Ferrari Power Unit Hardware & Software

[godlameroso]

afaik
extra harvest bypasses the 2 MJ limit K charging ERS but not the 4 MJ limit (K+H) charging ERS
it's value is increasing (by the extra) the H energy available for directly driving the K

You know, if you think about it, the extra harvest business...In a regular car that uses a belt driven supercharger, you sacrifice engine power to make boost. What you gain is higher than what you lose. These F1 cars have the capacity to drive the compressor exactly like a supercharger, except that the belt is electrical, and being electrical, can overcharge the compressor and store the excess in the battery. If the compressor takes 100kW, then the MGU-K can just brake the engine by that much and drive the compressor straight off the K, you have a potential 20 extra kW to harvest. You could charge the ES by the difference of the K energy and blowdown needed to keep the compressor running at optimum, and you'd have the advantage of low pumping losses due to having the wastegates wide open.

Here is where it gets tricky, if you have the wastegates wide open and are driving the compressor from a combination of electric belt supercharging and blow-down exhaust, will overall output be greater than if you could dedicate all 120kW of the K directly to the crank, while using the MGU-H as recovery to control high RPM overboost(surely increasing back pressure and pumping losses)? Do maps transition through these parameters through the rev range? Or is there an ideal happy medium?

[Zynerji]

afaik
extra harvest bypasses the 2 MJ limit K charging ERS but not the 4 MJ limit (K+H) charging ERS
it's value is increasing (by the extra) the H energy available for directly driving the K

You know, if you think about it, the extra harvest business...In a regular car that uses a belt driven supercharger, you sacrifice engine power to make boost. What you gain is higher than what you lose. These F1 cars have the capacity to drive the compressor exactly like a supercharger, except that the belt is electrical, and being electrical, can overcharge the compressor and store the excess in the battery. If the compressor takes 100kW, then the MGU-K can just brake the engine by that much and drive the compressor straight off the K, you have a potential 20 extra kW to harvest. You could charge the ES by the difference of the K energy and blowdown needed to keep the compressor running at optimum, and you'd have the advantage of low pumping losses due to having the wastegates wide open.

Here is where it gets tricky, if you have the wastegates wide open and are driving the compressor from a combination of electric belt supercharging and blow-down exhaust, will overall output be greater than if you could dedicate all 120kW of the K directly to the crank, while using the MGU-H as recovery to control high RPM overboost(surely increasing back pressure and pumping losses)? Do maps transition through these parameters through the rev range? Or is there an ideal happy medium?

I imagine these maps to be Tensor-ish. There are too many parameters (IMHO) to have set maps for every occurrence of each variable and their interactions.

https://www.tensorflow.org/

https://www.tensorflow.org/get_started/eager

I always figured they used this type of software to machine - learn the best operating parameters for any given load scenario.

[godlameroso]

afaik
extra harvest bypasses the 2 MJ limit K charging ERS but not the 4 MJ limit (K+H) charging ERS
it's value is increasing (by the extra) the H energy available for directly driving the K

You know, if you think about it, the extra harvest business...In a regular car that uses a belt driven supercharger, you sacrifice engine power to make boost. What you gain is higher than what you lose. These F1 cars have the capacity to drive the compressor exactly like a supercharger, except that the belt is electrical, and being electrical, can overcharge the compressor and store the excess in the battery. If the compressor takes 100kW, then the MGU-K can just brake the engine by that much and drive the compressor straight off the K, you have a potential 20 extra kW to harvest. You could charge the ES by the difference of the K energy and blowdown needed to keep the compressor running at optimum, and you'd have the advantage of low pumping losses due to having the wastegates wide open.

Here is where it gets tricky, if you have the wastegates wide open and are driving the compressor from a combination of electric belt supercharging and blow-down exhaust, will overall output be greater than if you could dedicate all 120kW of the K directly to the crank, while using the MGU-H as recovery to control high RPM overboost(surely increasing back pressure and pumping losses)? Do maps transition through these parameters through the rev range? Or is there an ideal happy medium?

I imagine these maps to be Tensor-ish. There are too many parameters (IMHO) to have set maps for every occurrence of each variable and their interactions.

You always have set maps, a map is nothing but fuel and ignition tables referenced to sensor parameters like O2 IAT MAP ECT EGT pressure traces in the cylinder, knock count, boost/rpm level, wastegate opening etc and so forth.

I bet these tunes wouldn't make sense to normal tuner guys, the way the fuel is delivered and the ignition timing set, their first thought seeing what they're using now is, "why hasn't it blown up yet".

[Zynerji]

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

Seems like an obvious thing...

[godlameroso]

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

Seems like an obvious thing...

Sure there is, Mercedes is on record saying they have such technologies. Just putting around and the machine works out the best deployment strategies based on their requirements. There's always some bloke poking around though, machine learning isn't perfect. Machine learning is dependent on data, and sometimes you just don't have that data until you get out there and send it.

[Zynerji]

I imagined the dyno giving the real time data feedback to a Tensor unit, that just keeps cycling through parameters to brute force the rainbow table of possible outcomes, then the ecu pulls the image from that table based upon rpm and torque demand.

[godlameroso]

I imagined the dyno giving the real time data feedback to a Tensor unit, that just keeps cycling through parameters to brute force the rainbow table of possible outcomes, then the ecu pulls the image from that table based upon rpm and torque demand.

How many engines do you have to blow up before the machine starts making good decisions?

[PhillipM]

Far too many.

[AJI]

How many is too many? I'm just curious as to how many PU's get munted on a dyno or VTT before they're happy enough to put them in a car that has to last 7 race meetings.

The PU numbers we do know are how many are allocated per season, which for Ferrari and Merc and Renault is 18, say ~24 to include a spare for each car, plus another 6 for testing, so, roughly 30.
In 2014 they produced way more than that for racing/testing, so we know they have they have the capacity, and that can have only increased over time.
I can't see a problem with the chief engine designer allocating 10 of what he/she considers to be the final iteration of the release spec PU to a ML department and telling them to have at it? If they blow them up then they blow them up, but if they find a better strategy along the way...
Ironically I can see Ferrari and Merc doing exactly this, and Renault, who should be doing it, not bothering.

[sosic2121]

Ferrari are essentially doing a "Reverse Extra-Harvest."

Instead of only bypassing the 4MJ charging limit allowed from MGUK, via the MGUH, like every manufacturer does, Ferrari are also going from battery to MGUH to MGUK using this method in reverse.

You heard it here first!

Yes, it seems that is possible and legal.
(Rules do say that
ES-MAX - ES-MIN < 4MJ)

I think this mode can only be useful during Q. I don't think that during the race, there is enough energy available for this kind of trickeey.

[PlatinumZealot]

Yes. . In my previous post i got the numbers mixed up but that is beside the point.

Take a look here at my explanation of "Extra deploy":

After being fleshed out in the honda thread we know "extra harvest" uses the mguk> mguh> battery energy path which bypasses the 2MJ charging limits. This mechanism is explained elsewhere in this forum so lets move on to the next point.

In the background, the rules originally intended only 4MJ to go from ERS>MGUK. The rules also intended for energy to go from TURBINE/MGUH> MGUK unlimited. Good so far?

Now... What Ferrari are doing is using the same "extra harvest" pathway in reverse simultaneously with the turbine. And going ERS/turbine > MGUH > MGUK. UNLIMITED. Which bypasses the 4MJ easily if that now higher battery demand is supported with extra harvest. This is called "extra deploy"

Note that "extra deploy" needs "extra harvest" in order to be sustainable in the race or else you end up with an empty battery.

remember the battery charge limits on the MGUK to ERS pathway do not apply here. This is a different pathway that is allowed to be unlimited so u can easily empty the battery here.

[aleks_ader]

https://uploads.tapatalk-cdn.com/201805 ... c5010./jpg

remember the battery charge limits on the MGUK to ERS pathway do not apply here. This is a different pathway that is allowed to be unlimited so u can easily empty the battery here.

You post is ok and very plausible. Heck i sad that is what all manufacturers do in Q modes or in starting modes.

But i cannot agree with last statemend. Battery store is limmited to max dif 4MJ SOC. So pathway in unlimmited but not total energy at your disposal.

[nzjrs]

I imagined the dyno giving the real time data feedback to a Tensor unit, that just keeps cycling through parameters to brute force the rainbow table of possible outcomes, then the ecu pulls the image from that table based upon rpm and torque demand.

That's barely machine learning.

[trinidefender]

https://uploads.tapatalk-cdn.com/201805 ... ec5010.jpg

Yes. . In my previous post i got the numbers mixed up but that is beside the point.

Take a look here at my explanation of "Extra deploy":

After being fleshed out in the honda thread we know "extra harvest" uses the mguk> mguh> battery energy path which bypasses the 2MJ charging limits. This mechanism is explained elsewhere in this forum so lets move on to the next point.

In the background, the rules originally intended only 4MJ to go from ERS>MGUK. The rules also intended for energy to go from TURBINE/MGUH> MGUK unlimited. Good so far?

Now... What Ferrari are doing is using the same "extra harvest" pathway in reverse simultaneously with the turbine. And going ERS/turbine > MGUH > MGUK. UNLIMITED. Which bypasses the 4MJ easily if that now higher battery demand is supported with extra harvest. This is called "extra deploy"

Note that "extra deploy" needs "extra harvest" in order to be sustainable in the race or else you end up with an empty battery.

remember the battery charge limits on the MGUK to ERS pathway do not apply here. This is a different pathway that is allowed to be unlimited so u can easily empty the battery here.

I'm pretty sure Honda already stated that they already do this when they have the energy to spare. Both an extra harvest and an extra deploy tactic with the MGU-H acting as the middle man between the ES and the MGU-K.

Pretty sure that all the PU manufacturers do this.

[sosic2121]

https://uploads.tapatalk-cdn.com/201805 ... ec5010.jpg

Yes. . In my previous post i got the numbers mixed up but that is beside the point.

Take a look here at my explanation of "Extra deploy":

After being fleshed out in the honda thread we know "extra harvest" uses the mguk> mguh> battery energy path which bypasses the 2MJ charging limits. This mechanism is explained elsewhere in this forum so lets move on to the next point.

In the background, the rules originally intended only 4MJ to go from ERS>MGUK. The rules also intended for energy to go from TURBINE/MGUH> MGUK unlimited. Good so far?

Now... What Ferrari are doing is using the same "extra harvest" pathway in reverse simultaneously with the turbine. And going ERS/turbine > MGUH > MGUK. UNLIMITED. Which bypasses the 4MJ easily if that now higher battery demand is supported with extra harvest. This is called "extra deploy"

Note that "extra deploy" needs "extra harvest" in order to be sustainable in the race or else you end up with an empty battery.

remember the battery charge limits on the MGUK to ERS pathway do not apply here. This is a different pathway that is allowed to be unlimited so u can easily empty the battery here.

In qualifying normally they can use 4MJ from battery + what ever turbine>H can produce.
So, what ever they can regenerate during the lap, they can also deploy.
They regenerate up to 1.5MJ during braking, surly they use motor generating ICE>K to MJ limit, and from there on extra harvest, or hot blowing turbine(what ever is more efficient)

Since when teams are doing this? Was this a reason for clipping at the end of the straights?

During the race, is there enough energy going around for this?