Trust the Data (and Occasionally the Driver): Haltech Report

Introduction
It has been a while since I have sat down to write about Oxford Universities Motorsport Foundation’s (OUMF) recent data logging endeavors but it’s time to right this wrong. As you may recall in my second update, we had just left the team’s first test at MIRA which ended with a potentially destroyed limited slip differential. Well, good news, the LSD was completely fine,and it was just the team being overly cautious. With the car back together, the next 3 weeks were looking to be the busiest of the year. They consisted of the Goodwood Revival, followed by the HRDC meeting at Castle Combe, and culminating in our second test with the Elva at MIRA. 

The Revival and Castle Combe 
This year was the Riley 1.5’s turn to compete at the Revival with a new guest driver at the helm: BTCC star Charles Rainford. As this is primarily a data logging focused report, I will leave the full account of OUMF’s Revival experience to Sophie Rugg’s excellent write up on our website. Rainford was both surprised and elated to find we could go through his laps after qualifying and was particularly interested in his minimum corner speeds and gear choices throughout the lap. In the race Rainford was competing at the sharp end of the grid, which was then sadly cut short due to a mechanical DNF.

Figure 1: OUMF Team at Revival

---

​Figure 2: Castle Combe Race Circuit

The HRDC meeting at Castle Combe went on to show the practicality of the Haltech/Racelogic. We were again running the system in the Riley 1.5. A standout moment to me was after qualifying, Ding brought the car home 2nd in class, which was great, but what brought me the biggest smile was his eagerness to look at the data! He has been trying out different gearing, particularly through the final corner which leads on to a long flat out run to turn 3 (Quarry). Ding was experimenting with downshifting into turn 12 (Camp) or keeping it in 4th.
​
When looking at the data, downshifting into 3rd allowed for better pull out of the corner leading to the next lap’s Sector 1 time being consistently faster. This can be seen in the figures below, where on lap 8 he downshifted into turn 12 allowing the subsequent sector 1 to be 0.25 seconds faster! 

​Figure 3: OUMF Riley Quali Sector and lap time. Focus on Sector 1 on laps 9, 10, and 12

---

Figure 4: Speed, Delta-t, and RPM traces for braking zone of turn 12

Figure 2 shows the act of the downshift. On top of the time gained after turn 12, Figure 4 shows, when downshifting, there is a significant drop in delta-t. This is most likely due to the now combined efforts of engine braking and foot braking allowing for a later braking point. Sadly, all this effort in qualifying would be snuffed out in the race due to poor communication from the marshals leading to a black flag for our car. All in all, a disappointment as we had a quick car, but very educational on both the tools now at our disposal and processes we must have in place if the black flag flies for OUMF again.

MIRA Round 2
The team’s first attempt testing at this famous facility was a dress rehearsal. We thought we were prepared. We were not. We thought we knew how to test suspension. We did not. We thought the plan was achievable. It most definitely was not. But boy did we learn a ton! We now had procedures in place regarding assigning responsibilities, spanner checks, data pull, and what the driver needed to do. We were going into our second chance at testing with a sense of poise. We were also going into our second test with a new crop of freshers! We needed to show as many as we could what OUMF was about and that we (sort of) know what we are doing. 
However, the amount of interest we got surpassed all expectations and to our surprise, choosing who came with us became a part of planning the test. 

​Regarding actual test planning, we cut dampers out of the equation completely. This test would solely focus on springs and steady state cornering, saving damping experiments for another time. We needed this test to go well both for our sake and to prove to MIRA we were 
not wasting their state-of-the-art facility. Cutting down our planned runs to just 4 seemed withing our capabilities, if not a bit conservative. The run list can be seen below; each run would be labeled by the British Royal Navy Phonetic alphabet. Why? Because shouting “Get ready for Butter!’ got a chuckle out of even the most cynical of the team and (quite seriously) kept each run memorable for future reference. 

Figure 5: The Itinerary, Short but Doable

We took 19 people, 7 of which were freshers who would shadow an experienced member. When we arrived at MIRA, we set up, calibrated, and simply followed the itinerary. Ding was still getting used to being a test driver, but he knew we were here for data and not pole position, so he (for one of very few times) listened to me about keeping every input consistent. If this test had any philosophy, it was “one thing at a time” so we ran the car at the same cornering speed, same turn in points, same corner, and different springs. Our spring 
selection was varied although there were some large jumps in stiffness, we were still thankful to have multiple sets. The choice of where to start came from the previous setup (Non Bilstein Dampers) as well as minimizing the delta between front and rear calculated ride frequencies, which can be seen below for each spring rate kindly provided to us from Eibach and Bilstein. 

​Figure 6: Elva Spring BOM

At the time we thought we were going to look for a metric called understeer gradient. Put simply, understeer gradient is how much angle of the steering wheel to achieve a certain amount of lateral force. It is measured in degrees per G [°/G]. The keen observer would realize if we kept speed, turn in point, and corner the same then our lateral G will never change, thus keeping our understeer gradient the same each run. We (well, I, quite stupidly) did not realize this on the day but have now found ISO standard 4138 ‘steady state circular driving behavior–open-loop test methods’ so we now all know how to test understeer gradient properly. Even with this misguided thinking, we were able to gather plenty of data rather efficiently. So much so, we got through the itinerary and were able to continue through run (F)reddie. Also, our driver gave us great subjective feedback. Ding was able to keep a clear mind, and the old dog did indeed learn new tricks. During the test planning phase, Ding was hesitant to go stiffer from the (A)pples spring rates, but as we ran the test, he found on his 2 permitted ‘fast’ laps the car was more controllable and easier to read. 

Sadly, as our time was running out and the car came back from run (F)reddie, we opened the bonnet to a missing front pulley! As people were not permitted to walk the track without supervision and the sun was setting, we knew our day was over and we would have to find it another day (a story for another time!). Leaving MIRA, my laptop was full of damper potentiometer data and lateral G traces, great success!

The Real Data
As classes resumed after the test, and I once more hit the books, I found exactly the objective variable we now had data for: roll angle and roll gradient.

1. As the reader probably knows, oversteer and understeer balance when cornering is decided by the load distribution between front and rear axles (Segers, 2014). When cornering, the roll angle at the front or rear can tell an engineer where the load is located and the amount front or rear. Roll angle is decided by 3 modes: Roll angle of the wheel and tire (Unsprung mass) 

2. Geometric roll angle based on roll center location

3. Roll angle due to lateral weight transfer (sprung mass)

We will be focusing on the 3rd point primarily as springing is a primary variable in tuning this mode. Using our Haltech linear potentiometers and Racelogic’s software we can create math channels to calculate front and rear roll angles. Taking the average of these two values, we get a total roll angle of the car. Then plotting this roll angle value against lateral G, we create a plot like the one seen below: 

Figure 7: Run (D)uff total Roll Angle. Y axis is Angle and X axis is Lateral G

Then if we plot a line of best fit and find the equation for that line we can find our roll gradient. The exact value of Roll Gradient is found as the slope the best lines equation (in this case 1.3453). The roll gradient is determined by that axle’s resistivity to roll. Below is a chart of standard roll gradients for various cars: 

Figure 8: Typical Roll Gradients (Segers, 2014)

In the (A)pples run we had a total roll gradient of 1.75 °/G and our final run of the day ended at 1.36 °/G. Below are charts of the progression of our roll gradient throughout the testing day as the Elvas subsequent first race at a wet Silverstone:

Figure 9: Total, Front, and Rear Roll Gradients

From these charts it is clear the front and rear axles have very different roll gradients, but with subjective feedback we found Ding was happiest at run (C)harlie values. This gives us a target roll gradient value. At races, using the roll gradient value and static roll center’s location we are now able to calculate roughly what the axle stiffness should be. Axle stiffness is determined by spring and anti-roll bar rates, meaning we can now roughly calculate necessary components. A very powerful tool indeed.

Silverstone: Putting it Together 
The time finally came; we were going to put the Haltech/Racelogic equipped Elva up against other cars. All our preparation, learnings, and newfound ideas were going to be put to the test at last! I won’t go too in depth into the weekend as you should read Elliot Nash’s report on our website, but I’ll give you a brief resume! With a rebuilt engine and some fiddly clutch issues solved, we loaded the Riley and Elva - now both equipped with data logging - to make the short drive to our ‘local’ circuit. The weather was not on our side as it looked like our day would start dry and quickly work its way into a rainy mess. 

Importantly this weekend, we were loaned a very interesting piece of equipment: a live streaming v-box unit from Racelogic. We decided it was best for the live streaming to go on the Riley where it could be full showcased. Qualifying went well for both vehicles, with the only minor issue being a loose connection in Elva’s data logging. On the other side of the garage, the live streaming team came up with a unique solution to an issue regarding sim cards… just send one of our phones in the car! While a little odd, it worked great and for the first time we could watch in-car footage for a whole session, brilliant!

With the rain pouring down for both races it was decided to go with a softer setup on the Elva. It proved difficult to drive anyhow with all the standing water, but Ding was able to bring the car home 4th in class behind some very fast Lotus Elans. They have a target on their backs for next year. The Riley also ran a wonderful race, and it was brilliant seeing some brave overtakes on the live stream! All in all, the rain put a damper on things, but the systems worked and we now have a better understanding of where to go from here. 

Conclusion 
It’s safe to say my time working with both the Haltech and Racelogic systems have taught me more than I could have imagined both about systems engineering and race engineering. With the additional support from Bilstein and Eibach our adventures into Vehicle Dynamics are turning out to be quite fruitful and fun. 

My most sincere thanks to Andy Owen for lobbying so successfully for us to be allowed the use of the MIRA test track, as it has been key in figuring this all out and I look forward to learning even more there soon hopefully! This whole experience has been a blast, and I can’t wait to keep having a crack at it next season (hopefully in the dry!) with a team of fresh and old faces. Onto the next race - and Happy New Year.

-Charlie Lemme 3rd year BEng Motorsport Technology 
President of Oxford Universities Motorsport Foundation