Regen Experiments

Progressive Regenerative Braking

I'll begin by saying that, in my opinion, the only need for regenerative braking is to increase a vehicle's range.  The mechanical brakes work just fine for retarding forward motion.  However, that opinion may not be shared by everyone.  Cindy loves variable regen braking as she feels it gives her much finer control than the rear brake offers.

Control of regenerative braking can be divided into two categories: Vehicles that have a mechanical clutch and those that don't. I developed regen braking for the latter class of vehicles and described it fully in the EM 5.7 section.  Vehicles with mechanical clutches offer new challenges and opportunities.  The ePure Race is such a vehicle.

EM's optional solution is shown below.  It's called “PRB R” and mounts in place of the clutch master cylinder reservoir cover and plugs directly into the SiliXcon controller.  It is EM's P/N TC01Q-60403-00-00 with a US retail price of $279.  This actually seems like a reasonable price to me – especially compared to the cost of some EM parts.

It does seem like an elegant solution, but far from optimal for trials riding.  I imagine transitioning your “clutch finger” to the PRB lever and back would not be smooth.

Credit: Electric Motion PRB R video

The active electronic component inside the device is a linear Hall effect sensor manufactured by Littelfuse (P/N 5100-AP-02-A).  These cost less than $20 in single-piece quantities, but as of September 2022, there was no stock anywhere in the US.  The expected delivery was a year out.

Performance Metrics

How can we evaluate the effectiveness of regen braking?  The intangible “optimal feel” is important, and likely differs greatly among users.  But riding time/distance is not intangible.  The Cycle Analyst has been a valuable tool in helping quantify my intuition.  Prior to having any data, I guesstimated that as much as a 10% additional range could be achieved on the EM 5.7 via regeneration.  Maybe so, but it's not likely for my riding.  Based on data from the Cycle Analyst over a few rides in a single location, 5% would be a more achievable goal.  That number is the proportion of regen Wh versus driving Wh.  But just because you put “X” Wh into the battery does not mean you will get the full “X” Wh back out.  Based on the battery charge/discharge testing I conducted earlier, I would estimate the process is about 88% efficient (e.g., it took 1828 Wh to charge the battery, but only 1610 Wh came back out).

PRB Input

The adjacent table shows EM's wiring scheme for  Progressive Regenerative Braking.  These signals are accessible via a 3-pin JWPF connector into the SiliXcon controller.  I must say, the controller's regen behavior is much better than that of the Kelly controller on the 5.7.

Push Button Regen

The prior owner of my bike had wired a regen push button near the left handgrip.  It routed +5 volts directly into the controller's analog regen input. Pressing the button produces maximum regeneration (which still seemed fairly weak to me).  In retrospect, this was actually a pretty good solution.  But I wanted to get a feeling for any advantages variable regen could offer.

Credit: eBay

“Finger Throttle”

I replaced the push button with a Chinese “finger throttle” (I use my thumb instead) intended for use on an electric bicycle.  This is a non-contact Hall-effect device that produces 0.8 to 4.2 volts depending on lever position.  It does work but is really awkward/uncomfortable if you are trying to simultaneously one-finger the clutch lever.  It did show one interesting feature in that at 0.8 volts, some “automatic” level of regeneration was present when the throttle was fully closed.  This proved to be desirable.  These Chinese finger throttles come in two cable lengths – the long one is only pennies more than the short one which is too short.  Although I did find myself varying the lever position somewhat, maximum regen was usually what I wanted.

Aux Potentiometer

A better solution was discovered by a friend who also owns an ePure Race (two of them, actually).  It's also intended for an electric bicycle and can be mounted on the handlebar near the left grip and spun with the thumb.  It is a $25 accessory manufactured in Canada by Grin Technologies.  It can be used to vary the regen level from 0.5 volts to 5 volts.  Their cable is just long enough – only the connector needs to change.  Although this may be the perfect solution for my friend (he has lots of long downhills in mountainous terrain) I found it inconvenient and uncomfortable to adjust on the fly.

Credit: Grin Technologies

Potentiometer used as voltage divider provides a small amount of automatic regen on closed throttle.   Push button then provides maximum regeneration.

Hybrid Design

A hybrid design uses a push button and a fixed resistive divider or potentiometer.  Basically, the fixed divider always generates something around 1 volt and produces regeneration when the throttle is completely shut.  This provides some minimal automatic engine braking.   At about 1.1 volts there is too much regeneration for my taste.  Much beyond 1.1 volts has the side effect of diminishing the throttle response.

The push button is used to short out the upper resistor in the divider thus allowing the full 5 volts into the controller's regen input.  I found this to work nicely, but the old Leonelli kill button I had laying around was unreliable.  I would suggest using an OE kill button from a Japanese MX bike instead.

The controller's regen input impedance is 68k ohms and will load the voltage divider's output.  I made a JWPF Pickoff Connector (described in the section on the ePure's wiring) to adjust the voltage divider while it is under load. 

Integrated Rear Brake Regen

For me, the perfect regeneration method would be incorporated into the rear brake.  To that end, I made a special banjo bolt for the rear caliper.  It accepts a 1/8" NPT pressure transducer.  I started with 19mm round stock.  The thread is M10-1.0.  There is a 14mm hex for tightening.

Chinese pressure transducers are available very inexpensively (about $14) on eBay in the following full-scale ranges: 5, 15, 30, 100, 150, 200, and 300 psi.  They all have a burst pressure in excess of 600 psi and use Sensata's connector standard.

When I experimented with this idea for the electronic clutch lever on the 5.7, I found that the clutch master cylinder would produce 200 psi, but the finger effort was extreme.

I selected a 100 psi transducer for initial testing.

Custom banjo bolt to accommodate pressure transducer

Pressure transducer and adapter installed 

Installation

Adjacent is a photo of the pressure transducer installed on the bike.  

I would have liked to get this tucked out of the way near the rear master cylinder, but there was no room.  

When powered by a 5 VDC reference, full-scale pressure (100 psi) produces an output of 4.5 volts, whereas zero pressure is 0.5 volts.  Thus the span is 4.0 volts (4.5 – 0.5) for a 100 psi change, or a slope of 25 psi per volt.  

The supplied connector is a bit bulky.  Assuming it all works, I plan to just solder on some wires once everything is sorted.

Difficulties Bleeding

Unfortunately, I found it very difficult to bleed the rear brake!  There was a lot of air in the system when I first attempted to take a pressure measurement.  At first, I could only generate 30 psi.  More bleeding, then 50 psi.  More bleeding and finally in excess of 100 psi.  At 50 psi there is a minor drag on the wheel, but you can easily turn it by hand.

None of my usual tricks for bleeding worked.  I backfilled the system by injecting brake fluid into the bleed nipple.  I used a Mity-Vac to suck brake fluid through in the usual direction.  Of course, I tried manual bleeding (while elevating and tapping on various parts of the system while doing so).  Because the brake hose is routed through the swingarm, it's not worth the trouble of removing the swingarm in order to elevate different parts of the system during bleeding.  I had the same trouble with the 5.7 (which has the same rear brake hose routing).  Because the 5.7 had lain on its side returning from Texas, there was zero rear brake when we got it home.  I had to literally hang the rear of the bike from the ceiling in order to get all the air out.  I'll probably have to do the same with the ePure.  I can't really blame EM for this since it is a Sherco swingarm.

Signal Conditioning

I expect to have some analog signal conditioning electronics (just an op-amp hopefully) between the pressure transducer and the controller's PRB input.

Ideally, pressing on the brake pedal will initially produce regen braking with no or minimal mechanical braking.

I'm hoping that some combination of pressure transducer full-scale range, post-signal amplification/limiting, and air in the brake system will give me the feel I want.  A cheap Chinese eBay product may be useful here.  I have a hydraulic ABS attachment for motorcycle brakes on order.

Unfortunately, I don't expect to be able to field test a prototype before the weather turns.  Some lab experimentation will be necessary anyway.

Encouraging Initial Results

The weather was favorable for some driveway testing - although I had to rush a bit to get everything ready (was in the middle of a tire swap).  Unfortunately, my Chinese ABS reservoir got delivered to the wrong buyer, so I won't be trying that anytime soon.  Not sure how long it will take to get a replacement.

My first test in the driveway was using a rock-hard pair of Jitsie / GoldFren 173 brake pads.  I originally bought these for my OSSAs, but they were completely unusable.  I considered throwing them away numerous times but figured they might be good for something someday.  That day has not yet come.

I thought the hard pads might work because I wanted the regen to do most of the initial deceleration.  Between the hard pads and having air in the system, the stopping power was insufficient.  I did however see up to 25% regen on the Cycle Analyst with those pads.

Subsequent testing with the stock pads yielded only about 10% regen and the stopping power is almost good enough (there was still air in the system).  Other numbers from the Cycle Analyst: Maximum driving current 45A.  Maximum regen current -13.2A.  Peak speed of 13.6 mph, an average speed of 5.3 mph.  Total distance 0.81 miles and 41 Wh/mile.

I'm encouraged by these preliminary results.  My biggest fear was that I would lose the ability to simultaneously use the rear brake and throttle.  I sometimes play one against the other in very tight, very slow speed turns.  After a certain level of regen (command voltage) is reached, the SiliXcon controller no longer takes input from the throttle.  I tested that a lot in my driveway and had no difficulty.  Because of the ePure's mechanical clutch and its precise throttle, it's easy to achieve very slow, very smooth low-speed turning and I had no need to simultaneously drag the rear brake.

MCP6001 SOT23-5 op-amp on adapter PCB

Op-Amp

The initial test was performed with no post-transducer signal processing.  Although the transducer is specified to deliver 0.5 volts at zero pressure, it's more like 0.4V.

I'm thinking about using an op-amp with a gain of 1.7 (but probably making it variable).  A suitable op-amp would operate from a single 5-volt supply and have a rail-to-rail output.  I had the Microchip MCP6001 on hand.  It is only available in a small SOT23-5 package so I used a little SMD to DIP adapter PCB that's sold on eBay.

This amplification would yield about the maximum allowable regen with a closed throttle, and full regen at around 85 psi system pressure.  If those endpoints turn out to be sub-optimal, I could build a more complex circuit.

Chinese ABS Module

A replacement ABS module finally arrived.  I mentioned this item briefly in the 5.7 section on the design evolution of my electronic clutch.  At that time, I had assumed it worked something akin to a hydraulic accumulator and would introduce some sponginess into the feel of the brake system.  Nope.  It's actually a pressure-limiting valve as discussed below.  I'm not sure if it (or something like it) will be useful or not.  No instructions accompanied the unit.  I had assumed the gold knob” was an adjustment.  Wrong again.  It's just a cap - the setscrew it fits onto can be used to provide some preload to the spring.  But the O-ring is the only seal and the gold cap needs to bear against it.  So if there is any adjustment, it's quite limited. 

ABS Module Inner Workings

It appears the inlet is open to the outlet when the pressure is below a threshold.  Changing the spring would affect this shut-off pressure.

The spring rate calculates to 194 lb per inch.

8 mm piston diameter = 0.077 sq. in.

Spring measured about 0.156" to coil-bind

194 lbf/inch * 0.156" = 30.26 pounds force

30 pounds force / 0.077  sq. in. = 393 psi

One problem is that the expected caliper thread is M10 x 1.25.  Trials calipers are M10 x 1.0 so I need to make a new adapter.  Note: The module end of the adapter has an M10 x 1.0 left-hand thread. 

Chinese ABS module, showing inside parts

April 2023 Testing

After purging all the air from the rear brake (a process I detail in the section on Bleeding the Rear Brake) I was ready to test again.

Basically, the system worked (fairly well) for about a minute.  Then under a very hard stop, I suddenly had zero rear brake.  Based on the circumstances, I had a pretty good guess as to what had happened.  I electrically tested the pressure transducer and found it to be completely kaput (no output whatsoever).  I surmised that the pressure-sensing element had ruptured, and this increased the volume of the hydraulic circuit significantly (hence no mechanical brake either).

Later inspection confirmed this, and I photographed the results for posterity.

Left: Ruptured Transducer.  Right Intact Transducer. 

I had used a 100 psi transducer for a variety of reasons even though I assumed the pressure could be well above that.  The $14 Chinese sensor was rated for a burst pressure of 600 psi.  I now realize the failure pressure is considerably below the burst pressure.  Several things were learned:

1. It is very difficult to bleed the rear brake with the pressure transducer as installed.

2. The presence of the transducer makes removing the rear wheel more difficult and I could see it being damaged in a crash.

3. A much higher-pressure transducer will be required, plus some signal conditioning electronics to make this design work.

4. If the transducer does mechanically fail, you will have zero rear brake. 

I have ordered a 1000-psi transducer and will build some signal conditioning electronics to produce a 5-volt output at the appropriate pressure (which would have to be determined).   I'll report back on this topic as progress is made.

Preferred Embodiment, For Now

The adjacent photo shows the regen solution I've settled on, for now.   It is the Chinese Finger Throttle (described earlier) in a location that allows me to actuate it with my thumb while still being able to quickly transition to one-fingering the clutch lever.   This also provides some fixed automatic regen (although slightly less than I find ideal).

The only downside is that with any fixed regeneration that's greater than about 0.6 volts, the SiliXcon controller sees this as an error condition at bootup.  The result is that the motor no longer produces a tone when the map button is pressed.

Location of thumb-actuated regen control

2024 Regen Solution

When I returned to using a lanyard kill switch, the tether interfered with my operation of the thumb-actuated regen lever so I again implemented the fixed automatic regen described below.  This has the side benefit of removing a cable to the handlebars.

I also found that using large amounts of regen in conjunction with my 10T front sprocket caused a funny sound.  I think the sound may have been due to the action of the chain tensioner, but did not investigate.

Potentiometer for fixed regen.

Automatic Fixed Regen

I configured a 10k ohm potentiometer as a voltage divider to provide a constant fixed voltage into the PRB input.  I adjusted the potentiometer to provide 1.0 volts.  This produces an automatic regen whenever the throttle is closed.  It provides engine braking that's not quite as powerful as a typical 2T trials bike in 3rd gear.  I like the familiar feeling but do not expect it to extend the bike's range.   The Cycle Analyst showed only 0.2% regen (equivalent to 0.2% more range) during a 39-minute test ride.  Although the Cycle Analyst calculates this number, I checked it against other values it provided of -0.0151 Ah during regen versus 5.40 Ah during driving.