G-Sum - questions for the proponents
#46
Race Director
Originally Posted by Premier Motorsp
Sunday is right. If you have speed data from a faster driver to compare with, then looking at Gsum is secondary to that. If the car was really well instrumented, looking at Gsum would become less necessary yet, as the faster driver's technique to gain speed would be obvious from the other graphs.
Am I missing something here?
At least with sector times and speeds you can piece together (or rather your software can) your own theoretical best lap to compare against.
#47
Addict
Rennlist Member
Rennlist Member
Hmmm...
I like this "virtual best lap" function. I need it! It seems I am rarely able to get a good lap for many reasons. Cranial Flatus; just forgetting or spacing out, traffic, people spinning, black flags, etc. Seems you have to be a really good driver, that you must CONCENTRATE on driving consistently, probably have to drive within your means (perhaps 90% or less), and you need luck... all of this just to get a baseline!
I was looking at my DL1 data from VIR over the weekend with someone who knows how to navigate the software (ANOTHER big learning curve). We were overlaying speed charts from different guys. It seemed readily apparent that differences is speed were only useful if you could figure out what the cause was. I gather that rather gradual deviations are more traceable to the driver, but what about the larger more sudden changes? It seemed to me that this data was only marginal if you did not have the visual reference to check for external factors.
For instance; we overlayed my data with a buddy in a 968. Our speeds were virtually indentical up the Bridge Straight and into the Esses. Suddenly, my speed dropped off while his only leveled fairly gradually as he went into the hill. Then, by the 3rd Ess I caught back up to him, without his graph changing much. What does this tell us?
- I braked into the Esses? PUH, Not Hardly!
- My 147HP out-muscled his 235 in the space of 75yds? Same answer.
Without a visual (video) to use as a reference, we might never know.
Seems to me you have to have a VERY jaundiced and discriminating eye to make REAL use of this stuff!
I like this "virtual best lap" function. I need it! It seems I am rarely able to get a good lap for many reasons. Cranial Flatus; just forgetting or spacing out, traffic, people spinning, black flags, etc. Seems you have to be a really good driver, that you must CONCENTRATE on driving consistently, probably have to drive within your means (perhaps 90% or less), and you need luck... all of this just to get a baseline!
I was looking at my DL1 data from VIR over the weekend with someone who knows how to navigate the software (ANOTHER big learning curve). We were overlaying speed charts from different guys. It seemed readily apparent that differences is speed were only useful if you could figure out what the cause was. I gather that rather gradual deviations are more traceable to the driver, but what about the larger more sudden changes? It seemed to me that this data was only marginal if you did not have the visual reference to check for external factors.
For instance; we overlayed my data with a buddy in a 968. Our speeds were virtually indentical up the Bridge Straight and into the Esses. Suddenly, my speed dropped off while his only leveled fairly gradually as he went into the hill. Then, by the 3rd Ess I caught back up to him, without his graph changing much. What does this tell us?
- I braked into the Esses? PUH, Not Hardly!
- My 147HP out-muscled his 235 in the space of 75yds? Same answer.
Without a visual (video) to use as a reference, we might never know.
Seems to me you have to have a VERY jaundiced and discriminating eye to make REAL use of this stuff!
#48
Sunday says: (My comments in blue)
Here is what I have done and would do:
1) Follow that car in practice/qualifying to see where and why they are faster.
This is not realistic. When I get beat is it by a few tenths, not 5 seconds. You can't see a few tenths in 1:30 lap
2) Find someone faster to drive your car to get their data. There are always some very talented pros around who can do this becuase huge driving talent does not equal financial success in racing. For example, I know of a few Mid Ohio instructors who do this at MO.
Easy in testing, impossible at a race. Since I can't afford testing...
3) Get coaching from a pro.
See above about budget
4) Look inside yourself and ask, for each part of every corner - am I terrifying myself? If not, then there is probably more speed there.
This is crazy. I am NEVER EVER terrified, and rarely even concerned. How can you drive terrified?
5) Look to the data. If you look at you data after every day, then you will have a feel for what it should look like. I think the ability to use data to find more speed is one of those 'art' things. It is not always scientific. But tools like eclectic lap (where the system builds a theoretical lap based on best segment times) can give you a pretty good target for improvemet.
Vague. Gsum is not vague
Just trying the make the point that Gsum is a useful tool because it can easily tell you where you are not using the car's performance. If I knew how to post a @#$%^@#^ing screen capture then I'd show an example.
Sector times are sometimes very misleading. The eclectic or best possible lap is even more misleading. You can run very very fast sector times one at a time if are willing to blow the previous and following sectors. I not saying sector times are useless, but you really have to look at your best sectors with a little skepticism. Only if the great sector is bracketed by decent sectors do I consider it legit..
Speed overlays are great. But they are mostly useful when SOMEBODY ELSE sets the goal for you. If I had the data from guy who was beating me, you are right, Gsum wouldn't be that useful, the speed trace would tell the story. But since I NEVER have this, Gsum is the only other way I can think of to do it!
Chris Cervelli
Premier Motorsports
Here is what I have done and would do:
1) Follow that car in practice/qualifying to see where and why they are faster.
This is not realistic. When I get beat is it by a few tenths, not 5 seconds. You can't see a few tenths in 1:30 lap
2) Find someone faster to drive your car to get their data. There are always some very talented pros around who can do this becuase huge driving talent does not equal financial success in racing. For example, I know of a few Mid Ohio instructors who do this at MO.
Easy in testing, impossible at a race. Since I can't afford testing...
3) Get coaching from a pro.
See above about budget
4) Look inside yourself and ask, for each part of every corner - am I terrifying myself? If not, then there is probably more speed there.
This is crazy. I am NEVER EVER terrified, and rarely even concerned. How can you drive terrified?
5) Look to the data. If you look at you data after every day, then you will have a feel for what it should look like. I think the ability to use data to find more speed is one of those 'art' things. It is not always scientific. But tools like eclectic lap (where the system builds a theoretical lap based on best segment times) can give you a pretty good target for improvemet.
Vague. Gsum is not vague
Just trying the make the point that Gsum is a useful tool because it can easily tell you where you are not using the car's performance. If I knew how to post a @#$%^@#^ing screen capture then I'd show an example.
Sector times are sometimes very misleading. The eclectic or best possible lap is even more misleading. You can run very very fast sector times one at a time if are willing to blow the previous and following sectors. I not saying sector times are useless, but you really have to look at your best sectors with a little skepticism. Only if the great sector is bracketed by decent sectors do I consider it legit..
Speed overlays are great. But they are mostly useful when SOMEBODY ELSE sets the goal for you. If I had the data from guy who was beating me, you are right, Gsum wouldn't be that useful, the speed trace would tell the story. But since I NEVER have this, Gsum is the only other way I can think of to do it!
Chris Cervelli
Premier Motorsports
#49
Rennlist
Basic Site Sponsor
Basic Site Sponsor
Originally Posted by Premier Motorsp
Sunday says: (My comments in blue)
4) Look inside yourself and ask, for each part of every corner - am I terrifying myself? If not, then there is probably more speed there.
This is crazy. I am NEVER EVER terrified, and rarely even concerned. How can you drive terrified?
Chris Cervelli
Premier Motorsports
4) Look inside yourself and ask, for each part of every corner - am I terrifying myself? If not, then there is probably more speed there.
This is crazy. I am NEVER EVER terrified, and rarely even concerned. How can you drive terrified?
Chris Cervelli
Premier Motorsports
__________________
Larry Herman
2016 Ford Transit Connect Titanium LWB
2018 Tesla Model 3 - Electricity can be fun!
Retired Club Racer & National PCA Instructor
Past Flames:
1994 RS America Club Racer
2004 GT3 Track Car
1984 911 Carrera Club Racer
1974 914/4 2.0 Track Car
CLICK HERE to see some of my ancient racing videos.
Larry Herman
2016 Ford Transit Connect Titanium LWB
2018 Tesla Model 3 - Electricity can be fun!
Retired Club Racer & National PCA Instructor
Past Flames:
1994 RS America Club Racer
2004 GT3 Track Car
1984 911 Carrera Club Racer
1974 914/4 2.0 Track Car
CLICK HERE to see some of my ancient racing videos.
#50
Rennlist Member
More DAS questions on G-sum. I started using a Racepak G2X system this year. Simple and inexpensive, but pretty good software (g smoothing, etc.). Segments and charts are very effective in mapping g-plots along with a friction circle. I have seen an increase in corner speeds by increasing trail-braking ( left foot mostly). Very valid point to include video (next on list). I think analyzing this some more during the cold season along with video overlay will make this an effective tool. The theoretical lap times can be elusive: complex sections can be a matter of throwing away the beginning to pick up exit speed eg: Summit Point, VIR. I have run both of these scenarios on alternating laps and of course the "ideal lap" ends up being quite impressive eg: below 1:24 at Summit Point in a near stock D class car. Not sure I buy it as more than optimistic. I will do some homework on my data and post an opinion based on the results........
#51
Race Director
Originally Posted by Premier Motorsp
5) Look to the data. If you look at you data after every day, then you will have a feel for what it should look like. I think the ability to use data to find more speed is one of those 'art' things. It is not always scientific. But tools like eclectic lap (where the system builds a theoretical lap based on best segment times) can give you a pretty good target for improvemet.
Vague. Gsum is not vague
Just trying the make the point that Gsum is a useful tool because it can easily tell you where you are not using the car's performance.
Vague. Gsum is not vague
Just trying the make the point that Gsum is a useful tool because it can easily tell you where you are not using the car's performance.
I can see a g-sum plot telling you if you're smooth, but without comparison data (like with sector times) I don't see how it can tell you anything of real value.
#52
Addict
Rennlist Member
Rennlist Member
Join Date: Dec 2001
Location: MA
Posts: 1,010
Likes: 0
Received 0 Likes
on
0 Posts
Originally Posted by Premier Motorsp
Just trying the make the point that Gsum is a useful tool because it can easily tell you where you are not using the car's performance. If I knew how to post a @#$%^@#^ing screen capture then I'd show an example.
#53
I've said this before in more detailed ways but somehow it gets lost in the mix. I'll try again...
G-sum is extremely useful, but it has a couple flaws:
-----------------------------
In itself, G-sum doesn't account for different driving lines. New DAS systems are supposed to have some of this capability using GPS and other positioning systems. Why is this important? Because centripital force = mv^2/R. When your R changes (driving line) then so can your forces.
G-sum is relative. What is the limit? What is a good g-sum? You need a comparison either to a extremely good driver or emperically derived or modeled car performance data for various situations.
-----------------------------
Still, if I had David Donohue jump in my car and do 20 laps at Road America and I had solid data with his driving lines then I could easily compare his g-sum to mine and see where I am leaving something on the table. Can I differentiate less than 1% of difference in a given section? Probably not. Could I see where I am leaving significant time on the table? Absolutely.
As for the question of where g-sum came from; I heard it first from an PhD in mechanical engineering who owns an auto data aquisition and analysis company. As a mechanical engineer myself, the concept makes perfect sense.
Oh BTW, there is a lot of nonsense floating around about sector times. If you think about it, these are inherently built into a g-sum. By twice-integrating accelerations over a sector time you would get a sector distance (of course, given a particular line).
G-sum is extremely useful, but it has a couple flaws:
-----------------------------
In itself, G-sum doesn't account for different driving lines. New DAS systems are supposed to have some of this capability using GPS and other positioning systems. Why is this important? Because centripital force = mv^2/R. When your R changes (driving line) then so can your forces.
G-sum is relative. What is the limit? What is a good g-sum? You need a comparison either to a extremely good driver or emperically derived or modeled car performance data for various situations.
-----------------------------
Still, if I had David Donohue jump in my car and do 20 laps at Road America and I had solid data with his driving lines then I could easily compare his g-sum to mine and see where I am leaving something on the table. Can I differentiate less than 1% of difference in a given section? Probably not. Could I see where I am leaving significant time on the table? Absolutely.
As for the question of where g-sum came from; I heard it first from an PhD in mechanical engineering who owns an auto data aquisition and analysis company. As a mechanical engineer myself, the concept makes perfect sense.
Oh BTW, there is a lot of nonsense floating around about sector times. If you think about it, these are inherently built into a g-sum. By twice-integrating accelerations over a sector time you would get a sector distance (of course, given a particular line).