View Single Post
      08-23-2008, 06:36 AM   #1
Lieutenant General
swamp2's Avatar

Drives: E92 M3
Join Date: Sep 2006
Location: San Diego, CA USA

iTrader: (1)

PRELIMINARY M-DCT shift times - results and data

I have done some preliminary measurements of the M-DCT shift times. The numbers are not too much of a surprise and are quite consistent with my previous estimates and rumored and quoted values for other DC vehicles. However, the differences between modes with and without surge are quite surprising.

For those that don't want to read the remaining details nor actually look at the data the shift times likely vary from about 30 to 200 ms.

Thanks again to the members who helped fund the purchase of the equipment to accomplish these measurements, T-Bone, lucid, TLud and OC Kid. The thread that proposed this is located here. The equipment used is listed here as well. We would still like to test many more cars listed on that thread, if you or someone you know would volunteer their car (preferably in the Southern California or Boston area) please PM me. Of course the process to get better, cleaner and more accurate data will be refined before moving forward with a lot more testing.

Since these measurements involve ONLY the time for the actuation of the clutches they DO NOT include the delay which has already been noted in S modes between the actuation of the paddle/stick and the beginning of the clutch actuation (and hence any measureable changes in acceleration). So until that can be better instrumented and tested these are shift times in D modes or partial shift time NOT INCLUDING any shift lag.

We know already that not all gears, all modes and all rpms produce equal shift times and these are the approximate measured range. When including the 1->2 shift in LC mode and the 6->7 shift I would expect that range to widen (just an educated guess here to about 20 to 400 ms. The former figure being maybe D2 6-7th at low throttle and the latter for 1-2 under LC as it really slurs this shift, while not sacrificing acceleration.

  • This raw data clearly needs better filtering or averaging due to the fairly high signal to noise ratio.
  • Contributing factors to the signal to noise ratio include:
    • A crude preliminary zip tie mounting of the accelerometer on the left upper door hinge bracket (I could indeed get my door shut without damaging the accel nor its cable ). Proper mounting will require the fabrication of a small aluminum bracket that can be bolted to a seat rail or some other equivalently rigid part of the vehicle.
    • A prior overload of the input voltage to the accel due to an error on documentation included with it.
  • Deciding exactly where a shift begins and ends is subject to a great deal of interpretation especially with the preliminary nature of the raw data. I would estimate about a 25% error is possible here.
  • Data thus far has only been taken for shifts 1->2, 2->3 and 3->4 at medium to full throttle. Mostly to avoid ridiculous speeds on public roads. The aforementioned high signal to noise at this point prevents measurement of shift times at low throttle.
  • Although I feel the shift times are fairly consistent across all of the shifts I observed in the complete data and there has been no "cherry picking" of results to provide better or lower times, results shown below absolutely have been "cherry picked" to provide the most clarity of the distinction of the shift itself amongst the fairly noisy data. There are some shifts where you just can not see anything useful.

Data and results:

First graph

Mode: D3
Power: Normal
Throttle (qualitative): medium to high, close to WOT
Shift: 3->4
Data sampling rate: 300 Hz
Vertical axis: acceleration in g's
Shift time: 32 +/- 5 ms

Second graph

Mode: S4 (recall this means the transmission will exhibit surge or more precisely stated a noticeable jerk)
Power: Normal
Throttle (qualitative): medium to high, close to WOT
Shift 2->3
Data sampling rate: 500 Hz
Vertical axis: acceleration in g's
Shift time: 200 +/- 50 ms


The 30 or so ms observed in D modes is consistent with figures discussed for VAG DSG. Many sources quoted 8 ms but a more credible source provided a very similar figure. 8 ms has always seemed way too low given it is still a mechnical and hydraulic system. There are about 10 individual sample points across the duration of the shift in the first graph. The acceleration profile shows the steady state acceleration drop as the first clutch is disengaged very quickly followed by an equal duration surge of a higher magnitude as the second clutch engages.

The S4 shift time is a big surprise. They certainly feel faster than the S3 shifts but it is psychological. The mathematical surge/jerk in S3 is actually much greater (a large jump in g force over a shorter duration) compared to the more plateau like positive and lengthy acceleration profile in S4. I think what is happening is that the jerk in S3 is simply so short we just don't feel it, whereas the much more lengthy profile, all at an increased acceleration compared the the steady state, both before and after the shift provides this jerk/surge/"clunk" that we know so well from modes S4->S6.

A quick calculation of the extra speed that this surge provides compared to the wash in D3 is given by:

ΔV = a x t = 3/8 g x .15 s ≈ 0.6 m/s ≈ 1.2 mph

This is the relative gain comparing the mph after the 2->3 shift in S4 vs. D3. So again footie, my friend, S4->S6 with surge does not shift faster than D modes but it is faster. The former result I believe is quite novel but the later one was what I and others have been saying all along.
Attached Images