I’m sufficiently disoriented at this point to scarce recall what I expected of the Tesla Model 3. For years we knew the progression was from the top down starting with very premium high margin low volume vehicles and working our way down toward higher volumes and lower prices.
This made perfect sense to me and I rather scorned the Nissan approach of taking a really lame lower end car, the Nissan Versa, and adding technology until it was quite unaffordable in an effort to produce a really lame electric vehicle for the masses. I much preferred Tesla’s top down migration as it covers a multitude of sins and missteps along the way, but also because of the upscale imprimatur it stamps on the electric vehicle more generally. My view of the mission from the beginning was that this was not a technological problem. It is about acculturation of the public to the idea and concept of electric cars. They should be shown how to DESIRE one. The actual building of one is sufficiently simple minded that a 54 year old unemployable working half days falling down drunk in yellow shoes could make a car in three months that would go 110 miles on a charge and hit 94 mph. Don’t ask me how I know.
I did not purchase a Roadster. By the time they were widely available the price had risen to about $126,000 for a two-seat sports luxury car and I just couldn’t quite get my head around all that as a personal value proposition. But of course I did indeed pony up $107,000 for the $55,000 Model S using Musk pricing metrics – that never was really about $55,000.
So it didn’t faze me particularly to cough up $56,000 for the $35,000 Model 3, using the same ElonMath I was by now accustomed to.
But I guess what I was expecting was a tamed down/toned down Model S – smaller, less power, but basically a miniS for the masses. And I was hoping the disappointments would be minimal and mostly about things I didn’t care about too much.
When push comes to shove, I do care. But it’s a mixed basket. Some things seemed actually improved – like the brilliant solution to the $400 car key thing and I was delighted with the air conditioning and heat, the visibility, and actually the performance if you must know. The interior seems easily as large and spacious as the S. And actually the weeny door geometry of the S was improved a very important have inch for ingress and egress. The car looks great and the little bit of additional range over the S is actually key to a St. Louis urban area 120 miles away. The S just won’t really get there and back without charging, and the Model 3 will. I’ve done three roundtrips with at least 50 miles remaining on the clock on my return. The 310 miles is very real if you keep it 70mph or under.
True, the ride is stiff and the controls and display screen, I’m sorry they are just comically bad. My initial experiences with autopilot, autopark, summon, and even the garage door opener have been poor. I actually view autopilot not as improved safety as Musk hopes, but perhaps fatal in even the proper hands. Autopilot is a bent arrow and it’s hard to predict where it will strike in the end. But it isn’t looking good. It’s actually CAUSED me three harrowing near misses on excellent Interstate highway in severe clear weather. I’ve been an aviation pilot for years, and rather think I drive a car very conservatively but with precision and I’ve a driving record of some 45 years now without a reportable accident. I just don’t have close calls.
This past few weeks we have received our first salvaged Model 3 allowing us to do a bit deeper dive than I was quite willing to make in my brand new daily driver. And the thot plickens most curiously.
So to give you the end punch line BEFORE the reasons why, the Tesla Model 3 is not at all a tamed down/toned down Model S. In fact, despite my personal choice of color and wheels to match my S, it simply isn’t an S at all and I’ve been viewing this car through entirely the wrong lens. Broadly speaking, it has nothing to do with a Tesla Model S at all and I can’t find any indication internally that it is even manufactured by the same company. It doesn’t even use the same LOGO on the components. The familiar Tesla T has been replaced with a Model 3 logo that I don’t even know what to make of, which kind of matches my impression of all that we are seeing within the car itself.
I don’t even know what it is but it shows up EVERYWHERE on components within the vehicle. It looks like a Mason Jar used to “put up” peaches but with a cape that has a stylized 3 on it. If anyone can explain where THAT came from I’d love to hear about it. What is it SUPPOSED to be or evoke? Peach Technology? SuperPeaches?
But the really stunning thing is that ALL electronics and technology in the Model 3 are VERY next generation with NOTHING familiar from the Model S in any way shape or form. I haven’t seen ANYTHING even vaguely Model S or X in the vehicle. NOTHING. No DNA found. And the improvement, from what I can tell at this early stage, is not at all incremental. It’s like its from another planet.
I feel like we are examining a crashed spacecraft from another world. Picture Star Trek with an annoying ping ping tweet sounding every few seconds and an overly dramatic William Shatner ever ready to make huge drama from nothing as a Tesla Model S. And Star Trek Next Generation with Patrick Stewart as the Model 3. I’m down in engineering and it just doesn’t look the same here with Jordy and Commander Data as it was with Scotty ever trying to give Kirk “more power” somehow. In the Model 3, we resolve all irresolvable issues in the final five minutes with Tachyon Waves, which we never do quite have to explain but MAN are they some hot shit.
And in a way, I am shocked. It’s as if everything we thought we knew about the Model 3 is totally wrong. And the story of this remarkable device is simply a not yet revealed secret. And with all the hoopla about the Model 3, no one has ventured anything on this at all. What are they doing over at TeslaMotorsClub forums? I was aware they were kind of simple minded, but they do go on and on and on and I would think SOMEBODY would have actually looked under the hood or something.
We’ll be looking further at a number of items in coming weeks, including inside a fascinating drive unit, an astounding charging technology, and much more. And I rather doubt I’m going to have all the answers, and mostly likely some of the answers won’t even be correct. I trust you’ll mention it in the comments.
I guess I fear you’ll not find this moving with sufficient speed. People seeking sub 4 second 0-60 times are not a patient lot generally. But generally the Model 3 technology is beyond my paygrade. Which is cool.
I’ve kind of been bored with electric cars in recent months and increasingly thinking about solar. But I face a huge learning curve on the Model 3 and that makes me all dancy and chipper again.
Inn this episode we removed the heart of the melon, the battery and took a peek inside.
In recent months we’ve been doing a lot of work on Tesla Model S battery modules and packs with an eye toward using them for Solar Energy Storage. We did develop a little control kit to use the Tesla Battery Modules and the BMS board that is mounted on them and we can scarce keep them in stock at this point. http://store.evtv.me/proddetail.php?prod=basicsmodulecontroller I rather had aspirations of starting the investigation into the Model 3 battery for a very good reason.
In all of Tesla put together since 2004, the company has made some amazing strides in initiating the conversion of our personal transportation to electric battery powered magnetic drive. No surprise that I rather favor that as a strategy. But understand that in ALL that time and ALL put together, they have sold 300,000 vehicles world wide.
Today they stand with 450,000 orders in hand, give or take a hundred thousand as the whim and the mood of the apostolic following’s mood varies, striving heroically to reach a 5000 unit per week production rate. That’s 250,000 vehicles PER YEAR. So with the X and S, they intend to manufacture more vehicles in the next year than their cummulative total manufacturing to date.
In the Q1 Analysts call, Elon was a little short with a guy asking where the 450,000 orders stand. I initially also thought it odd but on reflection, he was actually very nice about it. They are all over him as to when he’ll be able to produce enough, WHILE Tesla car sales in China have ballooned in the last year to over $2 billion representing 19% of their total automotive revenues. Musk is staring down the maw of a monster trying to eat him if he doesn’t paddle faster producing cars and here is some freaking moron wanting to talk to him about DEMAND for a car that’s never been advertised anywhere and which he can’t possibly make enough of for about five years coming NO MATTER what he does or how well he does it. At some point irony begets absurdity and absurdity ventures into moronic chant, and that’s where he was in a very exasperating moment. I’m sorry Wall Street, but you’re going to have to up your game or sit this one out. It won’t be like you’ve never missed a big move in technology before.
My point is, California drivers being what they are, it’s going to soon be RAINING Model 3 batteries in the junk yards. But we don’t know how to use them. And I haven’t heard of anybody else that does either.
So I’m going to share what we learn as we go along hoping to advance the token more widely as well. But if you notice something in the photos and descriptions, please feel free to jump in and suggest something. Remember, there ARE no stupid questions, only stupid people asking them. Note to stock analysts.
The first thing I would note is that in this vehicle, Tesla has integrated MOST of the vehicle electronics INTO the battery assembly which is put together in the Gigafactory in Nevada. The charger, BMS controller, junction box, DC-DC converter and all protection circuitry has been moved into a hump on the end of the battery termed for no apparently good reason the “penthouse”. This can actually be accessed for service by removing the back seat as it projects upwards exactly where the seat cushion is for the back seat. You would have to remove the seat bottom, a couple of braces, and then a bunch of star head screws, but it pops off and allows access to the fuses and contactors and so forth pretty readily.
The combined charger/dc-dc converter is a thorough MARVEL inside. We will be looking more closely at this in the future but it again has NO apparent relationship to any previous Tesla charge circuitry, and they squeeze 10kW of charger into a remarkably small and cool running package.
Contactors and fuses and so forth are most notable in the compact arrangement eliminating dozens of feet of expensive cabling and connections. They did add a pyro-disconnect. Picture a fuse you can blow up with a digital signal totally and most assuredly disconnecting the pack from everything in the event of say, welded contacts on the control relays. A total failsafe and kind of on the principle of an airbag detonator.
And no evidence of a SINGLE part in this assembly that has ever been near a Model S. Entirely different contactors. Of course a different charger/dc-dc converter which is now a “power conversion unit” and the BMS board is now entirely different and referred to as a “high voltage controller”.
It didn’t take long to strip this down and remove all that and we decided to learn to deal with a Model 3 battery as a solid assembly in the future and probably with another wreck. For this one, we opened the can to get right to the modules.
The modules are much larger than in the S or X. They are long narrow logs of cells with 23 cells in two of the modules and 25 cells in the other two.
The two 23 cell modules are at each edge of the pack and 67 1/2 inches long, 11 1/2 inches wide and 3 1/2 inches thick and weigh 191 lbs.
The two 25 cell modules are located in the middle of the pack and 73 inches long, 11 1/2 inches wide and 3 1/2 inches thick and weigh 207 lbs.
And of course what caught our immediate attention was the very very different battery management system boards located on each module and connected by a tiny two wire cable in daisy chain fashion and back to the high voltage control board.
I’ve been explaining battery management systems for 10 years now. It’s been a shitshow. And an annoying one. The number of BMS experts available in the world who know absolutely NOTHING of which they speak is astounding in quantity and awe inspiring in their insipid lack of comprehension as to what they know and think they know with almost all of it completely dicked up. And they pose a danger to battery owners everywhere. I gave UP trying to find one for EV’s and also gave up designing one of my own and developed survival techniques using LiFePo4 batteries without them entirely. Unfortunately, those techniques are simply impractical with the more energetic NCA cells used here and with Tesla’s packaging techniques.
The issue is that measuring battery cell voltages is deceptively simple. And unbelievably hard. And for simple minds, the simple part seems to be the part they like. One thing worse than no BMS is a really badly designed amateurish ass clown BMS that you THINK is working and controlling your charge cycle, but it isn’t, and your charger burns up your battery pack while you snooze peacefully one floor above the car in your now burning house. And then of course some times the amateur BMS’s themselves simply catch fire and find a ready fuel supply in the batteries they are attached to. And if none of that happens, the spaghetti wiring connecting the BMS to the battery catches fire. The end result is the end result. And it’s not a good outcome.
The problem is common mode voltage. It SHOULD be evident, but it apparently is not apparent to many, that the cells are in a stack that results in voltages 100, 200, or 400 volts higher than the return terminal. And while there are hundreds of circuits available to measure 3.6v or 3.2 v very nicely and with great accuracy, most of the semiconductors simply can’t stand to do so at voltages hundreds of volts above ground or above the potential of their own power supply.
And so you have a double problem in that you have to power the BMS circuitry FROM the battery cell, while measuring them, and you have to isolate each measurement from the others – usually with some sort of isolated communications bus linking them all together, or some combination of all of that. And so the very simple looking task quickly becomes a monster. I’ve known PHD level EE who have simply dicked such a thing up beyond all possible recognition and at times catastrophically.
The nature of one off EV builds is that EACH is unique and so EACH is a beta test site for something designed by someone else for something different. No possibility of harm there.
And so imagine my horror when BOTH the EV community AND the solar community began using salvaged EV batteries for their own concoctions, and generally without the BMS at all or using something designed for something entirely else?
But the economics are apparent to anyone. Half-priced lithium battery cells. This is what everyone has wanted all along. So it is GOING to be done. How to enable them to do it safely. Or at least more safely than hoping for good things…? And this in a world where MOST of the BMS designers out there suffer from delusional fantasies of competence.
Indeed Tesla has had six or eight totally catastrophic vehicle fires with essentially no known cause. But out of 300,000 cars, you know what? That’s not too bad all things considered. In fact I would hold it up as a model to aspire to.
Tesla has a multilevel engineered approach to battery application design and most of the parts are actually quite important. The first line of defense is the fusable links they use to connect each cell to the current collectors. If any cell does short out, it will immediately receive current from the brethren spirits adjacent and of such quantity that it burns up the little connecting wire. That takes it OUT of the circuit and it cannot receive further current from the pack or provide any to it. It’s been command ejected from the cockpit electrically.
They do cool the individual cells with coolant and heat them as necessary to prevent charging below 0C/32F – freezing. This is a major cause of cell death. But temperature control need not be terribly high tech.
The main BMS board and software really is about controlling contactors and monitoring high voltage safety interlocks and threats from the outside world. That’s pretty easy stuff.
But the cell stack monitoring and balancing is where the “don’t know and don’t know they don’t knows” hit the fan. Inherently, the semiconductor devices are already LESS reliable than the battery cells. And when you put them in hostile common mode voltage floats they just can’t cope. If you connect them to a bunch of unshielded antennas (connecting wires) in a noisy automotive environment, you court armageddon.
So our strategy is to handle ordinary housekeeping such as DISCONNECTING THE ENTIRE THING IF “ANYTHING” is even slightly amiss. And using the highly engineered and somewhat road proven designs Tesla has onboard for the cell stack. We want to TALK to that over an isolated bus, learn what it knows about the battery cell voltages and temperatures, and otherwise connect NOTHING to it or to the cells at all.
We’ll overlay a more configurable contactor and charging control layer on top of it. But cell stack monitoring must remain intact and unmolested.
The Tesla Model S modules feature a dedicated stack monitoring BMS board on each module, securely wired to the individual cells and two temperature sensors. This board uses a Texas Instruments cell stock monitoring board that is designed to link vertically using an SPI bus. Serial Peripheral Interface.
But Tesla didn’t use that because SPI is a bit twitchy in noisy environments. They used an RF isolation chip on each board to allow it to daisy chain the boards on a simple two wire serial connection to the master BMS board. We figured out how to wire to that isolated bus and talk to the Ti chipset over it to get all the cell voltages and temperatures and trigger cell balancing and all that.
So note well the humble philosophy of this. We are not going to attempt to design a safe BMS for you. We’re going to steal one that is already proven somewhat safER by those educated in the art, supervised by the responsible, and with a shitpot to lose if they don’t get it right. But even in theft, it comes mounted ON the module you’ve paid for, dug out of the car someone already paid for, so it’s not much of a theft. We like to euphemistically refer to it as “repurposing.” The rag and bone man’s ancient right. And even there, care is required to keep from defeating important but not readily apparent safeties.
So there really are safety issues regarding the use of these batteries, and they have nothing to do with the child minds and their preoccupation with rubber gloves and safety glasses. Feel free to fondle the end terminals. They tingle but are mostly harmless. Like a warm stove, you’ll figure all that pretty quickly. Fires and explosions are not.
The Model 3 battery stores sufficient energy to hurl a 3500 pound car with a couple of, shall we say, “well fed” adults over 300 miles. A sudden release of that energy is non-trivial.
And so we look with great interest to the BMS board mounted on the Model 3 cells. And did Tesla build on their experience and expertise with the Ti chipset? Well no. It is ENTIRELY different. Advanced, next generation stuff.
The modules feature what is basically a six foot long flexible printed circuit board with copper runs insulated by a translucent insulating layer and connecting to each “brick” or “cell” of the battery pack plus a couple of temperature sensors. There is one of these copper run “tapes” running down EACH edge of the module. The bricks or what I call “cells” is actually made of of 46 cylindrical cells 21 mm in diameter and 70mm long. They are nominally 3.6v at 5 amp-hours capacity with the 46 connected in parallel. So 3.6v x 230 Ah per cell and 25 cells for 90 volts on the long ones and 23 cells for 82.8volts This gives us a nominal pack voltage of 345.6volts at 230 Ah for a total capacity of 79,488 watt-hours. It has been LISTED at 80.5 kWh elsewhere and that COULD be true if the amp hour capacity were 232.92 Ah indicating 5.0636 Ah per cylinder. I’m good with that but we’re parsing the fine print here.
There are TWO 38 pin chips connecting to the left side copper tape and two 64 pin chips connecting to the right hand copper tape. One of the ongoing little never discussed issues with measuring cell voltage is that if you are pumping any current through the same wires you are measuring, the drop of the wires causes perturbation of the measured voltage. So whenever you “balance” these cells, you throw off the voltage measurement.
It is my belief that this generation chip has a real measurement resolution as small as 0.0022volts or 2.2 millivolts. So very very small resistance values can cause errors. Note the three tiny wires connecting to each copper land. And I guess I think we are doing cell balancing on one side of the module, while simultaneously monitoring voltage on the OTHER side of the module entirely out of the current path. This is a stunning solution and very difficult to implement normally because it means twice the numbers of wires and connections. But the flexible PCB tape approach makes this quite feasible. I’ve never seen anything like it. It’s like technology from another planet.
Our interest is of course in communicating with them. And handily there are two connectors on the board apparently just for that featuring precisely TWO pins on each connector. It would be wonderful if this were CAN but of course it isn’t. On the board you will note four modules labelled XFMRS. If you can’t spell very well, and you say it very quickly, it comes out like TRANSFORMERS. And indeed there appear to be two in each module, but they aren’t very powerful at that size.
And I am curiously drawn of course to the 64 pin chip and it is quite unusual. First there are TWO of them necessary but one of the modules has 25 cells. And so the previously largest 12cell stack monitors can’t quite cover the necessary 25 cells. Collin assures me that there are indeed 3 or 4 different cell stack monitor chips out there in a 64 pin configuraiton but I can’t find them.
Linear Technologies was acquired last year by Analog Devices for a handsome $46 in cash and .232 shares in the result. In 2008 they introduced the LTC6802 Battery Monitoring chip and has done pretty well with it. They have updated it nearly annually and have recently introduced the LTC6813-1 which they refer to as 5th generation in this series.
If we blow up our image of the chip, we don’t see much. So Collin etched off the conformal coating and did some light and angle on it with a microscope to determine a little of the logo and part number.
I did a little photoshop magic on it to further enhance that.
And so what we THINK we see there is 1722 T2-A2 PZ38984. Which there is no reference to such a number on planet. This chip is indeed from outer space.
But there is also a curious little figure on the chip. And it looks like our goofy jar of superpeaches – the Model 3 logo.
And so we’re feeling a little dead ended here. A custom chip made for Tesla. And how are we going to reverse engineer a custom chip? If they KNOW they are going to sell 500,000 Teslas, and each board has TWO of these chips, and each battery has FOUR of these boards, that looks like smells like 8 million chips. I guess almost anyone would make about anything they like and print anything on it they wanted to see for forty or fifty million dollars.
Now all of this had to be done two years ago. But last year Analog announced the LTC6813-1 in a 64 pin package. Single units are $22 a pop. But I’m guessing Tesla paid less and once it was designed, LTC had the right to also introduce it separately as the LTC6813-1. So I’m going out on a limb here to say LTC started working on this chip with Tesla before it was acquired and was allowed as part of the deal to release it as their next gen chip as long as they didn’t identify it as the same as the Tesla chip. Or something like that. I hope. Otherwise, I’m nominating a chip that wasn’t announced prior to pencils down at Tesla.
The reason I like this theory is that LTC several years ago introduced an isolated 1Mbps serial bus on TWO wires they call Isolated Serial Peripheral Interface of ISOSPI. And it uses tiny external TRANSFORMERS to provide isolation. It also steps the standard 4-wire SPI to this single pair of conductors.
So my hope is that if we use the LTC isoSPI communications protocol and the LTC6813 register commands, that we can talk to this alien beast through the “universal translator” and try to make peace between our species. As you see, I’m quite willing to throw the prime directive under the bus on this one.
So we are going to gen up a little hardware to speak isoSPI to the board, and try to at least get a wakeup call to bring it to life and talk to us a bit.
As the last remaining humanoid on the Internet who COULD actually BE wrong, I have to confess I’m making all of this up out of pretty rarified air approaching lab vacuum. Hopefully some knowledgeable person will correct me in useful fashion and of course we’re open to thoughts on the 38 pin chip as well. They are very much in the daisy chain but my guess is we’ve separated the balancing from the measurement, at unusual and extreme expense and difficulty.
And so my assertion in the video and more emphatically so at this point, given the level of advanced technology and integration I’m seeing in this battery assembly, and assuming the same level of effort from the Panasonic team at the chemistry/cell level, this is the BEST most ADVANCED large scale lithium battery every produced on planet and is YEARS ahead of anything currently in work. Tesla appears to be saying to the rest of the industry, “We’ve upped our game… so UP YOURS!”
Like a 7-year-old on Christmas morning, I am in absolute DELIGHT at what we are learning about this remarkable car. The kid in me is going ape shit by what we are seeing in the charger and particularly in the motor drive unit. Across the board, this ain’t your daddy’s Oldsmobile. And I can pretty much assert that everything you THOUGHT you knew about the Tesla Model 3 is basically a miss. It’s going to be a delight, but a bit of a longer more drawn out affair to explore just what IS in this remarkable device.
Youtube teardown of 2170 cell https://youtu.be/_uKpn3zflBE
71 thoughts on “Tesla Model 3 – NextGen Battery”
Munro & Associates claim that Tesla named chip sets Batman and Robin, (the caped mason jar being Batman) but did not state their source.
Pingback: Elon Musk may never deliver a $35,000 Model 3 - Page 3 - Chevy Bolt EV Forum
While I see what you mean about the logo looking like a jar (maybe of peaches) I have always thought and continue to think that it looks like a battery cell – maybe not exactly like an 18650/2170 but definitely a generic battery cell of some kind. It leads me to believe that there is probably a team of engineers within Tesla that is responsible for the battery module (including penthouse) and they consider this the superhero component of the Model 3 – and they might not be wrong.
It’s a play on words. Batman = battery man = battery manager
And what is Robin’s purported roll, other than to say “Holy Smokin Rat Shit Batman!”????
Fascinating Jack. I raise my hat to your electronics sleuthing. For those interested here is a 2013 article on ISOSPI in the context of BMS’s: https://www.eetimes.com/document.asp?doc_id=1280373
Jack you state:
“The first line of defense is the fusable links they use to connect each cell to the current collectors
The first line of defense is the CID and PTC fuses within the top of the 2170 cells, there are these two
fuses before you get to the thin wire fusable links!
Fascinating. Please provide a link or reference to this information.
While this is true for normal marlet cells, my understanding of the Model S batteries is that they omit the CID and PTC inside the batteries entirely because they rely on the fusable wire links outside of each cell. I expect this is the same on the model 3. See this page for info on both PTC and CID as well as someone else confirming they don’t have these features in their cells (near bottom of article) https://batterybro.com/blogs/18650-wholesale-battery-reviews/18306003-battery-safety-101-anatomy-ptc-vs-pcb-vs-cid . There are other results when searching for “Tesla Cell PTC CID” and its confirmed that Roadster cells had these features, but nothing concrete other than people on forums claiming that Model S cells don’t have them.
Here is a link to a model 3 cell dissection. There are three holes for venting and insulators on the top and bottom but cannot tell if there are CID and PTC fuses.
I’ve watched this video before, and have just re-watched it. As you say it is hard to tell if there is a CID or PTC in this battery because he doesn’t take apart the positive pole area, but as far as we know they aren’t present in the Model S batteries and I wouldn’t expect them to include them in the Model 3. They have really reduced the battery down to as minimal as it can be (vents to prevent explosion and insulators to prevent shorting) in order to maximize battery capacity per volume and per weight and are relying on the rest of the system to provide safety. My guess is that the fusable wires weigh less than CID and PTC would, as well as being a more reliable mechanism and easier to diagnose.
I agree completely. I have personally disassembled Model S 18650s and they absolutely do NOT have PTC in them. Or anything else for that matter beyond the jellyroll of aluminum foil. I can’t imagine them adding such nonsense to the Model 3 2170’s. But of course I could be wrong… ergo the request for the link that doesn’t seem to be forthcoming…
I’ve added a link to the bottom of the blog for a teardown of Tesla 2170 cell. Total bullshit. None of that is in the cell.
Here is a pretty good interview with Sandy Munro on the Tesla Model 3. Sandy is super impressed with the electronics and drive but not so much with the body and manufacturing methods.
I thought it was a little bizarre. He thought the NVIDIA autopilot card was made by Tesla. He went on and on about how the body was overbuilt. Duh. Apparently missed the Tesla video with the Volvo S90 and Model 3 doing comparative side impacts. The overbuild is very deliberate and I think stellar. As no one has ever manufactured such a car, what would he know about manufacturing methods?
This video clamshell’s for us technological disruption. Old school meets the disruptive new technology and they wind up scratching their heads. THey continue to itch until the office furniture disappears at 10 cents on the dollar.
Nice things to say about the suspension work. But really a huge miss on the NVIDIA board and the body structure. I understand. I’m just now coming to grips with this thing myself – kind of the topic of the blog this week. It’s hard to get your head around.
Well the card is based on Nvidia chips but addition of extra GP106 to the standard dual Parker chips for PX2 is more than doubling the compute power of the card.That gp106 gets them additional 1280 the cuda cores of vs 512 that 2x Parker gets.
Parker has 1.5 TFLOPS while single GP106 reaches 3.8. This is a very significant divergence from PX2
Given the huge dedication and direction of NVIDIA currently with a line of basically supercomputers with an Autopilot development suite, the possibility that Tesla designed that board are ridiculously below zero percent. Indeed, much of Tesla’s autopilot talent have left to work at NVIDIA. NVIDIA is currently ground zero in autopilot development. It would be a profound act of hubris for Tesla to NOT use an NVIDIA design. Musk doesn’t have that gene in his DNA.
It cant be confirmed or denied if there is a cid or ptc, until someone dismantles a tesla 2170 lid and posts
the info, it looks unlikely, but I’m hoping someone such as the video Jack has linked can do that work and post results. Its a small piece of information that seems to missing at this point in time.
The video I linked DOES make it clear there is neither.
this video seems to suggest no ptc or cid ( no close up to confirm is it is actually tesla 2170 cell)
Thank you Jack Dickard!
Just one question: with their new motor, we now know that Tesla has a better electrical engine, munching around 2% in efficiency. But what about other parts of the power electronics chain? How efficient is the total Model 3’s electrons-to-torque conversion, compared with a Model S or X? That’s quite interesting. Of course we have the EPA numbers and other tests but I meant in a detailed fashion, piece by piece.
The title is NextGen Battery. Not motor.
We are looking at the motor and I think you’ll be shocked at the size and format of the Silicon Carbide inverter power electronics. But the measurements you seek will never be available. Your 2% is quite off base wherever you got it.
The overall drive train is 6-7% more efficient as discussed in an earlier video. And I would attribute almost all of that to the motor/inverter. It is kind of an eye opener but we won’t discuss it until we complete a video examination of it.
Thank you for the reply (and sorry for the typo, just realize now, can’t edit though).
Yes that was indeed a related question, not directly to the battery’s electronics, but since you were astonished by the level of it, I was wondering of the state of it in other parts of the car.
The 2% efficiency gain on the motor alone was retrieved from research papers comparing in theory both motors, without knowing exactly what Tesla really achieved, which would already be impressive considering the already high numbers achieved by the good old AC induction. I just noted your measured 89% efficiency (vs 83% in the S) in your video https://www.youtube.com/watch?v=qfmgj8nB3Z0 which I didn’t watch until now. Close to 90% would be very, very impressive indeed… Can’t wait to see this tech in a fully refreshed Model S/X platform which will eventually happen in the next years.
PS: I never watched your videos (too long, sorry), just knowing you by name and just discovered you had a blog.
The actual efficiency of the motor would be interesting. But very difficult to separate from the gear drive assembly and the costs of a test setup to actually perform that analysis would be considerable for not very much information.
And do understand the calculations for the video you cite are a very rough wag. But I’m convinced there is a significant gain there that could not be attributed to a simple permanent magnet motor of any design with which I’m aware. As you might imagine, very small increases in efficiency at that level come very dear.
On May 19, Musk admitted in a tweet that the Model 3 does indeed employ a PMSRM hybrid. The literature I’m reviewing now is promising. But no such motor has actually been deployed at these power levels and in production quantities. This represents a significant and ballsy risk on the part of Tesla Motors. As I’m driving one, I would posit it paid off.
We’re “too long sorry” for a very specific reason. I see in this case it didn’t work. But we have a new affiliate it would appear that does a summary for child minds you might find more viewable.
Thanks again for the reply. By too long, to be completely honest, I also meant that it was not dense enough and I prefer spending my time reading technical papers than watching 1/4 info and 3/4 ramblings… Sorry to say it this way but this my honest take. Too bad really because a shorter format with more dynamism and content could attract a lot more people out there, Tesla watchers being the first ones.
Also I still think that a independent test of the motor would prove very valuable, not just for the efficiency question but also what is made of… and the software! The infamous PMSRM must have asked a lot of deep knowledge in this field and from what I read, there are a lot control programming that must have demanded years of development. (and also, how much of it is being updated via OTA?…) This is watch so cool about EVs: software is eating the car world too indeed!
Well clearly I’m not as dense as you are. But my initial analysis of your comments remain. You prefer to read very important scholarly journals. You are probably a very SMART and very IMPORTANT and very BUSY man who walks briskly and carries a lot of VERY IMPORTANT papers around. Actually I’m intimately familiar with that type. And I gather your purpose here was to stress that fact because it works entirely best if everyone or at least someone recognizes that you are very intelligent and very important and very busy.
I recognize and acknowledge it. You clearly are WAY past needing anything here buried within the ramblings and ravings of an old guy in yellow shoes. So what are you doing here?
Again, I think you would find Jehu’s summary much more dense. I think his density and yours might just be a better match.
In terms of hybrid permanent magnet / reluctance motor, you should look at the BMW i3 (it has 97% peak efficiency). The Tesla M3 rotor magnet layout looks like a less sophisticated version of BMW patent 2012/0267977 figure 7.
And where did you see the Tesla M3 rotor magnet layout?
“And where did you see the Tesla M3 rotor magnet layout?”
Tesla Model 3 motor disassembled:
Thank you Mark. Yes, Steve Bakker’s article of course. I talked with him at length about that. I was pretty certain it was an internal geometry for air gaps and magnets, but only had general geometries from published papers on the specfiics. Munroe’s showing the actual laminations helps muchly.
The SiC Bakker puzzles so over have been available for a year or so by Cree – I think Wolf Creek is the name of the distributor now and they are SiC MOSFETS. MOSFETS have always been much superior to IGBTS for power switching in most ways, but were difficult to make to handle high voltages. Cree’s SiC MOSFETS are now capable up to 1200 volts. They allow much higher switching frequencies, but their hugish advantage is much much lower ON resistance values – often 10 milliohms or better. This dramatically reduces the heat loss and also makes it so you can use smaller devices and smaller heat sinks for much higher power levels. I have had the inverter off the motor and knew this. But taking the motor apart to the point of removing laminations I am loathe to do. I want to be able to put it back together and potentially run it.
“The SiC Bakker puzzles so over have been available for a year or so by Cree”
It was reported that Tesla is using STM SiC chips in Model 3.
“STMicroelectronics was first to market with a new class of highly energy efficient 1,200-volt silicon-carbide chips, which brokerage Liberum said helped it win the initial deal to supply power chips for the Model 3. But it predicted Infineon would become a second source of such chips as volumes ramp next year.”
If you have readable chip markings on the inverter, I’d love to see them.
Retired electrical engineer here. I designed a simple BMS for a 6 cell battery using the LTC6802 back in 2009. Two things I’d like to point out. The differential IsoSPI bus is really nice in noisy environments as any EMI causes a similar change in voltage on both sides of the differential, so the noise is converted to common mode which a differential receiver tunes out (only looks at the difference in voltage between the two signal lines). Which is why CANbus is also differential signaling.
Second, the LTC6802 had tiny internal fets to accomplish cell balancing during charging by diverting some current through these internal fets. However, this wasted power generates heat internal to the chip, so the Linear Technology engineers allowed you to drive external fets with these pins so the heat could be dissipated externally in larger fets with more robust heat dissipation specs.
The problem with all this fet controlled cell balancing is that you are wasting power to keep the cells balanced during charging. But what if the cells were so tightly controlled during production that there was very little imbalance in the first place? (As Sandy Munro has discovered with his cell balance statement of less than 0.2 mv difference). With such small balancing currents needed, you would greatly reduce the needed heat dissipation which may even allow you to go back to very small internal fets on the main chip which would be the cheapest solution. My conclusion: Tesla’s making their own cells in the Gigafactory with Panasonic may be their biggest advantage over their competitors, none of whom seem to want to invest the huge amounts of money needed to build their own captive battery factories.
You would appear to confirm the central tenet of this blog. Designing BMS systems is a lot harder than it first appears.
I think some gains can be made in evening capacity by going to a larger cell format. But designing batteries is a lot harder than it first appears as well. Applying cathode and anode coatings in uniform fashion is devilishly difficult at any realistic production speed. I have to believe the slight improvement in cell balance, a bit overblown by Munroe, is actually achieved BY balancing. Now I might preface this with the fact that I do NOT know what I’m talking about. But the expense of running TWO complete sets of connections to every cell in a battery simply is not warranted for “redundancy”. From direct and repeated personal observation, it is quite hard to accurately measure a voltage through the same connection used for current draining or charging. Indeed, we often run a set of large leads to drain or charge and a separate set of small wires for voltage measurement using high impedance meters.
And so I think we have two chips on one side to monitor cell voltages and two different chips on the other side used to balance them using entirely separate connections. And this I would theorize accounts for the finer balance of the cells, not the precision manufacture of the cells themselves.
As the identity of the one chip is at this point speculative and unproven, and we haven’t a clue on the second chip, I could be completely reading messages from God in cloud formations here. Even the speculation on the first chip, in case any of you missed it, implies Tesla was using a chip at pencils down that was not even announced, documented or available at the time. That would be very unlikely, UNLESS the entire board and concept was actually designed and built by Linear Technologies themselves, and the chip LATER released as an LTC/Analog Devices product. I think it quite likely that LTC is actually the supplier of the entire module BMS board.
Running separate leads for operating mode and measurement is a standard Kelvin connection: the two measurement wires to the high impedance meters do exactly what you think: very little current drawn by the meters means very little voltage drop along the measurement leads and therefore more accurate measurements. This is also done in dc power supplies where the measurement leads are used to remotely sense the voltage close to the point of use of the dc power and thereby compensate for the voltage drop in the high current leads from the power supply output to the point of use. And the sense leads are usually twisted pair, sometimes shielded, to protect the measurement from EMI.
As for the cost of those flexible circuits, well they aren’t as expensive as they used to be. Plus, when ordering in quantities of a hundred thousand, well the price drops significantly. So they might be not very expensive compared to lots of other things in the car.
Which brings up another point: when it comes to electronics, Tesla doesn’t seem to pinch pennies. Rather, they seem to understand improved performance and are willing to pay for it where it makes a difference. Unlike many other companies who seem to think saving a few pennies with a marginal design is a good idea. Until it isn’t.
Well, one thing I can confirm about the Model 3 BMS is that the 38 pin chips we don’t know the model of connect to the cells through 10k resistors. There is no other connection so those chips can’t really be doing any appreciable balancing. They have SPI right next to them and labeled such that one has to believe the SPI is for programming. If that’s the case then maybe those chips are MCUs of some sort. But, no proof of anything other than that the chips connect to the cells and seem to also connect to the ISOSPI lines.
The 64 pin chips are a different story. We’ve got a pretty good idea of what those chips are and how to interface with them and talk to them. Those chips connect to each cell a bit differently. Each cell has two connections. One goes through a resistor I haven’t gotten a good enough picture of but it looks like maybe 47k. That goes to a capacitor (usually) then through another resistor. This looks like an RC filter network and would be for cell voltage monitoring. In parallel to this is a 1k resistor that runs seemingly straight to the 64 pin chips thereafter. That connection must be for balancing. And I find that very interesting because that would suggest that they balance cells through a 1K resistor. Of course, that would yield something around 3-4ma worth of balancing. Even if all cells were balancing that would only be a max of about 50ma of balancing current per 64 pin chip – quite manageable even though Tesla is using the built-in FETs for balancing. The only way 4ma of balancing could have any hope of working is if the cells are EXTREMELY close in capacity and characteristics. This would seem to suggest that they’ve got insane control over the battery characteristics and basically no balancing current is required to keep them in line.
I must say, 4ma balancing is the lowest I’ve ever seen. The Model S most certainly uses a whole lot more balancing current than that.
So one of the proposed modes of usage of the LTC680x chips was a pairing – One full AFE (the thing measuring and ballancing the cells) and one simple monitor. Won’t look it up now, but I remember seeing it in the LT’s app note – the second chip was simpler, cheaper, but would signal something is amis _immediately_ by pulling a wire (as opposed to only learning about hazardous condition much later on your next query over SPI). That could be a viable explanation.
As for balancing, 1k seems a little high, but the balancing currents used in EVs are generally pretty low. Model S AFAIK balances at about 100mA, Leaf at less than 10mA. Since long term, the only thing you fight is the uneven self-discharge (the electrons have nowhere else to go otherwise in the serial cell arrangement), you have hella lot of time to get the pack in order…
So what did they recommend for the second chip?
Love to have that app note.
It was late night so I posted before finishing the thread and w/o realizing someone mentioned it later on – LTC6801 + LTC6802 combo back in the days. 6802 is 44 pin full featured AFE with balancing, 6801 is 36pin monitor only. The 6802’s datasheet (http://www.analog.com/media/en/technical-documentation/data-sheets/68021fa.pdf) shows a redundant configuration at page 18 (with both sides 6802). This article (https://www.edn.com/Pdf/ViewPdf?contentItemId=4012539) explains using 6801 as the alternative.
As for split harness for kelvin connect, I don’t think so. That’s what time division is for – the LTC chips are already smart enough to turn off balancing before taking the voltage sample, but can have it both ways and also with extra sink current for self test.
I have a question though – with everything in this LTC line being 12-cell and with 23S short modules and 25S long modules, is the BMS really handling 25 cells on the long modules, or did Tesla go with 4 virtual 24S stacks handled by 8 chips as a single monolitic 96S stack? That is, is the 25th cell on the long stack actually managed by the BMS of its shorter brother?
Could you tell us the dimensions of the entire battery unit as it was pulled from the car? Do you think there’s any hope of the DIY community like you guys figuring out the charger etc so we could use the entire battery assembly as is in a conversion?
As to figuring out the charger etc. so you could use the entire battery in a conversion, I would say its not only doable, but inevitable.
Your writing reminds me of reading “Ignition!” by John Clark. Very entertaining!
(can’t reply your last message, so I write a new message)
Ahhh don’t take it so wrong Mr. Dickhardt! The crux of my post was to say that you could reach a much greater audience, that’s all. I really think that you could adjust your video’s gear ratio you can achieve a 1:10 conversion in number of followers, without giving up the rambling (I like it moderately and so do a lot of people) and the yellow shoes (I prefer yellow boots but that’s just me). I watched older shows, like the one from June 22, 2012 for the launch of the Model S and that was good! https://www.youtube.com/watch?v=k_xg0XukRCk It gives me some motivation to watch more regularly (also, I have to admit, the screeching mic was too a reason for giving up on the show, it’s like Youtube kryptonite to me).
All in all, I’m pretty impressed that you didn’t ban me already, which shows how a fine gentleman you are with good humor (like in your ramblings too by the way). Good humor and good intelligence to also recognize my superior cognitive abilities! Finally I found someone admitting it! (…and a bit of sarcasm maybe? I hope not.)
— Rouget, a ‘important busy smart’ person (JR,
oilman from Dallas, EVman from Cape Girardeau)
PS: Jehu like in Jehu Garcia? https://www.youtube.com/user/jehugarcia Didn’t understand at first that you meant him. Sure he also mentioned your long videos, right. I see a trend, just saying.
Weird! I can’t post new message here!
“Mason jars with a handle or wine stem cause me to lose faith in humanity“
Wow. Thank you a lot for post. I am an EE working in the energy storage field and have experience with LT (now AD) 6813 and 6804/11 BMS chips.
At first I though the primary (64-pin IC) BMS would be indeed LTC 6813 and secondary LTC 6811. But looking closer, the secondary BMS chip does not have the correct pin count (LTC 6811 has 48 pins, this only 38). What does not make even more sense is no cell voltage filtering (usual LPF to filter out the differential noise on the pack) and cell balancing bleed resistors on both chips. I am curious if it is double sided board and these components are placed on the other side. Excessive energy bleeding could be done inside the BMS chip as well, but the current would be severely limited in that case.
Nevertheless this design, like you put it, is out of this world. As the battery pack of Model 3 is not easily changeable, they need to have the BMS as reliable as possible (in addition conforming to functional safety standards).
Anyway, I might be completely wrong. I am just happy to see that somebody goes the extra mile to document all of this stuff. Good stuff!
We had the same thought. We did disconnect the board and lift it and found no active components of any type on the reverse side. There were a number of 603 style resistors.
There are eight transformers and so all four chips use isoSPI but oddly the Robin chips have full SPI broken out as well. For testing? Not sure.
ANy further thoughts on the 38 pin chip are welcome. I’m thinking FPGA at this point but Collin indicates it can’t do much current through the external resistors.
If we can draw some kind of parallel to laptop’s batteries (“SmartBatteries”), the 38 pin chip could be a “2nd-Level Overvoltage Protection”, although it isn’t usually connected to anything else rather than the cells and the 3 pin fuse – Smart Batteries also have a fuse (Three-Terminal Fuse, they call it) that can be digitally blown to permanently disconnect the cells from the computer.
An overview: http://www.ti.com/lit/ml/slyp087/slyp087.pdf
You might want to break the habit of resting tools on top of the modules such as drills and anything over a pound or so in weight. Those ribbons running the length of each side of the modules have a habit of breaking much like the ribbon cables used for LCDs.
Second chip is probably backup battery surveillance for functional safety
In case the main supervision chip goes dead it can be detected with the secondary chip and the system can go to a safe state.
Second chip would be less accurate and manufactured in a different process.
This is a usual a process in highly safety critical applications
That’s a jar, but not for peaches.
I saw the other image on the left side of the “flexible” PCB that looked like a duck.
Started thinking… Maybe it’s a mallard?!? Could it be that the jar was for… Marmalade?
Marmalade, Mallard, Model 3??? Could they be code names that sound like Model 3?
But then I read the speculation about BatMan and Robin. Could that Mallard be a Robin? Maybe not a marmalade jar but a Bat jar? or a Battery Jar!!!? Old batteries used glass jars for each cell. Here’s a picture of one.
Could it be that the right side of the PCB goes to the Battery Management circuitry (Batt Man) and the left goes to monitoring circuits? Didn’t Robin just kind-of keep an eye on BatMan? You know… monitor him???
Pingback: Teardown reveals Tesla Model 3 has "best battery ever built to date" - Technology News Your Way To Success- Pradeep.win
You couldn’t send the chip board to Chipworks for them to take a look? They can do X-rays and such on chips. That’s their specialty.
I had a beer with a former employee who knew just a little about this (not an EE). What I could gather:
The icon is Batman- does high resolution battery monitoring and balancing. It’s a purchased chip but not sure the supplier- your guess of Linear seems reasonable. The other side of the board is Robin- it is a completely Tesla designed chip and is a low cost, low resolution backup incase anything on Batman side fails. Batman is 16bit ADC vs 12bit on Robin. He thinks Tesla’s goal is to replace Batman with their own chip to lower the cost, and the base model battery will only have one set of sensing. He doesn’t think Robin even works yet, which may be why SPI is broken out on it- some future over the air update will enable it.
Hope this helps.
Thanks for the excellent info. Sounds entirely plausible if a little discouraging. Well actually encouraging in a way. Collin had posited the backup concept. I was a little hesitant on that because of TWO complete sets of contacts to the cells. That’s kind of a big deal just for redundancy. But it would make sense if you wanted to measure the voltage to extremely high precision.
The discouraging part is if Tesla is designing their own ASICs, it will make reverse engineering enormously more difficult.
That makes sense. I could find no existing chip that matches the “robin” chip. It seemed to be custom to me but I didn’t know why Tesla would do that when there already exist many good options. Of course, buying singles of the BMS chip is something like $22 so even though Tesla would have a big volume discount they might still find it kind of expensive. If you’re going to build a million cars and each has 8 BMS chips then you have the volume to potentially do it yourself. The problem, seemingly, with the Robin chip being the only one on some models is that it has no balancing capability. There simply aren’t enough pins to allow it to balance. So they’re always going to need something else that does balancing. It’s possible that they’ve included this current chip just to test things out on everyone’s car. Since the more advanced BMS chip is there they can rely on Batman while fiddling around with Robin to perfect things. Then they could replace both with another chip based on Robin but with balancing. But, that sounds a bit weird. It’s a waste of money to put chips on boards and not use them or just use them for testing. They can test locally at their facility or with cars that their staff drive. There’s no reason to add cost to cars they’re selling when they’re bleeding money like a pig at a slaughter house.
If nothing else, I’d believe that it is a redundant voltage reading chip and that they use it as a backup to ensure that things are going well. Still, the Model S doesn’t have redundant voltage reading and it seems to work fine. It’s a little weird any way you look at it. This is supposed to be a cheaper car and generally you don’t go overboard with redundant BMS on your cheaper design while leaving the old single scheme on all your more expensive models.
Any way you go, it seems like these Robin chips aren’t strictly required so it should be possible to control and monitor just using the 64 pin chips we know about.
I think the “robin” IC is simply a custom secondary protector, like the LTC6801. It’s just an extra hardware comparator based check of OV/UV limits without the need for a sample command. It helps them reach a higher level of functional safety in the system.
Perhaps. What leads you to “think” that?
It’s inarguable that two is better than one if redundancy is for some reason desired. But it is a lot of board real estate and chip to accomplish that.
The Model S boards were not redundant and I’ve never heard of a battery failure or problem from the use of those boards. So what were they trying to improve or fix in a less expensive car?
While it’s their less expensive car, it also marks the general evolution of their design towards better reliability; which will presumably find its way into their more expensive cars too at some point…
Beside general monitoring of the correct operation of the balancers, I can think of some other possible uses for the side-kick chips:
– When the Bat(t)man chips or their sensing lines blow up, the Robins might serve as a fallback to ensure safe operation in some sort of emergency mode until the battery can be replaced, instead of leaving you stranded.
– Measuring an unexpected potential difference between the left and right sensing lines might help in detecting problems within the modules, such as individual cells or cooling channels failing.
I’m not sure whether the latter is actually a realistic proposition — but it would certainly be desirable, considering that inability to detect faults in individual cells is sometimes cited as a safety downside of designs using many small cells. Indeed in the recently reported Model S battery fire, apparently bystanders already saw smoke emanating from the car, while the BMS didn’t notice anything being amiss…
The voltage monitoring hypothesis seems plausible. Noting some other things:
The SPI header has a pin marked VPP. That usually means programming voltage, which suggests SPI is an input. I’ve never seen a microcontroller which is programmable by SPI, but some peripheral chips can be configured that way. This connects to another pin marked VDD-5V – in other words it’s a general supply rail and also the way the programmer powers it.
The chip has power rails marked VDDD, VDDA and VCC, suggesting an analogue chip with separate analogue and digital supplies.
There’s some pins marked FSK-A-P, FSK-A-N, FSK-B-P, FSK-B-N. FSK traditionally means ‘frequency shift keying’, but I can’t see how that would work here. However it does suggest two differential pairs called FSK-A and FSK-B on each chip
Given that there’s already four pins elsewhere marked ISOSPI-A-P and similar, I might take a hunch that the Robin chips are also speaking IsoSPI and it might be worth scoping them to see if the voltage levels are plausible. Also, from the SPI connector, it’s possible to work out the length of the SPI chain (give it a 1 on MOSI and see how many cycles it takes to appear on MISO) which might help narrow down if it’s speaking any protocol used by another chip.
I also note it’s in TSSOP38 which is a slightly unusual size. Most popular with TI for MSP430 microcontrollers and BMS chips. So if I were putting money on an OEM I’d put it there, but it’s not impossible they fabbed their own and TSSOP38 was what the packaging house had.
To pile on (a little late) the redundancy would be a part of ISO26262 and ASIL D which most certainly a battery would fall under. The secondary circuit sole job would be to trigger the thermal fuse (I think I saw one) in case the primary failed to prevent the laptop battery fiasco of a few years ago. I’m sort of surprised they didn’t go with https://www.nxp.com/products/power-management/battery-management/battery-cell-controllers/14-channel-li-ion-battery-cell-controller-ic:MC33771 but perhaps politics got in the way. Good to see that Tesla is getting some automotive engineers in the mix.
I was watching a YouTube video on how SpaceX uses commercial chips on their space control systems rather than much more expensive hardened “Space Qualified” low volume parts. They use multiple levels of redundancy which was explained. It is possible he is using a similar approach. It might be worth watching what SpaceX is doing to help understand what he is doing on the Tesla control systems.
Could it be a Leyden jar?
Interesting article. At the risk of being a downer here I have a different perspective to share. The architecture shown here is another attempt to copy of the early A123 systems battery management architecture which went into 4,500 hybrid busses starting in 2007, and then into 1KV multi megawatt grid energy storage systems around the world shortly there after. (A123 Patent number: US 7990101) LTC wanted the chip business so they pulled a bunch of requirements from us at the time and essentially copied the architecture. TI did the same. They couldn’t meet out price targets or our demonstrated accuracy of .001V @20hz so we never bought the chip versions and implemented it with discrete components. The chip manufacturers did win however because they sold the rest of the world on the architecture.
Tesla may have done a good job integrating the battery pack electronics but to describe it as next generation from another planet goes a bit far ; )
Interesting point. Thank you. Yes, my enthusiasm was expressed as perhaps hyperbole. Sue me. To my backwoods aborigine mind, this looks very advanced. TO your of course superior intellect and experience, it is perhaps not magic.
We are struggling a bit at the moment as it appears the tiny wires and lands are ALUMINUM and cannot be soldered. Spot welded. Very inconvenient I just say.
My own enthusiasm wanes just a bit. The modules are long, heavy, and at a very different voltage. And they have a very fragile BMS board on them. Repurposing them poses challenges somewhat different from MOdel S modules.
Hi Jack. I just found your blog and videos for the first time. Amazing work! Congratulations. It might be off topic but actually I was looking on internet for state-of-the art way to connect and combine the individual cells (e.g. 18650) to each other in parallel. There are lots of examples out there which uses spot welding but replacing cells in such batteries is a nightmare. Unfortunatelly your great video about Model3 battery ends before this question was answered. 🙂
From this problem another question came to my mind. There are ready to use solutions to check the series connection of battery packs using chips like LTC6813. But in a 14S15P configuration (my case) how can I identify problem inside 15P block?
Pingback: Expert Says Tesla Model 3 Battery Pack Is Most Advanced Ever Produced
Pingback: Tesla Model 3 Has "Most Advanced Large-Scale Lithium Battery Ever Produced," Battery Expert Notes | CleanTechnica
I can lend some credibility to your Linear Tech (now Analog Devices) theory. I am an electrical engineer in the Dallas area and approx 18-24 months ago I was at the Linear Tech offices in Plano, TX. I had driven my Model S to show to one of the FAE’s that I know. During a brief tour of their facility, he pointed out an engineer who was working on designing a battery management or monitor IC for Tesla. I didn’t get any additional details, but am almost 100% certain that is the chip you have found in the Model 3 battery pack.
I’m a big Tesla fan, and an engineer at heart, and this is the most interesting article I’ve ever read regarding the model 3. Thank you so much for your time in writing this.
I love your investigations. You do what i can’t. I fit into the don’t fully understand but really like the potential crowd. We’ve lived off grid for 7 years. I just swapped our lead acid batteries for chevy volt packs which i adjusted to 14s. They work really well but took a lot of work and testing to finish. If someone clever like you could supply an off grid power system at 173v nominal, using 1/2 a tesla model 3 battery pack, people like me would buy it. I haven’t seen anyone make use of the model 3 battery module voltages yet but look forward to watching it happen. Duncan.
I just bought a 25s Model 3 module and plan on using it in a Sprinter Van RV camper build. Three key solutions make it possible:
– 96V chinese inverters (80 to 105VDC) and ELCON 6.6KW charger are/is on order.
96V chinese (MakeSkyBlue) solar charger has arrived.
– Mounting the module on it’s side, on a side wall of the Van will be done with proper support, strong wood or aluminum supports in the X,Y and Z directions.
– Orion BMS2 with CAN bus charge and discrete safety controls. Charging will be through solar, alternator, J1772 EV or RV shore power (120 VAC or 240 VAC split phase).
I know there is risk in a one-of-a-kind build, but this is not my first rodeo. This design should last ten years. In 2030, hopefully the battery size will be cut in half.
Thanks for your detailed analysis of the battery and BMS board.