Return to Main Menu

eAssist Mechanisms



eAssist Theory

Slap a motor on a vehicle, and just GO. That's the easy part. This page covers the factors that allow a person to get on a vehicle, take it for a spin, and gave a grin on their face when they get back after their very first 'test' ride. The following post by Justin on ES in 2013 summarizes the issues to consider.

Post subject: Re: Weight Sensing Longboard with Inline Wheel Motors
PostPosted: Thu Apr 25, 2013 11:31 pm
Location: Vancouver

... I used to believe I preferred throttle control over a torque sensing pedalec style ebike. But then, once I started riding more ebikes that had nicely integrated pedalec I realized that for the most part this was a nicer and more natural control interface between the human and the bike, since it's using the same input (your leg power) that you already use. Similarly, since leaning is such a core part of how you maneuver a skateboard (or surfboard, snowboard etc) I think that it would be a more natural control extention for motor power than moving one of your digits across a throttle. If you've ridden a throttle controlled board, you learn quickly that you need to lean forwards first before you engage it or the board will shoot out from under you. So the leaning already has to take place, and the brain has to learn to do that first in anticipation of you hitting the throttle.
In any case, our brains have a great way of figuring out control schemes and turning them into an extension of ourselves regardless of how they're implemented, so pretty much anything can work. Like gamers who achieve all kinds of delicate feats from pushing buttons on a keyboard or control pad, even though those button presses have nothing to do with the action going on the screen. We are used to pushing foot pedals to accelerate and brake a car, but why not instead have little throttles and buttons on the steering wheel for that? Or why have a steering wheel at all when you could use your feet to turn and then use your hands on a joystick to accelerate? I don't know if anyone's actually studying what is the 'best' way for a human to control a 2000lb vehicle, but given that we're no longer constrained by having to do everything with mechanical linkages like in the old days it's a question worth asking.



PAS(Cadence) eAssist Overview

Refer to Grin's comprehensive PAS page for a great selection of "off-the-shelf" PAS hardware and how to use it. This page also covers CA3 Aux Input Devices and how to best use them with PAS setups, as well as setting up basic PAS sensors with CA3.1 Firmware. If there are issues with PAS and the CA, this page should be your first stop.



Mixing: Throttle, Torque, PAS(Cadence)

Credit for this outline is given to many members on the ES forum, such as Justin (above), and specific posts and threads including this one by user "Avitt".

An "ideal" e-assist for lightweight electric (assist) vehicles (LEV's) will need to consider the following conditions, and would likely involve a "mix" of all three sensor inputs:

1) Takeoff from stop
For both convenience and safety reasons, getting an LEV rolling smoothly and predictably from a dead stop is critical. When the light turns green, you want to be clear of that intersection as quickly as possible.
s With a PAS sensor alone, there is some delay before the power comes on. If the LEV is using pedals, this is typically a half-revolution of the cranks. A LEV is heavier due to the addition of the motor, controller, battery, and "beefed up" components. If 'downshifting' is available, it will typically need to have been done before stopping -- not always an option used. This condition is best addressed by use of a throttle or torque sensor. Note that many control systems using a throttle require the LEV to be already moving (2-3MPHP, or "walking speed") for safety reasons. This severely limits the use of the throttle to assist in getting going from a dead stop. There is no such limitation for torque sensors.

2) Maintaining desired speed
Once safely underway, there are many radically different expectations that need to be met by the LEV's assist sytem. The rider may want a lot of exercise, or none at all. The only way to meet the needs on this spectrum is to have an assist system that can be 'dialed in' from fully "OFF" to "WOT" (wide-open throttle).
If using torque sensing only, the rider must be constantly supplying some amount of force to maintain the desired speed. If using a 'mixing computer' such as the Cycle Analyst, there is a lot of room for adjustment, but because most sensors are limited in their detection range, this amount of required force is often relatively large.
A "throttle only" system can meet this condition in that a rider can use as much assist as desired, but it requires constant attention and decision making. (Some systems support a "cruise control" mode in which speed is maintained at a level determined by holding the throttle at a constant level for several seconds.)
A PAS only system requires that the rider is constantly pedaling, or rowing, or moving whatever mechanism this particlar LEV is using for propulsion. It may not require much effort, depending on the level of assist that is selected, but it does require constant motion. The advantage of PAS is its intuitive interface -- the faster you move, the faster you go. The amount of exercise you get is dialed in via the "assist level".
In some jurisdictions (countries), this PAS "Pedalec" system is a legal requirement, so the rider has no choice about its use.

At this point it should be fairly obvious that my preference for a universally ideal system (if it could exist) would be an LEV that included sensors of all three types: manual throttle, torque, and cadence (PAS). A programmable computing device, such as the Cycle Analyst, will accept the input from all sensors and mix/blend/condition them into a single signal going into the controller, which determines when and how much the motor(s) spin. Note that 'garden variety' controllers for BLDC motors, such as the "Infineon" type and their derivatives, will typically have limited inputs, including 'throttle only', in which case a 'mixing' device is mandatory if you want to use multiple sensors. Having the ability to combine all sensor inputs upstream from the controller, and giving the controller a single 1-4 volt DC "throttle" signal, may make the tuning process for the LEV much simpler. The computer determines the performance characteristics of the LEV, and optionally provides levers, buttons, etc. to the user/rider to customize their riding experience based on circumstances or their preferences on that particular ride.


See also: Cycle Analyst (v3) for CA specific throttle setup.
A throttle can be either resistive (Magura) or Hall Effect (typical chinese twist grips). No matter which system is used, any throttle device will supply the controller's "throttle" input port with typically 1-4 volts DC representing a fully closed or wide open ("WOT") throttle.

Regen Braking

Justin answering ES questions: Justin You really want the hand opposite the throttle to be the one that activates the regen button, ideally just via an ebrake swtich on the brake lever, but otherwise a stand alone push button on the left side would do the trick if you have the throttle on the right. other: Having gone through the various alternatives in my head, I feel that a second left-handed thumb throttle would be the best way to implement proportional regen. Justin That would also work just great. There is no problem at all parallel connecting two throttles together and combining their outputs. Since the throttle output signal can source current but it can't sink current, then whichever throttle has the higher voltage signal is the signal that dominates. So simply split your throttle wires t


Strain Gages

("gage" is generally considered the American spelling and "gauge" the British form) The reason that strain gages are typically used in pairs is that you can combine both tension and compression in one strain gauge, and it also doubles the signal strength (one sensor's resistance is changing up, while the other's goes down, and the connections are made in a way to read the total change).
Justin's advice is "For the best sensitivity with strain gauges, you want to build a full bridge with two gauges in compression and two in tension."
From the same ES forum thread referred to above, he says the following: (Posted: Thu Mar 19, 2015)
Strain gauges are just resistors etched in a zig zag grid, and when the surface of the metal is under tension, it stretches, elongating and thinning the resistor wire causing its resistance to increase tinyest amount (less than 1%). Similarly, when the metal surface is under compression, the resistive wire gets shorter and fatter, lowering its resistance by a fraction of a percent. It's one of the most deceptively simple and ridiculously accurate electromechanical sensor devices, but it does require a high precision low drift instrumentation amplifier to turn the bridge output into a 0-5V signal.

I already detailed the amplifier circuitboard and strain gauge operation on this thread here

I got most of my strain gauge education from Richard Nakka's rocketry website way back in the day. 15 years later his website is still there and still a great reference:

Strain gauges are manufactured by Vishay (very high end, expensive), Omega (good, but "cheaper"); Revere; Micro Engineering (the kind we got). For a very extensive reference to all things "strain gage" -- see this online PDF reference by Omega Engineering: PRACTICAL STRAIN GAGE MEASUREMENTS

The measure of the gage's resistance change with strain is GAGE FACTOR, GF. Gage factor is defined as the ratio of the fractional change in resistance to the fractional change in length (strain) along the axis of the gage. The gage factor is a dimensionless quantity, and the larger the value, the more sensitive the strain gage. Gage factor is expressed in equation (see original PDF). The gage factor for Constantan and nickel-chromium alloy strain gages is nominally 2, and various gage and instrumentation specifications are usually based on this nominal value. This means that twice the strain results in twice the change in resistance -- a strain gage is "linear".

Close-up of a Strain Gage


SR-4 Strain Gages

Same model as used on the original moon lander in 1969
photo of SR-4 strain gages

MOUNTING Strain Gages

This section covers the mechanical process of "gluing" strain gages to some substrate.
Useful material from Keith J. Wakeham, B.Eng, M.Eng, re Strain Gages on a Bicycle Crank Keith has a history of using strain gages in the bicycle world, and somewhere around 2014 ended up working for, making "power meters", which do just what we want int terms of measuring the rider's torque and cadence input, but just *report* this data to the rider instead of translating it into a throttle signal.

Solder on fine wires [directly to the foil of the strain gage], usually 30--36 AWG wires [I can use the very fine pre-tinned (silvered), insulated wire I have in 6" precut lengths] to the pads. You'll see my new version using bondable terminals as it's easier to bond small wires to gauges and big wires to the terminals. It also provides strain relief. Two gauges are wired in series and measured against the mid point of a voltage divider creating a half wheatstone bridge arrangement. [I think Keith put one strain gauge, at 120 ohms, on the top edge and one on bottom of a crank arm, connected in series to make up one side of the wheatstone bridge, and two additional 120 ohm resistors (small ones) in series to complete the bridge] Well, I don't want to break the gages so lets protect them. Epoxy, 5 min kind. Very flexible and soft compared to metal and the superglue glueline. [looks to be a Permatex brand, general purpose two part expoxy, "Perma Oxy"]

Youtube Videos

35 min video - great!
I also looked at the following to "confirm" strategy:
burg0183 for UofM - "Strain Gage Tutorial"
queenesfahan - U of Concordia, Montreal; Strain Gauge Installation Part 1
part 2 of above video
Univ of Jordan

Suggestions for installing a Strain Gauge

This is a compilation of Suggestions, mostly from Keith Wakeham:
- work on stack of clean paper, discard layers as you go along
- thoroughly clean area - mask off area (optional) with good tape
- degrease with acetone, wipe off in one direction;
- degrease before sanding to prevent forcing grease into sanding grooves
- sanding to bare metal, not much finer than 400 grit, use wet sandpaper (vishay recommends mild/diluted phosphoric acid)
- some recommend sanding in circular patterns to prevent "channels" that can affect strain readings
- neutralize with light ammonia water (ok on skin) - "neutralizer"
- find a clean piece of plastic/glass as a 'staging area' (this is the surface you use to stick gage to tape before application to host)
- do not grab a gage by the fine resistance lines -- grab corners near foil/solder pads, and only with tweezers
- put strain gage on staging area, top side up
- use "strain gage instrumentation tape" (i.e., weakened, transparent, 'reusable' tape)
- put the tape over the gage lying on the staging area, starting on one end and "laying" tape over gage so no air bubbles are trapped
- when top surface of gauge is bonded to the sticky side of the tape, peel back one end using shallow angle
- this is intended to remove both tape and gage from the staging glass without the gage separating from the tape or bending
- align the gage over the host, using the 'alignment marks' if necessary to get the right position
- press one end of the tape down and using your thumb, smooth the tape down forcing out all air bubbles
- verify position of the strain gage on the host
- at shallow angle (again), peel back the gage with its carrier to expose the bare metal of the host
- take 'off-the-shelf' superglue (cyanoacrylate) [vishay of course has special stuff for all of this]
- put a line/bead of glue at the beginning of the gage space, and then 'roll out' the gage/carrier over the glue [much like Polaroid film packs used to be processed by being squeezed between two rubber rollers which evenly distributed the chemicals -- the key here is to get just the right amount of glue to evenly cover the gage and not have too much/excess squeezing out, making an ugly glue line]. You're aiming for an 8 micron layer of glue...
- apply steady pressure over the gage for at least a minute -- maybe 5 mins; test will be if edges stay down if trying to pick up with pick
- after glue has set (allow 24 hrs for *full* cure), peel back carrier tape, this time at oblique angle, since gage stays on metal host
- pull back the tape just barely past the solder pads, leaving the rest of the tape to protect the gage while attaching the wire leads
- solder fine wires to gage solder pads; Keith uses 34 ga. copper "magnet wire" which has a fine (enamel) insulation layer
- you do NOT want leads to touch bare metals, since it will skew the ohms reading, plus pick up lots of RF, since metal acts as antenna
- do not use any wire that's already been bent!
- HBM says use solder tip temp of 250-270 degrees Centigrade
- all parts to be soldered must be "wetted" with flux -- which is colophonium dissolved in "methylated spirits" (alcohol?) ...
- solder has "colophonium core"?
- "tin" both the solder pads and the wire ends; Keith applies liquid flux directly to gage pads before tinning them
- when solder on pads is cool, remove excess flux with rosin solvent (alcohol?)
- flux is mildly conductive, so it can act as a bridge/shunt/short between solder pads, which kills the gage's accuracy!
- flux cleaner can also be conductive, so allow it to completely dry before testing
- check for continuity between gage and ground (host bare metal) -- you do NOT want ANY continuity here (more than 10K megaohms!)
- protect the strain gage using air dry polyurethane; still have masked off area ... press wires down against metal to keep a low profile
- maybe use a second layer of polyurethane; you don't want to have any conductivity for this protective stuff (5 gigaohm resistance is ok!)
- I wonder if the black plastic self-bonding tape (Duluth Trading) would be good for protecting strain gage on round handlebars?

Load Cells

A "load cell" is simply a substrate with one or more strain gages already attached to it.
Typical wiring for load cells: (if only three wires, the negative input/output wires are common)
INPUT wires: (Excitation, typically 10VDC) Red+, Black- OUTPUT (Signal) wires: Green+, White-
These load cells are from postal scales rated for 1Kg and 5Kg. Because they have 4 wires, we know they use a full bridge strain gage configuration.
photo of load cells

Instrumentation Amplifier ICs and modules

Integrated "in-amp" ICs (e.g., IN-128)

These chips are differential amplifiers with a very high input impedance. They include the TI INA-122, INA-125, INA-128, INA-129, and the AD623, AD8553, AD8556, AD8221; MAX4194, LT1167.
Digikey sells the TI INA128P [Part Number INA128P-ND] for $11.31 (Feb 2014)

Deb got 2 of these, but for $4.68 each via US seller on ebay. These chips are standard 8 pin DIPs.
photo of invoice
An instrumentation amp is a package containing two op-amps to filter/buffer/clean up the input voltage/signal, and another op-amp that multiplies the cleaned up input by a factor of "x" (gain), which is controllable by the user of the instr. amp via an external resistor (R-gain).
Another property of an instrumentation amplifier is that it has a high common mode rejection ratio, which means that if there are very quickly changing values in BOTH SIGNAL INPUTS, it dampens/smoothens/averages these changes.
The idea is to focus the measurement on the small differences (voltage swings) between the two inputs., not on any larger signals common to both inputs. [Greg Lewin:]

Tringa designed/built Instrumentation Amplifier module (INA)

See dedicated page to Tringa's INA

Because the eRowBike1 does not use PAS as part of its e-assist strategy, we decided to design a simpler (fewer inputs and outputs) strain gage amplifier circuit. We assumed that we'd obtain our torque signal from a Wheatstone bridge strain gage configuration that was excited at either 5 or 10VDC, either which is available from the CA. The amplifier module's output would be mapped into the 1-4VDC range, maing it compatible with a typical ebike throttle.
We selected an Instrumentation Amplifier chip that only needed a single voltage (rail) input, and would allow sufficient gain and filtering to utilize the expected 4-5 millivolt (mV) signal coming from the strain gage array.

photo of Tringa INA board with strain gage inputs

Justin's Complete Instrumentation Amplifier module

We also got a completed Instrumentation Amplifier board from Justin at in Vancouver, BC for $40. This board will support both a half-wave and a full-bridge strain gage configuration. It uses a high quality dual op-amp (Justin had these on hand for CA manufacturing).

Has onboard linear regulator for use with 6-16V DC supply Can be run from 5V by putting jumper on R11 Default gain of 1000x. Set by resistors R3/R6 and R8/R5
For Half Bridge, connect signal to S2 For Full Bridge, remove resistors R9 and R10, and connect bridge signals to both S1 and S2
photo of GRIN Amplifier board with strain gage inputs
Online documentation for this board

Correspondence with Justin - Re Strain Gages, etc.

Weight Sensing Longboard with Inline Wheel Motors
Another project where Justin used strain gauges and a CA3 to control a "vehicle"

Torque Sensors

TDCM bottom bracket

Harry Wigglesworth
Head of European Operations
Tel (TW): +886-9-81-529-661 | Tel (EU): +49-178-239-0088
Skype: harry.wigglesworth |

diagram of TDCM in bottom bracket

TDCM torque bottom bracket Manual (PDF, 26 pps)

The following summary is based on ~ 05/04/2015 emails with Grin and Harry W.

Excitation voltage; current consumed by TDCM BB

+5/6 VDC Input Voltage, 25 mA Power Consumption
NOTE: Grin Tech ( "define and specify specific specifications to work with their Cycle Analyst controller, based on a 12V input voltage." The CAv3 connector describes it as 10VDC.

Signal voltage returned from BB

The effective voltage span (signal size) is ONE VOLT (1V), falling into the ~ 2.4V to 3.4V DC range.

Maximum torque measured on pedals

The maximum force applied to the petals, beyond which there is no change in the output signal voltage is ~40Nm

Connectors and conductors

TDCM uses different pin defines and cable numbers depending on the specification - see page 6/7 of Manual.

TDCM female connector specs (PDF, 1 pps)

TDCM male connector specs (PDF, 1 pps)

Emails between Tringa, Scott O., Grin Tech (Justin et al.), and TDCM (Harry W).

Return to Top of Page
Return to Main Menu