Ferrari Power Unit Hardware & Software

[wuzak]

@wuzak, here’s a site with Baku 2017. https://www.formula1.it/news/5175/1/gp- ... l-circuito

It says 19% of the lap spent braking. Around 20 seconds, which is above the average.

Magneti Marelli used to publish energy recovery data in a similar form, perhaps sanitised Ferrari data. From that it was possible to deduce a braking energy recovery efficiency, around 50%. That makes Baku recovery 1.2 MJ.

Brembo used to publish the circuit identity sheets to journalists. http://www.brembo.com/en/company/press/ ... -formula-1 they were originally more detailed.

Didn’t you post some info on recovery efficiency in the early days of the fotmula?

Oh and @sonic boom and @Tommy Cookers had a conversation about available brake energy in this thread here viewtopic.php?f=4&t=21958&p=762082&hilit=1000#p762082

Edit: I didn’t see the subsequent posts. But I think the links are still useful.

Yes, I did do an analysis a few years ago. Found the braking recovery possible at all the circuits (at that time). Based on a spreadsheet someone had done from Brembo data.

I was just pointing out that PZ's calculation for energy recovery was off, and used the Brembo file he linked to do some quick calcs.

[djones]

Have the FIA (or anybody) said this will actually be investigated, or is it just a theory going around at the moment?

[Muniix]

So, there is no machine learning being used to determine those maps?

Seems like an obvious thing...

They just don't have the compute onboard they're only using ancient RISC CPU's, 22 terraflop is now available in 20 Watts designed for automotive use.

[Phil]

So, there is no machine learning being used to determine those maps?

Seems like an obvious thing...

They just don't have the compute onboard they're only using ancient RISC CPU's, 22 terraflop is now available in 20 Watts designed for automotive use.

I'm very doubtful you need 22 terraflops for that...?

[Muniix]

So, there is no machine learning being used to determine those maps?

Seems like an obvious thing...

They just don't have the compute onboard they're only using ancient RISC CPU's, 22 terraflop is now available in 20 Watts designed for automotive use.

I'm very doubtful you need 22 terraflops for that...

But a lot more than dual 600 mhz powerPC cell RISC cores in the regulation ECU. You need dozens to hundreds of cores to do this, GPU shader cores and TPU Tensor PUs. I'm doing optimisation based on cylinder pressure sensor data, with AI ML/DL in JuliaLang, the language used to locate the binary neutron stars in 11 mins after LIGO detection of the gravity wave using 625,000 cores/1.25 million threads.

[Muniix]

Not that it's possible but if teams could store 10MJ of energy in the battery they would.

No. Battery storage is limited to 4MJ, defined as the difference between the maximum state-of-charge and minimum state-of-charge during a lap.

There is no limit, however, to how much energy you put to the ES or take from it over a lap, except for the case of the direct path to and from the MGUK. However, the MGUH can send and take any amount of energy to and from the ES.

Battery temperature limited unless of course you want to release it's energy all in one go!!! It has a never above temperature for obvious reasons.

[PlatinumZealot]

@wuzak, here’s a site with Baku 2017. https://www.formula1.it/news/5175/1/gp- ... l-circuito

It says 19% of the lap spent braking. Around 20 seconds, which is above the average.

Magneti Marelli used to publish energy recovery data in a similar form, perhaps sanitised Ferrari data. From that it was possible to deduce a braking energy recovery efficiency, around 50%. That makes Baku recovery 1.2 MJ.

Brembo used to publish the circuit identity sheets to journalists. http://www.brembo.com/en/company/press/ ... -formula-1 they were originally more detailed.

Didn’t you post some info on recovery efficiency in the early days of the fotmula?

Oh and @sonic boom and @Tommy Cookers had a conversation about available brake energy in this thread here viewtopic.php?f=4&t=21958&p=762082&hilit=1000#p762082

Edit: I didn’t see the subsequent posts. But I think the links are still useful.

Yes, I did do an analysis a few years ago. Found the braking recovery possible at all the circuits (at that time). Based on a spreadsheet someone had done from Brembo data.

I was just pointing out that PZ's calculation for energy recovery was off, and used the Brembo file he linked to do some quick calcs.

I used the total braking energy over the race - which they quoted as 181 kWhrs. Then I divided that by the number of laps to get an average. Then apportioned a percent to the rear brakes and a portion of that to MGUK.

The power at the MGUK shaft is not limited to 120kW. Only the torque is limited to 200Nm and the angular speed to 50,000 rpms. The 120kW limit is power along the direct path to the ERS. So this does not limit extra harvest power from the MGUK.

[henry]

@wuzak, here’s a site with Baku 2017. https://www.formula1.it/news/5175/1/gp- ... l-circuito

It says 19% of the lap spent braking. Around 20 seconds, which is above the average.

Magneti Marelli used to publish energy recovery data in a similar form, perhaps sanitised Ferrari data. From that it was possible to deduce a braking energy recovery efficiency, around 50%. That makes Baku recovery 1.2 MJ.

Brembo used to publish the circuit identity sheets to journalists. http://www.brembo.com/en/company/press/ ... -formula-1 they were originally more detailed.

Didn’t you post some info on recovery efficiency in the early days of the fotmula?

Oh and @sonic boom and @Tommy Cookers had a conversation about available brake energy in this thread here viewtopic.php?f=4&t=21958&p=762082&hilit=1000#p762082

Edit: I didn’t see the subsequent posts. But I think the links are still useful.

Yes, I did do an analysis a few years ago. Found the braking recovery possible at all the circuits (at that time). Based on a spreadsheet someone had done from Brembo data.

I was just pointing out that PZ's calculation for energy recovery was off, and used the Brembo file he linked to do some quick calcs.

I used the total braking energy over the race - which they quoted as 181 kWhrs. Then I divided that by the number of laps to get an average. Then apportioned a percent to the rear brakes and a portion of that to MGUK.

The power at the MGUK shaft is not limited to 120kW. Only the torque is limited to 200Nm and the angular speed to 50,000 rpms. The 120kW limit is power along the direct path to the ERS. So this does not limit extra harvest power from the MGUK.

At the beginning of a high speed stop the braking power needed at the rear axle is approx 1000kw. (@Tommy Cookers provided this figure with which I agree) the MGU-K is limited to 120kw so at the beginning of the stop they can harvest 12% of the available energy at best. So a calculation based on the full available braking energy is bound to be a considerable overestimate.

The MGU-K is definitely limited to 120 kw, from 5800 rpm up, no matter what the source. It’s on the power flow diagram. It is verified by measuring the electrical energy into and out of the K. The 200 nm is measured at the crank, not the K shaft. At 5800 rpm 200 nm is 120 kw, at 10500 rpm the allowable torque has reduced to 110 Nm.

So extra harvest power is limited to 120 kw but extra harvest energy is unregulated.

For interest if the MGU-K were to produce 200 Nm at 50000 rpm that’s around a megawatt. The actual motor torque will be in the order of 25 Nm at max revs.

[kalinka]

They just don't have the compute onboard they're only using ancient RISC CPU's, 22 terraflop is now available in 20 Watts designed for automotive use.

I'm very doubtful you need 22 terraflops for that...

But a lot more than dual 600 mhz powerPC cell RISC cores in the regulation ECU. You need dozens to hundreds of cores to do this, GPU shader cores and TPU Tensor PUs. I'm doing optimisation based on cylinder pressure sensor data, with AI ML/DL in JuliaLang, the language used to locate the binary neutron stars in 11 mins after LIGO detection of the gravity wave using 625,000 cores/1.25 million threads.

I think he meant that the AI part of the computation is done back in factory from dyno data, and the ECU is only using the AI-generated maps to select the closest matching one for a given engine mode.

[Muniix]

The wording on the SOC restriction is “any time the car is on track” and not per lap. So they might cycle between 5 and 1 but I doubt they cycle between 10 and 6.

That’s my reading of things anyway.

The 'any time the car is on track' is the confusing part. The implication is that once the car is on track the SoC can't vary by more than 4MJ, ever, but they always refer to being able to use the K 120kW for 33.333 seconds per lap at maximum output, so, 4MJ per lap...
I know this is a very simplistic way to look at it and they are doing all sorts of tricky things to circumvent this, but I believe it was the spirit of the rule.

The reality is we don't know the capacity of the battery as it is only limited by weight, but 10MJ is totally realistic for the weight, particularly for F1 tech levels.
They may cycle between the numbers you suggest and certainly there are comments of de-rates from the drivers, so they clearly hit the minimum threshold occasionally, but is it the SoC threshold or is the battery simply out of juice?

"de-rate" could be cells low on thermal head room so less power available.

How often do they replace ERS or Li-ion cells? Degradation from thermal stress can be high but cells also need to be heated when operated at high power levels.
Is there any super capacitors in the ERS?
graphene cells produced here in Oz at near 50 w hr/l and exceeding 10kw/litre though this may drop to 10 to achieve the goal of 60 in energy. They weigh next to nothing and near 100% efficient.

[AJI]

The wording on the SOC restriction is “any time the car is on track” and not per lap. So they might cycle between 5 and 1 but I doubt they cycle between 10 and 6.

That’s my reading of things anyway.

The 'any time the car is on track' is the confusing part. The implication is that once the car is on track the SoC can't vary by more than 4MJ, ever, but they always refer to being able to use the K 120kW for 33.333 seconds per lap at maximum output, so, 4MJ per lap...
I know this is a very simplistic way to look at it and they are doing all sorts of tricky things to circumvent this, but I believe it was the spirit of the rule.

The reality is we don't know the capacity of the battery as it is only limited by weight, but 10MJ is totally realistic for the weight, particularly for F1 tech levels.
They may cycle between the numbers you suggest and certainly there are comments of de-rates from the drivers, so they clearly hit the minimum threshold occasionally, but is it the SoC threshold or is the battery simply out of juice?

"de-rate" could be cells low on thermal head room so less power available.

How often do they replace ERS or Li-ion cells? Degradation from thermal stress can be high but cells also need to be heated when operated at high power levels.
Is there any super capacitors in the ERS?
graphene cells produced here in Oz at near 50 w hr/l and exceeding 10kw/litre though this may drop to 10 to achieve the goal of 60 in energy. They weigh next to nothing and near 100% efficient.

Interesting points.
There are only 2 ES allocations for this year without penalty. They must be robust.
When I hear 'de-rate' from a driver I assume he has not managed the ES levels properly, i.e. in the heat of the battle the driver has used the ES 4MJ prematurely...
We don't know what the battery tech is, but it's pretty safe to assume that it's state-of-the-art. Graphene, supercapacitors, the sky's the limit.
The metaphorical 'burger with lot'. It's F1...
The next logical step is NASA. They might have a million times the resources, but they don't have the PR value.

[godlameroso]

If they're using carbon they're still limited to a theoretical maximum 420wh/kg which says nothing of charge/discharge rates, or thermal requirements. I'd be surprised if they're pushing more than 300wh/kg, giving them benefit of the doubt, the cell capacity at 25kg would be ~7kw/h or ~25MJ of energy. Even if it's 200wh/kg it's still 5kw/h or 18MJ.

If the total battery capacity is only say 12MJ you have huge head way in regards to charge/discharge rates and thermal requirements. The lower the capacity the higher the head way in other important areas. Battery performance isn't just about holding the highest charge, it's charging and discharging rate as well as how well it handles heat(internal resistance), as well as efficiency.

[63l8qrrfy6]

Keep in mind that battery chemistry is a compromise between energy density and power density. There is little point in exceeding the 4MJ/20 kg if it limits the ability for full power simultaneous K and H usage.

Also battery internal resistance and its efficiency are pretty much the same thing. There are secondary mechanisms of heat loss (enthalpy changes and electrochemical polarization) but they are insignificant compared to RI^2 losses.

[Muniix]

Keep in mind that battery chemistry is a compromise between energy density and power density. There is little point in exceeding the 4MJ/20 kg if it limits the ability for full power simultaneous K and H usage.

Also battery internal resistance and its efficiency are pretty much the same thing. There are secondary mechanisms of heat loss (enthalpy changes and electrochemical polarization) but they are insignificant compared to RI^2 losses.

The current/voltage maps are effected by state of charge, below 30% SOC the voltage drops off rapidly with load. Charging above 80% SOC at high current will bring thermal limits (from memory) the maximum charge temperature, the never above temperature is less than in discharge.

High power density cells use thicker current carriers and this reduces their energy density typically the highest power density cells will only have 140 watt hr/kg down from the highs of 260. While there are some advanced anode and cathode materials such as Nano Nouvelle's Lumafoil that handles higher current with less heat, an Australian company but the battery companies are not really interested, the not invented here issues and they're now concentrating on the medical industry for their polymer materials.

Graphene super capacitors are todate very low in power density measured in watt hours per litre, with the best available in the few watt hours a litre. Just being produced in limited pre-production achieving very near 50 watt hours/l with the target of 60 and a 10kw power rating, currently well exceeding this power but likely to drop to reach the goal of 60.

Their are yet more issues but you get the point. It's 3am here and my brain is semi-shut down.

Managing chemical energy storage is just as difficult as operating an ICE to meet EU6 emissions and where AI is really useful/necesory to obtain the ideal outcome which is what it's good for. For instance "How much degradation can one afford". And degradation is effected by heat, LIB's, Li-Ion cells i'm talking about have an ideal operating temperature depending on load, to drive them hard you want to follow the ideal heat profile, but if you want lifetime with less degradation you want to keep them cool around 25 Celsius. Hence Tesla's in Denmark where the mean temperture is 5 Celsius you have a good 20 C delta from the ideal for effective cooling and they're getting up to 400,000 km with not much degradation. If your only power source is from batteries and you need power to manage their temperature this is real parasitic power overhead a hybrid can with clever AI manage this a lot better.

Even if the ICE has no variable valves etc.

[henry]

Keep in mind that battery chemistry is a compromise between energy density and power density. There is little point in exceeding the 4MJ/20 kg if it limits the ability for full power simultaneous K and H usage.

Also battery internal resistance and its efficiency are pretty much the same thing. There are secondary mechanisms of heat loss (enthalpy changes and electrochemical polarization) but they are insignificant compared to RI^2 losses.

The current voltage maps are effected by state of charge, below 30% SOC the voltage drops off rapidly with load. Charging above 80% quickly has thermal limits the maximum charge temperature is less than in discharge.

High power density cells use thicker current carriers and this reduces their energy density typically from the highs of 260 down to 140. There are some advanced anode and cathode materials, Lumafoil that handles higher current with less heat from research company Nano Nouvelle but the battery companies are not really interrested and their now concentrating on the medical industry.

Their are yet more issues but you get the point.

Managing chemical energy storage is just as difficult as operating an ICE to meet EU6 emissions and where AI is really needed to obtain the ideal outcome. How much degradation can one afford?

Your numbers would suggest using an 8 MJ ES operating between 2.4 and 6.4 MJ would be a minimum requirement. Other posters have suggested this is easily attainable within the current regulations.