Go Back  R/C Tech Forums > General Forums > Electric On-Road
Dyno, Homemade, Using a Novak Sentry Data Logger, Continued, The Experimental Thread. >

Dyno, Homemade, Using a Novak Sentry Data Logger, Continued, The Experimental Thread.

Dyno, Homemade, Using a Novak Sentry Data Logger, Continued, The Experimental Thread.

Old 05-24-2009, 08:52 AM
  #166  
Tech Addict
iTrader: (1)
 
BobW's Avatar
 
Join Date: Mar 2004
Location: Kingston, Ontario
Posts: 633
Trader Rating: 1 (100%+)
Default

Thought you guys might find this little program useful. I call it the RCVirtualDyno. What it does is allow you to compare staright line acceleration performance of two different car configurations. The primary inputs are the electric motor and car. Lots of other variables can be changed as well.

You can create both car and motor models. The motor model is very simple but complex enough for car level simulation. If you know the Kv, peak power and peak efficiency you can build a motor model.

I used an Eagle Tree datalogger and recorded speed, pack voltage and pack current both on track and on the bench. That data was used to validate the models and resulted in pretty decent correlation.

Screen shots attached.

You need at least Excel 2002 to run the program. There is a crude users manual as well. You can find the program here if you are interested.

http://www.mediafire.com/file/tyntna...lDynoV1102.zip
Attached Files
File Type: pdf
CarScreenC.pdf (385.2 KB, 309 views)
File Type: pdf
MotorScreenC.pdf (310.8 KB, 224 views)
BobW is offline  
Old 05-25-2009, 04:39 PM
  #167  
Tech Adept
 
Join Date: Mar 2009
Location: Metro New York
Posts: 139
Default

BobW,

The software looks interesting. A first look provokes these questions.

1. I note what appear to be slider controls in the CarScreen. Are these added via VisualBasic or C# code? Or standard Excel tools?

2. In the MotorScreen I note a 10.5BL motor with Dyno curves that look like you are applying the theoretical model for a conventional brush motor. Are you calculating the resistance values from other measurements? If so, what parameters did you actually measure?
SystemTheory is offline  
Old 05-26-2009, 03:47 AM
  #168  
Tech Addict
iTrader: (1)
 
BobW's Avatar
 
Join Date: Mar 2004
Location: Kingston, Ontario
Posts: 633
Trader Rating: 1 (100%+)
Default

Sliders are Excel ActiveX Controls controlling values in Visual Basic routines.
Lots of VBA code in this program.

As far as I know the characteristic model for a DC motor is the same wheather the commutation is via brushes or by electronic means. The primary difference is efficiency (thermal and electrical). There are motor models in the program for 3.5 to 21.5turn motors. I built the models from manufacturers published data for Kv, peak power and efficiency (where I could find it). Essentially you build a motor model this way.

Step 1 - Enter Kv (controls max rpm)
Step 2 - Adjust winding resistance until the peak power matches.
Step 3 - Adjust no load current until efficiency matches

It will be interesting to see how the dyno values compare. Give it a try.
BobW is offline  
Old 05-26-2009, 01:50 PM
  #169  
Tech Adept
 
Join Date: Mar 2009
Location: Metro New York
Posts: 139
Default

I've dabbled with Microsoft's free Visual Express C# edition. I think graphic support is not as good as within Excel, otherwise I can picture standalone software with similar features, avoiding the need to launch via Excel. I think a lot of functionality could be achieved with properly designed slider controls.

The problems with a software product like this, as I see it, are designing a good interface, helping the user understand the system model, and explaining the sofware limitations. Since the predictive value is probably on the order of +/- 15%, it should become a tool for understanding systems and how to get tuning improvements.

As I mentioned in a post above, re Inrush Current Limiting, it appears from the Dyno curves that the lower power motors on 8.2{V} Lipo output (with zero advance?) show curves similar to a conventional brush motor, while the higher power brushless motors on 8.2{V} Lipo probably require inrush current limiting that reduces starting torque to protect the low resistance coils in the motor and the low resistance FETs in the ESC. This distorts the curves that might otherwise be similar to the theoretical brush motor performance.

That said, I applaud your effort, particularly on the CarScreen interface.
SystemTheory is offline  
Old 05-26-2009, 04:07 PM
  #170  
Tech Rookie
 
Join Date: May 2009
Posts: 19
Default

Originally Posted by BobW
As far as I know the characteristic model for a DC motor is the same wheather the commutation is via brushes or by electronic means. The primary difference is efficiency (thermal and electrical). There are motor models in the program for 3.5 to 21.5turn motors. I built the models from
Don't forget about ESC 'profiles' or variable forward advance algorhythms / parameters. We've seen what a significant difference this can have on dyno tests earlier in this thread. With brushed, you set forward advance at the motor in a fixed fashion only, typically trading off between power and efficiency. There was no such thing as a variable profile.
Access is offline  
Old 05-26-2009, 05:47 PM
  #171  
Tech Addict
iTrader: (1)
 
BobW's Avatar
 
Join Date: Mar 2004
Location: Kingston, Ontario
Posts: 633
Trader Rating: 1 (100%+)
Default

As I said the motor model is very simple and provides good results for car level simlations. It doesn't include coil inductance, PWM frequency, or timing advance as a function of RPM or load. I'll try and post some of the graphs comparing telemetry data to the program for the 10.5 motor I was running.

My purpose was to provide a tool to look at the effect of changing mass, pinion, spur, tire dia., rotating mass and the multitude of other variables there are to consider on straight line performance. Of course it can be improved upon in many areas.

The motor test results that are posted here would certainly help improve the accuracy of the motor model. I just don't have the time right know.

If this program gets developed further then it would be better to create a standlone program in C# or VB as System Theory suggested. This would take a lot of effort particularly in the area of graphing which Excel does easily.
I wrote the RollCentre/Camber Gain program in VB which has limited graphics and it was not easy. But then I am an amatuer in the programming field.
BobW is offline  
Old 05-27-2009, 04:52 PM
  #172  
Tech Addict
iTrader: (1)
 
BobW's Avatar
 
Join Date: Mar 2004
Location: Kingston, Ontario
Posts: 633
Trader Rating: 1 (100%+)
Default

Some predicted versus measured graphs attached. This was recorded indoor on a 80'x42' carpet track with my Losi R, rubber tires and a 10.5 Novak.
Attached Files
File Type: pdf
TrackTestValidation.pdf (324.0 KB, 250 views)
BobW is offline  
Old 05-27-2009, 07:25 PM
  #173  
Tech Elite
Thread Starter
iTrader: (27)
 
Join Date: Aug 2002
Location: Houston, Texas
Posts: 3,780
Trader Rating: 27 (100%+)
Default

Very nice! Please describe the action on the track. Coming out of a corner, onto straght up to slowing point for next corner?
John Stranahan is offline  
Old 05-27-2009, 07:41 PM
  #174  
Tech Addict
iTrader: (1)
 
BobW's Avatar
 
Join Date: Mar 2004
Location: Kingston, Ontario
Posts: 633
Trader Rating: 1 (100%+)
Default

Tight corner onto the straight. The sharp initial rise in speed is a little wheel spin. Unfortuntely the model doesn't simulate wheel spin. Track is 80' long so the driving line on the straight is maybe 60'. Tight turn at the end of the straight.

This is just a portion of a lap.
BobW is offline  
Old 05-29-2009, 05:51 PM
  #175  
Tech Adept
 
Join Date: Mar 2009
Location: Metro New York
Posts: 139
Default

I agree, nice graphs.

I would look for maybe less accuracy in the model-data layover when running a high power motor and/or a car with taller gears.

It is a curious zig-zag data pattern but on average fits the theoretical curves for this particular system.

Do you suspect any wheel hop could be impacting the rpm and current samples?

What is the data sample rate?

Did you compute the best g-rate of acceleration?
SystemTheory is offline  
Old 05-30-2009, 06:32 AM
  #176  
Tech Addict
iTrader: (1)
 
BobW's Avatar
 
Join Date: Mar 2004
Location: Kingston, Ontario
Posts: 633
Trader Rating: 1 (100%+)
Default

Originally Posted by SystemTheory

I would look for maybe less accuracy in the model-data layover when running a high power motor and/or a car with taller gears.
Not sure why you would say this. The physics don't change. The key is a motor model that is close. Taller gears shouldn't affect anything. Downforce and drag are modeled.

Originally Posted by SystemTheory
It is a curious zig-zag data pattern but on average fits the theoretical curves for this particular system.

I'm installing the data recorder and will try it on our outdoor track. I'll post the results when I get them.
Not sure why. The current is measured on the battery leads. So I would expect the battery capacitance to smooth out anything to do with the PWM frequency or switching. May just be something to do with the data recorder itself.

Originally Posted by SystemTheory
Do you suspect any wheel hop could be impacting the rpm and current samples?
No probably just inaccuracies in the A/D conversion.

Originally Posted by SystemTheory
What is the data sample rate?
As fast as it could go. 40 samples/sec (25 msec). Faster would be better

Originally Posted by SystemTheory
Did you compute the best g-rate of acceleration?
Program graphs accel, speed, distance, and battery voltage and current versus time. Download and try the program.
BobW is offline  
Old 05-31-2009, 06:03 PM
  #177  
Tech Adept
 
Join Date: Mar 2009
Location: Metro New York
Posts: 139
Default

Bob thanks for the information.

If the model doesn't account for current/torque limiting during electronic commutation in the speed controller, which I suspect is necessary to protect the coils and transistors during the low speed full throttle application, I would look for evidence of this error in the data/predicted comparison. If starting velocity is high this effect disappears, as it would occur at the lowest rpm.

Damping in the chassis reflects back to the motor in proportion to 1/G^2. As G goes down with taller gears, any error between the model input value versus the actual value in the system gets magnified. Since damping is hard to measure exactly, I expect a bit more divergence with taller gears.

I don't run Office/Excel on my system at present. If it runs in Calc I may take some time to try it. Right now I'm busy researching an Ubuntu Server build ...
SystemTheory is offline  
Old 06-01-2009, 08:34 PM
  #178  
Tech Addict
iTrader: (1)
 
BobW's Avatar
 
Join Date: Mar 2004
Location: Kingston, Ontario
Posts: 633
Trader Rating: 1 (100%+)
Default

There is a current limit feature in the ESC parameters. What I have seen in track testing is the car itself limits the current as there is only so much torque you can apply at low speed before you get wheel spin. Once wheel spin is induced motor RPM climbs rapidly and the motor current drops equally as fast so the time spent in the extreme high current region is short.

Regarding chassis damping I assume you are referring to inherent damping in the drivetrain due to bearing friction, drive belts and so on. This effect is partially covered in the model by the drivetrain mechanical efficiency term and a friction term. You are right in that there should be another term related to the RPM of drivetrain components. As you pointed out this is difficult to determine. It could be measured by removing the pinion and getting the car up to speed and then recording the deceleration curve. Subtract out the rolling resistance and aero drag and what’s left is the drivetrain friction and damping.

From what I have seen so far this effect is probably minor in relation to drag and rolling resistance. If you know of any sources for Lift and Drag coefficients for different RC bodies that would be very helpful.

I’ve installed the eagle tree datalogger back in my Losi Type R and will do some testing tomorrow at our club race. The model predicts it should reach a speed of 40 mph at the end of the straight. Max speed is 42 mph. Gearing is 46/118 (43.3 rollout , 4.57:1 final drive ratio), Novak GTB and 10.5SS Pro.
BobW is offline  
Old 06-04-2009, 06:26 PM
  #179  
Tech Addict
iTrader: (1)
 
BobW's Avatar
 
Join Date: Mar 2004
Location: Kingston, Ontario
Posts: 633
Trader Rating: 1 (100%+)
Default

Test results are attached. There are three test with three graphs in each PDF file. Speed, current and voltage. Car motor, battery and gearing are all the same for each test. Current data is still very noisy, not sure why.

Test 1 is a bench test with the wheels up and pinning the throttle. I just ran this test so the battery volts are a little low. To get a comparison with the model the car mass, drag and lift coefficients are set to zero since the car is not moving. I also adjust the battery zero load voltage to match the battery state of charge. This test gives a good indication of the drive train losses and rotating mass. Looks like I need to work on my drive train as the last time I did this test the terminal speed was almost exact. Now it is out by about 1.5mph.

Test 2 this is a track test I recorded Tuesday. It is scary how close the speed graphs are. Again all I adjusted was the battery zero load voltage. The straight at this track is slightly down hill which may explain why my the recorded top speed is closer to the predicted than in the other two tests.

Test 3 is a road test to see how close the top speed is. This was run on a long flat section. Top speed in this run is off by 1 mph.

All in all not bad I would say.
Attached Files
File Type: pdf
Test1_Bench.pdf (21.4 KB, 199 views)
File Type: pdf
Test2_Track.pdf (30.6 KB, 180 views)
File Type: pdf
Test3_TopSpeed.pdf (32.6 KB, 191 views)
BobW is offline  
Old 06-05-2009, 01:03 PM
  #180  
Tech Adept
 
Join Date: Mar 2009
Location: Metro New York
Posts: 139
Default

Noise in the three phase switching system of a brushless motor occurs on both the power supply bus (inductive voltage spikes) and due to radio frequency induction (antenna effects picking up radio signals on unsheilded wires). This may explain some of the zig-zag data patterns.

Current is switched in three phases and may not be smooth on the battery. It also may be limited via the ESC in a choppy on-off pattern. One trick to smooth your current sample would be to apply a true RMS filter and see what you get back.

http://www.dataq.com/applicat/articles/freq_rms.htm

I've never run such a calculation but this looks like a correct approach.

The RPM signal looks pretty good, especially at high speeds where wheel spin is unlikely, and in any case one is not likely to get better than 10 or 20% expected difference between the computer model and sensor measurements without a NASA budget.

Engineering papers list 13% ripple for back-emf voltage, torque, and current for brushless motors designed with trapezoidal back-emf waveforms. These are the least expensive systems and apparently the ones used in RC.

Last edited by SystemTheory; 06-05-2009 at 01:06 PM. Reason: Ripple
SystemTheory is offline  

Thread Tools
Search this Thread

Contact Us - Archive - Advertising - Cookie Policy - Privacy Statement - Terms of Service -

Copyright © 2024 MH Sub I, LLC dba Internet Brands. All rights reserved. Use of this site indicates your consent to the Terms of Use.