Is a rigid chassis needed w/ actively adjustable suspension?

[DaveW]

... and using real position sensors (surely the actuators are more naturally position controlled devices).

It is true that hydaulic actuators are normally position controlled, but only by virtue of a position control loop. Intrinsically, they generate an actuator velocity proportional to drive current.

In our case, the control code was arranged to generate actuator velocity demands for each corner. These were used to drive idealised model actuators, the position outputs of which fed positions back into the control code. As a separate task, model velocities & positions were used the drive the actuators themselves. There two reasons for this strategy: it minimzed possible coupling between the control code and temporary actuator shortcomings, & it provided a way of detecting actuator faults by comparing differences between model & actuator positions.

I think it all depends on from where the feed-back, or "is-value", is taken, ground-clearance sensor or something else, but either way, the draw-back with hydraulics is always speed. Ideally electro-magnets would probably the way to go, which is something that has been tampered with on valve-trains, but at a terrible energy-price.

The system transport delays were something of a problem, and electromagnetic actuators would have made life easier. They have two problems, however, the size & weight of the power amplifiers & the fact that they are not good at supporting steady state loads.

Total delay of our system was the sum of transducer filters, code delays (the time required to response to an input), and actuator delays. Transducer filters were set at 100 Hz (2-pole), and the actuator delays corresponded to filters of around 200 Hz (roughly). I worked hard on control code layout & achieved an additional transport delay of around 150 microsecs. The system was reasonably responsive, repeatable & robust.

The real problem is having at all times the right (position?) demand signal for each actuator ? This will take some development I think. This was the critical factor with the Lotus ?

The '99T worked well on "mechanical" circuits (it won at Detroit & Monaco in the capable hands of Senna). Performance suffered in high downforce circuits (not working the tyres hard enough, we decided - probably not helped by lack of aero). We should have won at Monza (a low-downforce circuit), but Senna fell off overtaking a lapped car. He was adamant that was caused by the system - it is possible that the lateral balance algorithm caused the car to three wheel at the limit (it was known to have a poor natural balance). That race was won by a jubilant Williams team - their first win using an active system. Senna recovered from the sand trap & finished a distant second.

We ran 3 cars (2 & a spare) for a whole season. Whilst we experienced the odd problem in testing (some self-inflicted), we only suffered one DNF. I recall that was caused by an electrical failure after an errant piston severed a cable loom. We suffered a major oil leak courtesy of a factured coupling at Imola - Senna finished second after running out of fluid 6 laps from the end.

Edit: I should down-play my DNF claim - I think it was 1 active system-related DNF, apologies.

[Belatti]

Dave, your insights about active suspension beginings in F1 are priceless to me.

[bhall]

[...] Terrain Following. This was a system originally developed by Ferranti for TSR-2, & refined on Tornado.

I didn't know about that part.

(Better late than never.)

[gixxer_drew]

Dave, absolutely love your posts, every time.

I am curious about your comments what things you saw teams doing to compensate for too much compliance in the chassis. Did you see any patterns to what they were doing or just a lack of response from the chassis to them?

What an interesting post. I would like to put my own interpretation on it.

It is worth pointing out that there is no such thing as a rigid chassis.

In the years I have spent rig testing a wide range of race vehicles, I have encountered several vehicles that have suffered from excess compliance. Such vehicles are almost never competitive. The signature of a problem is often what the team has done with suspension set-up, presumably to try to mitigate the consequences of the compliance. It is usually the case that those "enforced" changes do not actually improve performance.

I have also rig tested some good "stiff" race vehicles. Such vehicles respond logically to set-up changes and, others things being equal, are normally competitive.

I think true to say that chassis compliance is not often measured, & when it is, interpretation is somewhat primitive. Gaining control over compliance is not easy, & neither is it "free" (it costs weight & or volume). Largely because of this, I suspect, many designers ignore compliance, & those that do impose limits, do so in a fairly "global" way.

Without question, I would suggest that devoting some of the megaflops currently used on CFD to the design of a well controlled chassis would pay dividends. This would not be a completely straightforward task, because "well controlled" needs to be defined in detail, and details matter. For example, the upright-upright torsional stiffness of a open wheeler is divided, typically, more or less equally between the chassis and the suspension links.

[xpensive]

Aren't we looking at a high force, lowish speed situation, ideal for hydraulics ? They will beat anything else for all round performance'. Really the hardware is the easy bit, trust me !
...

If you by "hardware" mean transporting hydraulic fluid through the orifices of the valve, lines and in/out of the actuators,
I'm afraid I don't trust you at all, said physical transportation is to my humble mind likely to be a crucial parameter here?

Dave W; What hydraulic system-pressures and actuator-areas are we talking about here anyway?

[DaveW]

I am curious about your comments what things you saw teams doing to compensate for too much compliance in the chassis. Did you see any patterns to what they were doing or just a lack of response from the chassis to them?

The most common signature of a compliance problem is a suspension set-up that is too stiffly sprung &/or too lightly damped.

It is a commonly held view mainly by Aerodynamicists (being unkind), that excessive compliance can be corrected by increasing spring stiffness. From a mechanical control perspective, it is important to maintain the ratio of stiffnesses. For example, when playing with active suspension we used a rule that installation stiffness should be greater than 3 times the tyre stiffness.

Compliance that is in series with a damper affects damper performance in a way that varies with frequency. More damping is required for control at low frequencies (the heave mode) but less is required at high frequencies (the hub modes and sometimes the pitch mode), where the compliance allows the damper progressively to "lock out". It is quite common for an F1 vehicle (because they are often subjected to aggressive weight reduction exercises) to have under-damped heave modes (sometimes by a factor of 2) in order to retain some control over the hub modes. I find it frustrating to identify and explain the problem to test engineers, knowing that nothing will be done about it.... The mantra "mechanical set-up never gives lap time" is, in a sense, self-fulfilling.

It is also true, I think, that compliance that can be felt by a driver often results in increased stiffness (e.g. poor rear toe control is often felt as "roll").

[DaveW]

Dave W; What hydraulic system-pressures and actuator-areas are we talking about here anyway?

Pressure is easy, 210 bar, with supply side accumulators at each axle. I don't recall actuator areas, but they were probably between 0.5 & 0.75 square inches (apologies about the units). Each actuator was controlled using light-weight EHSV's modifed by Moog to around double their normal capacity. They were quite lively, before we modified the accelerometer location, they would quite happily drive the A-arm fundamental bending mode into instability.

Whilst searching for info, I came across this video. It is interesting to compare it with footage from today's vehicles.

[xpensive]

@ DaveW:
Right, if you utilize half of 210 Bar over a 22 mm piston, 3.8 cm^2 or 0.6 sq.in, that's 4000 N, which seems plausible enough.

Where the acutators in line with the push-rod, whithout any ratio, as I recall the Williams layout?

[gixxer_drew]

I am curious about your comments what things you saw teams doing to compensate for too much compliance in the chassis. Did you see any patterns to what they were doing or just a lack of response from the chassis to them?

The most common signature of a compliance problem is a suspension set-up that is too stiffly sprung &/or too lightly damped.

It is a commonly held view mainly by Aerodynamicists (being unkind), that excessive compliance can be corrected by increasing spring stiffness. From a mechanical control perspective, it is important to maintain the ratio of stiffnesses. For example, when playing with active suspension we used a rule that installation stiffness should be greater than 3 times the tyre stiffness.

Compliance that is in series with a damper affects damper performance in a way that varies with frequency. More damping is required for control at low frequencies (the heave mode) but less is required at high frequencies (the hub modes and sometimes the pitch mode), where the compliance allows the damper progressively to "lock out". It is quite common for an F1 vehicle (because they are often subjected to aggressive weight reduction exercises) to have under-damped heave modes (sometimes by a factor of 2) in order to retain some control over the hub modes. I find it frustrating to identify and explain the problem to test engineers, knowing that nothing will be done about it.... The mantra "mechanical set-up never gives lap time" is, in a sense, self-fulfilling.

It is also true, I think, that compliance that can be felt by a driver often results in increased stiffness (e.g. poor rear toe control is often felt as "roll").

That is interesting because I am aerodynamicist and one of the first things I tell my teams is that downforce makes it even more critical that the chassis is as stiff as possible. Because we will end up with a stiffer car to support the aero platform and all the loads increase... so the compliance issue just gets to be a big deal. You can look at any well performing car in history that developed increasing aero loads and they eventually had to rethink their chassis.

The people I always have a problem with is the constructors who have no experience with high downforce, they swear their standard way to build a roll cage is the pinnacle of strength and stiffness. I've spent months designing chassis stuff only to have it totally ignored and a cage built as the fabricator saw fit.

I think one problem is that a lot of this stuff is often left out of simulation code, so you figure out where to look next for the low hanging fruit in terms of total performance. Aero almost always wins in the simulator and then in the real world you find things like the chassis is flexing so bad the car sucks and you have to fix that before you can do anything else experience teaches you those kind of thing, I know I'm always learning.

[DaveW]

... which seems plausible enough....

Thanks, X! I recall no issues with the rear motion ratio (hence probably fairly linear & close to unity). The front suspension was a problem, though, rising fairly rapidly to >2 if not controlled properly.

That is interesting because I am aerodynamicist...

Oops, ...but I didn't condemn all aerodynamicists... Apart from having to make that apology, I agree with your sendiments.

[xpensive]

... which seems plausible enough....

Thanks, X! I recall no issues with the rear motion ratio (hence probably fairly linear & close to unity). The front suspension was a problem, though, rising fairly rapidly to >2 if not controlled properly.
...

While 210 Bar is of course 3000 psi, didn't see that at first, but please identify "motion ratio", is that a geometrical something?

[DaveW]

.. but please identify "motion ratio", is that a geometrical something?

Wheel velocity/Actuator velocity. Hence actuator loads are wheel loads * motion ratio.

[xpensive]

.. but please identify "motion ratio", is that a geometrical something?

Wheel velocity/Actuator velocity. Hence actuator loads are wheel loads * motion ratio.

Got you W, was that a constant or progressive/degressive?

[WilO]

Dave,

Active suspension is by far (at least by my definition), the most interesting topic on this or any other forum, due to your contributions. Thank you for taking the time to share your knowledge/experience with us. Learning about the development process and the logic behind the decisions that were made is fascinating,and I think much can be learned about the operation of conventional suspension through this discussion.

I'm wondering if you'd permit me dumb question; How critical are wheel alignment factors given the use of active suspension? I would think that changes could be made to camber and toe settings in particular.

Thanks again for the insight.

Wil

[DaveW]

How critical are wheel alignment factors given the use of active suspension? I would think that changes could be made to camber and toe settings in particular.

I can't help thinking that I am about to reveal my lack of knowledge of the finer points of suspension geometry.

First, it is possible to program an active suspension to simulate a passive suspension more or less exactly. Then the subtleties of suspension geometries would be replicated in full, but that would be an expensive way of achieving not very much.

I guess most people, given the opportunity, would try to manipulate the way the sprung mass moves in response to maneuvering loads. That is certainly what we explored in the prototype vehicle (a Lotus Esprit). We thought little about camber gain, etc., but learned much in a very short time (or at least I did, being a fairly dumb aerospace engineer). For example:

The warp mode was very important (a mode I omitted initially, I am ashamed to admit).
Making warp offset a function of lateral acceleration controls lateral balance.
Nulling the dynamics of the sprung mass response in maneuvers is by a margin the best strategy (roll out or roll in are equally bad, ditto for pitch).
Using hub velocity as a feedforward signal in moderation has a substantial effect on ride.

After around 3 months of making it reliable and playing with control loops (at that stage the controller was a bespoke analogue computer, and our test track was the local roads around Cranfield) we spent a day at Snetterton (a local race track) with Elio de Angelis & Nigel Mansell. The "standard" was provided by a passive Esprit (Peter Wright warned me not to be upset about the outcome, because the passive Esprit was a very quick car & our objective was not actually to improve on its performance).

The day was a defining moment. The active car was slow in a straight line (the active system was over-designed by a margin, the gearbox-driven pump consumed too much power & made it almost impossible to change gear quickly). However, overall lap times were very similar. Both drivers praised the handing of the active car. The boss emerged smiling after being driven around the circuit, & approved the next phase.

All of that happened without anybody (to my knowledge) worrying about suspension geometry - apart from agreeing actuator strokes. Perhaps we were lucky, or perhaps, active suspension allows more important parameters to be controlled better. If that is the case, then there could still be opportunities to re-optimize suspension geometry.