As we continue the tale of Elon and Martin, we pick up from the initial involvement of Elon Musk in Tesla Motors. Musk has put in the single largest chunk of money that has allowed Tesla to develop and now start building an electric car. Here we delve into the subject of his role in design decisions. As some in the mainstream media have taken to referring to Musk as the "creator" of the Tesla Roadster, this is a particularly contentious subject.
The whole issue of the choice of transmission suppliers is a particularly thorny one. Musk was insistent that Tesla should build a car worthy of the price tag. Anyone who has ever worked in the auto industry knows that is almost always a lot harder than it looks. The friction that clearly existed between Eberhard and Musk from very early on in their relationship definitely didn't help matters. (Note:If you missed Part 1 check it out first).
ABG: So, how did you get to the first stage, of the Roadster, the specs that debuted in 2006 with the first prototypes? How did that come about?
Read on after the jump.
ABG: So, how did you get to the first stage, of the Roadster, the specs that debuted in 2006 with the first prototypes? How did that come about?
Elon: That was basically an iteration between myself, Martin, and JB.
ABG: From your perspective, at least, how did you get to the point where you had a car with a two-speed gear box and the lower power motor that it started off with?
Elon: Technically it's a lower torque motor. The motor power stays about the same but the torque increases (with the upgraded motor now being tested). The problem with the AC Propulsion motor is that when you go from a kit car, the tZero, to a production car that actually has all of the safety systems and resists all the crash issues and actually has all the amenities like a real stereo system and the thing adds a fair bit of weight. So in order to have good performance, you either have to upgrade the motor torque and the current capability of the power electronics because the vehicle weight has increased, or you have to have a two-speed transmission. If you don't do that, you end up with a car that does not have sports car performance. So in fact, you're going to end up with a car that is worse than the Lotus Elise, which is a car that's half the price.
Again Eberhard's perspective on this was slightly different:
"The 2-speed transmission was the first major edict to come from Elon, and though I thought it was an unnecessary risk for the first model year's cars, I was certainly willing to be a team player and support Elon's edict. I knew there were risk and cost associated with the decision, but by themselves, I felt we would be able to manage them.
Keep in mind that with a 1-speed, I was not proposing performance that sucked. See the executive summary that I sent you. AC Propulsion's tzero was turning out 0-60 in about 3.6 seconds with a single-speed transmission. We thought that we might just break 4 seconds with Tesla's additional weight, with our "worst case" estimates coming in around 4.8 seconds. This is still EXTREMELY quick for a sportscar, and would have been a great car."
Elon: Less than half the price, actually. So it's a very tough sales proposition to tell people "please pay twice the price for a car with worse performance." It's not that nobody would buy it, that car. There would always be some people that would buy that car. But could you make a business out of it, would you sell enough to actually show that the business works. And if you can't show the business works with the Roadster, you'll have a real hard time convincing investors that to give you money for car number 2 if car number 1 is a flop.
So, the approach that I wanted to take was, the right architecture in my view is, and JB's view actually, is let's upgrade the motor power and have a single speed transmission. It's not even a transmission, it's an rpm reducer, with a differential. There's not even a clutch or anything. It's very light; it's very cheap; and it's very efficient because you're not spinning any unloaded gears, you don't have a wet clutch or any of that stuff. So you have greater energy efficiency through the transmission.
Martin actually said he doesn't want to deviate from the AC Propulsion power or torque level in the motor and he said that, we should do a two-speed transmission instead. He actually put forward a proposal for a two-speed transmission of his own design. The thing I was insistent on was, we must have compelling sports car performance. But actually, my preferred path was not a two-speed – it was a single speed with an upgraded motor.
It was Martin who insisted that we go to the two-speed route in order to achieve that outcome. And Martin did say, "Well, what about if we just do a single speed for the first year or so, and suffer that loss of performance?" I said, "I think it's really tough to crack that perception of poor performance, and say, 'yes, it's coming down the pike or something like that.' And unless you tell me this is a critical path item, you know, that this will fundamentally delay the schedule, then we should do the two-speed. If it's just a matter of investment, we should do it. If it's a matter of delaying a critical path, then we should think about it."
Here's an another example of how different people see the same events through their own personal filters. Musk acknowledges that he pushed for the 2-speed transmission from Job 1 in order to meet the original performance targets. From a December 2007 interview in Inc. magazine:
None of this is disputed by either man and yet the motivations and interpretations are very different. Eberhard tells ABG that:
"JB was the guy in charge of the motor and inverter, and he did not have the skills or resources to redesign the ACP system. JB and I and the rest of the team discussed this approach and concluded that we (JB especially) could not achieve a higher-power motor/inverter in time.
JB was actually quite scared of deviating from the ACP design in the beginning because he did not understand it. Marc and I had to push him to eliminate the old analog control and switch to a DSP-based control, and even then, he did not do this until he had hired a substantial team of engineers, 2 years later. Again, this is not to denigrate JB. The ACP design was quite tricky and tweeked, and it was very poorly documented. JB was trying to make the right decision based on the available resources - as were we all.
Note that Tesla's new high-power motor depends on very new silicon from International Rectifier - silicon that only became available this year. No such silicon (IGBT transistors to be specific) was available to us in 2004, and we knew it."
As an engineer myself and having picked up responsibility for other peoples undocumented designs and tried to develop them I can certainly understand Straubel's reluctance to mess with something that at least worked up to a point. In this context the decision to follow the two-speed path with the lower torque motor certainly made sense at the time. Hindsight is of course 20/20 but as Eberhard explains, the electronic hardware to make the change apparently wasn't available at the time anyway.
Elon: And Martin said it would not delay the critical path. It was just a matter of investment. So I said, "OK, Let's do it." And that's true. If we had picked the right supplier for the transmission – the two-speed transmission is not inherently difficult. Unfortunately we did not pick a good supplier. The initial pick was XTrac, which is basically... they make transmissions for track cars.
They're very expensive and they don't really how to make consumer stuff. So they screwed the pooch. And then, instead of going from them, to someone who could really do the job, we went from them to Magna. Martin assured me and the rest of the Board that there was no one better to do this transmission than Magna.
Here Musk's opinion of XTrac may be overly harsh. Major automotive suppliers are often reluctant to deal with low-volume manufacturers. The engineering cost of developing a component for 100 cars a year is the same as 1,000,000. Variable costs go down with volume but up front development cost doesn't. That means that unless a low volume automaker is willing to pay those costs either in separate engineering expenses or exorbitantly high piece costs (that factor in engineering costs) they are usually out of luck.
"Regarding the choice of transmission suppliers: It was not a matter of having picked the wrong supplier. The problem was that real transmission suppliers - those that made production volumes of transmissions - were simply not interested in selling transmissions to a company that buys so few of them. The first model year's production of Roadster transmissions is fewer than a prototype run for a real production transmission. This is why Lotus, for example, uses absolutely off-the-shelf Toyota transmissions in the Elise. But because a motor is so different from an engine, no off-the-shelf transmission was at all suitable for our needs.
I put Mac Powell in charge of selecting the transmission vendor, and he hired a team of ex-Lotus engineers to spec the transmission and to find a supplier. These were all people experienced with producing cars and experienced with sourcing components for a low-volume car manufacturer. They scoured the landscape and came up with only 2 suppliers who were willing to make transmissions for Tesla - they decided that Xtrac was the best of the two, and I respected their experienced opinions. Nobody was comfortable with Xtrac, but they were the best we could find."
As Tesla's VP Darryl Siry told us some time ago the issues with XTrac weren't even entirely of their own making. The design intent to manage the motor torque during clutchless shifts apparently proved unworkable necessitating a change to a different design and supplier. So again here Musk maybe overstated the case with regard to XTrac.
Elon: What he neglected to figure out was that Magna is a build-to-print manufacturer; they're not a designer. Particularly when it requires... this requires a first principle design. You know, you have to actually understand the theory because you can't simply derive this from some existing five-speed gasoline gearbox.
According to Siry, Magna was chosen by Eberhard without any competitive bidding. Due to the late decision to change suppliers, Magna was apparently chosen because an engineer at Tesla's Rochester Hills facility was familiar with the company and had a contact there.
ABG: So if you didn't have the fundamental expertise within Tesla to provide them (Magna Powertrain) the design that needed to be built, they didn't have the capacity just to do that on their own.
Elon: Right. And that's where we got into the bloody of Magna fiasco. We spent huge amounts of money in Magna and then in the, sort of, 13th hour, Magna brought in Ricardo, because basically the design failed and Ricardo told us that the entire thing had been incorrectly designed... that we'd have to basically do a complete refresh on the thing, soup to nuts, in order to have a reliable transmission.
So that came to a head in November of last year and we said, "Well, how long is it going to take to do that?" And Ricardo said, "Well, we went through the whole timeline. It would have been like November of this year before we'll actually have a working two-speed. And the bloody things would cost a huge amount of money and your cost would be crazy high. I'm like, "Wow," so the huge development cost with a nominal timeline of, like November 2008 and a high unit cost. That's pretty shitty.
So, on the flight back from Detroit, Ricardo US is based in Detroit. JB and I were talking and said, "Hey, look, why don't we look to going back to the single speed approach and upgrading the motor power, and the power electronics current capability?"
And so we iterated on that for a bit and decided to go that path. Essentially take the development burden on something that we knew versus something we didn't know. We understood motor and power electronics a lot, but we didn't really understand transmissions. So we decided to shift the burden of the problem to something we know.
So that's what we decided to do and that's just what we're doing. We have what we call Powertrain 1.5, is working right now. We've been driving around in one of the interim prototype cars. It delivers 3.9 seconds 0 to 60, 125-mile top speed. It's actually a superior product than what we originally promised people because the 0 to 100 time is significantly improved. There's no shift delay. There's no shifting. And it improves the range slightly too.
In a followup email exchange with Musk he elaborated on what probably should have happened in the wake of the problems with the XTrac gearbox:
"The problem with Magna Powertrain (USA) is that they are primarily a build-to-print maker of transmissions. They don't know how to design a transmission from scratch, particularly if it is outside the norm. We should have contracted with a company like Ricardo for design and then with Magna for manufacturing or gone to someone like Borg-Warner that can do both design and manufacturing.
What we did after the Magna fiasco was actually both. We are working with Ricardo on a joint Tesla/Ricardo single speed, which is already working in a car now, and Borg-Warner is contracted to do a transmission for us on a completely separate path. If both designs work well, we will pick the lower cost option."
"I had essentially no involvement in choosing Magna as the transmission design house. Martin told me and the rest of the board that they were the best in the business and we could not hope for a better supplier. I only stepped in when the Magna deal started falling apart in the late summer last year and they demanded huge sums of money from us, despite failing to deliver a working product."
The discussion continues in Part 3 coming tomorrow where we discusses the lessons that Musk and Tesla have learned about starting a car company and learn more about what to expect from WhiteStar.
The whole issue of the choice of transmission suppliers is a particularly thorny one. Musk was insistent that Tesla should build a car worthy of the price tag. Anyone who has ever worked in the auto industry knows that is almost always a lot harder than it looks. The friction that clearly existed between Eberhard and Musk from very early on in their relationship definitely didn't help matters. (Note:If you missed Part 1 check it out first).
ABG: So, how did you get to the first stage, of the Roadster, the specs that debuted in 2006 with the first prototypes? How did that come about?
Read on after the jump.
ABG: So, how did you get to the first stage, of the Roadster, the specs that debuted in 2006 with the first prototypes? How did that come about?
Elon: That was basically an iteration between myself, Martin, and JB.
ABG: From your perspective, at least, how did you get to the point where you had a car with a two-speed gear box and the lower power motor that it started off with?
Elon: Technically it's a lower torque motor. The motor power stays about the same but the torque increases (with the upgraded motor now being tested). The problem with the AC Propulsion motor is that when you go from a kit car, the tZero, to a production car that actually has all of the safety systems and resists all the crash issues and actually has all the amenities like a real stereo system and the thing adds a fair bit of weight. So in order to have good performance, you either have to upgrade the motor torque and the current capability of the power electronics because the vehicle weight has increased, or you have to have a two-speed transmission. If you don't do that, you end up with a car that does not have sports car performance. So in fact, you're going to end up with a car that is worse than the Lotus Elise, which is a car that's half the price.
Again Eberhard's perspective on this was slightly different:
"The 2-speed transmission was the first major edict to come from Elon, and though I thought it was an unnecessary risk for the first model year's cars, I was certainly willing to be a team player and support Elon's edict. I knew there were risk and cost associated with the decision, but by themselves, I felt we would be able to manage them.
Keep in mind that with a 1-speed, I was not proposing performance that sucked. See the executive summary that I sent you. AC Propulsion's tzero was turning out 0-60 in about 3.6 seconds with a single-speed transmission. We thought that we might just break 4 seconds with Tesla's additional weight, with our "worst case" estimates coming in around 4.8 seconds. This is still EXTREMELY quick for a sportscar, and would have been a great car."
Elon: Less than half the price, actually. So it's a very tough sales proposition to tell people "please pay twice the price for a car with worse performance." It's not that nobody would buy it, that car. There would always be some people that would buy that car. But could you make a business out of it, would you sell enough to actually show that the business works. And if you can't show the business works with the Roadster, you'll have a real hard time convincing investors that to give you money for car number 2 if car number 1 is a flop.
So, the approach that I wanted to take was, the right architecture in my view is, and JB's view actually, is let's upgrade the motor power and have a single speed transmission. It's not even a transmission, it's an rpm reducer, with a differential. There's not even a clutch or anything. It's very light; it's very cheap; and it's very efficient because you're not spinning any unloaded gears, you don't have a wet clutch or any of that stuff. So you have greater energy efficiency through the transmission.
Martin actually said he doesn't want to deviate from the AC Propulsion power or torque level in the motor and he said that, we should do a two-speed transmission instead. He actually put forward a proposal for a two-speed transmission of his own design. The thing I was insistent on was, we must have compelling sports car performance. But actually, my preferred path was not a two-speed – it was a single speed with an upgraded motor.
It was Martin who insisted that we go to the two-speed route in order to achieve that outcome. And Martin did say, "Well, what about if we just do a single speed for the first year or so, and suffer that loss of performance?" I said, "I think it's really tough to crack that perception of poor performance, and say, 'yes, it's coming down the pike or something like that.' And unless you tell me this is a critical path item, you know, that this will fundamentally delay the schedule, then we should do the two-speed. If it's just a matter of investment, we should do it. If it's a matter of delaying a critical path, then we should think about it."
Here's an another example of how different people see the same events through their own personal filters. Musk acknowledges that he pushed for the 2-speed transmission from Job 1 in order to meet the original performance targets. From a December 2007 interview in Inc. magazine:
"The most controversial of Musk's edicts involved the transmission. Martin Eberhard, Tesla's co-founder and then-CEO, argued that it would be quicker and easier to build the car with a single-speed transmission. Musk ordered a two-speed model so that the Roadster would be able reach a top speed of well over 100 miles per hour."
None of this is disputed by either man and yet the motivations and interpretations are very different. Eberhard tells ABG that:
"JB was the guy in charge of the motor and inverter, and he did not have the skills or resources to redesign the ACP system. JB and I and the rest of the team discussed this approach and concluded that we (JB especially) could not achieve a higher-power motor/inverter in time.
JB was actually quite scared of deviating from the ACP design in the beginning because he did not understand it. Marc and I had to push him to eliminate the old analog control and switch to a DSP-based control, and even then, he did not do this until he had hired a substantial team of engineers, 2 years later. Again, this is not to denigrate JB. The ACP design was quite tricky and tweeked, and it was very poorly documented. JB was trying to make the right decision based on the available resources - as were we all.
Note that Tesla's new high-power motor depends on very new silicon from International Rectifier - silicon that only became available this year. No such silicon (IGBT transistors to be specific) was available to us in 2004, and we knew it."
As an engineer myself and having picked up responsibility for other peoples undocumented designs and tried to develop them I can certainly understand Straubel's reluctance to mess with something that at least worked up to a point. In this context the decision to follow the two-speed path with the lower torque motor certainly made sense at the time. Hindsight is of course 20/20 but as Eberhard explains, the electronic hardware to make the change apparently wasn't available at the time anyway.
Elon: And Martin said it would not delay the critical path. It was just a matter of investment. So I said, "OK, Let's do it." And that's true. If we had picked the right supplier for the transmission – the two-speed transmission is not inherently difficult. Unfortunately we did not pick a good supplier. The initial pick was XTrac, which is basically... they make transmissions for track cars.
They're very expensive and they don't really how to make consumer stuff. So they screwed the pooch. And then, instead of going from them, to someone who could really do the job, we went from them to Magna. Martin assured me and the rest of the Board that there was no one better to do this transmission than Magna.
Here Musk's opinion of XTrac may be overly harsh. Major automotive suppliers are often reluctant to deal with low-volume manufacturers. The engineering cost of developing a component for 100 cars a year is the same as 1,000,000. Variable costs go down with volume but up front development cost doesn't. That means that unless a low volume automaker is willing to pay those costs either in separate engineering expenses or exorbitantly high piece costs (that factor in engineering costs) they are usually out of luck.
"Regarding the choice of transmission suppliers: It was not a matter of having picked the wrong supplier. The problem was that real transmission suppliers - those that made production volumes of transmissions - were simply not interested in selling transmissions to a company that buys so few of them. The first model year's production of Roadster transmissions is fewer than a prototype run for a real production transmission. This is why Lotus, for example, uses absolutely off-the-shelf Toyota transmissions in the Elise. But because a motor is so different from an engine, no off-the-shelf transmission was at all suitable for our needs.
I put Mac Powell in charge of selecting the transmission vendor, and he hired a team of ex-Lotus engineers to spec the transmission and to find a supplier. These were all people experienced with producing cars and experienced with sourcing components for a low-volume car manufacturer. They scoured the landscape and came up with only 2 suppliers who were willing to make transmissions for Tesla - they decided that Xtrac was the best of the two, and I respected their experienced opinions. Nobody was comfortable with Xtrac, but they were the best we could find."
As Tesla's VP Darryl Siry told us some time ago the issues with XTrac weren't even entirely of their own making. The design intent to manage the motor torque during clutchless shifts apparently proved unworkable necessitating a change to a different design and supplier. So again here Musk maybe overstated the case with regard to XTrac.
Elon: What he neglected to figure out was that Magna is a build-to-print manufacturer; they're not a designer. Particularly when it requires... this requires a first principle design. You know, you have to actually understand the theory because you can't simply derive this from some existing five-speed gasoline gearbox.
According to Siry, Magna was chosen by Eberhard without any competitive bidding. Due to the late decision to change suppliers, Magna was apparently chosen because an engineer at Tesla's Rochester Hills facility was familiar with the company and had a contact there.
ABG: So if you didn't have the fundamental expertise within Tesla to provide them (Magna Powertrain) the design that needed to be built, they didn't have the capacity just to do that on their own.
Elon: Right. And that's where we got into the bloody of Magna fiasco. We spent huge amounts of money in Magna and then in the, sort of, 13th hour, Magna brought in Ricardo, because basically the design failed and Ricardo told us that the entire thing had been incorrectly designed... that we'd have to basically do a complete refresh on the thing, soup to nuts, in order to have a reliable transmission.
So that came to a head in November of last year and we said, "Well, how long is it going to take to do that?" And Ricardo said, "Well, we went through the whole timeline. It would have been like November of this year before we'll actually have a working two-speed. And the bloody things would cost a huge amount of money and your cost would be crazy high. I'm like, "Wow," so the huge development cost with a nominal timeline of, like November 2008 and a high unit cost. That's pretty shitty.
So, on the flight back from Detroit, Ricardo US is based in Detroit. JB and I were talking and said, "Hey, look, why don't we look to going back to the single speed approach and upgrading the motor power, and the power electronics current capability?"
And so we iterated on that for a bit and decided to go that path. Essentially take the development burden on something that we knew versus something we didn't know. We understood motor and power electronics a lot, but we didn't really understand transmissions. So we decided to shift the burden of the problem to something we know.
So that's what we decided to do and that's just what we're doing. We have what we call Powertrain 1.5, is working right now. We've been driving around in one of the interim prototype cars. It delivers 3.9 seconds 0 to 60, 125-mile top speed. It's actually a superior product than what we originally promised people because the 0 to 100 time is significantly improved. There's no shift delay. There's no shifting. And it improves the range slightly too.
In a followup email exchange with Musk he elaborated on what probably should have happened in the wake of the problems with the XTrac gearbox:
"The problem with Magna Powertrain (USA) is that they are primarily a build-to-print maker of transmissions. They don't know how to design a transmission from scratch, particularly if it is outside the norm. We should have contracted with a company like Ricardo for design and then with Magna for manufacturing or gone to someone like Borg-Warner that can do both design and manufacturing.
What we did after the Magna fiasco was actually both. We are working with Ricardo on a joint Tesla/Ricardo single speed, which is already working in a car now, and Borg-Warner is contracted to do a transmission for us on a completely separate path. If both designs work well, we will pick the lower cost option."
"I had essentially no involvement in choosing Magna as the transmission design house. Martin told me and the rest of the board that they were the best in the business and we could not hope for a better supplier. I only stepped in when the Magna deal started falling apart in the late summer last year and they demanded huge sums of money from us, despite failing to deliver a working product."
The discussion continues in Part 3 coming tomorrow where we discusses the lessons that Musk and Tesla have learned about starting a car company and learn more about what to expect from WhiteStar.
Sign in to post
Please sign in to leave a comment.
Continue