DSG or Manual

[swamp2]

Originally Posted by esquire

measurements on the VW DSG come in at .03-.04 seconds per shift. and while the M-DCT shift times might be substantially shorter than a manual throw, characterizing its throw time as "effectively zero" is an overstatement. assuming a good manual throw spans .3 seconds, the DSG/DCT shift times would be 10-13% of manual throw times... still a substantial percentage.

All claims I can find put VW DSG shift times at 8 ms! It does sound a bit too good to be true. What is your source showing .03-.04 s? Compared to 1/4 s for MT shift times that is a factor of 30! Or worst case scenario if you are right, still a factor of 6! Furthermore in 0-60 times that are around 5 seconds I would certainly call 8 ms close enough to zero to call it "effectively zero". Even worst case scenario .04 s is way less than a single tenth of a second which is about the finest scale reported for 0-60 times and almost down to a margin of mesurement error.

In conclusion I would say either 8 or 40 ms is "effectively zero" in this context.

[swamp2]

Originally Posted by southlight

As much as I love manual that can't be true. I did read the article and it's quite confusing and irritating. So I remembered an elder article in german "autozeitung" about the benefits of the "new" SMG III. That article did also show a chart comparing SMG II and SMG III. I tried to redo such a chart. If I do understand that right (original chart didn't have any shift times) the 65ms BMW states for the quickest shift are indeed for the entire shift! Don't know if anything went wrong on the test you mentioned!?

Best regards, south

The method used in the test I referenced was to look at the onset of deceleration at the begining of the shift then the onset of acceleration after the shift has been completed (in a speed vs. time curve). They admit a margin of error in the process and state it to be quite small, maybe .02 seconds. This estimation of error goes a long way in my book! When do auto journalists ever report error estimates (scientists almost ALWAYS do!). Of course the SAME test method was used for both SMGIII and for MT. Seems like a reasonable method to me. I get your point about SMGIII > SMGII and I buy that, however, 80 ms is close enough to 65 ms to me to call it mostly "a wash". However 80 or 65 ms vs. 1/4 s is in no way a wash. Without further contradictory test evidence I am inclined to believe the numbers in this test. They are further supported by better times to speed involving multiple shifts. If SMGIII really saved (1/4 s - 65ms) per shift shouldn't this show up in the overall times?

[esquire]

Originally Posted by swamp2

All claims I can find put VW DSG shift times at 8 ms! It does sound a bit too good to be true. What is your source showing .03-.04 s? Compared to 1/4 s for MT shift times that is a factor of 30! Or worst case scenario if you are right, still a factor of 6! Furthermore in 0-60 times that are around 5 seconds I would certainly call 8 ms close enough to zero to call it "effectively zero". Even worst case scenario .04 s is way less than a single tenth of a second which is about the finest scale reported for 0-60 times and almost down to a margin of mesurement error.

In conclusion I would say either 8 or 40 ms is "effectively zero" in this context.

source: http://car.kak.net/Article2767.html

sentence right above the picture of the navigation screen.

[paulyd]

I think some people are looking at the time savings wrong. Esquire, using your source, lets assume .03 sec shift time for DSG. And lets also assume .30 seconds for a manual shift.

0.30 - 0.03 = 0.27 shift time savings

Here's the problem, I think swamp is looking at this like every shift in say a 0-100mph run, will shave 0.27 sec off the end time. This is not the case if I am not mistaken. If a car must shift 3 times, lets say, a DSG will not defeat a manual transmission by 3 X 0.27sec (0.81sec).

The only way this would be true is if the car physically stopped during the time it takes to shift. It does not. It coasts while you are shifting. So the difference is that a MT car is coasting/decelerating for those 0.27 sec while the DSG has already completed the shift and is again accelerating for those 0.27 sec. The net result is much smaller than 0.27 sec. You don't gain 0.27 sec per shift.

I posted this earlier in this thread. I guess you would call it physics...

If everyone already knew this, I apologize.

[GregW / Oregon]

Originally Posted by chitown08

I think some people are looking at the time savings wrong. Esquire, using your source, lets assume .03 sec shift time for DSG. And lets also assume .30 seconds for a manual shift.

0.30 - 0.03 = 0.27 shift time savings

Here's the problem, I think swamp is looking at this like every shift in say a 0-100mph run, will shave 0.27 sec off the end time. This is not the case if I am not mistaken. If a car must shift 3 times, lets say, a DSG will not defeat a manual transmission by 3 X 0.27sec (0.81sec).

The only way this would be true is if the car physically stopped during the time it takes to shift. It does not. It coasts while you are shifting. So the difference is that a MT car is coasting/decelerating for those 0.27 sec while the DSG has already completed the shift and is again accelerating for those 0.27 sec. The net result is much smaller than 0.27 sec. You don't gain 0.27 sec per shift.

I posted this earlier in this thread. I guess you would call it physics...

If everyone already knew this, I apologize.

No, you're right, that's a very valid point that many might not think of. The differences are MUCH smaller than it would seem. You ar just minimizing times without additional acceleration. Probably amounts to a couple tenths of a second in the 1/4 at most.

[swamp2]

Originally Posted by chitown08

I think some people are looking at the time savings wrong. Esquire, using your source, lets assume .03 sec shift time for DSG. And lets also assume .30 seconds for a manual shift.

0.30 - 0.03 = 0.27 shift time savings

Here's the problem, I think swamp is looking at this like every shift in say a 0-100mph run, will shave 0.27 sec off the end time. This is not the case if I am not mistaken. If a car must shift 3 times, lets say, a DSG will not defeat a manual transmission by 3 X 0.27sec (0.81sec).

The only way this would be true is if the car physically stopped during the time it takes to shift. It does not. It coasts while you are shifting. So the difference is that a MT car is coasting/decelerating for those 0.27 sec while the DSG has already completed the shift and is again accelerating for those 0.27 sec. The net result is much smaller than 0.27 sec. You don't gain 0.27 sec per shift.

I posted this earlier in this thread. I guess you would call it physics...

If everyone already knew this, I apologize.

When looking at time to obtain a particular speed you do really get (theoretically, with all else equal and assuming a zero shift time with DCT) time improvements even greater than the sum of the shift times of the MT!

Here is the correct physics based reasoning. Draw a curve that represents speed vs. time. At any given short interval of a second or so the curve is contstant slope meaning rougly contstant acceleration. Makes "seat of the pants sense" right? Acceleration tapers off slowly but around any given second or so the "pull" is about constant. If the MT car has no power delivered to the wheels it will only have a constant velocity in vacuum with no rolling resistance. Because of the latter two effects when shifting a MT car it actually slightly decelerates and lowers it's speed. For the sake of argument though we can just assume it's speed is constant during the shift time. Meanwhile the DSG car enjoys an almost constant acceleration along its v vs. t curve (probably actually the average of the acceleration before and after the shift but again this is a refinement of the basic idea). Now draw a flat line representing a given speed, the time gain made by DCT to reach a given speed is indeed the time gain made for the shift! No matter what the slope of the line the times add up.

What I think you meant to say is that the distance gained by one vehicle is not equal to the speed of the vehicle at a shift multiplied by the shift time difference. This is true becuase one is more or less rolling at a constant speed but the other is accelerating. Here you need the more complicated distance = 1/2 acceleration x time x time formula.

Should be clear in the sketch below!

Real world evidence. A3 DSG faster than MT bt .2 sec 0-60. GTI DSG faster than MT by .1 sec. Also look at the chart I posted here long ago. Sure these are not the same car but look VERY equally matched in lower speed acceleration. The DSG car just keeps eating away at the MT until some high end power at the end of the graph seems to let the MT catch up quite a bit at the end.

Sure you won't see exactly .27 seconds x number of shifts in all cases. There is the whole effectiveness of the DCT launch to be concerned about as well as gearing differences between DCT/DSG and MT models. However, as a rough but accurate first approximation:

The benefit in time to speed is the number of shifts x shift time advantage.

[paulyd]

fabulous. my bad.

I was looking at it from a finish-line 'behind' perspective. After one shift, the MT would not be 0.27 sec 'behind' the DSG. (right?) As you said, this is a matter of distance. So what would a 1/4 run look like (ignoring launch or gearing differences) on a graph?

thanks for the correction and the graph... interesting.

[swamp2]

Originally Posted by chitown08

fabulous. my bad.

I was looking at it from a finish-line 'behind' perspective. After one shift, the MT would not be 0.27 sec 'behind' the DSG. (right?) As you said, this is a matter of distance. So what would a 1/4 run look like (ignoring launch or gearing differences) on a graph?

thanks for the correction and the graph... interesting.

"behind"?: Let's state it this way: if exactly after the MT finished its shift if the DCT went immediately to constant speed but the MT kept accelerating it would take the MT shift time amount of time to catch up. Again this assumes a perfect coast at constant speed during the entire shift time. Not exact but not a terrible approximation either.

In general a DCT vs MT v vs. t comparison should look like the first three gears of the Eos vs. C70 I posted here: http://www.m3post.com/forums/showpos...37&postcount=9 . Then just repeat the widening gap for each additional shift. I don't know off hand if the new M3 MT will get the 1/4 mi in 3rd or 4th and could only guess if the DCT will get it in 3rd or 4th as well (gear ratios not known yet). Either way the Eos vs. C70 should have the same general features and apperance of the M3 DCT vs. MT, with the gap widening again from the shift to 4th. Again I am going to empahasize this is counting only MT vs DCT effects and given identical launches! I am not going to go so far as to predict the DCT will consistently beat the MT by .25 x 3 or .25 x 4 seconds. But I suspect it will be somthing like that.

Remember the VW GTI with DSG doesn't get 9% better city mpg by doing something totally mundane/simple. DSG is responsible by "always" having power at the wheels.

[bernhtp]

The only meaningful time for a shift includes the entire period during which power is not being applied to the wheels. This obviously includes all clutch-in time. I'm also very skeptical of reports of near-instant shift times (e.g., 8 ms). Do people consider what happens to the engine momentum in such shifts? With six gears (5 shifts) and an 8500 RPM red line, you will lose about 3000 rpm with each gear change. With an instant shift where there is no time for the engine to spin down a bit a being closer to a rev match, the full 3000 rpm momentum of the flywhee, crank, etc. will be slammed into the rear end of the drive train. WOT shifts in S6 mode on an M5 make a head-snapping and fearsome jolt as is. It's hard to imagine what it would feel like with the shift time compressed to about zero.

Tom

[paulyd]

Originally Posted by bernhtp

Do people consider what happens to the engine momentum in such shifts? With six gears (5 shifts) and an 8500 RPM red line, you will lose about 3000 rpm with each gear change. With an instant shift where there is no time for the engine to spin down a bit a being closer to a rev match, the full 3000 rpm momentum of the flywhee, crank, etc. will be slammed into the rear end of the drive train. WOT shifts in S6 mode on an M5 make a head-snapping and fearsome jolt as is. It's hard to imagine what it would feel like with the shift time compressed to about zero.

Tom

Interesting... how is this problem sorted out?

And thanks swamp for the additional info, it seems DCT really does give you a 0.25 sec advantage per shift... cool! (and better gas mileage!)

[enigma]

Its funny, I heard the same thing about SMG back in 2001 when I ordered mine. How "real drivers" should drive manuals and how I would be bored with it.

5+ yeas later I still love it. When I take the Elise to the track, SMG is one of two things I miss from the BMW.

Here is the long acceleration graph for one of my better shifts in the Elise at Laguna Seca. Stock 6spd gearbox.



I have never had the logger in the M3 at the track but it consistently quicker. I doubt its a huge advantage comparing best to best, but I cannot consistently do that shift every single time. SMG can.

DSG will be even quicker and never dip into the negative acceleration.

[JEllis]

Yeah DSG is going to be faster. My concern is that the drivers experience will be lost a little with the shift to an auto tranny. I may get the 6spd just to keep the driver experience a little more intact. I guess I can give up .2 sec

Jason

[swamp2]

Originally Posted by enigma

Its funny, I heard the same thing about SMG back in 2001 when I ordered mine. How "real drivers" should drive manuals and how I would be bored with it.

5+ yeas later I still love it. When I take the Elise to the track, SMG is one of two things I miss from the BMW.

Here is the long acceleration graph for one of my better shifts in the Elise at Laguna Seca. Stock 6spd gearbox.



I have never had the logger in the M3 at the track but it consistently quicker. I doubt its a huge advantage comparing best to best, but I cannot consistently do that shift every single time. SMG can.

DSG will be even quicker and never dip into the negative acceleration.

Just a hair under .25 seconds. Nice. Do you feel like the Eilise has a tranny where you can really get a minimum speed shift out of it? I sure would LOVE to see the long. accel in your SGM M3 in S6! Next time hopefully. Thanks!

[swamp2]

Originally Posted by bernhtp

The only meaningful time for a shift includes the entire period during which power is not being applied to the wheels. This obviously includes all clutch-in time. I'm also very skeptical of reports of near-instant shift times (e.g., 8 ms). Do people consider what happens to the engine momentum in such shifts? With six gears (5 shifts) and an 8500 RPM red line, you will lose about 3000 rpm with each gear change. With an instant shift where there is no time for the engine to spin down a bit a being closer to a rev match, the full 3000 rpm momentum of the flywhee, crank, etc. will be slammed into the rear end of the drive train. WOT shifts in S6 mode on an M5 make a head-snapping and fearsome jolt as is. It's hard to imagine what it would feel like with the shift time compressed to about zero.

Tom

I think I am a bit skeptical of the 8ms (VW DSG) figure as well. However, I do think it will substantially faster than SMG. The jolt from shifting will also be less. Here's why. In MT or SMG as well the clutch must totally disengage, then the gear is disengaged, the next gear is engaged and finally the clutch regages. Four basic steps, sure there is some overlap between the processes but not that much. With DCT there is only one step! Why? For one the next gear is already engaged and two there are two clutches acting simultaneously. As one opens the other simultaneously closes, that is THE only step. I'm not sure how much time in an SMG shift is spent clutching vs. moving the gears but assume each part with SMG takes roughly an equal amount of time 60ms/4 = 15ms. If the clutch travel and actuation speed can be further optimized by it's design (remember SMG is a retrofitted manual and std. clutch) you can then shave a bit more off of that figure. 8ms, maybe, maybe not.

Smoothness: The reason MT and SMG are jerky and violent is because the car undergoes a dramatic series of accelerations and decelerations during a shift. Here is the sequence: accelerating hard at WOT, clutch then causes deceleration from wind/rolling resistance, then when clutch is engaged and full power is reapplied you go to large positive acceleration, finally settling in to an acceleration less than where you were in the previous gear at WOT. When this all happens fast you really feel it. See enigma's graph. With DCT the second clutch is engaging as the first is disengaging - automatically smoothing things out. Also there is probably some very short duration throttle/power management by the DCT computer to further reduce the clutch stress. The clutches certainly would not last if every shift was like a launch with a large rpm drop.

BTW an optimal 1-2 shift is likely to drop about 3500 rpm, each successive shift will drop less and less ending at around 1000 drop from 5->6.

[joema]

Originally Posted by JEllis

My concern is that the drivers experience will be lost a little with the shift to an auto tranny...

Yet Formula One drivers use both SMG-type transmissions and DSG-type. They're the best drivers in the world, but they have no problems using a paddle-shifted single or dual-clutch transmission. Supposedly the DSG-like "seamless transmission" gives about 0.4 seconds PER LAP advantage over the SMG-type F1 cars.

As a street or race driver you have finite bandwidth. Instead of focusing on heel-and-toe rev-match downshifts, if SMG/DSG does that for you, it clears up mental processing for other things.

I think the main complaint (among enthusiasts) may stem from equating paddle-shifted SMG/DSG with sluggish Tiptronic or similar torque-converter automatics.

If the paddle-shifted transmission does instantly what I command, no more and no less, I'd be happy.

[gmund1948]

MMAM

Make mine a manual

[joema]

Originally Posted by enigma

..Here is the long acceleration graph for one of my better shifts in the Elise at Laguna Seca....DSG will be even quicker and never dip into the negative acceleration.

Your graph shows approx 400 milliseconds before acceleration reaches the previous level, about 125 mS of which was deceleration.

The effective "shift time penalty" might be about 250 mS (just a guess). And that's on an Elise with a rifle-bolt-like shifter. Everybody thinks they can shift fast, but we must accept the actual data.

Let's compare this to SMG/DSG. If an E90 M3 can do 0-60 (26.8 m/s) in about 4 sec, that's an average acceleration of 6.7 meters/sec/sec.

A 250 mS manual shift during that period would cost you about 6.7 * .25 or 1.68 m/s -- that's 3.75 miles per hour penalty from a single shift !!!

By contrast SMG would cost about 6.7 * .08 = .5 m/s, or 1.1 mph. DSG would cost about 6.7 * .008 = 0.05 m/s or 0.1 mph.

At higher speeds, acceleration would be less but air drag greater, causing more deceleration during the non-powered shift interval.

This simple analysis shows why Formula One is so interested in DSG-style transmissions, even they they already use SMG-style. In a performance-critical racing application each DSG up shift gives an advantage. The cumulative advantage over a race is significant.

For street use I'd think the driveability advantages of DSG are probably more important.

Has anybody ever done a side-by-side ("heads up") drag race between two e46 M3s, one with manual the other with SMG?

[M2002]

Originally Posted by joema

Has anybody ever done a side-by-side ("heads up") drag race between two e46 M3s, one with manual the other with SMG?

I would like to see the results of that race as well.

[swamp2]

Originally Posted by M2002

I would like to see the results of that race as well.

Look at the link in my post #111.

[JEllis]

Quote:
Originally Posted by JEllis
My concern is that the drivers experience will be lost a little with the shift to an auto tranny...
Yet Formula One drivers use both SMG-type transmissions and DSG-type. They're the best drivers in the world, but they have no problems using a paddle-shifted single or dual-clutch transmission. Supposedly the DSG-like "seamless transmission" gives about 0.4 seconds PER LAP advantage over the SMG-type F1 cars.

As a street or race driver you have finite bandwidth. Instead of focusing on heel-and-toe rev-match downshifts, if SMG/DSG does that for you, it clears up mental processing for other things.

I think the main complaint (among enthusiasts) may stem from equating paddle-shifted SMG/DSG with sluggish Tiptronic or similar torque-converter automatics.

If the paddle-shifted transmission does instantly what I command, no more and no less, I'd be happy.

What I was trying to say has nothing to do with performance. To me the standard 6MT will be more driver involving than the M-DCT tranny. I do suspect the DCT tranny will provide a better performing car but will also make the driver feel further detached from the traditional M driving experience. Of course this is all subjective.

[joema]

Originally Posted by swamp2

Look at the link in my post #111.

OK, thanks, missed that. Facinating article, I now see your point about true SMG shift time not being dramatically better than manual. There may indeed be a difference between the "gear change" portion of an SMG shift (say, 80 mS), vs the entire shift from throttle off to throttle on. The later would also include the clutch and de-clutch operations, and from the above ref'd article is apparently closer to 250 milliseconds, only marginally faster than an expert human driver.

That was SMG II, which I understand is essentially a bolt-on electrohydraulic mechanism to the 6-speed manual. I think SMG III is custom-designed from the outset as a clutchless manual. Does anyone know if shifts any faster?

However it's a moot point if BMW is going to M-DCT.

[ruff]

I read that Porche is having trouble perfecting a smooth DSG like transmission for their high reving engines and thus have not incorporated it into their cars yet. I have also read the dual clutch transmission in the Audi R8 has some kinks in it as well. It must be an engineering enigma to figure out how to make a smooth dual clutch gearbox in automatic mode while still having it run full bore in manual mode. Hopefully, the M boys are on to something.