Hypermiling a turbo GDI engine (VW TSI)
 

I've been hypermiling for almost 15 years now (I recall hanging out at CleanMPG around 2007-2009 or so), but I just decided to create an account here because my current car has been leaving me puzzled. I posted something similar at the CleanMPG forums back when I bought the car, but apparently the forum is not very active currently.

As everyone knows, the latest trend is in downsized, turbo direct injection engines. I bit the bullet last year and bought a VW up! TSI sporting a 1.0 l, 3-cylinder turbo engine with 105 hp and a manual transmission.

While I'm getting nice figures (about 12-13 km/l tank averages with E100 fuel, all-city driving, with very short, 5-10 km trips), I get the feeling I could be doing better. Especially from the kinds of mileages that I see reported elsewhere

I obviously follow the usual hypermiling tips: reduce speed, properly inflate tires, remove weight from the car, time traffic lights, avoid putting yourself in a position where you'll need to brake, no air conditioning, etc.

As for driving style, I use P&G with engine on (trust me, this car doesn't like to do EOC -- I'm quite familiar with EOC as I used it on my previous cars), try to stay in a 1700-2700 RPM band more or less, and target 70-80% of max load during pulses, which means using some boost. I installed an analog boost gauge on the car, which connects to the MAP sensor of the car (not to OBD-II), and seeing as the car can reach about 0.7-0.8 bar of boost, I try to run it at 0.2-0.4 bar of boost. I've seen people say you should avoid boost like the plague on turbos, but on the other hand, the BSFC maps that I've seen appear to contradict that advice -- sadly I've yet to find one for my engine.

A few days ago I decided to look at the throttle position reading using an OBD-II app for my smartphone and found that, at about 0.2 bar boost or a little bit over that, I already reach the WOT condition at 88%. Further pressing the throttle pedal will take boost to 0.7-0.8 bar like I said, while the throttle position stays fixed at 88% -- I believe the wastegate valve may be controlling the exact amount of boost. I've always understood that a partially closed throttle reduces efficiency through pumping losses, so it appears to me the 0.2 bar boost for pulses may be a good choice.

Also, I've got a lambda gauge (rich/lean) on my OBD-II app, and monitoring it as well as PID 03 appears to indicate that my car stays in closed-loop, lambda=1 mode even at WOT and high boost values of 0.7 or 0.8 bar. I've mostly tested this at low to mid RPMs, which is the range of interest to hypermilers anyway. From this it appears that fully flooring the car (at least at low RPMs) shouldn't harm fuel efficiency.

I welcome comments on what I'm currently doing and advice on what I should be doing better. Also, if anyone who advocates avoiding boost on turbo engines could explain to me the technical reasons behind this, I'd also be grateful, because I'm really not convinced that I should be avoiding boost -- at least that's not I see in BSFC maps for turbo engines.

Did you ever notice any influence of weather to the fuel-efficiency?


You might remember when direct injection was quite a rocket-science feature. Then, either resorting to a large-capacity intercooler (or water injection) or to enrich the air/fuel ratio was a requirement to prevent knock. Remember when Hyundai introduced the 1.0 turbo to the previous-generation HB20 retaining the port-injection and was plagued with excessive fuel consumption? No wonder it has turned to direct injection for the current generation, the same way it had been with any other overseas Kia or Hyundai with the 1.0 turbo engine before its introduction to Brazil.

A few years ago, I was driving a rental car in Finland. A Golf with the 1.2L TSI. When doing P&G, I noticed the instantaneous fuel consumption was the same at WOT and other high-load pedal positions. So I just floored it for the pulses! I can't say if it was the most efficient way, but it was fun and easy. This was all at lower RPM.

Definitely, although as I mentioned, I mostly do short trips, so a big factor is the starting temperature and how long the engine takes to heat (although the engine in my car is designed for quick heating, but still, since trips are short, it will spend a non-negligible amount of time in a cold condition).


I'm not sure I get the point you're trying to make here. Could you be more explicit?

Maybe you're saying turbo was an issue in the past? Like I said, however, instrumentation indicates that my car runs stoichiometric regardless of what I do, so unless the OBD-II data is not accurate, then enrichment is not an issue -- indeed, I've thought of trying to read the voltage of the oxygen sensor directly somehow, but haven't gone to the trouble of doing that yet.

Do note that I see changes in lambda sometimes: for instance, after a pulse, when I shift the car into neutral, the mixture is enriched for a few seconds, then overshoots a little into lean territory for a couple of seconds, and then settles into stoichiometric again. So it's note a case of a "broken" ECU reporting a constant lambda=1 all the time.

But perhaps there are reasons to avoid boost other than fuel enrichment? That's what I'm looking to be enlightened with.

With direct injection, it's mostly pointless trying to get out of boost.

I can't speak for your specific vehicle, but engines under boost have higher cylinder pressures. To prevent detonation (also called "knock"), the typical strategies involve going rich and retarding ignition timing.

Going rich reduces combustion temperatures by dumping in extra fuel which won't be burnt to make power, but instead just absorbs heat and is flushed out the exhaust.

Retarding ignition timing starts combustion later, and effectively dumps part of the useful energy that could have been extracted from combustion, out the exhaust.

If your engine has the proper fuel and spark maps to take advantage of E100's higher knock resistance, it may not need to go rich or pull any timing under boost, but that isn't a given.

Turbo engines usually have lower static compression ratios, which makes it so they don't need to dump as much fuel or pull as much timing under boost, but it also makes them less efficient outside of boost.

~

12-13km/L all-city driving doesn't seem bad at all to me, frankly, but there aren't many (any?) cars sold in the US which can match those figures without hypermiling, that aren't hybrids.

Nice observation, which reminds me: are these higher cylinder pressures harmful to the engine? In that case, if I'm trying to balance fuel savings with maintenance cost, trying to make the engine last as long as possible, maybe I should prefer less over more boost? Unless more boost is especially efficient.

I live in Brazil and the absolute majority of cars sold here are flex fuel straight from the factory (and to be clear, my car is one of those). Indeed, you'll find both E100 and E27 on nearly every gas station in the country.

You do remind me, however, that I should perform the same tests on E27 as well. Maybe I'll switch fuels for the next tank to test that.

The results are as shown, though: lambda=1 and closed loop operation under all conditions that I tested (save while in DFCO, evidently, and when switching the car into idle after a pulse).

My car's engine in particular has a 10.5:1 compression ratio. Not sure you'd consider that low; my previous car was NA and had the same ratio.

Still, is it the case that the thermodynamic efficiency increases with boost? I've always thought that, in a way, boost achieves the same effect as increasing the compression ratio. If, at a given moment inside the engine, there is a full 1 bar of boost over atmospheric pressure, then isn't that equivalent, in terms of efficiency, to having a NA engine at WOT with twice the compression ratio?

In that case, and assuming the car is indeed running stoichiometric under all conditions, then I would assume the most efficient operating point would be at max boost.

Indeed these are good figures (the local equivalent to EPA figures are 9.6 km/l city and 11.1 km/l highway), and I can get even better figures when conditions help.

For instance, today I took a somewhat larger trip (about 10 km each way), most of which was on the peripheral highway around my city. Recall it's summer here, and I don't use A/C. Because it's Sunday and there were few cars on the road, it was safe to P&G around 50-60 km/h, sometimes going a little higher. Even with a small patch of city driving, a few traffic lights, etc. I was able to achieve 19 km/l according to the trip computer, which after correction should be closer to 17 km/l, but still, it's an excellent figure.

The reason I'm looking for ways to improve is that I've seen people report figures as high as 25-28 km/l (of E27) on the highway, and these people probably aren't taking any heroic measures to save fuel (i.e. they're probably targeting a speed of 100 km/h or more, A/C on, etc.) Now I'm well aware that there are differences on the quality and performance of two engines coming out of the same plant, due to e.g. manufacturing tolerances (and never mind that I haven't yet put 10.000 km on my car, so the engine is probably a little rough still). Also, knowing my fellow countrymen, I wouldn't be surprised if they're "rounding up" the numbers a bit; probably using trip computer values rather than actual measurements at the pump; claiming a record best, once in a lifetime figure is an average; etc. Still, maybe there's some truth to these figures, in which case I have a lot to learn.

Probably not. The turbo has a lifespan, and being in boost more frequently might increase piston ring wear somewhat, so maybe it'll start burning oil a little sooner. I don't have a good reference to go by for small turbo VW engine lifespan though.


Even under boost? That's surprising, but suggests much lower economy losses under boost.




Same as my car, approximately. The Insight stock engine was 10.8:1. That was 20 year old port injected tech though.

Mazda's SkyActiv direct injected engines are 13-14:1, but they're on the upper end of what's typical.

OTOH, higher compression does show diminishing returns. I think at ~10:1 it's something like 2.5% efficiency for an additional point of compression, but less going from 12:1 to 13:1.


I would think so, yes. There are some other confounding factors though.

For example with my K24 engine (10.5:1 compression), I have to pull timing at WOT below 3000rpm, or I get knock, even with 93 octane fuel. Undoubtedly timing is being pulled with cylinder pressures literally twice as high in a turbo engine. Some of those efficiency gains are being lost due to not being able to use MBT ignition timing.

Turbo boost is also not *entirely* free, it's just far less costly than a supercharger. The exhaust gases leaving the cylinder have to push through a turbine, and that steals a bit of energy from the piston trying to evacuate the gases from the cylinder.


With my Insight in perfect functioning order, with some small aero mods, before my engine swap, I was able to achieve around 100mpg @ 50mph, in good weather, which is around 43km/L. Even with the much bigger and less efficient ~240hp 2.4L K series engine under the hood, I'm still seeing ~26km/L at lower highway speeds.

I'm certain there's some more you can squeeze out of your car. I'll be looking forward to seeing what you come up with!

Yes, even totally flooring the car. At least with E100. And like I said before, under some specific situations I can see it report lambda values other than 1 (under DFCO and coming out of a pulse), so it's not a case of a broken ECU that only reports lambda=1.

I'm reading lambda from PID 34 (hex)/52 (dec), which is reported as supported by my car. But if there are suggestions of other PIDs which are more reliable for reading lambda, I'm all ears.

It looks like ignition timing is something I'll have to learn about. I'll start monitoring that gauge during my drives and see if I can learn something.

Wow, I'm thoroughly impressed. Does it use a manual transmission? Are these figures for E0, or E10, or what?

My previous car was a Honda Civic Si (not sure if the engine was a K20Z3 or K20Z5), and I believe I got, only once, 20 km/l out of it, driving very very slowly, using E27 -- that car wasn't flex fuel. In the city, 12 km/l was an excellent mileage.

Just for the record, there's some help from your car's aerodynamics: according to Wikipedia, the 1st generation Insight had a Cx of 0.25, and the frontal area (if it's as simple as multiplying width and height) is about 2.30 m^2, so a Cx*A = 0.574. According to one source that I found, the up! has a Cx of 0.367 and a frontal area of 2.08 m^2, so Cx*A = 0.763 (about 33% more than your Insight). Also, I'm avoiding very high tire pressures on my car: I'm using about 35 psi -- not sure about you.

Regardless of these differences, these are very impressive figures.

Have you documented your driving style somewhere in the forum? If not, could you very briefly summarize it here? I know these are very different cars and engines, but I could probably learn a lot.

So I went out for another drive today, paying attention to ignition advance -- PID 0x0E (hex)/14 (dec).

So indeed, at lower boost, ignition advance is higher (more degrees before TDC), while at higher boost, ignition advance is lower (fewer degrees before TDC). Either way, while pulsing, it's always positive (before TDC). I assume none of this is news.

So you claim ignition is retarded (pushed closer to TDC) to avoid knock, and that makes sense: if you ignite too early, there is the risk of the flame front meeting the piston head while it's still compressing rather than expanding.

What I'm trying to understand is how that "dumps part of the useful energy that could have been extracted from combustion, out the exhaust". I imagine early ignition, when the flame front meets a compressing piston, besides being destructive (through knocking), would also waste useful energy by pushing the piston head in the opposite direction of its movement, therefore doing negative work. But if the flame front meets the piston head when it's already expanding, then isn't that doing positive work as expected? Where exactly is useful energy being dumped out the exhaust?

Sorry if it's a stupid question, but I really want to understand this. It would help me understand why my car could have a BSFC sweet spot at part load, and why it would be better to pulse at lower boost pressures rather than higher ones.

Those figures are using E0, so I have a 50% energy density advantage.

I can summarize a few key points I've found in getting good economy with my car:

-Cold starts kill economy. After maybe 15-20 minutes of running, it reaches peak efficiency, but economy can increase by as much as 50% on long trips vs short ones. While there are some things that can help get up to temperature more quickly (grille block, block heater) we often can't choose our driving conditions, so it's important not to get too wrapped up in comparing with other drivers who have different driving conditions

-I find I get better economy the lower I shift. I've heard others report getting better economy by raising their shift points and accelerating briskly, but I personally haven't observed it. I'll throw my car into the highest gear it can take, as soon as I can, if I'm driving for economy.

-On the highway, slow and steady wins. With the stock motor it was very easy to keep it at 50mph and ignore other drivers. The new engine is capable of accelerating much more quickly and requires a lot more self control. I find myself really needing cruise control to keep my speeds down. For my vehicle, above ~50mph / 80kph, economy seems to drop off quickly. Below that, it seems more susceptible to environmental conditions, but slightly higher numbers are possible.

-it probably doesn't need to be said, but around down, conservation of momentum yields large gains.

-Modifications provide very small but incremental improvements that add up. I'm aware of one driver on Insight Central who can consistently get 150mpg+ tanks out of his car. To do this, he runs 120psi+ in his tires (which are mounted on donut wheels), his fuel line wraps around his catalyst to preheat it, and for a while he had completely removed his front brakes (because discs have a lot more drag than drums), plus dozens of other modifications (no mirrors, I think?). I think he's nuts, personally, but it goes to show that there's always more fruit hanging.

That said, even the best mods' gains can disappear behind changes to driving habits. Pick the mods that work best for you, drive as well as you can, and don't sweat the rest.

Great observations!

You're right, it isn't news that ignition timing is more retarded (or less advanced) at higher boost. Just looking at timing numbers though, it's nearly impossible to glean much.

You're right about igniting things too early as having a negative impact on economy. There's a concept, called "MBT" (mean best torque), which refers to the perfect ignition timing at a given load and RPM, to get the most power out of combustion. Ignite the charge any earlier, and you get more negative work, with combustion pushing on the rising piston. Ignite it much later, and the flame is chasing the piston down the cylinder, never actually getting to push on it. It's a continuum, of course, but the idea is that there's a "perfect" time to ignite the charge to get the most power out of combustion.

It sounds like you already know a fair bit of this, so forgive me if I write a lot that isn't news to you, but the reason you can't glean much from just looking at timing numbers is that MBT is different for literally every point on the map, and for every engine. Further, combustion isn't instantaneous - it's a flame front that spreads from the point of ignition, and takes time to complete.

As more fuel burns, the heat (and therefore pressure) in the cylinder rises. If that pressure gets too high before combustion is complete, the remaining unburned fuel can auto-ignite, either in points or all at once. When the two flame fronts crash together, this is audible as a "knock" or "ping", and the resuoting shockwave can damage the engine. It also reduces fuel economy (and power) so even if you build an engine strong enough to handle it, you still wouldn't want it (with some notable exceptions).

Keeping cylinder pressure from getting too high is managed in several ways. Higher "octane" fuel is more resistant to detonation, but typically has some drawbacks (refining, cost, energy density), and what kind of fuel one can get varies from place to place. Others ways include: Reducing intake air temperature, dumping extra fuel to cool the chamber, intercoolers. Some combustion chamber designs are simply better at preventing detonation ("hemi" designs are among the worst, pentroof among the best). Lowering boost (dynamic compression) lowers peak pressure. Reducing static compression. Also, starting combustion later (retarding ignition timing) moves the point at which combustion competes to a later time, when the piston crown is farther down the cylinder, meaning the instantaneous chamber volume is larger when the maximum heat and energy has been released. Unfortunately, retarding timing to prevent knock may also start ignition too late to "capture" as much combustion energy as possible (past MBT), resulting in more combustion energy passing into the exhaust as heat, rather than doing useful work.

Some other interesting bits:

When cylinder pressure is higher (more air and fuel, or boost), the combustion event happens more quickly. One can think of it as the fuel and air molecules as being closer together, so it spreads between them faster and more easily. Likewise with a higher static compression ratio, the flame front literally has a smaller area to fill before complete combustion. Because of this, less ignition advance is needed for MBT (you can start it later because it finishes earlier), and a smaller portion of combustion happens while the piston is still rising, meaning there's less negative work that happens.

This is one of the reasons higher loads are generally more efficient. In most engines, and especially at low load, combustion needs to be started long before TDC. I believe at certain points in my engine's map, I have ignition starting as far as 48 degrees before the piston reaches the top of its compression stroke. At higher loads, it's more like 12-25° depending on RPM (flame speed is mostly constant, so if the piston is moving faster, you have to start combustion sooner).

Different fuels also burn at different rates. Alcohols generally burn faster than gasoline, so they need less ignition advance even with all else being equal.

There are also a few tricks engineers have found with engine design to help with this. One example is an offset crankshaft, where TDC doesn't occur at 0° crank rotation. While it results in an unbalanced engine, it helps sidewall loading on the combustion stroke, but more importantly, it offsets ignition timing so that much less negative work happens. Downsides include a lot more vibration and harshness, and the need for much lower redlines. Possibly engines built this way are also less able to take high boost, though I'm not certain.

There are probably some things I'm forgetting, and some things I haven't explained well, so I might post a bit more later if I think of anything.

Wow. Lots of information from the last couple of posts. It'll take me a while to digest all this. I'll probably come back to them quite a few times. For now:

My car has a shift indicator, and it appears that, as long as the current speed is such that next gear will be at 1500 RPM, it wants me to shift. For a while now, I've delayed shifting so that I'm closer to 2000 RPM rather than 1500 RPM at the next gear. The turbo lag is especially noticeable under 1500 RPM, with a transition band from around 1500-2000 RPM, and at 2000 RPM I can get full torque.

I was basically afraid of lugging the engine (note this is a 3 cylinder engine), or of being in an inefficient region of the engine map. Maybe I should experiment with following the shift indicator's advice.

Do you do P&G, or just cruise at a low load?

I've gotten so used to P&G that I just do it automatically. If I have the car in gear for too long, I start getting jittery. When I got this car I learned how to do a proper double clutch so I can engage back after a glide with minimal jerking. It's just too bad the car doesn't tolerate EOC.

However, I'm not sure how helpful P&G actually is -- and one of the things I had in mind with starting this thread was to figure out the optimal way to perform a pulse. This car is underpowered at low RPMs without boost, and the gearing is very tall (it sure could use a 6th gear), so maybe P&G is not really necessary -- especially considering I can't EOC it.

This is something I do religiously. Maybe I'm doing everything wrong in terms of how I'm accelerating the car, but I'm sure I make up for it by properly conserving momentum.

The Insight's original engine is a 1.0L 3 cylinder non-turbo with lean burn at part throttle, and very tall gearing - around 1700rpm @ 80kph. Because it lacks a turbo, its efficiency zone may be very different than your vehicle's. Manufacturers very rarely publish BSFC maps, and even doubly so more recently - I don't think I've ever seen one for a modern GDI turbo engine, so it's pretty hard to advise based on published data.

That said, Honda went to great lengths to maximize economy, often at the expense of all else.

Its shift indicator light has some logic which seems to take load into account, but basically it comes on if it could possibly be in a higher gear above idle speeds, or above maybe 2000rpm with accelerating.

This is slightly oversimplified, but when cruising, it leans out to ~24:1 AFR, which allows very high loads when cruising, often above 75%. Even slight grades may need downshifting to maintain speed. It will attempt to hold onto lean burn until maybe 90% load, at which point it drops back to stoichiometric for more power. Below maybe 2500rpm, one of the two intake valves only opens a hair.

The strategy seems to be to get load up as high as possible, as much as possible. It will happily cruise or even accelerate (albeit very slowly) as low as 1000rpm and not suggest a downshift until the pedal reaches the floor, and even then not for a few seconds.

I can't advise if avoiding boost is more or less efficient, but the shift light can probably be trusted.


With the stock engine, I found P&G without EOC to provide minimal gains, but P&G is largely a tactic to get load up, and secondly to minimize other engine losses (allow average engine RPM to be lower), and Honda already did a good job getting load up. However, there were significant economy gains to be found if I killed the engine any time there was a downgrade where I could coast without power for more than maybe 5 seconds.

With my new engine I have an even taller top gear, but the amount of torque available makes it such that load isn't very high when cruising, often at or below 25%. I see moderate gains from P&G, and large gains from EOC. I'd say under the right conditions, the improvement can approach 50%. It's kindof a pig, and the less time it spends running, the better.

With my previous car, gearing was very bad for economy. In that one, I could see gains as high as +80%. So, I'd say P&G shows diminishing returns the taller your gearing is, or the smaller your engine is.

Let's hope someone else can speak from personal experience about boost. I wish I could advise on that. With older engines, boost was definitely to be avoided, but that very well may not apply anymore.

Considering it's built that way at the factory, and the very same engine has been successfully fitted to larger vehicles for a while in other markets, it doesn't seem to be so much of a matter of concern. Besides one engine which was plagued by a defective phasing sensor, the overwhelming majority of 1.0TSI engine failures I'm aware of had been on engines with aftermarket mods.


Did you notice some engines with a smaller unitary displacement at each cylinder have a higher compression ratio than comparable engines with a larger displacement?