TMC is an independent, primarily volunteer organization that relies on ad revenue to cover its operating costs. Please consider whitelisting TMC on your ad blocker and becoming a Supporting Member. For more info: Support TMC
Start a Discussionhttps://teslamotorsclub.com/tmc/tags/

"Dataray"... was that a typo? (tech talk)

Discussion in 'Tesla' started by bcsteeve, May 8, 2017.

Tags:
  1. bcsteeve

    bcsteeve Member

    Joined:
    Jul 18, 2015
    Messages:
    430
    Location:
    Kelowna, BC Canada
    Short version: WTF is "DataRay"? (from here)

    Long version: I listened to the recent earnings call, and something caught my attention. At one point Mr. Musk was speaking about the Model Y and how it would not be built using the same platform at the 3. That alone was quite shocking to me, but a detail slipped out that was even more interesting to me professionally.

    He went on to half-mumble a few sentences on that point. He clearly said that the Model S has 3km of wiring, the 3 will have 1km and then the Model Y should have around 100m of wiring. To do that, he mused, you'd have to change the system bus. Whether or not they'll be allowed to do that I'm not sure (in the US and Europe, there are certain regulations regarding this but I'm sure he's more up to speed than I am), but I'd be more interested in HOW. He said that CAN bus (current system) doesn't have the bandwidth (which isn't completely correct, but it is if he wants to get down to 100m).

    Now... that's all I heard. There was something else inaudible that I didn't catch. Apparently some media source did (or did they?).

    I saw this article where they say:
    Wait, what now?!?! Where'd they get that from? And what is that? I immediately googled it, and other than a couple other references to that same Tesla earnings call (and probably all reporting off each other) I see NOTHING in terms of a data bus called "DataRay". There are various other unrelated things by that name.

    I consider myself relatively in touch with this sort of thing. I've worked on several automotive parts projects and I'm familiar with CAN and its ubiquitous use in the automotive field (again, due to regulations - as well as it being a reliable, robust and proven topology). I'm aware of a few potential future candidates for a CAN successor, but I have never heard of DataRay.

    Was this an ignorant reporter's typo? Or did I somehow miss something in my own field and my GoogleFu is not powerful enough?

    Anyone have any leads for me? If "DataRay" exists and Tesla is considering using it, I'd like to get out ahead of this.

    ps. From the name alone, I'm wondering if the author meant FlexRay. If so, that's by no means revolutionary or new. It has advantages and disadvantages vs. CAN and I don't really see it helping Mr. Musk reach a 100m goal. But the name is similar in form?
     
  2. WarpedOne

    WarpedOne Supreme Premier

    Joined:
    Aug 17, 2006
    Messages:
    3,149
    Location:
    Slovenia, Europe
  3. Brando

    Brando Member

    Joined:
    Sep 27, 2016
    Messages:
    215
    Location:
    Bainbridge Island, WA
    Elon also said 12 volts was totally wrong.
     
  4. kurdakov

    kurdakov Member

    Joined:
    Jul 25, 2016
    Messages:
    55
    Location:
    Moscow Russia
  5. MikeBur

    MikeBur ManualPilot

    Joined:
    Dec 8, 2014
    Messages:
    738
    Location:
    Seattle, WA
    Sounds like he said, and makes sense, that this was "high-speed datarate bus".

    Currently the fastest parts of the CAN bus is running at 1Mhz and is, effectively, a dinosaur in data communications. The issue is likely regarding separation of safety critical systems in a secure manner that would enable a much faster comms standard.
     
  6. Drewflux

    Drewflux Member

    Joined:
    Aug 25, 2015
    Messages:
    174
    Location:
    Australia
    I have a feeling its a updated/ported version of Flexray CAN system. The specs on say its 10Mbits/second vs normal high speed CAN at 1Mbit/second.

    This system is currently in bmw,audi,lambo,merc,land rover and some Asian manufactures highend models.
     
  7. bcsteeve

    bcsteeve Member

    Joined:
    Jul 18, 2015
    Messages:
    430
    Location:
    Kelowna, BC Canada
    "data rate"... OK, that makes way more sense. So the author just misheard it and figured he'd camel case it and make it sound like a brand "DataRay"... and other outlets ran with it.

    Incidentally, CAN is still very much relevant and can (and does) handle everything a Tesla can throw at it. This comes down to a cost/benefit analysis rather than a speed one, I'm sure. Moving to a wider bus will allow them to structure the wiring such that much less wire is required, which greatly reduces the weight and physical commodity count. That has manufacturing and performance benefits. If Model 3 proves the volume viability, then the Y can economically move to optical cable. I suspect that's where he's heading with his 100M (or was it 300M? I head 100 but the audio wasn't great. I see some reports saying 300M) target. I don't think that's achievable with copper.
     
  8. bcsteeve

    bcsteeve Member

    Joined:
    Jul 18, 2015
    Messages:
    430
    Location:
    Kelowna, BC Canada
    Yeah, I caught that too. That's a different matter, of course, but interesting.

    What is unclear is what he meant by "wrong". From a weight/cost reduction w.r.t. wiring standpoint, it would make sense to up the voltage to lower the copper requirements. And that's what media have widely been reporting - that he means a 45+ volt system. I'm not sure that's what he meant though. From a sensor/microprocessor integration standpoint, lowering the voltage makes more sense since most of them operate at 1.8V to 5.5V. Going from 12V down to sensor voltage at every sensor is wasteful and expensive. Doing so from 45+V is even more so. Having a lower V power bus eliminates that, but at the expense of requiring heavier/thicker wiring - which is counter to his "100M" statement. Or is it? Most of the wiring has nothing to do with power, but rather signal. So there could be a single higher gauge 3V or 5V power bus powering everything and an optical (or whatever) bus for signal.

    A lot of it depends on whether they want off the shelf components or they intend to customize everything. Which they can do with the volumes they expect. I wouldn't be surprised if Tesla's next acquisition and/or partnership is a semiconductor company so they can more efficiently (and safely) build their own ICs. Then they can use whatever voltage makes sense for them and deliver signals however they want.
     
    • Informative x 1
  9. mongo

    mongo Member

    Joined:
    May 3, 2017
    Messages:
    354
    Location:
    Michigan
    12 to 5/3.3/1.8 is only inefficient if a linear regulator is used, they are not. Switching DC-DC converters are 95+% efficient, even starting from 45V.

    200 W at 1.8V would be over 100 Amps. Not reasonable to run that much current in a vehicle cable, especially considering voltage drop/regulation requirements. Even less so if you design against fault handling and ESD concerns.

    -former Tier 1 automotive electronics designer
     
    • Helpful x 1
  10. bcsteeve

    bcsteeve Member

    Joined:
    Jul 18, 2015
    Messages:
    430
    Location:
    Kelowna, BC Canada
    I don't understand your 200W at 1.8V statement. I don't mean the math... I mean what sensor/etc is 200W? I'm just not sure what you're referencing so I can't really respond to that.

    As far as switching vs linear... of course switching regulators are more efficient (in most cases), but do you know of any vehicle that has a switching regulator on the load end of every sensor/display/etc? That would be not very cost effective and the electrical noise would be an engineering nightmare to deal with. From a traditional vehicle point of view, it would be incredibly unnecessary, so I can't imagine there's ever been a case prior to the last few years. When you say "they are not", how you know? Are you on the Model Y design team? Because that's really the only way you can say that with any authority. From a 100M goal point of view, you'd have to have the regulator at the load end. I can't see them doing that with noisy/expensive switching regulators. Can you imagine? What, 100? More? Every single everything with a switcher on it noisily switching back and forth interfering with each other.

    In my view, it would be FAR more practical to operate on a lower system voltage bus and use simple, cheap and QUIET linear regulators as necessary. And they wouldn't even be necessary if they can run everything on a single voltage - which is certainly possible. You'd just have one regulator at the battery side and that's it.

    Other than reducing copper AWG, there's really no advantage to running at a higher voltage. And if the goal is to reduce DISTANCE of the wiring, then it really isn't necessary to also reduce the weight of that shorter wire. Going from 12V to 45V can save them 20% of total weight but going from 3km to 100m saves them FAR more than that, even if that 100m weights (per m) even 100% more.

    Of course I could be wrong. Unlike you apparently have, I have no insight into Tesla's design ideas. But if he's figured out some way of drastically reducing distance then I don't see any benefit to also increasing system voltage and suspect he actually means lowering it.
     
  11. mongo

    mongo Member

    Joined:
    May 3, 2017
    Messages:
    354
    Location:
    Michigan
    Hi,
    Sorry for the previous terse message, mobile version was not being friendly. And nope, not on the Y team, all just guesses.

    200 watt 1.8V was a guess at a PX 2 type unit. After further searching, it looks like it is more like 250 Watts and the Pascal type GPUs run at 1.0 V. I was trying to show that distributing microprocessor level voltages is not practical. FWIW data centers are looking to using 380V-400V DC buses to power their racks (which matches the pack voltage, but doesn't seem like a safe idea for general automotive).

    To get to the 100M, they need control at the load, not meters away. I can sure see individual converter at each load. Here's why: to control a load requires a switch. The switch is already microprocessor controlled, if the micro uses a pulse width modulation (PWM) channel, and you add an inductor, diode, and some capacitors, you have a switching converter (sub $5 per lower power load). Discrete ICs to provide same function are also available if the processor is limited. High brightness LED assemblies like headlights are likely already using switching converters to boost efficiency and lifetime by providing constant current to the LED string. Linear works well for dropping the last bit of power, but if your LED voltage is far off of your supply, that's a lot of wasted power/ heat. Heat that you have to get rid of some how, and energy that could have been more range.

    If they are designed well, the converters would not interfere with each other, if they are step down (buck) type, the higher the vehicle supply voltage, the less modules on the same bus will interact. And as for noise, the main drive is a three phase umpteen kW switching converter.

    I don't see a lower voltage bus working for three reasons. The first being that linear regulators can only drop voltage, not raise it. So anything that needs a higher voltage will require a boost switching converter to operate. The second is that all the cables will need to increased in size to handle the higher current. And that is just for the loads they originally had, if things are combined on one power bus to reduce wire length, the gauge goes up even more. Finally, loads like window motors. power steering, antilock brake module, power seats, heaters, and air conditioning compressor would all prefer higher than 12V.

    Regarding wire size vs wire count: A high percentage of wire harness cost is the raw material, so half as many wires that are twice the cross-section will cost close to the same. However, the connector aspect is more complicated since connector systems are specified by their maximum current. Twice the connections at a lower current could cost more or less.

    12V to 42 V is ~3 times the voltage so circuits would carry 1/3 the current. To keep the same voltage drop, the wire could be 1/3 the cross section. That's a 67% weight savings, not 20%. Or, if they combine load power to maximize wire usage, they can cut the circuit count by 67% instead (with local control) and still have 67% weight/ cost savings assuming current is the limiting factor due to connectors. If the connector are able to handle higher currents, and the self heating of the wire was the limiting factor, then a wire 1/9 the cross section would work for the original load (power = current ^2 * resistance) or 9 centrally controlled loads could turn into one power feed to 9 locally controlled loads.
    Conversely, if the voltage dropped from 12V to 6V, current doubles. To keep the same voltage drop would require twice the cross section. However to limit heating of the wire, (maintain power dissipation), the wire would need 4 times the cross section.

    So let's talk overall circuit count/ wire length. To reduce the total length of wire, either the number of end-points (loads/ switches) need to be reduced, or the length of the individual circuits needs to be reduced. Since the number of individually controllable lighting loads has a minimum dictated by automotive regulations, the only way I see to get the wire length down is to reduce the wire dedicated to each individual load.

    For example, the rear of the vehicle has left turn, right turn, left stop, right stop, center high mount stop, left park, right park, license plate, left reverse, right reverse. In the typical topology, there would be at least 5 wires (2xturn, brake, reverse, park) running from the switches (older vehicle) or the body control module (BCM). In either case these circuit start near the firewall. Unless the control of the loads is moved rearward, we're stuck with this length of wire (a bit over 3m each).

    So let's move control back. Now we have 1 or 2 power feeds and a network connection of 1 wire (lighting is not data intensive) (assuming the brake circuit can be pure SW control, often has HW failsafe), for a total of 3 wires. Not a huge savings, but the network connection can be smaller gauge. More gains are achieved if there are separate circuits for a trailer connection. That replaces another 3-5 connections with 1 power feed. Non lighting loads like the rear defrost and power lift gate could also combine.

    Same thing goes for the front of the vehicle 2xturn, 2xpark, 2xhigh, 2xlow turn into 2 power and 1-2 network. I'm guessing the adaptive headlights are already smart to handle that function.

    If they take it to the minimal wiring extreme, the door lock/ window/ mirror control switch banks become smart. Each door would have a module to control the locks/ windows/ mirrors, but the individual control lines would be eliminated. Each door has only a power and network connection (plus however they do the audio).

    As for the higher data rate, I'm guessing the cameras provide a network feed vs individual connections to the PX 2 derivative. Same with ultrasonics.

    Will be interesting to see what their optimization point is regarding end load complexity vs wiring simplicity.
     
    • Informative x 2
  12. bcsteeve

    bcsteeve Member

    Joined:
    Jul 18, 2015
    Messages:
    430
    Location:
    Kelowna, BC Canada
    Yeah... you put a lot more thought into that than I did.

    And you're right, of course. In my automotive work, I only deal with interior μC applications. That's always <6V. Somehow in my thought processes above, I developed tunnel vision and completely ignored motors (windows, A/C) and high draw exterior lighting. You know... the biggest power hogs in the whole system besides drivetrain lol. Talk about forest for the trees... yeesh.

    So forgive me. And thank you for spelling it out. I feel stupid, but that's better than continuing on in ignorance.

    Next time I'll attempt to think thinks out a bit more before posting. *sigh*
     
    • Like x 1
  13. mongo

    mongo Member

    Joined:
    May 3, 2017
    Messages:
    354
    Location:
    Michigan
    Mostly just living through the design trade offs as OEM requirements changed. One body control module had each bulb on its own PWM channel to provide fixed voltage (also allowed for smaller wires with more drop).

    Unless we had a high intensity LED back light or VFD to keep happy, everything was on a linear regulator or constant current BJT circuit. A lot simpler and cheaper than a switcher.

    Then I started doing fuel cells with high efficiency power convertors that could handle 12V or 24V battery banks. Plus BLDC motor drives operating directly off the variable supply voltage.

    The other benefit of running 42V (could have multiple voltage outputs) as power trunk lines is that the upstream switching regulator can be current limited, so no need for down stream fuses. They shift to the feeds to the converter to protect against true hardware faults (or become non-replaceable fuse-links). Guessing this will be water cooled (or diverted for cabin heating, Humm does PX 2 provide heating?) and integrated into the high voltage junction box.
     

Share This Page