Stupid question of the day: What did Volkswagen actually do with its emissions control software?

The news media is full of sound and fury regarding something that Volkswagen did with diesel engine emissions control software, e.g., “VW Is Said to Cheat on Diesel Emissions; U.S. to Order Big Recall” (nytimes). The journalists either can’t or don’t bother to explain what Volkswagen actually did.

Readers: Can you help explain this? What does the software in question control? The Times says “The Environmental Protection Agency accused the German automaker of using software to detect when the car is undergoing its periodic state emissions testing.” What does this mean? Only when something is plugged into a connector under the hood? Then “During normal driving situations, the controls are turned off, allowing the cars to spew as much as 40 times as much pollution as allowed under the Clean Air Act, the E.P.A. said.” Can this be true? What is there in a VW that can cut or increase pollution by 40X?



22 thoughts on “Stupid question of the day: What did Volkswagen actually do with its emissions control software?

  1. see EPA letter link I worked at a company that supplied microcontrollers to a major US auto manufacturer that years ago did the something similar with gasoline engines, using an 8051(!). Emissions and fuel economy tests are artificial driving scenarios and follow a particular pattern. I am not sure with diesel engines what magic they did but for gasoline engines you can control the timing, lean out the air fuel ratio for fuel economy and then set it to the optimum (stoichiometric ) ratio for the emissions test. With diesel I think they were concerned with the by-products of diesel combustion. I seem to recall that VW advertised their diesel engines as not requiring a catalyst or urea exhaust treatment additive like Mercedes does (bluetec).

  2. It sounds like it carefully monitored steering wheel position, throttle position, barometric pressure, and temperature, and compared them against the known inputs during the test procedure.

  3. And as for what they changed, most of the VW engines used during 2009-2014 did not have urea injection, which is what most German manufacturers use to reduce NOx emissions. Instead, what they likely changed was the amount of EGR (exhaust gas recirculation) used, possibly the timing of the fuel injection event, and possibly bypassing the DPF (Diesel particulate filter).

    The net effect of having these on all the time would be slightly worse fuel economy, measurably lower engine power, and probably worse engine longevity.

    For the 2015 engines, they did use urea injection, so the effect of these changes will be increased urea consumption, which is very expensive to replenish.

  4. The “40x as much POLLUTION” is just reportorial sensationalism or sloppiness. In the case of one particular pollutant – nitrous oxides , the bypass mechanism used would cause NOx to increase by a factor of between 10 to 40x (assuming the EPA itself is not exaggerating, which they probably are). There are 7 major categories of diesel pollutant (Hydrocarbons, carbon monoxide, particulates, etc.) regulated and this is one of the 7 – the other 6 were not affected. The Federally regulated level is 1/5th of 1 gram of NOx per mile so a 10 to 40 x increase would mean between 2 and 8 grams emitted per mile. 2 grams is about the weight of a US dime coin. In case you have not noticed, the world has not ended because a couple of hundred thousand VWs have been running around emitting slightly more than permitted by the EPA’s ever more stringent rules. What VW did was wrong but sensational press reports don’t help us to understand the nature of the violation, they just reinforce the cartoon “corporations are evil” meme. A free press is supposed to educate us, not inflame us.

  5. I ran a diesel engine test cell for about a year and a half while I worked on the software to control the dyno to perform the standardized EPA emissions tests. The main one is the Federal Transient Protocol, or FTP, test. As an aside, the test procedure is, from a theoretical point of view, extremely well written and thorough. I was impressed that it was the result of a government body. No matter the targets of the emissions standards, the test itself, and it’s requirements for results, data gathering, and statistical correlation, are a fine piece of engineering.

    Joshua says the VW engine years in question didn’t likely have an SCR, but the link Jay C. provided says specifically that this was the device being defeated. An SCR is a catalytic converter, aided by the “dosing” of urea, sprayed on the surface of the substrate as a fluid. The ECM noticed when the engine was being subjected to a standardized test (which are exactly prescribed for throttle and load, and thus easy for a computer to detect), flipped “the switch” (from the EPA letter), and turned the “doser” back on. I don’t understand the comment about the REASON for this being that urea is expensive. It only costs a few bucks a gallon, and you only refill the tank (for a car) once a year or so.

    So I know WHAT was done, but I don’t understand WHY. What’s VW’s strategy in designing and making people pay for an SCR, and then defeating it in everyday use? It’s not expensive to maintain. Did they have a known problem with wear on the dosing system? Did they have sensors that were getting clogged with urea, and didn’t want to redesign? There’s got to be a reason to risk the wrath of the USA EPA. Hope it was a good one, because they’re going to feel it now.

  6. Izzy, good point. I meant to get to that in my note, and forgot about it. I’m also certain that the EPA is inflating the number, but something everyone needs to realize is that the standards that have been set by the EPA are SO STRINGENT now, that ANY increase will look relatively large. Look at the reduction in allowable diesel NOx emissions over the past 20 years:

  7. David,

    My understanding is that VW had been insisting for a while that their particulate filters and permanent catalyst catalytic converters were sufficient to meet emissions standards, but that they added one for 2015 model years due to tightening regulations.

    Having not owned a TDI engine, I can’t confirm firsthand, but several VW Diesel owners on the internet have said that 2014 and earlier cars do not have an AdBlue tank, but 2015 and later cars do.

    As an aside, the only reason this is coming out now is that the EPA was threatening to not certify the 2016 cars for sale in the USA at all until VW could explain the discrepancies noticed. VW had apparently been stonewalling, offering explanations based on differences in road and test conditions, but the EPA and CARB didn’t accept them as adequately explaining the magnitude of the difference.

  8. David K – Joshua is correct – VW diesels prior to 2015 did NOT have urea injection. They were also notorious (and this is unusual) for getting BETTER real world mileage than their EPA rating (probably because of the defeat mechanism – without urea injection they must have used some mode to pass tests that reduced both NOx and efficiency. Turn off the mode and both NOx and mileage go up). Better mileage would lead to lower CO emissions so they may be polluting LESS than other cars, just not the EPA approved mix of pollutants.

    Now VW is in a pickle. Since these cars were built without urea injection they would have to retrofit at enormous expense (or else the firmware fix will permanently lock these cars in “test mode” where they run poorly – that’s why they engineered the workaround in the 1st place.. And if you are a current owner, you would have to be nuts to bring your car in for a recall that will reduce its mileage and performance (even if it is “better” for the environment). Or maybe there will be a class action seeking compensation for same.

  9. The EPA letter says the devices being bypassed were “specifically the selective catalytic reduction or the lean NOx trap”. I defaulted to what I was familiar with: SCR’s. From this article, I see what’s going on.

    VW said they had “advanced” combustion techniques in their ECM’s, and those cars didn’t NEED an SCR. (Which is stupid, because there’s no practical, realistic way to meet the standards without one these days, as we’ve just seen.) Well, their ECM’s were, indeed, programmed not to need one, so +1.

    What I’m still confused about is how the EPA is saying that the ECM would enable and defeat the devices when apparently the SCR’s just don’t exist on those cars at all. NOx absorbers (“lean NOx traps”) are expensive, but the wikipedia article on them says that VW was to bring them to market c. 2008, so maybe that’s what’s on those vehicles? But they’re even more expensive than SCR’s. Curiouser and curiouser.

  10. David,

    I’m guessing the NOx trap they installed has much lower capacity than the car needs, making it cheaper. They’d be relying on the fact that it wouldn’t be used most of the time. They also clearly had to have the on-board diagnostics aware of the fact that it was being bypassed and not throw a code.

    It’s going to be very difficult for VW to argue that this wasn’t a decision made very high-up, given the number of people who would need to be aware of it for this to make it to production vehicles for so long.

  11. Izzie,

    Perhaps the “40x pollution” versus “40x of one pollutant” was a slight exaggeration, but this sounds like pretty egregious cheating and should be justly censured. I say this not from a “corporations are evil” standpoint, but for someone who has seen first-hand the expense that other engine manufacturers go through to meet ever-stringent emissions targets.

    In a competitive market, cheating can be a competitive advantage if it is condoned. If cheaters are not shamed, then honest companies will be driven out of business and honest engineers will be unable to feed their families.

    Eventually, this tends toward a culture where many people put more effort into cheating than actually doing their work:×350/2290895036e2ab1a59aedfe917d362e2/indiacheatingap408013340760.jpg

  12. I don’t disagree that what VW did was wrong and should be censured. But sensationalized and incomplete press accounts are not helpful to public discourse in a democracy (though they are great for dictatorships as long as they follow the party line). As we have seen with the Ahmed case, when it comes to technical issues, our press has fallen to such a low level that it’s often nearly impossible to learn what is actually going on from just reading the general press accounts. If you read between the lines and consult other sources (often other bloggers) you can piece it together, but what you see in the mainstream press bears only a vague relationship to the underlying facts and is utterly lacking in detail (or else is just plain wrong – Ahmed – Inventor of the Digital Clock!). If reporters were just plain stupid or lazy then the errors would be random, but what they do involves combining stupidity with ideological bias .

    Headline writing has always been about attention grabbing (and nowadays we have “clickbait”) so go ahead and make “VWs Emit 40 time as much pollution!” as the headline, but at least tell us somewhere later in the story what is actually going on. In the old days you would put the “boring” stuff near the end so that readers who just wanted to skip that part could. But now they just skip the details altogether and it often appears that the reporter lacks enough technical understanding of what they are writing about to give the details in a comprehensible way even if they wanted to, which they don’t.

  13. Fair enough about sensational and inaccurate reporting.

    And for the record, I haven’t spent the time to decide whether what VW actually did falls into the category of “overt cheating” or “partially inadvertent curve-fitting”.

  14. Why is urea expensive to replenish? Did the Haber Process stop working? Also, I agree that it’s interesting to think about the tests, if you know that they’re going to be in a certain performance range, wouldn’t you make sure that the various systems were optimal within that range? (Isn’t that one of the operating theory of hybrid vehicles, that they allow the hydrocarbon-based power apparatus to operate in a more optimal range by saving the excess and supplementing the peaks?)

  15. Taking the comments about reporting under advisement, the best set of facts we have about this is the EPA’s letter, and the premise from that document is that code in the ECM specifically flipped emissions-controlling equipment when the FTP test was being run against the engine. The test calls for a very specific curve of engine speed, load, and throttle (all by percentages of top-rated spec), and would be detectable within a couple seconds. The “bad” results from the first couple seconds would wash out in the test’s 20 minute cycle and the statistical validation. Someone had to have coded that. That’s intentional. Clever, but intentional. If the EPA’s letter is functionally correct — and, given what I’ve seen from the FTP test documentation, I get the impression that they’re not the type of people prone to politicization in something like that, but I could be naive — this is definitely “overt cheating.”

  16. The Federal Transient Protocol test is 20 minutes of shifting speed, load, and throttle all over the place. Most of the test changes all 3 within a second or two. As I alluded to earlier, it’s a neat test. Allegedly, the EPA drove trucks through the mountains, recorded what happened, and then applied a reduction to the data to keep only the interesting bits. It’s like when you apply a load cycle in a fatigue test. You drive a vehicle around a track, record what happens, then throw out all the “boring” data, and keep only the potentially damaging events, then run that reduced data set on your endurance rig. Anyway, my point is that the ECM wouldn’t “optimize” it’s results based on special test conditions because they specifically are NOT in a “certain performance range” of normal driving. They are very precisely NOT normal driving. In fact, it would be impossible for a human being to control the engine to that test curve on the road.

  17. Maybe Volkswagen should replace the ECU with a unit closer (perhaps even multi-core) closer to one that can be run with a Linux kernel? I mean this is a huge opportunity for the incoming CEO to make a significant change. Not saying a regular CPU with a full blown kernel is necessarily the safest or wisest thing to do but making a change *in that direction* could be worthwhile.

  18. Excuse my English. I am french. What I think is that Volkswagen software increase fuel to air ration during epa testing in order to diminish nox. By doing so. the temperature of the combustion chamber decrease so is nox.
    Nox is produced from high temperature. It also explain why the majority of Volks owners have better mpg than EPA reading

Comments are closed.