justin_le wrote: ↑
Mar 11, 2018 3:31 am
What I really wanted to do was then try setting things up for a similar row up the mountain but with the CA set to a speed control loop instead of a constant power loop. In theory, that should provide extra power to the motor during the recovery stroke and reduced power on each pull stroke so that the total power (human + motor) remains more or less constant at whatever power is required to sustain the speed set point.
Sooo, with all the batteries properly secured and with the CA3.1 firmware now having the 10Hz data logging rates I finally had a chance to explore this.
What I did last week is take the rowing bike up the same hill used in the Brompton Statorade
video, which gives a sense of the elevation and grades, and compared the performance with the CA operating in constant power assist mode and constant speed assist mode. The results are pretty cool.
First, the benchmark climb without assist. At the base I could sustain about 10kph but that fell to 6-7kph during the steeper bits of the ascent. The data shows the actual speed fluctuation on each rowing stroke was only ~1.2 kph on average, but it sure felt like more than that. As you decelerate on the recovery stroke it really seems like you need to start pulling immediately or risk slowing down to the point of loosing stability. Around half way up the hill I had to pull over to let a car pass if you're wondering why it goes to zero.
Then the next bit gets interesting, I set the CA3 to be in a constant speed limit mode of 10kph and repeated the same climb. In order for the CA3 to have a decently responsive speed feedback loop with the Phaserunner I had to increase all the CA's PID gain parameters by about 3 fold from the default values.
Here, you can see the CA holds the speed fairly steady between 10-11 kph over the entire climb, the amount of oscillation in the speed is reduced to about 0.7-0.8 kph on average (different speed axis autoscaling with preview graph exaggerates this difference), and you see the current fluctuating up and down by about 5 amps (~180 watts) during the climb.
When we zoom in more closely on the data, you can see that the current spike occurs exactly at the dip in the speed. Initially when the hill isn't very steep, there is no power from the motor except during the recovery stroke. So it's like a perfect alternation of human power, motor power, human power, motor power keeping the bike at speed. There is even a bit of regen just before the 0.24km mark when I'm pulling faster than the target.
Further along as the grade increases, there is a net current required to keep the rowbike at 10 kph with the ~5A pulses at each stroke super-imposed on that. What I found interesting here is that you can see in in most of the strokes that there is a dip in the speed hump while I'm pulling the rowingbike, and this gets reflected by a double pulse of current. It's as though I'm pulling hard at the start of the rowing stroke, then relax a bit in the middle, and end with a firm tug. I'd love to have a force/torque meter on the system to confirm if this is indeed the case! Or if instead it is caused by the response of the CA3's feedback loop having an initial overshoot.
Anyways, the climb at 20kph is similar but with more net motor power. There's an overshoot to ~22 kph when the hill levels out in the middle
Meanwhile, the climb with constant power rather than constant speed behaves quite differently. Here it is with a 300 watt target.
The current stays at an average of ~8 amps, with brief downwards spikes at the the high speed point of each stoke. Meanwhile the speed of the rowingbike varies from 10 to 22 kph between the steeper and shallower sections of road. This works fine too, but overall I prefer the climb in constant speed mode, with the motor not only filling in between rowing strokes but also adjusting automatically for the grade hill.
All the data presented in the post is available here:
http://www.ebikes.ca/tools/trip-analyze ... rip=J98D1x