posted 01-07-2004 06:10 PM         

Okay, there I was, adjusting my gears while researching an idea for a future race; so for once I'm actually paying
some attention to the relative gear settings. First up, I use the standard method known as the "tranny trick"
[note 1]; I drag the 'Final' (henceforth 'F') all the way to the top, zip the auto-set Level (henceforth 'L') to
the bottom, and then drag F all the way back to the minimum. So far so good, but I'm a little surprised when the
car still makes mid-range revs in 6th on the straight, because I somehow expected it to run out of puff before it
got to sixth. Whatever. So I note my times and go back to the start to try another setting - but I notice that the
new setting, which should have been very similar to the original, now has sixth gear almost entirely off the
graph!

So; everyone knows that when you ask someone for their setup, the only bit that never works out is their gear
settings. The effect of setting L depends on what F was set to before you changed L, and often the settings
described by the originator (eg. ripped out by the MK's program [note 2]) do not include this detail. But even
when the details are all there, it often doesn't work out; and generally you say something like: musta dunnit
wrong; or FUAK (faulty user at keyboard); or that idiot gave me crap info.

But the secret of true science is in seeing the anomaly, in seeing that there is an unanswered question to be
studied. Newton wasn't the first guy to notice an apple fall off a tree ... he was just the first to realise that
there was a question attached to it!

Still reading? Then join me on a journey to the land of scientific investigation...

To avoid giving too much away, let's start afresh with a shiny new Daihatsu Cuore TR XX Avanzato R from the
showroom [11140 cr] and a brand new full racing tranny from the parts shop [10000]. Now follow me to "Run &
Settings".

The default setting on the tranny looks like:
 
	Final	Level	1st	2nd	3rd	4th	5th	6th
	-----	-----	---	---	---	---	---	---
	5.071	18	2.642	1.927	1.476	1.178	0.981	0.852	(1)
 
I've labelled these ratios (1) for future reference.
 
Next I set F to its maximum value, 5.999. I chose to set this exactly, in case there was a difference between 
stopping on the exact number as opposed to trying to run past the exact number but hitting the limiter on the F 
adjuster; however, that proved not to be relevant, so you can set it whichever way you like - use the R1 button 
rather than the D-pad right, as it is goes up in tens (ie. hundredths) then.
 
	5.999 (max)
 
Now I adjusted L down 1 to 17. 
 
		17	2.503	1.860	1.446	1.132	0.991	0.789	(2)
 
I'm gonna start skipping the idiot commentary now, and only comment on the interesting parts.
 
		16	2.422	1.789	1.379	1.109	0.929	0.810	(3)
		15	2.536	1.885	1.459	1.176	0.988	0.865	(4)
		16	(3) <- to save typing and make clear where the same ratios recur.
		15	(4)
 
	3.000 (min)
 
		16	2.724	2.000	1.515	1.199	0.989	0.849	(5)
		15	5.072	3.769	2.918	2.353	1.976	1.729	(6)
		16	4.844	3.577	2.759	2.218	1.857	1.621	(7)
		17	4.643	3.408	2.619	2.098	1.752	1.526	(8)
		16	(7)
		15	(6)
		16	(7)
		17	(8)
		18	4.466	3.258	2.494	1.992	1.659	1.441	(9)
		17	(8)
 
So what have we here? On the first change of L (15 to 16) after changing F, we got an unstable
result (5), but all subsequent times we set L to 16 we got the stable result (7).
 
	5.071 (the original default)
		18	4.063	3.003	2.326	1.881	1.586	1.260	(10)	! not (1)!
		17	2.747	2.016	1.549	1.241	1.036	0.902	(11)
		18	(1)
		17	(11)
		16	2.865	2.116	1.632	1.312	1.099	0.959	(12)
		15	3.001	2.230	1.726	1.392	1.169	1.023	(13)
 
(10) is the first change of L after changing F, and is an unstable new set of ratios, which
revert to (1) when changed and changed back.
 
	3.000
		16	3.223	2.365	1.793	1.418	1.170	1.005	(14)
		15	(6)
		16	(7)
 
so again, the first change of L after changing F yields an unstable set (14) which reverts
to the regular ratios for this F, L combination when changed again.
 
	5.999
		26	3.212	2.271	1.710	1.348	1.112	0.872	(15) 
 
(15) is the result of changing from F min, L 16 to F max, L26.
Note also that the first change of L after a change of F may be either +/- 1 (using D-pad)
or +/- 10 (using R1/L1); in this case, R1.
 
	3.000
		16	(5)
 
changing L from 26 to 16 after changing F from max to min, has yielded the same ratios as changing
L from 15 to 16 after changing F from max to min; this seems to imply that the resulting unstable ratios
are not affected by the value of L (26 or 15) when the change of F was made.
 
		15	(6)
		16	(7)
	5.071
		26	(15)
 
Very interesting! The result of changing L 16 to 26 after changing F min to F orig is identical to 
the result of changing L 16 to 26 after changing F min to F max ! So we are seeing a set of ratios which 
is independent of the current value of F !!! Didn't we start off by saying that the effect of changing L 
depends only on the current value of F ???
 
	3.000	
		16	(14)
 
and similarly, changing L 26 to 16 after changing F orig to F min yields the same result as changing
L 15 to 16 after changing F orig to F min - so again, independent of the previous value of L.
 
	5.071
		26	(15)
		27	2.074	1.432	1.061	0.823	0.668	0.568	(16)
		26	2.116	1.469	1.092	0.850	0.692	0.590	(17) the true ratios for 5.071/26
	3.000
		16	(14)
		17	(8)
		16	(7)
		26	3.577	2.484	1.847	1.437	1.170	0.997	(18)
		16	(7)
	5.999
		26	(15) repeat result of F min L 16 to F max L 26
		36	1.544	1.010	0.724	0.546	0.432	0.360	(19)
	3.000
		26	2.011	1.424	1.029	0.787	0.629	0.527	(20)
		16	(7)
		26	3.577	2.484	1.847	1.437	1.170	0.997	(21)
		16	(7)
		26	(21)
		36	3.089	2.019	1.449	1.092	0.865	0.720	(22)
	5.071	
		26	(15) same result from F min L 36 -> F orig L 26 and F min L 16 -> F orig L 26.
		16	2.865	(12)
	3.000
		15	3.376	2.487	1.894	1.563	1.243	1.071	(23)
		16	(7)
		15	(6)
		16	(7)
	4.000	
		26	(15)
 
Looks like F min to F anything, L 16 to 26 gives (15).
 
	3.000
		27	2.958	2.086	1.501	1.143	0.912	0.726	(24)
	4.000
		26	(15)
 
Looks like F min to F anything, L anything to L 26 gives (15) !!!
 
	3.000
		25	3.082	2.189	1.590	1.219	0.979	0.822	(25)
		26	(21)
		25	3.653	2.552	1.905	1.487	1.214	1.037	(26)
	3.001
		26	3.575	2.483	1.846	1.437	1.170	0.997	(27)
		25	3.652	2.552	1.904	1.487	1.214	1.037	(28) 
		26	(27)
 
so it is not "F min to F anything"; if 'anything' is close to 'min', the result seems to be a 
stable, repeatable correct calculation.
 
	3.000
		25	(26)
	3.010	
		26	3.565	2.476	1.841	1.433	1.167	0.994	(29)
		25	?
		26	(29) so 3.010 is still too close to 3.000 to generate an unstable result
	3.000
		25	(26)
	3.100
		26	3.461	2.464	1.787	1.391	1.133	0.965	(30)
		25	?
		26	(30) - still too close to F min to get a false calculation on the first change of L
	3.000	
		25	(26)
	3.500
		26	3.212	...	(like 15)		0.872
 
this is where I get to admit a lack of thoroughness; I started assuming that a 1st gear of 3.212 meant that
all the other gears were the same, so I didn't check them - but I was wrong. I did review a few after I noticed
that 6th was different, but I'll be more precise about the discrepancies a little further down ...
 
 		25	3.131	?
		26	3.066	?	so 3.500 is far enough from 3.000 to generate an unstable first ratio
	3.300	...
		26	3.251	stable in 1st
	...
	3.400	25->26	3.212	unstable in 1st
	...
	3.400	16->26	3.212	unstable in 1st
	...
	3.350	25->26	3.212	"
	...
	3.310	"	3.241	stable 1st gear
	...
	3.320	"	3.232	
	...
	3.330	"	3.222
	...
	3.335	"	3.217
	...
	3.338	"	3.214
	...
	3.339	"	3.213
	...
	3.339	36->26	3.213	2.271	1.710	1.348	1.112	0.896	(31)
 
note the similarities and differences vs. 
[		26	3.212	2.271	1.710	1.348	1.112	0.872	(15) ]
 
	3.340	16->26	3.212					0.896
 
so we have found the point at which 1st gear ratio transitions from stable (correct calculation at first 
change of L) to unstable (3.212 irrespective of how much further away the original F was). But note that
this is not the same threshold for sixth gear, which continues to be unstable at this range.
 
Assume changing back to F min (3.000), L = 25 between each of the following samples at L = 26:
 
	3.350							0.893
	3.345							0.895
	3.344							0.895
	3.343							0.895
	3.342							0.895
	3.341							0.896
 
so this is how sixth gear varies above 3.340, but actually we are more interested in where it reaches instability
 
	3.400							0.880
	3.420							0.875
	3.430							0.872
	3.425							0.874
	3.428							0.873
	3.429							0.873
 
so the threshold for sixth gear, from F min (3.000) is,  F = 3.430 when L changes to 26.
 
So far, so good; 1st and sixth are stable at this range ... but the gears between are not ...
changing from F min (3.000) to F (x) and changing L to 26, gives
 
	3.300		3.251	2.271	1.710	1.348	1.112	0.907
 
but 2nd is not stable here
 
	3.200		3.353	2.329	1.732	1.348	1.112	0.935
 
gives a stable 2nd, but still not stable in 4th or 5th
 
	3.160		3.395	2.358	1.753	1.364	1.112	0.947
 
is stable in all except 5th, which changes to 1.111 on a retry ...
 
	3.159		3.396	2.359	1.754	1.345	1.112	0.947
 
which is stable in all gears.
 

So, if we set F = min (3.000) and adjust the setting of L, and then we change F beyond the value of
3.159, we see that the first change of L after that will result in a gear ratio in one or more gears which
is spurious, and will not be repeated if we then change L away from that value and back again to that value.

Conclusion:

If we change F by a small amount, then adjust L, the individual gear ratios will be calculated correctly.
But if the change in F exceeds some threshold, then the next change in L will not yield the correct ratios
for the individual gears. The threshold value differs from gear to gear. Irrespective of how much we change F,
the second and subsequent changes of L will yield the intended gear ratios for the individual gears. Therefore,
when adjusting gears, if repeatability is desired, it is advisable to change L more than once after each change
in F in order to get a repeatable set of individual gear ratios.

Here is a question to determine whether you graduate from this class ...

If we set a start point of the stable setting for F = min (3.0000) and L = 17 (giving 1st ratio = 4.643), and
then we run the tranny trick ( F max, L min, F min ), what ratios will we end up with ?

Exercise: try it for yourself before reading further ...

Really, try it before you look at the answers. You'll regret missing the chance of seeing it for yourself.
Okay, now you've tried it, what results did you get? Here are some suggested results. First gear ratio equals
five point oh seven two ? First gear ratio equals two point five three six ? First gear ratio equals four
point six three seven ? Well, the first suggestion is a red herring. The second and third answers are correct.
If you chose to move from L = 17 to L = 15 by using the D-pad, you took two changes of L, so the second change
gave the correct (stable) answer, with first gear ratio of two point five three six; but if you chose to move
from L = 17 to L = min by means fo pressing L1, then you got there on the first recalculation of L - in other
words, you got the spurious unstable result on the first reclaculation, giving a first gear ratio of four
point six three seven. So there you have it ... the reason why my original two attempts at the tranny trick
yielded completely different results.

notes
-----
1. The "tranny trick": is a widely used method to give the most flexible set of gear ratios for a given car and
track. it involves setting the 'Final' division ratio (F) to its maximum value, then setting the Auto-set Level
(L) to its minimum value (or occasionally slightly higher if the min value causes the car to top out in sixth),
then bringing F back down to the low end and adjusting only F to obtain a set of ratios which suitably (ie. only
just) covers the maximum speed the car can manage (with drafting if applicable). The benefit of the tranny trick
is that it maximises the overlap of adjacent gears, allowing the car to spend more time at or near the peak power
of the engine, and allowing the driver greater flexibility to choose the right gears for cornering. or that's how
I understand it, at least.

2. MK's program: a free PC program which can interrogate a PS2 game save or replay file which has been transferred
to PC by X-Port or Sharkport, and show the settings of each car in the replay or garage; and can also edit garage cars.

posted 01-08-2004 04:28 AM         

Here's the brief version...

If you change the value of 'Final', and then change the value of the Auto-set 'Level', the program tries to calculate the right individual gear ratios relevant for the new 'Final'. But on the first try, there is some temporary limitation which stops it from changing each ratio more than a certain amount, so if you changed 'Final' by more than that limit, the individual ratios will not change all the way, but stop at some intermediate value. When you change 'Level' again, the limit has evapourated again, and so you get the right ratios on the second (and all subsequent) change of 'Level'.

The long version at the top is mostly just a detailed list of datapoints I recorded, which show how I reached this finding.

posted 01-08-2004 09:23 AM         

quote:

---

when you slide your fianal all the way to the right, and slide the auto 1 click to the right and back you get a brand new set of different gears.

---

That is what this thread is about. After you change 'Final' by more than a small amount, the next one click of 'Auto' (which I call 'Level') using either D-pad or L1/R1 will result in a spurious set of ratios as if you hadn't moved 'Final' very much; but after that, any further adjustments of 'Auto' will yield the ratios the programmer intended, ie. the correct ratios for your present value of 'Final'.

So if your execution of the tranny trick involves only one click of Auto, then you have not really executed the tranny trick; but if you click Auto more than once, then you really have performed the tranny trick.

quote:

---

I find these to be a better set BTW.

---

This is why you like it better when you click Auto more than once!

quote:

---

is there a better position to stop sliding the final to the right, short of max?

---

posted 01-08-2004 10:44 AM         

quote:

---

Originally posted by Ubiquitous:
The thing I find great about this is the fact that NC1 seemed fairly bored too, so he posted a really simple quick race. P2 got bored and decided to venture on a voyage of scientific discovery :P

---

Which shows that p2 is just bored. I'm both bored and lazy.

posted 01-09-2004 05:47 PM         

quote:

---

If we are doing the tranny trick and slide the Auto to the left, is there a better position to stop sliding the final to the right, short of max?

---

So, when you are gaily travelling along the track, your 'wheel revs' are equated to your 'engine revs', divided by the product of: (the gear ratio number of the gear you are in, multiplied by the ratio of the 'Final' gear setting).
ie. if you are in sixth gear with F=3.000 and 6th=1.729, then your wheels are turning at a rate of about one fifth (precisely 1 / 5.187) of your indicated engine revs.

Now, as we have been discussing ( well, I've been talking and you've all been listening, or yawning, or logging off :/ ), when we set 'Final' to some value, and then set the Auto-set level to some value, the program makes a calculation as to what is the right set of gear ratios to represent that combination. (Well, excluding the first change of 'L', which yields rubbish, and is the cause of the existence of this thread!)

Right. Now we need some data points to work with. I've taken the trusty Daihatsu Cuore again, and printed off a selection of samples from it ...

	Final	Level	1st	2nd	3rd	4th	5th	6th
	-----	-----	---	---	---	---	---	---
 
	5.999	15	2.536	1.885	1.459	1.176	0.988	0.865
		27	1.753	1.210	0.897	0.695	0.565	0.480
		39	1.503	0.965	0.685	0.511	0.402	0.332
 
	4.500	15	3.381	2.513	1.945	1.568	1.317	1.153
		27	2.338	1.614	1.195	0.927	0.753	0.640
		39	2.004	1.287	0.913	0.682	0.536	0.443
 
	3.000	15	5.072	3.796	2.918	2.353	1.976	1.729
		27	3.507	2.421	1.793	1.391	1.130	0.961
		39	3.007	1.931	1.370	1.023	0.804	0.665
 

Looking at how much 'longer' one gear is than the previous one, we simply need to know by how much less is the engine revs divided? So Going from first @ 2.536 to 2nd @ 1.885, 2nd is 2.536/1.885, or 1.345 times, longer than 1st.

Interestingly, we see that the steps between gears look like:

 
	5.999	15	    1.345   1.292   1.261   1.190   1.142
		27	    1.449   1.349   1.291   1.230   1.177
		38	    1.558   1.409   1.341   1.271   1.211
 
	4.500	15	    1.345   1.292   1.240   1.191   1.142
		...
 
	3.000	15	    1.346   1.292   1.240   1.191   1.143
		27	    1.449   1.350   1.289   1.231   1.176
		39	    1.557   1.409   1.339   1.272   1.209

So (apart from rounding errors) the relative ratios depend only on the auto-set level, and not at all upon the value of 'Final'. In other words, the amount of overlap of the gears is determined only by the value of the Auto-set Level.

As stated earlier, the relationship between wheel revs and engine revs is found by :

engine revs / (gear ratio * final ratio)

So if we factor in the 'Final' for a couple of these, we see that :

 
	5.999	15	15.214	11.308	8.752	7.055	5.927	5.189
 
	3.000	15	15.216	11.307	8.754	7.059	5.928	5.187

After you have set the Auto-set Level to 'minimum' to set the gears as close together as possible, thus maximising the overlap between adjacent gears (which is the objective of the tranny trick), you have actually selected a ridiculously short set of gearing for most cases. To avoid your engine topping out on the rev limiter on all but the shortest straights, you are going to need to stretch the gearings out a bit - which you achieve by reducing 'Final' until the gears become long enough to allow the car to avoid hitting the rev limiter on the track which you intend to run. If you have used the tranny trick with a low value of 'Final', you have little scope to widen the gearing by reducing 'Final'. To make the tranny trick work for you, the value of 'Final' which you use must be high enough to allow you to then reduce it by enough to get your sixth gear long enough to cope with your car's potential speed on the track you are going to run. Therefore: if your car is not capable of topping out when using the ratios provided by the full tranny trick, then there is no disadvantage in using a lower value of 'Final' while executing the trick; but nevertheless, there is no better value of 'Final' than 'max'. And if your car is such that it can top out in sixth when you use the full tranny trick, you are going to have to compromise your tranny trick by means of not going all the way to 'min' on the Auto-set Level, in order to allow 'Final'=min to yield long enough gears to cope with your car's top speed.

So:

1. Always set 'Final' to max in the tranny trick - there is no better setting!
2. Set 'Level' to min, making sure you change it more than one time to get there!
3. Reduce 'Final' until your gears are long enough to avoid topping out.
4. If the gears top out even at 'Final'='min', repeat the tranny trick using one notch higher on 'Level' each time, until you get gears which cope with your car's top speed capability.

Hope that helps.

p2

posted 01-10-2004 12:19 AM         

1) Set Final to MAXIMUM
2) Set Level to minimum
3) Reduce Final until the car doesn't redline in top gear
4) Increase Level by one step (preferably by going up 2, then down 1), then do steps 1 and 3 again (and 4 if necessary?)

------------------
Santiago22 // Gran Turismo 3: A-Spec
Straightaways are for fast cars. Corners are for fast drivers!
Eat asphalt. Drink fuel. Spit oil. Belch fire.
Race to live. Live to race. :)

Button's Brigade - The track is ours. You may have it when we're done. :mad:
#92

posted 01-10-2004 06:09 AM         

4. if (3) doesn't work, repeat steps (1) - (3) but using L = (min+1) in step (2), and keep repeating with L = (min+2), (min+3), ... in step (2) until you successfully avoid redlining with step (3)

The important bit which wasn't clear in your version is that step (2) should always be executed with Final at max. ie. after trying a value of L and finding that step 3 doesn't work, you have to go back to step 1.

But you are right about making sure you move L more than once while running step 2.

posted 04-18-2005 04:33 PM         

No wonder we are all missing P2. (best wishes, btw)

posted 04-18-2005 05:32 PM         

Now get some rest. I imagine you need it. Good to see another post from you