most of the low voltage stuff will run better at 5V or below, while a lot of the higher voltage stuff would benefit from going as high as possible. 12V exists because DC-DC conversion used to be expensive, and you had to make a compromise about the voltage based on losses, wire thickness, and picking a low enough voltage that all the low-voltage stuff doesn't suffer too much.

What's changed is that you can get a single-device DC-DC converter for really cheap these days. Cheap enough that you might as well put it in the light bulbs, and everywhere else that wants a low voltage.

[myself248]

12v exists because 6v was too low; wires were impractically thick for even the early accessories being added in the 1950s. The 6v-12v transition happened in 1955/56 for many cars. Some stuff like lightbulbs could be reused by putting two 6v bulbs in series in a 12v car, so it was a very cheap and relatively straightforward transition.

If they'd just had some foresight and gone 48v in 1955, we would've saved 50 million tons of copper in the years since. It's no harder to make 48v motors or lightbulbs or relays or anything else (and in fact, the telephone network contains plenty of exactly those things, and has, in staggering numbers, for over a century), but the automotive industry isn't exactly known for being forward-thinking.

[robocat]

A 12-volt battery typically has six cells. A 48 Volt Lead-acid battery would have 24 cells - I'm not sure how that would change the constraints on charge balancing and starter-motor stress.

I can say that the 24 Volt deisel vehicles I have used makes buying two batteries expensive.

[15155]

> If they'd just had some foresight and gone 48v in 1955

How would they have done this without cheaply-available high-speed switch-mode power supplies with low-DCR inductors/MOSFETs?

[myself248]

Why would they need those?

All the stuff that's natively 12 volts now could simply have been made natively 48 volts. You can make a 48-volt lightbulb as easily as a 12-volt lightbulb. You can make a 48-volt motor as easily as a 12-volt motor. Actually, motors for higher voltage tend to be smaller and lighter, which is why industry tends to go straight for 4160-VAC motors whenever 480VAC is inadequate.

What applications cannot be made to work at 48? I'm not aware of any. As I said in the comment to which you're replying, the telegraph and later the telephone network had been running similar DC systems since the 1850s or so at various voltages depending on the length of the telegraph line, with the telephone network taking over and 48 volts firmly entrenched by the 1910s. There was a huge manufacturing base producing 48-volt equipment, including motors and generators, indicator lamps, and a mindboggling array of switches, relays, stepping selectors, and their ilk, and all that was before WWI.

Furthermore, Charles Kettering who invented the automotive starter motor (and made it work at 6 volts), was around the same time making Delco-Light plants for rural electrification, which mostly ran at 32 volts DC. These supported a whole line of 32VDC appliances -- lights, vacuum cleaners, kitchen gadgets, irons, motors that could be attached to other machines in the shop. There was also a less common 110VDC version of the system but I can't find any contemporary literature discussing the differences, although I'm sure they would've quickly discovered that the 32V system was pretty docile while the 110V encouraged extreme care around open contacts.

As for why cars didn't use the higher voltages already in use and superior in many ways, my only guess is that a lead-acid battery with a high number of small cells must've been difficult or expensive to manufacture, compared to one with a small number of large cells. The Delco-Light plant used a large rack of 2-volt cells, whereas the starter motor used a single 3-cell packaged battery that fit easily under the hood. If they'd just figured out how to package more smaller cells together....

[15155]

> Why would they need those?

To convert voltages to useful levels without suffering massive losses in efficiency.

> What applications cannot be made to work at 48?

Basically every logic-level transistor will not work at 48V. It's nice that these last-century analog devices could be made to operate at different voltages: present-day semiconductors are not so conveniently flexible.

Simple physics dictates that required inductors to step between voltages increase in physical size (and weight, and material cost) as that voltage disparity increases. Capacitance required, etc. all increases with it. Efficiency plays into both of these as a triangle. Heat increases as this disparity increases. These properties are unacceptable for a myriad of use cases.

[pokey00]

You didn't have to convert voltages all the time for things you did in a car until relatively recently. All the typical 12V stuff could've run at 48V no problem. By the time we wanted to put computers in cars and charge our cell phones, switching supplies were readily available. The only part of the system now that really benefits from lower voltages are semiconductors.

[elihu]

I think another big part of it is that DC switches tend to get expensive above 12V. Cheap AC switches work fine at higher voltage, because the arc is self-extinguishing as it passes through zero twice per cycle, but DC doesn't do that so you can end up with an arc that doesn't extinguish itself, which, aside from not turning the thing off when you want to, burns out the electrical contacts.

[dreamcompiler]

Now we have MOSFETS and IGBTs that can switch DC without sparks. I suspect most DC switching in the Cybertruck is relayed through these. (Except the main contactor and pyro fuse of course.)

[myself248]

And the contactor is usually switched with no current through it. The only time it would open under load is in an emergency, and presumably it only does that once.

Which is to say, yeah, it's a non-issue for pretty much everything. Even in the 12v realm, new BCMs have so few relays anymore, almost everything's done with onboard FETs and software.

[elihu]

And given that this is a Tesla, it wouldn't really have had many mechanical switches in the first place. Maybe the turn indicator? Which could happily run at 5 volts or something and blink the lights via CAN bus messages.

[Kirby64]

None, unless you count the physical buttons as 'mechanical switches'.

The turn signals on a Tesla don't even make the 'clicking' noise if the Infotainment is rebooting, because it's literally just a noise piped through the infotainment.

[carabiner]

Have you been able to look at the Tesla document? Do you think it'll meaningfully help the EE's at other automakers redesign their architectures?

[panick21_]

Awesome to hear from an expert. Im looking forward to some teardowns to see how the set this all up.

[sowbug]

Thanks for this detailed answer. The chassis is normally grounded. Has anyone tried sending a positive charge through part of it? Combined with a Powerline-style signaling system, some components wouldn't need wires at all.

I already can think of several reasons why this wouldn't work, but I wonder whether there's a good idea in there somewhere.

[mcguire]

The rumor I heard was that the higher voltage resulted in lower switch lifetimes. Any truth to that?

[tgsovlerkhgsel]

Would there actually be switches switching 12 or 48V in a modern car, especially a Tesla? I'd expect the switches to only switch signal voltage/current, and power electronics (MOSFET? no idea, not an electrical engineer) switching the actual loads.

[stephen_g]

That's right, the switching concern was in older cars where they were switching the 12V straight up, so that is a reasonable point for why they never switched to higher voltages, but yes, in most modern cars (and basically all EVs) buttons and switches would mostly all just be signalling electronic units to do the actual switching.

[cogman10]

Yes, but not meaningfully. The higher the voltage you get, the more arching there is when a relay trips (also depends on if there's any sort of inductive load, think the sparks you see when you unplug a vacuum without turning it off).

But when you think about the impact that has on switches and relays, realize that in your own home you have 120V controlled by switches. Very cheap switches last decades (though admittedly not switched as often as something like a blinker).

[myself248]

Ahh, no.

AC is fundamentally different from DC when it comes to arcing behavior, because it has zero-crossings. If a switch arcs while switching AC, the arc goes out 1/120th of a second later. An arc would have to be pretty enormous to have enough thermal mass to remain ionized long enough for the next half-wave to re-energize it and sustain it. (HV AC transmission and distribution tends to have SF6-filled switches for this reason.) But around the house, your AC switches are really simple because they're not moving anywhere near that much power. And statistically, some fraction of switch openings happen with near-zero instantaneous current anyway.

DC, by comparison, is brutal to switch. It doesn't have zero crossings, so the arc has to be blown out by the design of the switch. That means nice wide contact openings, and on really large ones, magnetic blowouts to divert the arc into chutes that cool it.

If you look at a switch datasheet.... pulling up a randomly-selected one from Digi-Key now.... https://mm.digikey.com/Volume0/opasdata/d220001/medias/docus...

Look at the cycle ratings. It has a bunch of different ratings depending on the contact form (some that're forced apart, some that're sprung apart), but in all cases, the DC rating is equal or much lower current than the AC rating. And the DC ratings only go to 24V, this switch IS NOT RATED for use at 48VDC at all, despite happily going to 250V when switching AC.

So, if you're comparing apples to apples, if you had 48VAC for instance, that would be easier to switch than 120VAC. (At constant current, that is. If you want to move the same power, you need more current at the lower voltage, and it gets harder again.) But DC is oranges.

Yes, switching 48VDC is harder than switching 12VDC, but only at constant current. And it may require _different_ switches than 12VDC. Given that you only need a quarter as much current to move the same power, it's still a net win, but it's not at all comparable to switching AC.

[Kirby64]

You're correct in all of that in regards to mechanical relays, but none of this matters for solid state devices. I'm not aware of any actual mechanical relays in Tesla's. They do have contactors for the HVDC connections, but that's about as close as you get to relays. Hell, they don't even have fuses; they use self reset-able 'soft' fuses.

You have to worry about shutting down current quickly (i.e., inductor flyback), but that's a pretty trivial problem to solve.

[mcguire]

I thought about those, but the 120v AC switches are gigantic beefy things compared to most automotive switches.

[tootie]

Is Tesla's design here actually innovative or really just they're the first ones to put together a bunch of stuff that everyone knew and hasn't had the wherewithal to implement?

[ricardobeat]

> or really just they're the first ones to put together a bunch of stuff that everyone knew

Thats what 80% of “innovation” is, with the exception of applied science fields.

[stephen_g]

I've been in the telco/digital communications space for years and all this stuff has run from 48V for decades. So basically plenty of electronic parts are already available with margins suited to 48V already since it's extremely common in other industries like the one I'm in.

Automotive just tends to be a pretty slowly changing industry, but everything is ready for them to adopt 48V that other industries have been using for a long time, someone just needed to take the plunge I guess.

[myself248]

I haven't seen the document being referred to elsewhere, but I highly doubt that there's anything fundamentally new under the sun. The industry tried this before but got stuck in a first-mover-disadvantage situation, which doesn't affect Tesla as severely because they have relatively few parts in common with other cars in the first place.

So put me down in the "wherewithal" column.

That's not to discount it at all. There are some real challenges; most automotive fuses for instance, are only rated for 32-volt operation. (Fuse voltage has to do with the length of the gap opened when the element blows, and the structure's ability to withstand or staunch any arcing that may happen.) Telephone fuses would work here but they're not exactly cost-optimized, I'd love to see what they do in this space.

Switch and relay contacts too, may need different or thicker coatings to reliably break 48 volts at the number of cycles needed, but they'll be doing so at much lower currents so I think it's a net win. (Contact wear isn't my field of expertise, though.) However, mechanical switches are decreasingly relevant in the power path anyway, and FETs will definitely do better with the lower currents.

One thing I saw talked about last time, which is completely irrelevant now, is alternator load-dumps. You know, due to the lack of alternators. But in the past, with an accessory belt spinning an alternator, the power produced by the machine was dictated by the current in the field winding. Regulating the output was a simple control loop, sensing the system voltage and servoing the field current accordingly. The field winding has significant inductance so its field can't change quickly, but with a big battery sitting on the bus that didn't matter. However, if the battery lead became disconnected, and the power draw on the system decreased, the alternator would suddenly be producing too much current and unable to rapidly reduce its field, and with no battery there to absorb the overage, the result is the system bus voltage spiking as high as 120 volts, or at least that's what the load-dump test spec says you have to withstand for 400 milliseconds. In practice with incandescent bulbs and some other linear loads around, they'll typically clamp the transient to 40 volts or so, but that's still pretty harsh for stuff that's working at 14-ish.

The concern was that a 48-volt alternator could produce some truly terrifying load-dump transients. (Although I think this is also overblown; it's running at lower current so the field winding would be weaker and should be able to decrease its field faster, no? Hmm. I should do some math...)

But now that the 12v or 48v is produced by an electronic DC-DC converter running from the traction battery rather than an alternator spun by the engine, it's completely immaterial.

[jpm_sd]

Littelfuse makes some nice 58V rated blade fuses.

[brandonagr2]

Doing it first is innovative

[dreamcompiler]

Musk said the 48v stuff was not innovative at all; they were only doing the latter things you pointed out.

[brandonagr2]

Related to the weight of signal wires, Cybertruck also moved to using ethernet instead of traditional canbus, which significantly decreased the complexity of that harness