Tesla advantage
A deep dive into what makes Tesla Model 3 the best Sedan in the market
Tesla Model 3 is outselling every major American car in its category. Even though it has a higher price than the compared competition, consumers still choose it, and Tesla doesn’t spend any money on marketing ads. One of the reasons of its success is that its product was designed from first-principles, making many of the features great. In this post, I go over many of the designed features and parts of the car that make Model 3 one of the best sedans in the market.
Air conditioner
Traditional automotive vents have a low aspect ratio for their air conditioner vents, meaning that their width is relatively similar to their height. For example, circular or rectangular vents are common. These vents are generally positioned flush with the surface of the instrument panel. However, these point-like outlets are not optimized for distributing the air over a wider area, which necessitates the use of multiple vents for each passenger. Also, the appearance of the vents may be unsightly and can disrupt an otherwise uniform design of the instrument panel or other interior surface
Tesla HVAC vents don’t have moving parts. The direction of the airflow is directed by two perpendicular air streams. The software of the car determines the intensity of the air coming out of each stream so that the final direction of the air stream goes towards what the user selects through Tesla’s interface. That makes Its HVAC vents have less moving parts and with a high aspect ratio.
Cooling system
Tesla Model 3 has one packaging for its cooling system. Traditionally, ICE cars have two cooling systems, one for the motor and another one for its AC unit. Tesla can do that because its design and engineer team think about the car as one instead of having different departments for each part of the car. This unique cooling system was named the “Superbottle” by Tesla’s engineers. This is a small detail that shows Tesla attention to detail.
The Supperbottle has a few advantages to the usual cooling system:
- Increased modularity & packaging space due to the integration of pumps, actuators, and valves with the housing. (Components generally have space protection requirements, separating them could inflate those requirements depending on layout)
- Potential for increased serviceability as functional aspects of the cooling system on collocated
- Potential weight savings associated with portions of those components housing being integrated with SuperBottle. (Separate clamshell housing with standalone/off the self-pumps for a system with comparable functionality and functional requirements)
- Potential weight savings associated with lack separate pumps mounting brackets.
- Reduced final assembly costs, as that likely comes in as a complete module.
- Reduced final assembly time/labor due to component integration & use of quick disconnects is an assembly process enabler.
The video below shows how the cooling system works throughout the car.
Wipers
Like most premium vehicles today, Tesla has an automatic wiper system that automatically matches the speed of the wipers to the intensity of the rain or snow. However, unlike most other automakers, Tesla doesn’t use a rain sensor for its system, saving Tesla money.
Instead, the automaker is using its Autopilot cameras to feed its computer vision neural net to determine the speed for the wipers. The system is called deep rain as an internal joke for using deep neural networks for solving this problem, even though the neural networks are not so “deep”.
Electric motor
Tesla Model 3 has the best electric motor in the market. Compared to the next best selling electric cars, the BMW i3 and the Chevrolet Bolt, Tesla elecric motor is much better. Initially, there is the weight, here is a comparison of the weight:
- Model 3: 48 kg
- BMW i3, Chevrolet Bolt: 50 kg
Even though the Model 3 is lighter, it brings about 40 percent more power on the street.
The Model 3’s permanent magnets give a first clue of what makes the motors so powerful. Those rectangular magnets a bit larger than the size of a Bazooka chewing gum, had some small rills. They form what is known as the Halbach- Array.
This leads to the effect that the magnetic force on the side with the increased magnetic flux distribution is much stronger than the sum of the single magnets would allow. One plus one is more than two. That explains why a smaller and lighter electric motor such as in Tesla’s Model 3 can bring more torque and uses less energy compared to its competitors. This innovation was only possible because of its focus n engineering excellence.
Battery plus drivetrain
Tesla optimized for the cost of its battery pack plus EV powertrain. The price is still not on pair with the ICE equivalent. But it is cheapest battery train in the market. As Tesla ramps up production and improve battery technology, prices will keep going down. Right now the cost of the battery pack for Tesla Model 3 is around $12,200. The images below show the breakdown of the cost of the battery pack plus EV powertrain, and the the prediction of battery prices over time.
If the prices go down as expected by BloombergNEF research team, we will have prices at around $94/kWh for the battery pack in 2024. That would mean that the total cost of the battery pack will go down from $12,200 to $6,516 by 2024. That would be on par with the upfront cost of the equivalent Mass Market ICE car. Right now, some estimates state that the ownership cost for a Tesla Model 3 over a span of 3 years is already on par with a Toyota Camry due to the decreased ownership costs. Electric cars do not have as many moving parts as ICE cars, and therefore have less maintenance costs.
Tesla not only leads on the cost of its battery pack but also in its design. Where Tesla excels is in the Model 3’s skateboard-like, lithium-ion battery pack and related power electronics. “We are shocked by the advanced integration and advanced manufacturing techniques” used in the electronic control center, Chief Executive Sandy Munro stated.
Tesla uses a clean sheet design for its battery back and motor, and a rear wheel drive motor. It’s a very efficient packaging. Tesla uses 4,416 21x70 cylindrical batteries in its base model.
For cooling its battery pack, its car has aluminum extruded channels that run through the battery pack. Each one of the battery cells are placed such that these channels go through its batteries. With this design, Tesla can remove heat as the car is fast charging. Other car manufacturers didn’t design their battery packs in such a way that is easy to cool down. They used standard batteries that already existed in the market. Tesla’s design has a much lower profile compared to other battery packs of other cars.
Maxwell Technologies
Tesla is not sitting still and it is innovating in battery design. Tesla announced the acquisition of Maxwell Technologies, an American battery technology company. Maxwell has developed a new kind of battery and technology, the dry battery electrode technology. This new technology allows the development of new batteries without any liquid solvent in its process.
Usually in the manufacturing of batteries, the materials are sprayed with a toxic solvent, that needs to be dried and baked off. The solvents in the process create off-gassing and pollution. The D.B.E technology doesn’t need the toxic solvent.
That, in turn, makes the batteries using this tech much more dense. They also have a smaller weight and reduced production cost.
They are also showing longer longevity. Tesla’s older batteries show a lifetime of around 350k miles, and the new battery pack that Tesla is developing would last around 1M miles of operation. Tesla is going to hold a battery investor’s day this year to showcase all the work that it is developing.
Suspension
In true Elon Musk fashion, Tesla actually used concepts from NASA when it was refining the suspension settings of the electric sedan. The electric car maker based the Model 3’s suspension settings on a study by the space agency about how long the human body can be subjected to a certain frequency without feeling uncomfortable. Considering that the vertical frequency of a suspension’s movement affects comfort and drivability, Tesla engineers settled on a vertical frequency that is equivalent to a brisk walk or a slow run to give the Model 3’s chassis a comfortable, sporty feel.
Brakes
Model 3’s braking system is quite unique, in the way that Tesla opted to equip the electric sedan with more expensive four-pot brake calipers at the front wheels instead of a single-piston sliding mechanism. This gives the Model 3 superior pedal response, and it opened the door for the electric car maker to design its own piston seals that fully retract the brake pads after braking; thus, boosting available driving range and cutting drag. Such a system adds to the Model 3’s efficiency, which has proven superior to other premium electric vehicles like the Audi E-tron and the Jaguar I-PACE.
Wheels
The bulk of mass of an EV is usually located lower than in an ICE-powered car due to the battery packs being mounted under the floor. As a result, there is less vertical force build-up through the outside pair of tires to generate grip when they corner. To tackle that, Tesla focused on tread stiffness, developing new compounds to deliver the desired combination of cornering grip and low rolling resistance. The tires are filled with sound-absorbing foam to suppress noise amplified inside the tyre cavity.
Each rear wheel has six degrees of freedom — five links and one damper, similar to a double wishbone — but the links have been split to give better control over the forces transmitted through the tyre’s contact patch. The front suspension has also been designed to provide maximum protection in the stringent, small-overlap frontal collision crash test.
Safety
NHTSA tested Model 3 Long Range Rear-Wheel Drive as part of its New Car Assessment Program, a series of crash tests used to calculate the likelihood of serious bodily injury for front, side and rollover crashes. The agency’s data shows that vehicle occupants are less likely to get seriously hurt in these types of crashes when in a Model 3 than in any other car. NHTSA’s previous tests of Model S and Model X still hold the record for the second and third lowest probabilities of injury, making Tesla vehicles the best ever rated by NHTSA.
In addition to its near 50/50 weight distribution, Model 3 was also designed with an extremely low polar moment of inertia, which means that its heaviest components are located closer to the car’s center of gravity.
Like Model S and Model X, Model 3 benefits from its all-electric architecture and powertrain design, which consists of a strong, rigid passenger compartment, fortified battery pack, and overall low center of gravity. These safety fundamentals help to prevent intrusion into the cabin and battery modules, reduce rollover risk, and distribute crash forces systematically away from the cabin — all while providing the foundation for our superior front crumple zone that is optimized to absorb energy and crush more efficiently.
In frontal crashes, Model 3’s efficient front crumple zone carefully controls the deceleration of occupants, while its advanced restraint system complements this with pre-tensioners and load-limiters that keep occupants safely in place.
In pole impact crashes, in which a narrow obstruction impacts the car between the main crash rails, energy-absorbing lateral and diagonal beam structures work to mitigate the impact. This includes a high-strength aluminum bumper beam, a sway bar placed low and forward in the front of the car, cross-members at the front of the steel subframe that are connected to the main crash rails, and additional diagonal beams in the subframe that distribute energy back to the crash rails when they aren’t directly impacted. An ultra-high strength martensitic steel beam is also attached to the top of the front suspension to further absorb crash energy from severe impacts, and the rear part of the subframe is shaped like a “U” and buckles down when impacted.
Model 3 also has the lowest intrusion from side pole impact of any vehicle tested by NHTSA. Unlike frontal crashes, there is little room for crumple zone in a side impact, so we patented our own pillar structures and side sills to absorb as much energy as possible in a very short distance. These structures work alongside the vehicle’s rigid body and fortified battery architecture to further reduce and prevent compartment intrusion. With less intrusion into the cabin, our side airbags have more space to inflate and cushion the occupants inside.
Chargers
For electric cars to work, it is also necessary that there are chargers located around the world, to make its cars usable.
Tesla has around 16k superchargers around the world in a network, that allows drivers to drive freely across the United States and other countries. Other carmakers do not have their own chargers, and third-party providers do not provide fast-charging like Tesla’s superchargers do. Tesla is already at its third version of its superchargers. Its third version is water-cooled, delivering a whooping 250kW of power. That means that people can add up to 180 miles of range to a Model 3 Long Range in just 15 minutes.
Tesla superchargers are a huge moat to its business. Other automakers don’t have such infrastructure, which makes drivers have much less driving freedom.
Tesla’s software is integrated with its hardware, and as such Tesla heats up the battery before it starts charging, making the battery be at optimal temperature, making its charging speed increase.
Self-driving system
Tesla has the most advanced self-driving system out there. Its system is a combination of the best-in-class hardware with best-in-class software. Let’s break it down.
Hardware
Tesla developed its own processors for making neural network inference in its cars, which some analysts say is 4 years ahead of competition — based on Nvidia’s timeline. Tesla claim a factor of 21x improvement in frame per second processing versus its previous generation , which was powered by Nvidia hardware.
While the power consumption of the device has increased, it didn’t come anywhere close to the scale of the capability increase.
Chip engineer Pete Bannon, who now leads the development of chip for Tesla, stated that hardware costs about 20% less per car than the Autopilot hardware 2.5. And that the difference is going to pay for the development of the new hardware.
Elon Musk stated that they are already working on the next-generation of the chip and they expect it to be 3 times better than the current chip that just went into production.
Tesla is also already developing on their own hardware for training the neural networks before they are deployed to its cars, which they aim to increase the efficiency by an order of magnitude at a lower cost compared to an out-of-the shelf GPU sold by Nvidia.
Software
With a fleet of approximately 500,000 vehicles on the road equipped with what Tesla claims is full self-driving hardware, Tesla’s fleet is driving about as many miles each day — around 15 million — as the next self-driving competitor has driven in its entire existence. 15 million miles a day extrapolates to 5.4 billion miles a year, or 200x more than Waymo’s expected total a year from now. Tesla’s fleet is also growing by approximately 5,000 cars per week.
That allows Tesla to keep improving its software and test it out in different conditions. Let’s take a look the three key areas where data makes a difference:
- Computer vision
- Prediction
- Path planning/driving policy
Computer vision
One important computer vision task is object detection. Some objects, such as horses, only appear on the road rarely. Whenever a Tesla encounters what the neural network thinks might be a horse (or perhaps just an unrecognized object obstructing a patch of road), the cameras will take a snapshot, which will be uploaded later over wifi. It helps to have vehicles driving billions of miles per year because you can source many examples of rare objects. It stands to reason that, over time, Teslas will become better at recognizing rare objects than Waymo vehicles.
Tesla’s Director of AI, Andrej Karpathy, explains in this clip (taken from his Autonomy Day presentation) how Tesla sources images to train object detection:
Prediction
Prediction is the ability to anticipate the movements and actions of cars, pedestrians, and cyclists a few seconds ahead of time. Anthony Levandowski, who for years was one of the top engineers at Waymo, recently wrote that “the reason why nobody has achieved” full autonomy “is because today’s software is not good enough to predict the future.” Levandowski claims the main category of failures for autonomous vehicles is incorrectly predicting the behavior of nearby cars and pedestrians.
Tesla’s fleet of approximately 500,000 vehicles is a fantastic resource here. Any time a Tesla makes an incorrect prediction about a car or pedestrian, the Tesla can save a data snapshot to later upload and add to Tesla’s training set.
Whereas images used to train object detection require human labelling, a prediction neural network can learn correlations between past and future just from temporal sequences of events. What behavior precedes what behavior is inherent in any recording (video or abstracted). Andrej Karpathy explains the process in the clip below:
As with object detection, the advantage over Waymo isn’t just more data for predicting common behaviors, but the ability to collect data on rare behaviors seen in rare situations in order to predict those as well.
Path planning/driving policy
Path planning and driving policy refer to the actions that a car takes: staying centered in its lane at the speed limit, changing lanes, passing a slow car, making a left turn on a green light, nudging around a parked car, stopping for a jaywalker, and so on. It seems fiendishly difficult to specify a set of rules that encompass every action a car might ever need to take under any circumstance. One way around this fiendish difficulty is to get a neural network to copy what humans do. This is known as imitation learning (also sometimes called apprenticeship learning, or learning from demonstration).
The training process is similar to how a neural network learns to predict the behavior of other road users by drawing correlations between past and future. In imitation learning, a neural network learns to predict what a human driver would do by drawing correlations between what it sees (via the computer vision neural networks) and the actions taken by human drivers.
Tesla is applying imitation learning to driving tasks, such as how to handle the steep curves of a highway cloverleaf, or how to make a left turn at an intersection. It sounds like Tesla plans to extend imitation learning to more tasks over time, like how and when to change lanes on the highway. Karpathy describes how Tesla uses imitation learning in this clip:
Also as with prediction, no human labelling is needed once the data is uploaded. Since the neural network is predicting what a human driver would do given a world state, all it needs are the world state and the driver’s actions. Imitation learning is, in essence, predicting Tesla drivers’ behavior, rather than predicting the behavior of other road users that Teslas see around them. As with AlphaStar, all the information needed is contained within the replay of what happened.
Based on Karpathy’s comments about predicting cut-ins, Tesla can trigger a car to save a replay when it fails to correctly predict whether a vehicle ahead will cut into the Tesla’s lane. Similarly, Tesla may capture replay data when a neural network involved in path planning or driving policy fails to correctly predict the Tesla driver’s actions. Elon Musk has alluded to this capability (or something similar) in the past, although it’s not clear if it’s currently running in Tesla cars.
The inverse would be when a Tesla is on Autopilot or in the upcoming coming urban semi-autonomous mode and the human driver takes over. This could be a rich source of examples where the system does something incorrectly, and then the human driver promptly demonstrates how to do it correctly.
Other ways to capture interesting replays include: sudden braking or swerving, automatic emergency braking, crashes or collision warnings, and more sophisticated techniques in machine learning known as anomaly detection and novelty detection. (These same conditions could be also used to trigger replay captures for prediction or camera snapshots for object detection.) If Tesla already knows what it wants to capture, such as left turns at intersections, it can set up a trigger to capture a replay whenever the vision neural networks see a traffic light and the left turn signal is activated, or the steering wheel turns left.
