DZR & TDI: New Performance Tests & Results

Hi everyone,

As promised, here's a new measurement series with a higher sampling rate (measurement interval of approximately 0.45 seconds), and most importantly, measurements taken both forwards and backwards on the same track, with multiple repetitions. A total of nine measurements were taken, which I have also semi-automatically analyzed, and which I would of course like to share with you. The raw data can be found, as always, in http://www.informatik.hu-berlin.de/~richling/tmp/.

The actual data file is LOG-01-011-xxx-xxx.CSV; the other CSV files and the files labeled "versuch" are from a previous run. The serie2.xls file contains the calculations for the residence times (a simple proportional calculation is used to determine the times corresponding to 2000 and 4000, respectively), and the remaining files are images, which I will also include below in PS and PNG format. Explanations will follow shortly.

In the original CSV file, the first column contains "BRANDS," which makes it easier to find the acceleration tests, and also provides additional information.

The test setup was as follows:

Ibiza with ASV (Anti-Slip System), unladen weight 1240 kg. With driver (60 kg) and fuel tank 2/3 full, the actual weight is likely around 1230 kg, possibly a bit more due to the trailer hitch. Air temperature: 20 degrees Celsius. Coolant temperature: 91 degrees Celsius. Driven for a very long time to warm up. Mileage: 10300 km. AGR (Active Grille Roller) is active. No tuning modifications.
My track is about 2 km long, with a slight curve in the middle, and I made one attempt before and one after the curve. I'll call the first section "1" and the second section "2". The driving directions are a (roughly westward) and b (opposite). If there was wind (subjectively, there wasn't), then all the "a" attempts would experience it the same way, and all the "b" attempts would experience it in the opposite direction. Therefore, the names of the attempts are constructed as follows:

Direction number.

So, "1b2" represents the second attempt on track 1, heading east, etc.

The correspondence to the markers in the file is as follows (number: name - calculated time):

The times were determined using a proportion based on data from the periods around 2000 and 4000, and then subtracting the results, as shown in the file "serie2.xls".

1: Test marker, no meaning.
2: 1a1 - 6.182
3: 2a1 - 6,191
4: 2b1 - 6,144
5: 1b1 - 6,073
6: 1a2 - 6,090
7: 2a2 - 6.238
8: 2b2 - 6,276
9: 1b2 - 6,107
10: 1b3 - 6,144 (I quickly added this one to the remaining section of track 1 because I was worried that I might have lifted off the accelerator too early at 1b2, which turned out to be a mistake. So, 1b2 is also valid.)

As you can see, the variation is VERY small, which speaks well of both the measurement method and suggests that the subjective impression that the tracks were level and windless seems to be correct. Otherwise, such discrepancies would be noticeable when attempting the course in the opposite direction. However, there is nowhere where either option a or b is significantly worse.

@Ulf: This time, you can clearly see how the boost pressure behaves, and just like you saw before 2000, the overshoots are also clearly visible.

If anyone has any further questions about the car, the tests, or the files, please feel free to ask. Perhaps these measurement results can help unlock the mystery surrounding the increased power and the torque curve.

Here are 9 more diagrams, each showing the engine speed (rpm) and boost pressure (ld) of the tests. The boost pressure is absolute, as reported by VAGCOM.

The order of the diagrams is as shown above, following the order of the measurements (I drove the entire route back and forth twice, which is why the order is like this).

The curves in the diagrams also have corresponding names, so if I made a typo in the link, it will be apparent from that.

1a1:

2a1:

2b1:

1b1:

1a2:

2a2:

2b2:

1b2:

1b3:


Okay, that's all for this set of measurements for now. Anyone who wants to do some calculations with it, please feel free, and anyone who wants to use it for further analysis, please feel free as well.

@Rainer: Of course, you are free to use this data for the TDI website as you see fit.

Best regards,

Jan.

Hi again,

a small addition:

I kept the accelerator fully pressed in the 2b2 until the car's speed no longer increased (beginning of the red zone in third gear). You can see in the diagram how the turbo boost is reduced in that situation.

The upward spike at the end of the first graph (i.e., beyond 4000) is due to my somewhat sluggish Knarr clutch (the workshop might need to take a look again on Monday), as I wasn't quick enough with upshifting. It has no further significance.

Best regards,

Jan.

Hello Jan,

Okay, I'll get back to you about that when I have a chance!

Best regards, Rainer.

Hi Jan,

I don't have time to delve into that right now.

What immediately strikes me as interesting is the "documented" relatively high deviation of 0.13 seconds between the two attempts on section 2b.

Without having seen the original files: Could this be due to unfavorable timing points? Or, how did you calculate the 2000-4000 times to the hundredth of a second?

Otherwise, TDIs seem to have not only temperature-dependent, but also "spontaneous" strong and weak phases. . .

Hi Ulf,

Regarding the calculation: The measurement points are approximately 0.45 seconds apart, so I performed the following calculation:

Time for 2000 = A2 + ((A3 - A2) * (2000 - B2)) / (B3 - B2)

where:
A2: Recently, RPM was less than 2000.
A3: Initial period with RPM > 2000.
B2: RPM at A2.
B3: RPM at A3.

For 4000, then the analog value, and then the difference. All these calculations, using the numbers extracted from the data files, are located in the file "serie2.xls".

The assumption is that the acceleration within this interval is sufficiently linear to be calculated using the rule of three, according to the formula above. The actual measured values are provided by VAG-COM with an accuracy of 1/100.

Here's how it looks for the two 2b attempts:

RPM at 2000: 1869 and 2037, at 4000: 3948 and 4074.
The second set of numbers is: 1890, 2058, 3927, and 4053.

It's possible that the inaccuracies originate from this source, but I don't think it would explain the 0.13 seconds; that's already a quarter of the sampling rate.

My suspicion leans in a different direction: With one exception (the first measurement), the attempts in the second series are slower than those in the first series, which could suggest that slight thermal effects might be playing a role – such as the warming of the charge pressure hoses, etc. And the first test of the second series was preceded by the turnaround, during which the system could have cooled down somewhat. The turnaround between the series took longer than the transition from a to b in the same direction, so that could also explain it.

Incidentally, after the tests, the coolant temperature was 1.5 degrees higher than before, and the intake air temperature at idle (approximately 3 minutes and 1 km of slow driving after the last test) was 36 degrees. When measured on Monday, with a 2-degree higher outside temperature after 10 minutes of leisurely city driving, it was 29 degrees.

It's possible that the warmer components in the second series are responsible for this effect.

What do you think?

Best regards,

Jan.

Hi Jan,

I also used interpolation, as I understand it, to determine the time at 2000-4000 rpm, and I followed the same procedure in the results.

I definitely think the thermal effects are plausible.
However, in my experience, these factors hardly play a role: Outliers of more than 0.1 seconds are usually noticeable even when simply pressing the stopwatch button.
(As musicians would say, "missing the entry despite counting the beat"), and my carefully taken measurements on the same day consistently differ by less than 0.1 seconds.

Okay, there are approximately 2 minutes between my measurements (always in the same direction, because there's a junction at one end of the measuring section that I don't want to approach at 100 km/h), but I can see from the oil pressure that the engine is warmer afterwards than it was before.

On the other hand, the latest measurements might be favored by a "cumulative purging" of the engine compared to the initial runs, which could offset the thermal disadvantage (at least partially).

Hi Ulf,

Yep, that's called interpolation.

"Then it is indeed strange, because the measurement conditions were very consistent in my tests. At most, the starting point could be a few dozen meters apart, which could amplify the influence of any slight unevenness. Also, in the second 2b test, I had to make a small detour around a branch lying in the street (we had severe weather chaos here last night), but that shouldn't have had much of a measurable effect, as it was only a lateral offset of half a meter. It's probably more likely that the difference in starting points had an effect (it's not always possible to have them be exactly the same, you also have to consider other traffic)."

The stop/start technology itself should function with absolute precision, as VAG-COM reads the data directly from the engine control unit.

The possibility of a speed limit override exists, of course, but I doubt it. I drove about 30 km to the measuring point, of which about 15 km were on the highway. Although the speed was limited, I was still going 150 km/h at times, especially when accelerating, and I certainly wasn't being gentle with the accelerator pedal.

Wind could be another influencing factor that could potentially affect times by fractions of a second if a gust happens to interfere. It wasn't really windy (there was nothing noticeable when getting out), but it could still cause an asymmetrical distortion.

Well... overall, I think that the average values of the nine measurements should provide a reasonably accurate time for a standard ASV (which was the whole point of the exercise).

I might try to overlay the diagrams later, when I have more time, so that they share a common point at 2000. This would make the variations even more apparent.

Does anyone know where VAG-COM gets the timing information? Does the engine control unit provide this data, or does the computer calculate it itself based on system time signals? Because if it's the latter (the latter case), inaccuracies in the range of 0.1 seconds could indeed be explained in this way, since Windows 2000 is not a real-time operating system (and 9x even less so).

Best regards,

Jan.