Red Bull exceed fuel flow limit, Ricciardo DSQ at Australian GP

[langwadt]

You can't have 6 injections per rotation. This is a 4 stroke engine.

+ 1 on this, a cylinder only fires once every 4 rotations of the crank.

2 rotations; down,up, down,up => suck, squeeze, bang, blow

[thomin]

You can't have 6 injections per rotation. This is a 4 stroke engine.

Yeah, you're right, silly me...that was actually why I went for 36 rotations per 0.2 seconds, so that I can divide it by 4...but I somehow got lost in the numbers...

That would go on to increase the impact of an offset though...Instead of 36 injection events you'd have 9...or taking half rotations into account, instead of 72, you'd have 18.

That would translate to 105.555 kg/h in a cycle where you measure 19 instead of 18 injection events.

So on the extreme end, a team would have to reduce its internal fuel flow to 94.7368 kg/h in order to be on the safe side, which would be in line with what we heard form the various teams.

I think I got it still wrong as beelsebob had it right, that we have to disentangle all 6 cylinders. Depending on the offset, that could reduce the numbers again. I might take a moment later on and do some real math instead of embarrassing myself further with these off-the-cuff calculations. Also, I think some graphs might be helpful. I'll see what I can do.

But I still think that this phenomenon could be at the root of all this. In-sync readings may very well differ quite a bit from random, out-of-sync readings.

[beelsebob]

I think I got it still wrong as beelsebob had it right, that we have to disentangle all 6 cylinders. Depending on the offset, that could reduce the numbers again. I might take a moment later on and do some real math instead of embarrassing myself further with these off-the-cuff calculations. Also, I think some graphs might be helpful. I'll see what I can do.

But I still think that this phenomenon could be at the root of all this. In-sync readings may very well differ quite a bit from random, out-of-sync readings.

Just to be clear – I didn't mean to seem negative on this. I think you've hit upon a really great explanation for why a fuel flow meter might see very different values from what you might see at the injector level.

This actually makes it rather important that the FIA *must* specify a measuring time, because without it, you could end up with extremely high readings. If for example you sample for half the length of time that the injector opens, every so often, you'll get one extremely high reading, due to the injector being open, and then a bunch of near 0 readings while none is open.

[langwadt]

I think I got it still wrong as beelsebob had it right, that we have to disentangle all 6 cylinders. Depending on the offset, that could reduce the numbers again. I might take a moment later on and do some real math instead of embarrassing myself further with these off-the-cuff calculations. Also, I think some graphs might be helpful. I'll see what I can do.

But I still think that this phenomenon could be at the root of all this. In-sync readings may very well differ quite a bit from random, out-of-sync readings.

Just to be clear – I didn't mean to seem negative on this. I think you've hit upon a really great explanation for why a fuel flow meter might see very different values from what you might see at the injector level.

This actually makes it rather important that the FIA *must* specify a measuring time, because without it, you could end up with extremely high readings. If for example you sample for half the length of time that the injector opens, every so often, you'll get one extremely high reading, due to the injector being open, and then a bunch of near 0 readings while none is open.

afaiu the sensor samples at 1KHz that is then filtered (averaged if you will) to provide measurements at 5Hz
so that is 200 samples to provide one measurement over the period of 100 intake cycles (10K/60/2/5 * 6) at 10Krpm

[thomin]

Ok, I went over it again and the overall effect is much smaller than I originally thought. Let me try to take that same principle but at a slightly different angle:

Let's assume, that car X runs at 34 rotations per 0.19 seconds.

Given the nature of the engine, after two rotations, every cylinder has fired once. In order to keep it simple, let's assume that every 1/3 rotation, a cylinder fires. By the same token, every 1/3 rotation, a cylinder gets fuel injected.

So we have 102 times 1/3 rotations per 0.19 seconds, meaning 102 injections.

Assuming we want to get the maximum fuel flow of 100kg/h, this would mean 0.05174 g per injection

Now, our FFM measures intervals of 0.2 seconds. Now lets lust play this out for a couple of intervals.

After the first interval, the crankshaft has rotated 35.7895 times.
During that interval, we measured 107 injections and, crucially, we still have some rotation left, namely 0.1228.
Those 107 injections during the first 0.2 seconds equals a fuel flow rate of 99.6569 kg/h.

After the second interval, the crankshaft has once again rotated 35,7895 times, but this time we still have the 0.1228 rotations from the last interval on top which makes a combined 35.9123 rotations.
That gives us again 107 injections and a remainder of 0.2456 rotations.

For the third interval, we now have 36,0351 rotations which means 108 injections and 0.0351 rotations left.
Those 108 injections translate to 100.5882 kg/h.

Overall, it can never alternate beyond 108, because when you add up two remaining rotations, I can never get to 2/3 as each remainder is necessarily smaller than 1/3. Therefore it will only alternate between two numbers.

However, in real life, the injection cycle isn't neatly divided into 1/3 rotations. There are longer and shorter gaps and that definitely can change the outcome. I don't know how real-life gaps would look like in an F1 engine but I think in this example, 109 injections during a 0.2 second interval could certainly be a realistic outcome. That would translate to 101.5196 kg/h, which I think would be significant enough. Depending on the engine layout, we might even get to 110 injections, which would mean 102.4510 kg/h, though that would require a more extreme layout with some very odd firing.

[thomin]

Though we can see that while this timing effect isn't particularly awe inspiring (though not negligible either), had the FIA stuck to the 10Hz sampling rate, this effect alone could have had a big impact...

[bill shoe]

Don't know the details of the FIA fuel flow smoothing process. Typically the nominal frequency (5 hz in this case) is done on a rolling basis. So if you have raw data at 1000 hz, you take the average of samples 1-200, then you take the average of samples 2-201, then you take the average of samples 3-202, etc. Each averaged point is indeed the average of 0.2 seconds but each block center is only 0.001 second away from the previous block center, rather than the initial intuition of 0.2 seconds away.

In most situations the way you filter/smooth/average/whatever is determined by what you want to know from the data. The FIA tech people must have had something in mind when they came up with the 0.2 second figure, but I don't know what it was.

If you look at firing freq at 10500 rpm, then you get 105 injections per 0.2 second interval. It seems reasonable in a back-of-the-envelope kind of way to average a minimum of 100 injections to determine if the overall flow rate is compliant with a limit. In this manner, whether you happen to catch 104 or 105 injections per 0.2 second sample will make less than 1% difference to each individual smoothed data point. This may be what previous people were getting at in this thread.

I like the idea of cheating by aliasing the data, perhaps via a hydraulic resonance in the fuel system. However, there are two problems in trying to do this. First, the sensor samples its raw data very fast before sending a lower freq result to the FIA (and the FIA then smooths that down to 5 hz). Second, consistent long-term cheating via aliasing might require knowledge of the FFM sample phasing, which the teams don't have because the data is digitally encrypted and sent only to the FIA. Having said that, the sensor sounds like it may be very awkward at trying to sense fast transients so it may not matter what subsequent smoothing filter is used if you intentionally employ some kind of rough transient cycle at a fast enough frequency. I would think the top teams are trying to discover ways to take advantage of any transient inaccuracy in the sensor.

[hollus]

Excellent catch with the ~100 injections per sampling period, Thomin. Certainly that could be a factor.

What about upshifts? When you shift up from, say, 11000RPM in 6th to 9000RPM in 7th, with these "seamless" gearboxes that engage both gears at the same time, there might be a moment when you are effectively in 7th but at 10000RPM, and at full throttle, with clutch and powertrain taking the extra stress. If the ecu is not in perfect control there might be a moment when the actual revs are higher than the intended revs. While this transiently high flow for a few milliseconds would not be a big factor for running the car (and fudged out by calibration), it might be a big deviation for any given 0.2s period.
Or maybe I am talking out of my a## and the shift happens by first reducing the rhythm of injections and only then shifting?

[CBeck113]

Two quick thoughts from my side (I said somewhere else that I wouldn't do this, but I'm only adding to the discussion here:

1) The fuel flow is measured at all engine speeds, is there a chance that the problems could be worse at lower rpms?
2) Is there a limit on injections per cycle? Modern injection systems on street cars can spray up to five times per cycle, tayloring the combustion process, I just don't know if the FIA excluded this or not...wouldn't surprise me if they did though.

[jz11]

now you finally got to the point I tried to make 5 pages ago - if you go into high enough resolution ("instant" flow rates), then you are bound to run into these aliasing problems (whatever the real cause of error is), that is what I meant by saying - you cannot, with this type of sensor, measure a pulsating flow rate at infinitely small increments of time and get a consistent and true measurement, you will see either what thomin is trying to point at (which IMO is an oversimplification) or you there will be even bigger problems I described in the gill sensor topic related to the built-in error correction of the sensor itself and individual sample width (1kHz sampling rate doesn't equal to 1ms wide actual measurement)

this is why you cannot define a fuel flow at 100kg/h and say that whichever small fraction of time you decide to look at with this particular instrument, it must not exceed [100kg/h]*x, where x is your freely chosen measurement time period, thus the rule needs at least a clarification of the minimal time period that the flow rate must comply to, because as of right now - the minimal length of the sample is defined as 1 hour!

if the rule said - "maximum flow rate is defined at 100kg/h which will be measured 5 times a second", this whole debacle would have been avoided

[Cam]

MARUSSIA’S Graeme Lowdon

“My own view is, and it’s purely a personal view, that it would open the floodgates to all sorts of things if this appeal were to be successful because, yes, the ICA is the final arbiter, opinions given by the FIA outside of the Technical Regulations are meant for guidance and they are guidance, it doesn’t constitute part of the Technical Regulations as such, but I think some things should be reasonably straightforward.

So that does support a theory that Technical Directives hold no weight?

“If Red Bull chose to ignore the guideline from the FIA then they will have only done that to make their car go quicker - there’s no reason why they would have done anything other than that.”

Well, no. Red Bull are saying they did comply with the strict wording of Article 5.1.4 - which has been their position from the start.The fact that it made the car faster or slower is irrelevant.

This is the first real inside insight into what Technical Directives do. Interesting stuff.

Source: http://www.dailytelegraph.com.au/sport/ ... 6864180116

[FW17]

The discussion in the appeal court is going to be if the sensor was working or not and who is in a better position/informed to make that decision. As redbull would have had both sensor and injector data I think it was is fair for them to take the call that the CIA sensor is not working.

I am sure fia has a copy of the fuel maps for the race and could cross check the code for any foul play.

[bonjon1979]

The discussion in the appeal court is going to be if the sensor was working or not and who is in a better position/informed to make that decision. As redbull would have had both sensor and injector data I think it was is fair for them to take the call that the CIA sensor is not working.

I am sure fia has a copy of the fuel maps for the race and could cross check the code for any foul play.

This isn't the whole point though. Red bull aren't allowed to decide that they know best and can ignore FIA. I'll be amazed if they don't get some harsh criticism for their actions.

[Juzh]

This isn't the whole point though. Red bull aren't allowed to decide that they know best and can ignore FIA. I'll be amazed if they don't get some harsh criticism for their actions.

nope, It is exactly the whole point.

[turbof1]

I'm sure the discussion at the court will also handle the technical directive. I think that'll be the biggest issue: the fia has only this to defend its position, while red bull faces the task to proof it doesn't have any ground. The case effectively stands or falls with that.