Dyno, Homemade, Using a Novak Sentry Data Logger, Continued, The Experimental Thread.
#136
YYhayyim- I saw your post back there. Thanks. Glad to see you still at the hobby.
Access-Well you can teach an old dog new tricks. The GTB is in all green lines on this chart. The LRP is in all red lines. The highest curve for each is power, and is read off the left axis. The next highest curve is efficiency and is read off the right axis. Certainly efficiency is responsible for part of this power improvement on the LRP, but the motor also has characteristics of fatter wire (Peak power at higher RPM).
Mechanical timing is adjustable on the Novak Ballistic motor. There is no real telling if it is mechanically at zero. The Novak ballistic motor has a scale that is centered and symmetrical in both directions and has hash marks in either direction, but it may just point to the preferred timing. On the LRP I see no scale. On the older LRP there is a scale with numbers indicating 2. Don't plan to adjust these. Racing nears.
I retested my older LRP X11 with a new Nickel plated sintered rotor on the new SMC 6000 mA-h battery (28c) it was over the 700 mark as well at 735 Watts.
We are looking at tiny electric motors putting out about 1 Horse Power and over.
Access-Well you can teach an old dog new tricks. The GTB is in all green lines on this chart. The LRP is in all red lines. The highest curve for each is power, and is read off the left axis. The next highest curve is efficiency and is read off the right axis. Certainly efficiency is responsible for part of this power improvement on the LRP, but the motor also has characteristics of fatter wire (Peak power at higher RPM).
Mechanical timing is adjustable on the Novak Ballistic motor. There is no real telling if it is mechanically at zero. The Novak ballistic motor has a scale that is centered and symmetrical in both directions and has hash marks in either direction, but it may just point to the preferred timing. On the LRP I see no scale. On the older LRP there is a scale with numbers indicating 2. Don't plan to adjust these. Racing nears.
I retested my older LRP X11 with a new Nickel plated sintered rotor on the new SMC 6000 mA-h battery (28c) it was over the 700 mark as well at 735 Watts.
We are looking at tiny electric motors putting out about 1 Horse Power and over.
Last edited by John Stranahan; 05-07-2009 at 06:14 PM.
#137
Tech Adept
John, those graphs are looking fine.
Gamover posted the attached graph above (#113) and I repeat it here. A number of papers on brushless motor control/operation discuss ripple torque. The torque ripple shows up on the flywheel as wavy acceleration, which can show up on the rpm sensor as a slight glitch at very low rpm, and as a wavy reading at higher rpm.
Access, I agree an actual current sensor may not be integrated in the ESCs, and your listed soft start techniques would save the manufacturer some coin.
Also I note your comment on the correlation of efficiency with the power advantage. It shows up clearly in the graph John just posted.
One reason may be the power dissipation in a MOSFET has two components, one for the DC power loss and one for the switching loss at high frequency. If an ESC has MOSFETs with lower DC on resistance they tend to have a much larger device capacitance and thus greater switching losses with increasing PWM frequency. This tradeoff would partly explain the correlation. John or another Dyno expert might be able to test this hypothesis or pull the ESC specification for Rds(on) and see if lower value jives with getting the earlier advantage in efficiency/power.
Gamover posted the attached graph above (#113) and I repeat it here. A number of papers on brushless motor control/operation discuss ripple torque. The torque ripple shows up on the flywheel as wavy acceleration, which can show up on the rpm sensor as a slight glitch at very low rpm, and as a wavy reading at higher rpm.
Access, I agree an actual current sensor may not be integrated in the ESCs, and your listed soft start techniques would save the manufacturer some coin.
Also I note your comment on the correlation of efficiency with the power advantage. It shows up clearly in the graph John just posted.
One reason may be the power dissipation in a MOSFET has two components, one for the DC power loss and one for the switching loss at high frequency. If an ESC has MOSFETs with lower DC on resistance they tend to have a much larger device capacitance and thus greater switching losses with increasing PWM frequency. This tradeoff would partly explain the correlation. John or another Dyno expert might be able to test this hypothesis or pull the ESC specification for Rds(on) and see if lower value jives with getting the earlier advantage in efficiency/power.
#138
Tech Adept
if i look at the raw data (for all runs), the RPM numbers are the same numbers ie. quantization. I think an earlier post mentions a lookup table or similar. Example RPM data @ 10Hz:
29,850
29,850
30,150
30,000
30,456
30,150
30,456
30,000
30,456
30,769
30,456
30,456
30,769
30,769
30,769
30,456
It could be a ripple torque effect that is being amplified by the quantization, or it could also be a harmonic friction effect (slight flywheel imbalance etc.) starting to dominate as motor torque gets close to zero.
29,850
29,850
30,150
30,000
30,456
30,150
30,456
30,000
30,456
30,769
30,456
30,456
30,769
30,769
30,769
30,456
It could be a ripple torque effect that is being amplified by the quantization, or it could also be a harmonic friction effect (slight flywheel imbalance etc.) starting to dominate as motor torque gets close to zero.
#139
Tech Rookie
Also I note your comment on the correlation of efficiency with the power advantage. It shows up clearly in the graph John just posted.
One reason may be the power dissipation in a MOSFET has two components, one for the DC power loss and one for the switching loss at high frequency. If an ESC has MOSFETs with lower DC on resistance they tend to have a much larger device capacitance and thus greater switching losses with increasing PWM frequency. This tradeoff would partly explain the correlation. John or another Dyno expert might be able to test this hypothesis or pull the ESC specification for Rds(on) and see if lower value jives with getting the earlier advantage in efficiency/power.
One reason may be the power dissipation in a MOSFET has two components, one for the DC power loss and one for the switching loss at high frequency. If an ESC has MOSFETs with lower DC on resistance they tend to have a much larger device capacitance and thus greater switching losses with increasing PWM frequency. This tradeoff would partly explain the correlation. John or another Dyno expert might be able to test this hypothesis or pull the ESC specification for Rds(on) and see if lower value jives with getting the earlier advantage in efficiency/power.
As for switching MOSFETs, you have the PWM and in brushless you also have the phase switching. Once the ESC is out of 'soft start' and the RPMs are sufficient, the PWM might be at full duty cycle (esp. if you're at full throttle) so at this point the only switching going on is phase switching. You can see the transition if you hook up a scope on one of the brushless poles. I only have a Novak Havoc and a super stock (13.5) motor, but when I ramped the throttle, slowly, letting the motor spin up freely, I could observe the PWM on the scope and it went to full duty cycle around 70% throttle.
#140
Tech Adept
Gameover,
Just for clarity on the thread I'll describe generally the quantization. On the earlier development thread I noted quantization on the Sentry current sensor.
In a simple scheme, digital resolution is 2^N, where N is the number of bits.
2^8 = 256
2^10 = 1024
Let's say the current sensor presents a linear output on a range up to 5 volts for analog to digital conversion inside the Sentry:
5{V} / 256 = 19.5 {mV} |discrete resolution
5{V} /1024 = 4.88 {mV} |discrete resolution
and since the RPM sensor plugs into a discrete system it becomes quantized even though it may be a digital signal input rather than an A/D input (numbers are for example purposes only).
There could be a lookup table, which takes up more memory. Or just a conversion algorithm using up more cycles on the microprocessor. In general you use a lookup table in real time systems where speed is essential. You use more processor cycles to conserve memory where speed is not essential.
Access,
I'll put up a basic inverter schematic with my questions about your PWM versus phase switching when I get a chance.
I'll include the line-to-line resistance and inductance model. Have you tried to measure these values in a brushless motor?
Just for clarity on the thread I'll describe generally the quantization. On the earlier development thread I noted quantization on the Sentry current sensor.
In a simple scheme, digital resolution is 2^N, where N is the number of bits.
2^8 = 256
2^10 = 1024
Let's say the current sensor presents a linear output on a range up to 5 volts for analog to digital conversion inside the Sentry:
5{V} / 256 = 19.5 {mV} |discrete resolution
5{V} /1024 = 4.88 {mV} |discrete resolution
and since the RPM sensor plugs into a discrete system it becomes quantized even though it may be a digital signal input rather than an A/D input (numbers are for example purposes only).
There could be a lookup table, which takes up more memory. Or just a conversion algorithm using up more cycles on the microprocessor. In general you use a lookup table in real time systems where speed is essential. You use more processor cycles to conserve memory where speed is not essential.
Access,
I'll put up a basic inverter schematic with my questions about your PWM versus phase switching when I get a chance.
I'll include the line-to-line resistance and inductance model. Have you tried to measure these values in a brushless motor?
Last edited by SystemTheory; 05-08-2009 at 01:20 PM.
#141
Tech Rookie
1) Phase-only model.
2) Inrush current dropoff.
3) Phase + PWM model (one coil only -- incomplete)
4) PWM-only model.
5) PWM current-leveling due to inductance.
I didn't really try too hard, though, the models are very simplified. It's easier to just hook up the scope and observe from there. Things do look pretty distorted on the scope, but you can observe both the phase switching and when the PWM goes to full duty cycle.
http://www.youtube.com/watch?v=vqapT...e=channel_page
You can observe the PWM switching in this video, I don't think I've posted a video where you can observe phase switching on a scope but it's real easy to make one if anyone wants.
Last edited by Access; 05-08-2009 at 05:01 PM.
#142
Tech Adept
Access,
Thanks for the feedback on the values. It looks like a system model is pretty difficult to build with predictive accuracy.
It proves the need for a Dyno just as much with brushless as with the brushed motors for competetive testing.
I think one could measure 2*R with a 1{A} current regulator and 1{mV} sensitive DVM, where R is resistance per phase. Inject a variable AC signal and use a known capacitor value to measure 2*L, where L is phase inductance. The measurement of air gap constant k is also possible. We can discuss details off thread if you're interested.
These diagrams show DC/Battery voltage on VBus, and six control signals to the transistors. This suggests PWM signal is integrated with the frequency modulation signal within the phase control signals.
(Just saw your waveforms as I posted this question, maybe it is resolved there, check on it later).
If PWM is 100% at 70% throttle, does that mean 30% of the throttle is not accelerating the motor to greater rpm, or that frequency modulation takes over to enable further acceleration?
Or am I missing something?
Thanks for the feedback on the values. It looks like a system model is pretty difficult to build with predictive accuracy.
It proves the need for a Dyno just as much with brushless as with the brushed motors for competetive testing.
I think one could measure 2*R with a 1{A} current regulator and 1{mV} sensitive DVM, where R is resistance per phase. Inject a variable AC signal and use a known capacitor value to measure 2*L, where L is phase inductance. The measurement of air gap constant k is also possible. We can discuss details off thread if you're interested.
These diagrams show DC/Battery voltage on VBus, and six control signals to the transistors. This suggests PWM signal is integrated with the frequency modulation signal within the phase control signals.
(Just saw your waveforms as I posted this question, maybe it is resolved there, check on it later).
If PWM is 100% at 70% throttle, does that mean 30% of the throttle is not accelerating the motor to greater rpm, or that frequency modulation takes over to enable further acceleration?
Or am I missing something?
#143
Tech Rookie
Access,
These diagrams show DC/Battery voltage on VBus, and six control signals to the transistors. This suggests PWM signal is integrated with the frequency modulation signal within the phase control signals.
(Just saw your waveforms as I posted this question, maybe it is resolved there, check on it later).
If PWM is 100% at 70% throttle, does that mean 30% of the throttle is not accelerating the motor to greater rpm, or that frequency modulation takes over to enable further acceleration?
These diagrams show DC/Battery voltage on VBus, and six control signals to the transistors. This suggests PWM signal is integrated with the frequency modulation signal within the phase control signals.
(Just saw your waveforms as I posted this question, maybe it is resolved there, check on it later).
If PWM is 100% at 70% throttle, does that mean 30% of the throttle is not accelerating the motor to greater rpm, or that frequency modulation takes over to enable further acceleration?
With brushed the only way the ESC could control the speed of the motor was to vary the PWM duty cycle. Any PWM duty cycle other than 100% causes switching losses which means less efficiency. With brushless the ESC has more tools with which to control rotor speed, like variable forward advance / phase switching speed. So the idea in designing an ESC to work efficiently is to get to 100% PWM as early as possible and then to control farther acceleration with the latter two. Forward advance is important because coils do not 'turn on' or 'turn off' instantly, the inductance is a factor.
You also have Back-EMF in the picture and I'm honestly not exactly sure how that fits in or is properly modeled.
#144
System Theory's oval simulator project continues
Driveline Inertia Estimate
I did some experiments to determine the rotational inertia of a set of Pan car wheels and axles with the Sentry dyno. I used two cars. I placed the rear driven wheels on top of a set of front wheels. I adjusted cambers for flat running fronts. I used the center shock to just barely load the front wheels with the weight of the pod and little more. Friction was low; the wheels would coast a bit when turned by hand. The rears were freshly trued at 2.240 inch.
I used a 17.5 Novak motor which I dynoed previously on Matts steel flywheel. I ran the wheels up to full speed and collected the data. I used a 45 pinion and 85 spur. Then I used Chris's version of the Sentry Dyno Spreadsheet to analyze the data. It has a 6th order polynomial instead of a cubic. It also allows selection of a non zero start time as the dyno really starts in between your first data set and the previous point. This version of the spreadsheet works better with the light Aluminum flywheels on stock motors where insufficient data is collected.
I selected, manually, rotational inertias until the spreadsheet plotted my Peak power obtained previously with a steel flywheel. The inertia was 2.8 x 10^-5 kgm^2s^2. This is probably not that accurate a method as the wheels are very light, but it puts us in the ball park. It is not corrected for gear ratio.
More fun with the Sentry dyno.
System Theory may continue.
Driveline Inertia Estimate
I did some experiments to determine the rotational inertia of a set of Pan car wheels and axles with the Sentry dyno. I used two cars. I placed the rear driven wheels on top of a set of front wheels. I adjusted cambers for flat running fronts. I used the center shock to just barely load the front wheels with the weight of the pod and little more. Friction was low; the wheels would coast a bit when turned by hand. The rears were freshly trued at 2.240 inch.
I used a 17.5 Novak motor which I dynoed previously on Matts steel flywheel. I ran the wheels up to full speed and collected the data. I used a 45 pinion and 85 spur. Then I used Chris's version of the Sentry Dyno Spreadsheet to analyze the data. It has a 6th order polynomial instead of a cubic. It also allows selection of a non zero start time as the dyno really starts in between your first data set and the previous point. This version of the spreadsheet works better with the light Aluminum flywheels on stock motors where insufficient data is collected.
I selected, manually, rotational inertias until the spreadsheet plotted my Peak power obtained previously with a steel flywheel. The inertia was 2.8 x 10^-5 kgm^2s^2. This is probably not that accurate a method as the wheels are very light, but it puts us in the ball park. It is not corrected for gear ratio.
More fun with the Sentry dyno.
System Theory may continue.
Last edited by John Stranahan; 05-10-2009 at 12:22 PM.
#145
Tech Adept
Simulated Pancar Response from 2s Lipo Novak SS 17.5 Dyno Curve
1. Velodrome / Dyno Details Post #36:
Brushless 17.5 170{W} peak power at 8200{rpm} on Sentry Dyno
Velodrome Run Line 820{ft}
http://www.rctech.net/forum/5692381-post36.html
2. Velodrome Details Post #56 & #62:
Track Photo: http://www.rctech.net/forum/5702105-post56.html
Lap times discussed by SWTour for 21.5 stock motors.
http://www.rctech.net/forum/5704554-post62.html
3. Drag Model Post #45 & #53:
Model: http://www.rctech.net/forum/5697672-post45.html
Scale down numbers for a Dodge Viper applied in my simulator.
http://www.rctech.net/forum/5701064-post53.html
4. Acceleration Model Post #75:
Full throttle acceleration model described in some detail.
http://www.rctech.net/forum/5717938-post75.html
5. Input Data
Dyno Curve & Car Input Data in Standard SI Units ----------
T /TAB/ FILE=NovakSS17_5_2S;
Tw = T(ws); ::[N-m] torque-speed source
Js = 4.13E-7; ::[kg-m^2] source inertia
G = 85/45; ::[#] gear ratio
Je = 9.419E-05; ::[kg-m^2] eq. driveline inertia
J = 0; ::[kg-m^2] inertia per tire
r = 28m; ::[m] tire radius
m = 1.16; ::[kg] car mass
rho = 1.229; ::[kg/m^3] air density
Cd = 0.4; ::[#] drag coefficient
Af = 1.79E-02; ::[m^2] frontal area
b = 0; ::[N-m-s/rad] driveline damping
tI = 0; ::[s] initial time
xI = 0; ::[m] initial position
vI = 0; ::[m/s] initial velocity
tF = 12; ::[s] final time
xF = 820; ::[ft] benchmark distance
6. How I Calculate Equivalent Driveline Inertia (includes 4 wheels):
John measured the equivalent inertia at the source as Jes = ~2.8E-5
Je = G^2*(Jes - Js) = (1.889^2)*(2.8E-5 - 4.13E-7) = 9.419E-05{kg-m^2}
where this is the equivalent inertia at the rolling axles.
Note I set wheel/tire intertia J = 0 since John spun the wheels to measure Jes. I think John's measurement is accurate. I can't calculate precision. It could be off by +/- 20%.
7. Simulator Graphs
A. Time, Torque, and Power versus radian Shaft Speed at the motor. Time evolves left to right. Operating point indexed to time for acceleration to estimated top speed. The simulator uses this torque-speed curve like a lookup table to find torque at each velocity (a velocity feedback source).
B. Acceleration, velocity, and displacement versus time. Assuming top speed can be kept in the high banked turns, lap 1 time estimated where distance hits the 820{ft} run line. Here t1 = 11.6{s}. Top speed is an estimate due to assumptions in aerodynamic model and zero driveline damping b = 0. Actual amp draw probably closer to 15-20 Amps at top speed.
C. Acceleration, Power Curves, Air Drag Power versus time. The hump in the curves respresents power invested to spin mass and inertia up to higher kinetic energy during acceleration. The flat line represents power required to push air at top speed with zero acceleration. Again, values are approximate.
8. Lap Time Evaluation
Simulated:
1st Lap time: 11.6{s}
Top Speed: 54{mph}*1.47 = 79.4{ft/s}.
2nd Lap time: 820{ft}/79.4{ft/s} = 10.33{s}.
Posted: http://www.southwesttour.com
Blackburn Destroys Track Record
21.5 Stock Class
1st Lap time: 12.8{s} best holeshot.
Top Speed: 52.2{mph} average lap for race winner.
2nd Lap time: ~10.0{s} best lap.
I consider this a good simulation to illustrate system performance. Better input data would give more accurate simulation of lap times on the Velodrome.
1. Velodrome / Dyno Details Post #36:
Brushless 17.5 170{W} peak power at 8200{rpm} on Sentry Dyno
Velodrome Run Line 820{ft}
http://www.rctech.net/forum/5692381-post36.html
2. Velodrome Details Post #56 & #62:
Track Photo: http://www.rctech.net/forum/5702105-post56.html
Lap times discussed by SWTour for 21.5 stock motors.
http://www.rctech.net/forum/5704554-post62.html
3. Drag Model Post #45 & #53:
Model: http://www.rctech.net/forum/5697672-post45.html
Scale down numbers for a Dodge Viper applied in my simulator.
http://www.rctech.net/forum/5701064-post53.html
4. Acceleration Model Post #75:
Full throttle acceleration model described in some detail.
http://www.rctech.net/forum/5717938-post75.html
5. Input Data
Dyno Curve & Car Input Data in Standard SI Units ----------
T /TAB/ FILE=NovakSS17_5_2S;
Tw = T(ws); ::[N-m] torque-speed source
Js = 4.13E-7; ::[kg-m^2] source inertia
G = 85/45; ::[#] gear ratio
Je = 9.419E-05; ::[kg-m^2] eq. driveline inertia
J = 0; ::[kg-m^2] inertia per tire
r = 28m; ::[m] tire radius
m = 1.16; ::[kg] car mass
rho = 1.229; ::[kg/m^3] air density
Cd = 0.4; ::[#] drag coefficient
Af = 1.79E-02; ::[m^2] frontal area
b = 0; ::[N-m-s/rad] driveline damping
tI = 0; ::[s] initial time
xI = 0; ::[m] initial position
vI = 0; ::[m/s] initial velocity
tF = 12; ::[s] final time
xF = 820; ::[ft] benchmark distance
6. How I Calculate Equivalent Driveline Inertia (includes 4 wheels):
John measured the equivalent inertia at the source as Jes = ~2.8E-5
Je = G^2*(Jes - Js) = (1.889^2)*(2.8E-5 - 4.13E-7) = 9.419E-05{kg-m^2}
where this is the equivalent inertia at the rolling axles.
Note I set wheel/tire intertia J = 0 since John spun the wheels to measure Jes. I think John's measurement is accurate. I can't calculate precision. It could be off by +/- 20%.
7. Simulator Graphs
A. Time, Torque, and Power versus radian Shaft Speed at the motor. Time evolves left to right. Operating point indexed to time for acceleration to estimated top speed. The simulator uses this torque-speed curve like a lookup table to find torque at each velocity (a velocity feedback source).
B. Acceleration, velocity, and displacement versus time. Assuming top speed can be kept in the high banked turns, lap 1 time estimated where distance hits the 820{ft} run line. Here t1 = 11.6{s}. Top speed is an estimate due to assumptions in aerodynamic model and zero driveline damping b = 0. Actual amp draw probably closer to 15-20 Amps at top speed.
C. Acceleration, Power Curves, Air Drag Power versus time. The hump in the curves respresents power invested to spin mass and inertia up to higher kinetic energy during acceleration. The flat line represents power required to push air at top speed with zero acceleration. Again, values are approximate.
8. Lap Time Evaluation
Simulated:
1st Lap time: 11.6{s}
Top Speed: 54{mph}*1.47 = 79.4{ft/s}.
2nd Lap time: 820{ft}/79.4{ft/s} = 10.33{s}.
Posted: http://www.southwesttour.com
Blackburn Destroys Track Record
21.5 Stock Class
1st Lap time: 12.8{s} best holeshot.
Top Speed: 52.2{mph} average lap for race winner.
2nd Lap time: ~10.0{s} best lap.
I consider this a good simulation to illustrate system performance. Better input data would give more accurate simulation of lap times on the Velodrome.
Last edited by SystemTheory; 05-10-2009 at 06:15 PM. Reason: corrections
#146
Tech Adept
iTrader: (14)
Simulated:
1st Lap time: 11.6{s}
Top Speed: 54{mph}*1.47 = 79.4{ft/s}.
2nd Lap time: 820{ft}/79.4{ft/s} = 10.33{s}.
Posted: http://www.southwesttour.com
Blackburn Destroys Track Record
21.5 Stock Class
1st Lap time: 12.8{s} best holeshot.
Top Speed: 52.2{mph} average lap for race winner.
2nd Lap time: ~10.0{s} best lap.
I'm not sure using the data above is useful as is. The first lap at the velodrome is probably 150 feet shorter because the start is near the entrance to turn 1 and the loop is at the exit of 4. Also, the track is amazingly slick - so slick that I wasn't to full throttle until into turn 2. The second and third lap was also not full throttle for parts of it due to traffic.
jblackburn
1st Lap time: 11.6{s}
Top Speed: 54{mph}*1.47 = 79.4{ft/s}.
2nd Lap time: 820{ft}/79.4{ft/s} = 10.33{s}.
Posted: http://www.southwesttour.com
Blackburn Destroys Track Record
21.5 Stock Class
1st Lap time: 12.8{s} best holeshot.
Top Speed: 52.2{mph} average lap for race winner.
2nd Lap time: ~10.0{s} best lap.
I'm not sure using the data above is useful as is. The first lap at the velodrome is probably 150 feet shorter because the start is near the entrance to turn 1 and the loop is at the exit of 4. Also, the track is amazingly slick - so slick that I wasn't to full throttle until into turn 2. The second and third lap was also not full throttle for parts of it due to traffic.
jblackburn
#147
Tech Adept
jblackburn,
I appreciate your description of the practical limitations on the Velodrome. It accurately identifies limitations with the "lap time simulation" in these models.
These models must be simplified, to some extent, to promote clarity and insight. Then the assumptions and limitations of the simulation model must be explained. I've tried to do that here given the space limitations.
Full traction at full throttle with a full battery and fresh motor as on the Dyno is a set of assumptions that one makes to generate the "best case" acceleration. Also in the best case one would specify the minimum for car mass, driveline inertia, driveline damping, battery internal resistance, frontal area, and drag coefficient. An increase in any of these factors adds to acceleration time on the track. On high speed aerodynamic tracks, there is different set of trade-offs.
I offer the models primarily to learn more about the actual limitations from experienced drivers, so I might better explain these concepts to students in the event that I publish my models for science and technology education. I also might be able to improve model accuracy by discovering a little trick here or there, but it is looking like the relationship between throttle profile and available traction profile is too data intensive to build into a simplified simulator.
I appreciate your description of the practical limitations on the Velodrome. It accurately identifies limitations with the "lap time simulation" in these models.
These models must be simplified, to some extent, to promote clarity and insight. Then the assumptions and limitations of the simulation model must be explained. I've tried to do that here given the space limitations.
Full traction at full throttle with a full battery and fresh motor as on the Dyno is a set of assumptions that one makes to generate the "best case" acceleration. Also in the best case one would specify the minimum for car mass, driveline inertia, driveline damping, battery internal resistance, frontal area, and drag coefficient. An increase in any of these factors adds to acceleration time on the track. On high speed aerodynamic tracks, there is different set of trade-offs.
I offer the models primarily to learn more about the actual limitations from experienced drivers, so I might better explain these concepts to students in the event that I publish my models for science and technology education. I also might be able to improve model accuracy by discovering a little trick here or there, but it is looking like the relationship between throttle profile and available traction profile is too data intensive to build into a simplified simulator.
#148
Tech Elite
iTrader: (9)
Here's something from the VELO that might be of more "Performance" assistance.
At this years ALLSTAR SHOOTOUT held just a couple weeks ago, we did our (4) lap - Single Car qualifying.
This is (4) Laps "AT SPEED" - using the STAGGERED START scoring format. (The WARM UP Lap gets the car up to speed, so when he crosses the start/finish line to start the clock - he's running full speed)
This was the fastest TIMED Run from that event.
ALLSTAR SHOOTOUT 2009 04-25-2009
Shootout Qualifying
Pos Car Laps time name id avg.mph
1 1 4 0:39.72 Gary Hamilton 1 56.30
__1__
1 9.79
2 9.87
3 9.95
4 10.09
_____
laps 4
time 39.72
Min 9.79
Again, this was with the NOVAK 21.5 motor and the ORION 3400 Lipo Battery
At this years ALLSTAR SHOOTOUT held just a couple weeks ago, we did our (4) lap - Single Car qualifying.
This is (4) Laps "AT SPEED" - using the STAGGERED START scoring format. (The WARM UP Lap gets the car up to speed, so when he crosses the start/finish line to start the clock - he's running full speed)
This was the fastest TIMED Run from that event.
ALLSTAR SHOOTOUT 2009 04-25-2009
Shootout Qualifying
Pos Car Laps time name id avg.mph
1 1 4 0:39.72 Gary Hamilton 1 56.30
__1__
1 9.79
2 9.87
3 9.95
4 10.09
_____
laps 4
time 39.72
Min 9.79
Again, this was with the NOVAK 21.5 motor and the ORION 3400 Lipo Battery
#149
Tech Adept
SWTour,
Lap Time Deratings During Qualifying
1 9.79s 56.95mph fast lap
2 9.87s 56.40mph -0.55mph
3 9.95s 56.04mph -0.91mph
4 10.09s 55.26mph -1.69mph
In post #145 above note gear ratio G = 85/45 = 1.889.
I change to G = 1.735 and obtain vMax = 56.95{mph} in 11{s} assuming full starting traction at 2 g's. No other changes in the system assumptions.
Dissipates 73{W} to push air and gives back the fast qualifying lap time.
Drag force D = 2.85{N} on the nose of the 1/10 Viper at 56.95{mph}. This should not require any downforce to load the rear drive tires in the straights.
If banking is sufficient to hold one's line at vMax in the turns with a bit less downforce, aero tuning for the Velodrome requires streamlining by reducing frontal area and/or drag coefficient. This brings down the power and force needed to push air, and lets one gear a little taller, to run at greater vMax. Gear a bit taller for qualifying to go faster in 4 laps. Gear a bit shorter to spare battery volts and reduce motor heat build-up for the much longer main.
This ignores aerodynamics in traffic and the draft, which may be significant. John discusses aero stability at speed on a pancar thread (keeping the center of mass in front of the center of pressure, as for a model rocket).
Lap Time Deratings During Qualifying
1 9.79s 56.95mph fast lap
2 9.87s 56.40mph -0.55mph
3 9.95s 56.04mph -0.91mph
4 10.09s 55.26mph -1.69mph
In post #145 above note gear ratio G = 85/45 = 1.889.
I change to G = 1.735 and obtain vMax = 56.95{mph} in 11{s} assuming full starting traction at 2 g's. No other changes in the system assumptions.
Dissipates 73{W} to push air and gives back the fast qualifying lap time.
Drag force D = 2.85{N} on the nose of the 1/10 Viper at 56.95{mph}. This should not require any downforce to load the rear drive tires in the straights.
If banking is sufficient to hold one's line at vMax in the turns with a bit less downforce, aero tuning for the Velodrome requires streamlining by reducing frontal area and/or drag coefficient. This brings down the power and force needed to push air, and lets one gear a little taller, to run at greater vMax. Gear a bit taller for qualifying to go faster in 4 laps. Gear a bit shorter to spare battery volts and reduce motor heat build-up for the much longer main.
This ignores aerodynamics in traffic and the draft, which may be significant. John discusses aero stability at speed on a pancar thread (keeping the center of mass in front of the center of pressure, as for a model rocket).
#150
Tech Adept
Two Good Papers on Brushless Motor Models
Access,
First paper 7 pages includes time constants for brush motor and brushless motor, with an obvious error in Equation (25). I want to be able to derive the air gap equations for a brushless motor similar to equations (1) and (2) for a brush motor, and it should be possible based on this treatment, but I'm not sure how to isolate the equivalent brushless value for Ke and KT.
Second paper is a partial chapter from a brushless motor design book. The basic circuit model is shown in Figure 8-6 with regard to sinusoidal AC motor, but it is the same for the brushless motor with trapezoidal reverse voltage.
This is the essence of a SPICE simulation but as I said it needs a feedforward and feedback signal coupled to the mechanical system response built into the reverse voltage sources.
Access,
First paper 7 pages includes time constants for brush motor and brushless motor, with an obvious error in Equation (25). I want to be able to derive the air gap equations for a brushless motor similar to equations (1) and (2) for a brush motor, and it should be possible based on this treatment, but I'm not sure how to isolate the equivalent brushless value for Ke and KT.
Second paper is a partial chapter from a brushless motor design book. The basic circuit model is shown in Figure 8-6 with regard to sinusoidal AC motor, but it is the same for the brushless motor with trapezoidal reverse voltage.
This is the essence of a SPICE simulation but as I said it needs a feedforward and feedback signal coupled to the mechanical system response built into the reverse voltage sources.