The beauty of the Tesla Supercharger is that it uses the exact same chargers that are in the car, the only difference is where they are installed. So they will use the same cooling mechanism.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
How does a Supercharger work?
- Thread starter deckofficer
- Start date
-
- Tags
- Technical
The beauty of the Tesla Supercharger is that it uses the exact same chargers that are in the car, the only difference is where they are installed. So they will use the same cooling mechanism.
As the OP, I sure appreciate all the input to help me understand the Supercharger workings. I have another question. At a 120 kW charge on a very low battery bank, I'll assume the voltage to be somewhere around 300 volts, which would mean the amperage would be 400. With the amount of contact surface area of both the male and female charge connections, how is the heat generated at the connection handled? Does the charge plug have any kind of flowing median that removes the heat from the connection?
nwdiver
Well-Known Member
Not liquid cooled... yet but I've heard rumors that's coming... there are rumors about a lot of things though...
One thing is for sure... supercharging generates A LOT of heat...
One thing is for sure... supercharging generates A LOT of heat...
i have heard Elon say (I think it might have been at the original Teslive get-together in response to a question about adaptors) that his biggest problem with CHAdeMO is that in that protocol, the charger wants to be in control, but Teslas know what they want. I take this to mean that the car is in charge (haha).
Inside a SpC cabinet:
I forget who first took this, someone on TMC...
Of note:
- 12 independently connected chargers that look very similar to Model S charger
- Two large control boards, possibly one for "A" and one for "B" (group of similar components on each board may be I/O for each charger, CANBUS or similar)
- Two boards seem to be chained somehow so probably one "Master" and one "Slave"
- Can't see any switchgear for each channel, but if there is to be any, it is likely to be a single contactor capable of joining channels A and B together (each channel will also need an isolation contactor - this is in case someone picks up an unused connector "B" whilst "A" is charging)
- May be arranged in banks of three going by four wire looms - complete guess but this might indicate a channel A1,A2,B1,B2 and a car can use any granularity of these to get power in 30kW blocks?
- Each group of three will be tied off a single phase to phase connection, to get ~277V to the charger module.
- - - Updated - - -
It is actually extremely compact for a 10kW charging module.
It has to:
- Rectify incoming mains power (up to 40A) which requires beefy diodes
- Filter out a ton of switching noise from PFC and converter stage
- Boost incoming mains to ~400VDC for PFC stage
- Switch this 400VDC through a transformer to the battery voltage - the battery voltage can exceed the PFC bus voltage, so it needs to boost OR buck, depending on the current battery SoC
- Do all this whilst not setting the battery on fire or electocuting anyone (control PCB's job)
Magnetics wise: there's a transformer (isolation), inductor (for PFC boost) and line filter. All are heatsunk and water cooled. There's also an output inductor which surprisingly is just passively cooled. I'm guessing it's got low core losses.
I have to admit, I was surprised to see a separate PFC stage with the associated electrolytics (lifespan concerns come into play with electrolytic capacitors.) PFC is necessary to extract maximum line current, and allows the charger to achieve PF>0.99 (with 1 being ideal.) If I were to approach this problem, I would have considered modulating the battery charging current with the line current to keep unity power factor. Would save an entire PF stage improving the efficiency and cost. I am guessing Tesla did not do this for several reasons. Primarily, it would create a ripple current on the battery pack voltage, which might impact the a/c cooling/heating (causing the pump to beat with mains frequency) and perhaps affect sensitive systems like audio and lighting. Also, perhaps it would reduce the lifespan of the battery, not sure. And there is the possibility it would not play well with three-phase multi-charger stacks, though I'm not certain.
I have seen much bigger units only deliver 2-3kW. Typical power density feasibility limits are 10~30W/in^3...
I forget who first took this, someone on TMC...
Of note:
- 12 independently connected chargers that look very similar to Model S charger
- Two large control boards, possibly one for "A" and one for "B" (group of similar components on each board may be I/O for each charger, CANBUS or similar)
- Two boards seem to be chained somehow so probably one "Master" and one "Slave"
- Can't see any switchgear for each channel, but if there is to be any, it is likely to be a single contactor capable of joining channels A and B together (each channel will also need an isolation contactor - this is in case someone picks up an unused connector "B" whilst "A" is charging)
- May be arranged in banks of three going by four wire looms - complete guess but this might indicate a channel A1,A2,B1,B2 and a car can use any granularity of these to get power in 30kW blocks?
- Each group of three will be tied off a single phase to phase connection, to get ~277V to the charger module.
- - - Updated - - -
It is actually extremely compact for a 10kW charging module.
It has to:
- Rectify incoming mains power (up to 40A) which requires beefy diodes
- Filter out a ton of switching noise from PFC and converter stage
- Boost incoming mains to ~400VDC for PFC stage
- Switch this 400VDC through a transformer to the battery voltage - the battery voltage can exceed the PFC bus voltage, so it needs to boost OR buck, depending on the current battery SoC
- Do all this whilst not setting the battery on fire or electocuting anyone (control PCB's job)
Magnetics wise: there's a transformer (isolation), inductor (for PFC boost) and line filter. All are heatsunk and water cooled. There's also an output inductor which surprisingly is just passively cooled. I'm guessing it's got low core losses.
I have to admit, I was surprised to see a separate PFC stage with the associated electrolytics (lifespan concerns come into play with electrolytic capacitors.) PFC is necessary to extract maximum line current, and allows the charger to achieve PF>0.99 (with 1 being ideal.) If I were to approach this problem, I would have considered modulating the battery charging current with the line current to keep unity power factor. Would save an entire PF stage improving the efficiency and cost. I am guessing Tesla did not do this for several reasons. Primarily, it would create a ripple current on the battery pack voltage, which might impact the a/c cooling/heating (causing the pump to beat with mains frequency) and perhaps affect sensitive systems like audio and lighting. Also, perhaps it would reduce the lifespan of the battery, not sure. And there is the possibility it would not play well with three-phase multi-charger stacks, though I'm not certain.
I have seen much bigger units only deliver 2-3kW. Typical power density feasibility limits are 10~30W/in^3...
Last edited:
nwdiver
Well-Known Member
The 'business end' of power electronics is usually small... A grid-tie inverter looks big until you open it up and realize most of the size is really heat sink. A 10kW SMA inverter is 1 card measuring ~1 sq ft... Making these liquid cooled makes them MUCH smaller.
Inside a SpC cabinet:
I forget who first took this, someone on TMC...
Of note:
- 12 independently connected chargers that look very similar to Model S charger
- Two large control boards, possibly one for "A" and one for "B" (group of similar components on each board may be I/O for each charger, CANBUS or similar)
- Two boards seem to be chained somehow so probably one "Master" and one "Slave"
- Can't see any switchgear for each channel, but if there is to be any, it is likely to be a single contactor capable of joining channels A and B together (each channel will also need an isolation contactor - this is in case someone picks up an unused connector "B" whilst "A" is charging)
- May be arranged in banks of three going by four wire looms - complete guess but this might indicate a channel A1,A2,B1,B2 and a car can use any granularity of these to get power in 30kW blocks?
- Each group of three will be tied off a single phase to phase connection, to get ~277V to the charger module.
- - - Updated - - -
It is actually extremely compact for a 10kW charging module.
It has to:
- Rectify incoming mains power (up to 40A) which requires beefy diodes
- Filter out a ton of switching noise from PFC and converter stage
- Boost incoming mains to ~400VDC for PFC stage
- Switch this 400VDC through a transformer to the battery voltage - the battery voltage can exceed the PFC bus voltage, so it needs to boost OR buck, depending on the current battery SoC
- Do all this whilst not setting the battery on fire or electocuting anyone (control PCB's job)
Magnetics wise: there's a transformer (isolation), inductor (for PFC boost) and line filter. All are heatsunk and water cooled. There's also an output inductor which surprisingly is just passively cooled. I'm guessing it's got low core losses.
I have to admit, I was surprised to see a separate PFC stage with the associated electrolytics (lifespan concerns come into play with electrolytic capacitors.) PFC is necessary to extract maximum line current, and allows the charger to achieve PF>0.99 (with 1 being ideal.) If I were to approach this problem, I would have considered modulating the battery charging current with the line current to keep unity power factor. Would save an entire PF stage improving the efficiency and cost. I am guessing Tesla did not do this for several reasons. Primarily, it would create a ripple current on the battery pack voltage, which might impact the a/c cooling/heating (causing the pump to beat with mains frequency) and perhaps affect sensitive systems like audio and lighting. Also, perhaps it would reduce the lifespan of the battery, not sure. And there is the possibility it would not play well with three-phase multi-charger stacks, though I'm not certain.
I have seen much bigger units only deliver 2-3kW. Typical power density feasibility limits are 10~30W/in^3...
Good info and you know your stuff. I was wondering about caps used in a attempt to get a good power factor.
It certainly helps but the largest items by volume in the Tesla charger are probably magnetics rather than the electronics. The transformers, line filter(s) and PFC inductor are very large. The magetics are also liquid cooled and we know from diagnostic screens that they measure the temperature at the magnetics separately from the rest of the system.
https://technology.ihs.com/Teardowns/binary/devices/2611/photos/96475/binary?size=
https://technology.ihs.com/Teardowns/binary/devices/2611/photos/96472/binary?size=
https://technology.ihs.com/Teardowns/binary/devices/2611/photos/96469/binary?size=
Each of those large metal boxes = a magnetics component.
Had not seen that they have two large transformer-like devices. I wonder if they're in a parallel configuration or one is a resonant choke for an LLC/ZVS converter, or something similar.
Grid tie inverters have the advantage that they don't usually need to be isolated, as the solar PV system is considered to be a hazardous and isolated system. This means they can ditch the output transformers.
It's also interesting that Tesla do not use hybrid modules but instead discrete IGBT/diodes (just like in the Roadster & Model S power inverters.) This probably comes down to a low-ish volume thing (the Prius uses custom Toyota hybrid modules), as it's a more costly design choice.
The old 90/120 kW cabinets were 12 of the US-spec chargers, assigned 4 per phase @ up to 40A. The new 135 kW cabinets are 12 of the EU-spec charger modules, each module charging at 16A per phase across all three phases.
The Supercharger unit does not use the L-L voltage (400VAC EU, 480VAC US) but rather the L-N voltage (230V EU, 277V US). When used with 230V, the SC is limited in capacity and in many EU sites you'll find a 20% boost transformer (400->480VAC) providing the extra voltage to power the modules.
Last edited:
No, they are wired phase-to-neutral for 277V. Phase-to-phase would be 480V
It's more like 350V at the low end. Highest reported current is ~333A, so that seems to be the limiting factor.
- - - Updated - - -
It's less about off-peak vs on-peak rates, but rather reducing demand charges. Commercial users are charged for both total kWh consumed, and peak kW drawn (demand charges). The idea is that your connection to the grid needs to be sized by your peak draw. Short, bursts of 135kW still require the same infrastructure as if you pulled 135kW 24x7. Demand charges are intended to cover the cost of the connection.
Putting a battery between the superchargers and the utility shaves of the peaks, reducing demand charges significantly.
Good point about 3ph. I always get them mixed up. I wonder how well balanced it is between phases and if the modules communicate with each other (slight efficiency and component differences will lead to a phase current imbalance.) Might not be significant, but the greater the imbalance, the bigger the transformer needs to be.
For the most part, the balancing is fairly easy to do; even if they used granularity of 1A, that means balanced steps could be taken 831W at a time. The only out-of-balance conditions they'd run into would be a failure of modules (in the original cabinets). In the new cabinets, they could keep things balanced by disabling entire modules if one sub-charger experiences a problem.
That sounds like a good reason for the newer cabinets to use European charger modules - each of which has three 16 amp phases instead of the one 40 amp phase of the original US units. Then one failure will only knock one complete module offline.
Walter
One thing that we've learned from planning applications is the supercharging connector has a temperature sensor in it. This could be used to reduce power if the connector begins to get too hot. I don't think the UMC has this although they would require it more around the 14-50 connector than anything else.
I don't think that it's necessary to lose 30 kW of capacity for a single module failure (this applies to the old cabinets only, where one entire 11 kW module is tied to a single phase). You can operate in an unbalanced mode for a short period of time, unless they went really cheap on the transformer and wiring spec. I haven't seen their specs for transformers and neutral size, but with the correct transformer spec they could lose an entire phase and it would just shift the current to the neutral.
In the new cabinets, each module is wired to all three phases (each sub-module handles a phase up to 16A), so if they insisted upon power balance, you'd lose 13 kW of input power (277 * 16 * 3).
I don't see how they can use the EU spec modules, they are rated up to the same 277v max voltage as the US spec, they just add an additional leg. If you take 480v 3-phase wye service, you get 3 separate legs of 277v to neutral, but leg-to-leg is still 480v, so you could not connect that to the 3 inputs on the EU module. I suspect the 135kW upgrade is simply allowing the US spec modules to operate "harder" than the 10kW limit when used in the Model S.
If we assume that the efficiency is around 95%, then it would mean that you'd need an input of about 11.8kW for each of the 12 modules to produce a 135kW output. That would mean about 43A input per module at 277v. Not too big a stretch of "overclocking" if they are well cooled. Here's one of the fuses used on the AC input side, it's rated to 50A, so 43A should be fine:
If we assume that the efficiency is around 95%, then it would mean that you'd need an input of about 11.8kW for each of the 12 modules to produce a 135kW output. That would mean about 43A input per module at 277v. Not too big a stretch of "overclocking" if they are well cooled. Here's one of the fuses used on the AC input side, it's rated to 50A, so 43A should be fine:
In the supercharger, they don't use the L-L voltage at all. They use the L-N voltage - neutral is *required* if 400+ VAC service, it's only not required if you use < 277V L-L (some early EU superchargers were connected L-L). Inside the charging module, each sub-charger is connected to the shared neutral and one of the three phase legs. Power is balanced across the legs when the sub-chargers operate at the same power.
See Europe: Future Charging for Model S 1-phase or 3-phase? (Part 2) - Page 47 for more information. They do not use the old gen1 single-phase chargers in an "overclocked" configuration.
I was mainly talking about using the EU-spec modules in the US cabinets which is what I thought you said was going on in Gen2 SC's. Obviously it's different in the EU where transformers are wired wye, putting out 230v L to N and 400v L to L. My understanding is they use boost transformers to take the 230 to 277 for the SC, so I would imagine they are configured the same way, with a N and L to each module, and modules in groups of 3 to keep things balanced.
Regardless, I was trying to understand how a supposed 135kW SC can only put out 120kW per car. Today I found a diagram that shows the internal configuration:
If the SC cabinet is really wired this way, that means each A/B output can only be up to 9 modules, never the full 12. This would explain why on the Gen1 cabinets they were only capable of 90kW per car. There is no way to safely run all 12 modules for one output channel, otherwise the unused plug would be energized and dangerous.
Assuming the Gen2 cabinets are configured similarly, that would mean if each cabinet is rated at 135kW total, then the max output per car would only be 101.25kW. Aren't some clocked as putting 120kW out per car?
