Brompton - Cycle Analyst & Thun Torque Sensor

[dingrpdl]

Brompton - Cycle Analyst & Thun Torque Sensor Installation (Part 1)

I installed a Freedom Ebikes kit on my white Brompton in mid-2012. The original Tongxin motor worked well and I had no problems with it. At some point, I replaced the original controller with a KU63 from BMSBattery.com and that whole setup served me well for more than a year and a half.

A few months ago, I started to take the mods further and replaced the Tongxin motor with a smaller Keyde motor as well as installed a Cycle Analyst unit together with a Thun bottom bracket torque sensor. And to make the whole setup as stealthy as possible, I hid the two 2.3Ah 6S LiFePO4 packs and the KU63 controller within the frame of the Brompton Folding Basket.


Keyde motor swap

The replacement Keyde front hub motor (250W/260rpm/36V/28-hole) weighed in at about 1.45Kg (compared with 2.1Kg for the Tongxin). I got rid of the CR18 rim (thank God for that) and laced the motor with a new 2013 Brompton double-walled rim. I bought the angle-drilled version of the rim which works well with a 1-cross or 2-cross lacing. There is no spoke bending at the nipples at all. I ordered the 13-gauge spokes from BMSBattery.




FET anti-spark switch and charging harness

I used Jeremy Harris’ simple FET switch circuit with three IRFB4110 FETs in parallel. The components were soldered on a 1.5” x 0.6” piece of perf board and it, in turn, was wired to a small switch, battery connectors and charging connectors. With the switch in the one position, the battery pack is connected to the controller. In the other position, the battery is connected to the charging connectors. One IRFB4110 would have been fine for the KU63’s 15 Amp limit. But I decided to use three in parallel because I suspected that the FETs I bought were bootlegs and would likely have a higher RDS(on) resistance than the real ones. However, I found that the FETs were barely warm in normal usage. So two (or even one) would have been sufficient, in retrospect.




Concealed battery and controller

The two 6S LiFePO4 2.3Ah packs, the KU63 controller, FET switch harness and the Cycle Analyst current shunt were wired and strapped to the aluminium frame of the Brompton Folding Basket. The frame slips back into the Folding Basket and everything is concealed. Connection from the Folding Basket electronics to the bike was via a 9-pin connector — the same type used for connecting sensored motors to a controller. I chose this type of connector because it had the required 3 big conductors for motor power and the 6 signal conductors required by the Cycle Analyst. Slim and waterproof too.




<Continued on next post>

 

[dingrpdl]

<Continued from Previous Post>

Brompton - Cycle Analyst & Thun Torque Sensor Installation (Part 2)

Thun Torque Sensor

The Thun bottom bracket torque sensor was purchased with the Cycle Analyst V3 unit as a package. I chose the 120L/119.0mm model to match the 119mm spindle length on the original Brompton BB. However, I discovered after the purchase that the spindle on the original BB (on the pre-2013 Bromptons) had an ISO spec taper and that the Thun had a JIS spec taper. This meant that the crank arms would sit a few millimetres further out on the spindle when fitted on a JIS taper BB. Fortunately, as it turns out, the drive side crank ended up about 1.5mm further out and the non-drive side 4.5mm further out than in the original setup. The difference between the left and right side was because the original BB had a small offset built-in (with the drive side spindle longer than the non-drive side) whereas the Thun BB was symmetrical. So the 1.5mm change in the chain line did not cause any problems at all. And as a rider, I certainly could not tell the difference. But in retrospect, a 116.0mm Thun BB would have reduced the chain line shift to zero. I drilled the cable exit hole on the underside of the bottom bracket on the Brompton.




Cycle Analyst

The Cycle Analyst unit was mounted on the handlebar offset to the left. I could not mount it symmetrically because it would affect the folding. It’s not an ideal spot but will have to do until I find a better solution. Anyone who has worked on a Brompton electric conversion knows that cable routing is the trickiest and most time-consuming part of the conversion. I must have tried two dozen routing options until I got the cable lengths and cable bundling right. In the end, I replaced all the connectors on the Cycle Analyst with slimmer ones so that they would all stuff into a 22mm O.D. carbon fibre tube which I purchased from HobbyKing.




Brake Sensor

For the brake sensor, I used an inline hall sensor unit from Suzhou New Power. I had also purchased the Hidden Wire Brake Sensor from BMSBattery but I like the Suzhou one better because it was more compact.




Final comments

In operation, the Keyde motor emits a high frequency whine when under power whereas the Tongxin which it replaced was virtually silent.

On the positive side, the Cycle Analyst and Thun torque sensor combo improves the pedelec experience because it can be set up to regulate motor power in proportion to rider output. On the downside, it does increase complexity in that a lot of parameter-tweaking has to be done to get everything right. I am 95% there with some minor issues to work out. My setup does have a thumb throttle unit as well so I can override the pedelec mode. I set up the Cycle analyst to provide a current throttle function which I like better than the native speed throttle mode of the KU63 controller. But increasingly, I am finding I do not have to touch the throttle for a large part of my rides. Another negative is that the large display of the Cycle Analyst detracts from the stealthiness of the electric-assist setup.

The cost of the Cycle Analyst and Thun torque sensor is not insignificant and is probably quite unnecessary for most cyclists. I am not sure I would install a similar setup on my other bikes. But this conversion was a good exercise and will give me a baseline to work from in my future conversions which will likely be microcontroller-based.

Here are some final pictures of the completed bike.

 

[mountainsport]

Hi dingrpdl,

Lovely job done, can I please ask you, but do you think that the cycle analysis design should be made slightly smaller in size, or will it just do?

MS.

 

[jackhandy]

Very tidy

Rats' nests are not confined to Brommie conversions - The Ezee conversion of my Trek took longer than anticipated to do a tidy wiring job.

 

[jerrysimon]

Hats off, you really have taken the conversion to the next level!

Like you have have found the Keyde to be a little noisy under load. I have been using mine on my daily commute the last 4-6 months and I am not sure if it is because my head is less wrapped up now the warmer weather is here, but it actually seems louder than when I first fitted it!

My overall thought is that I would stick with the Tongxin unless your travels take you over fairly flat terrain. Note you can order the Tongxin with a 290RPM spec now. As stated also order either motor with 28H flange drilling to enable stock Brompton rims to be used.

PS I am also convinced that radial (zero crosses) is sufficient for both Tongxin and Keyde motors for the Brompton 16" rim. The big plus being its looks so cool and is so easy to lace!

Side by side comparison

Regards

Jerry

 

[mountainsport]

Sorry to bother you again dingrpdl, regards to your CAv3 is it possible to activate cruise control using one of the buttons on the display simultaneously along with the throttle ? Baring in mind that there are no switches/buttons on the throttle itself.
Thank you.

MS.

 

[Deleted member 4366]

I have a Thun torque sensor and a cycle Analyst too. I bought them for a project, but never used them because I changed my mind after thinking it through.

The problem is that a torque sensor can't work how you want it to with a speed control controller because the throttle relates to a speed, not a power. You need a current control controller if you want power related to how hard you pedal.

 

[dingrpdl]

Hi dingrpdl,

Lovely job done, can I please ask you, but do you think that the cycle analysis design should be made slightly smaller in size, or will it just do?

MS.

It could certainly be smaller. There are two circuit boards within the CA - the microcontroller board and the LCD display. The microcontroller board is fairly small. It's the LCD display that is large and that what the case is built around. But I think the LCD is larger than necessary. A smaller 16x2 display would make a more compact CA.

 

[dingrpdl]

Sorry to bother you again dingrpdl, regards to your CAv3 is it possible to activate cruise control using one of the buttons on the display simultaneously along with the throttle ? Baring in mind that there are no switches/buttons on the throttle itself.

MS.

No bother at all!

The cruise control function on the CA, when turned on, is an auto-cruise. it comes on when you hold the throttle steady for a certain period of time after which time you can release the throttle and the CA will hold the setting for you. You can specify that time period in the CA. Cruise control disengages when you next move the throttle or when it detects braking via the brake sensor.

I didn't really find the auto-cruise function useful so I don't have mine activated.

 

[dingrpdl]

The problem is that a torque sensor can't work how you want it to with a speed control controller because the throttle relates to a speed, not a power. You need a current control controller if you want power related to how hard you pedal.

Hi d8veh. The CA has four throttle modes -- pass-through, speed throttle, current throttle and power throttle. The CA works with a speed controller (like the KU63 and pretty much the majority of controllers) where the throttle signal controls the speed. It essentially 'converts' a speed throttle controller into a current throttle controller by using current feedback. It reads the current through its current shunt and then varies the throttle output signal to the controller in order to keep the current constant. So in this mode, the physical throttle position now relates to a current setting. And you essentially have a current controller system. When you hold the throttle in one position, you can see on the LCD display an indication of how the CA is varying the throttle signal to the controller in order to maintain a fixed current regardless of whether you are pedaling or not. Power throttle mode is basically the same as current throttle if the pack voltage is steady.

So that is why a CA mated with a speed controller 'works' with a torque sensor. It can now sense rider power input and then output a throttle signal to give a proportionate power output through the current feedback mechanism.

I hope that explains it adequately.

 

[dingrpdl]

My overall thought is that I would stick with the Tongxin unless your travels take you over fairly flat terrain. Note you can order the Tongxin with a 290RPM spec now. As stated also order either motor with 28H flange drilling to enable stock Brompton rims to be used.

Jerry

Hi Jerry.

I will stick with the Keyde motor for now as it is so much more stealthy visually. Pretty important here in Hong Kong where ebikes are not legal and the fun police do their occasional crackdown on ebikes. Too bad it is not as stealthy audibly. Another plus is that the Keyde has the motor lead exiting on the right side rather than on the chainstay hook side, as you know. I will definitely keep the good ol' Tongxin around in case I want to switch back.

 

[schiller]

Very interesting build, congrats. I haven't seen this thread while it was still active unfortunately.
I am interested in torque sensors too and i wonder if i could connect the output of the thun sensor directly to any of the inputs of the ku63 controller.
CA is a very nice and useful device, but to expensive for me -at least at present-.
What kind of signal does it output?
Thanks in advance for any suggestion,
Greetings
Konstantinos

 

[dingrpdl]

Sorry for the late reply as I have not visited this forum in quite a while.

The Thun sensor outputs the sensed torque as a 0.5VDC to 4.5VDC signal. 2.5V is the zero-torque level and the other voltage extreme correspond to the max positive and negative torque sensed. In addition, it outputs quadrature pulses so that a controller can tell the direction of rotation. So you can't simply connect the voltage output directly to a simple controller such as the KU63. Well, you could. But it wouldn't work properly.

 

[schiller]

Thanks, meanwhile i have found as alternative to the cycle analyst a circuit posted in a german forum and a solution with arduino.
I think i'll try the first one, although it seems a bit limited in terms of adjustability.

 

[Steven wong]

Does anyone have the wiring diagram for a thun torque sensor for Brompton

 

[matthewslack]

Does anyone have the wiring diagram for a thun torque sensor for Brompton

https://thun.de/wp-content/uploads/2016/02/THUN_XCell_2_0_datasheet_23.11.2018.pdf

This doc has a five pin plug diagram, which may be what you need.

The electrical spec is described above, and in the doc , and the two agree.