Sempu BB + cycle analyst assist not working well

vmbray

1 mW
Joined
May 8, 2018
Messages
12
Hi guys, I am working on a trike conversion for my wife who has some orthopedic issues and could use some hill assist. Have ezee hub motor, CA 3.1 and phaserunner, and sempu torque sensor. Having some random issues with the human watts being detected as negative and then as I pedal harder going positive. Suspect the sempu (already had an issue where the right side bb cup was threaded wrong, was right hand instead of left) but not sure maybe it is the ca. Earlier was getting pretty consistent human watts readings but no motor assist, and then after some help from the Grin guys the motor started working but was 100% as soon as torque was detected, as in as soon as you pedaled you would get turbo boost and quickly be going 15-18 mph. Which is awesome fun, but for her needs I need assist not turbo boost :)

It seemed like the CA was acting random earlier and after a settings change needed to be rebooted to take effect, maybe this is normal idk. So to sum up, had no action, then had full throttle as soon as motor engaged, then while troubleshooting that had no motor assist or random motor assist that was finally proportional but not consistent. Wondering if CA needs to be factory defaulted like it's corrupted or something. I found some old threads where the sempu is referenced as a one wire situation so I wonder about the default, but when I choose one wire I get no human watts reading in CA. Here are the pas settings from the CA:

Photo May 08, 9 39 33 PM.jpg

Some pics of the settings and trike for fun. Also in case it interests anyone custom fab on the grin torque arm to fit the trike, welded with stainless wire and tri-mix gas I had from various car exhaust projects and etc.

Photo May 08, 9 04 50 PM.jpg

Photo May 08, 9 05 08 PM.jpg

Photo May 06, 2 02 05 PM.jpg

Photo May 06, 2 20 36 PM.jpg

Photo May 06, 2 58 53 PM.jpg

Photo May 06, 3 55 31 PM.jpg
 
After going through the settings and with help from Grin, it is working intermittently now. When I start out, the human watts will read negative for about half a block and then will suddenly kick over to positive and assist will work well until I stop. Anytime I slow down to a near stop or completely stop it goes back to the negative human watts for a bit and then kicks in again. I added a preset for basic throttle assist (which I guess uses the cadence of the torque sensor instead of the torque reading) which works perfectly. This would be totally fine but one of the basic needs is torque assist so the rider is working for the assist. Any thoughts? Also loaded the latest beta software in CA and didn't make any difference.
 
vmbray said:
After going through the settings and with help from Grin, it is working intermittently now. When I start out, the human watts will read negative for about half a block and then will suddenly kick over to positive and assist will work well until I stop. Anytime I slow down to a near stop or completely stop it goes back to the negative human watts for a bit and then kicks in again.

Interesting. I've been wondering how well the Sempu works, and how much drag it may have. Really like our Thun BB's but they're no longer available. NTCE makes great torque sensors but I'm not too impressed with their BB implementation.

I kind of wonder if there may be some more setting refinement in order. Once you get it right, will be fascinated to hear your impression particularly the linearity of output relative to Human Watts input.
 
Is it zero’d out in the settings? I think mine is 1.49V, and I’ve got it set to 18 N/m / V, which feels accurate-ish for Hwt.

I don’t have the negative watts going on, but I’ve noticed that even when the Hwt display remains fairly steady, sometimes the motor wattage drops out or jumps up in unexpected ways. Maybe as a result of using a throttle while pedaling.
 
When I'm in the settings looking at the voltage and torque readings from the sensor, it looks pretty linear (granted I have no experience here) as in when I press harder the torque ramps up reasonably. It's interesting, when you take your feet off the pedals it will read a pretty steady 1.51v but as soon as you touch the pedals it starts fluctuating and reads 1.8 or more. It doesn't cause the motor to drive because the cadence isn't generating pulses I guess. When the assist is working it feels pretty good to me, I would say it's worth the upgrade from a cadence sensor for sure.

I'm wondering if the orientation of the bb is part of the issue. The trike has a stub tube that comes up to mount the derailer and there is a hole inside, so I fed the wire up that way. This puts the sensor about 150 degrees different than the engineers would have expected since the diagram shows the hole at the bottom of the bb, not sure if this matters but maybe is confusing the bb and it takes it some rotations to figure out what's happening? Not sure how much logic there is in the unit but there must be some if it can switch between one wire and two wire? Also there is the mysterious 'datum' mark on the square taper, mine is randomly sort of lined up with the crank arm as shown but for a recumbent maybe needs to rotate if it really matters. The fact that they bothered to show it in the diagram makes me think this could be part of the issue.
 
One user of our OpenSource firmware is trying Sempu torque sensor. We already use BMSBattery bb torque sensor and works perfectly as we were going to develop and adjust the firmware perfectly for it.

We implemented human power, simple torque signal or even PAS mode, all for the same torque sensor.
It can work as torque signal + PAS mode, were torque is mainly used at startup.

Look at our firmware that runs on the popular KT controllers and supports perfectly BMSBATTERY bb torque sensor and soon shoild support Sempu.
 
vmbray said:
When I start out, the human watts will read negative for about half a block and then will suddenly kick over to positive and assist will work well until I stop. Anytime I slow down to a near stop or completely stop it goes back to the negative human watts for a bit and then kicks in again.
This sounds like your Torque Zero Offset voltage is incorrect (too high) and so the CA is calculating a negative torque which propagates into negative HWatts. You need to zero that in the CA Console Setup (PASD->ZeroTorq).

  • Try riding the bike with the CA on the PASD setup preview screen and watch the voltage and torque values in real time. The voltage is what's coming from the Sempu and the torque is what the CA computes from your settings. If the voltage goes below your zero setting then it causes a negative torque. This should behave consistently for all pedaling situations, not suddenly or randomly change. If the voltage is all over the place or behaves differently in identical pedal force situations then your torque sensor may be installed incorrectly or damaged. If you are getting negative torque readings then your basic Zero torque setting is suspect.

vmbray said:
It's interesting, when you take your feet off the pedals it will read a pretty steady 1.51v but as soon as you touch the pedals it starts fluctuating and reads 1.8 or more.
The Sempu responds to any stress on the crank so proper installation is important (re: orientation of datum point) and simply putting your feet on the pedals will generate a non-zero torque signal.

  • See: https://endless-sphere.com/forums/viewtopic.php?f=2&t=93139
  • See: http://www.ebikes.ca/learn/pedal-assist.html

Also - although you do not mention it and there is no Aux shown on your screen grab above, You really need an Aux control to tune the PAS on the fly for different terrain and situations. Perhaps you have a DigiAux button set that just didn't appear on the screen grab....
 
Hi All

I found this thread as I have also been setting up pedal assist with the latest sempu torque sensing bottom bracket that I got from Grin tech. My bike is an HP velotechnik Speed Machine recumbent, I have a 52 V battery, phaserunner controller and CA3 running an ezee 350 rpm geared front hub motor on a 20” front wheel. Trying to set up the sempu I had some of the same issues described in this thread. Mainly the human watts sensing didn’t seem very accurate. It would swing unpredictably from 50 to 150 to 30 tho 400 watts as I was giving pretty even moderate pedal force. This caused the motor to speed up and slow down erratically and was a frustrating experience to ride. Reading this thread gave me the idea that the orientation could be off because of the recumbent pedal position, and that turned out to be right...

I followed the grin video to install the BB the first time, and when putting the cranks on I aligned the cranks at a 90 degree angle to the indicator dot on the sempu spindle. That works for an upright bike where the main pedal force is down, but on a recumbent the pedal force is forwards (horizontal). I took the pedals off and rotated 90 degrees, so now the crank arm is aligned with the indicator mark on the spindle. That completely fixed it! Now I get a nice smooth response, the readings are steady and proportional to my effort like I was expecting. So, apparently the sensor is designed to measure force in one orientation, and for recumbent bikes the cranks need to be rotated 90 degrees for the best torque sensing.

Hopefully this will help anyone else trying to set up a recumbent torque sensing ebike with the sempu. I have only gotten a short test ride in since fixing the problem, but the bike fees amazing now, can’t wait to get it out on a proper ride!
 
ebentrider said:
...for an upright bike where the main pedal force is down, but on a recumbent the pedal force is forwards (horizontal).

I took the pedals off and rotated 90 degrees, so now the crank arm is aligned with the indicator mark on the spindle. That completely fixed it! Now I get a nice smooth response, the readings are steady and proportional to my effort like I was expecting.

Hopefully this will help anyone else trying to set up a recumbent torque sensing ebike with the sempu.
I'm sure it will - excellent post!
I'm linking your post into the CA Beta thread "Posts of Interest" roll-up index.

Thanks!
 
I had a funny observation. Building up my Cetma cargo bike with a belt drive, I followed the gates tensioning guidelines. Once I finally hooked everything up I found that the torque sensor would have been zeroed out at 2.5v ! A mechanic had already mentioned to me that the Gates app suggests making it waaaay tighter than you need with the newer center track belts, so I backed it off, and it eventually settled down into a normal 1.5v.
 
ebentrider said:
I followed the grin video to install the BB the first time, and when putting the cranks on I aligned the cranks at a 90 degree angle to the indicator dot on the sempu spindle. That works for an upright bike where the main pedal force is down, but on a recumbent the pedal force is forwards (horizontal). I took the pedals off and rotated 90 degrees, so now the crank arm is aligned with the indicator mark on the spindle. That completely fixed it! Now I get a nice smooth response, the readings are steady and proportional to my effort like I was expecting. So, apparently the sensor is designed to measure force in one orientation, and for recumbent bikes the cranks need to be rotated 90 degrees for the best torque sensing.

Hopefully this will help anyone else trying to set up a recumbent torque sensing ebike with the sempu.

Thanks - that would be me! I installed that sensor on my Giro 26 recumbent (with CA3, Grinfineon 20 amp controller, and Bafang Mini 310 motor.) I noticed that the readings for the right pedal varied enormously with pedal pressure but with similar pedal pressure the reading for the left pedal hardly varied at all, but in my own limited testing it hadn't obviously been problematic. (Averaging covers many sins.)

A quick 90-degree rotation and now I'm seeing similar responses left and right. Thanks so much. Obviously, the video could do with a minor update.

Charlie
 
I am also having significant issues with my new eBike using the Sempu BB.
I noticed similar problems where I am occasionally getting negative Human Watt values. It starts out small at -5W, -40W, -150W and if I stop pedaling it will reset back to zero and when I start spinning again the values are positive again.

Similarly I sometimes see huge, unrealistic spikes in Human Watt measurements of between 500-600W. This doesn't happen at logical transition times but rather when I am pedaling along at a steady pace. I especially notice this when my PAS multiplier is around 2.00 and suddenly my bike jumps from 500W up to 1000W for a moment because it is sensing a non-existent spike in human effort.

I am going to try and pull my Bottom Bracket tonight and see if there is an orientation issue but I am pretty sure I did exactly what Grin recommended in the BB mounting video: https://www.youtube.com/watch?v=yg6cXP0oGQo

I did look at the hole in my BB shell this morning and it appeared that the BB had rotated a bit in the shell and could be pinching the wire at the edge of hole. I sure hope I didn't damage the wire!

For reference, here is my current CycleTruck with the Grin Tech kit installed.
Full Photo set here: https://www.flickr.com/photos/andy_squirrel/albums/72157695740591672
 
I also forgot to mention that I have zeroed out the PAS input a couple times and noticed that the zeroing out doesn't seem to be consistent with crank orientation. During unloaded zeroing, rotating left pedal forward vs right pedal forward gives me ~0.1V difference which seems like quite a bit when I am only getting about 1 volt swing when pressing hard on the pedal.
I also noticed that when I press the right pedal I am getting a change in the PAS voltage but pressing hard on the left pedal it doesn't really register. Does this seem typical?
 
Andrew_Squirrel said:
I noticed similar problems where I am occasionally getting negative Human Watt values.
There is always the possibility of some program glitch, so here I would recommend going to the PAS Device Setup Preview screen and watching the raw torque figure computed from the sensor. This displayed value should show what the sensor is sending. If it looks good, the firmware is having an issue. If it's all wonky, then you have a sensor issue.

Also - it's difficult to evaluate some of these issues without knowing the firmware rev you are running as reported on the splash screen.
 
Andrew_Squirrel said:
I am also having significant issues with my new eBike using the Sempu BB.
I noticed similar problems where I am occasionally getting negative Human Watt values. It starts out small at -5W, -40W, -150W and if I stop pedaling it will reset back to zero and when I start spinning again the values are positive again.

Similarly I sometimes see huge, unrealistic spikes in Human Watt measurements of between 500-600W. This doesn't happen at logical transition times but rather when I am pedaling along at a steady pace. I especially notice this when my PAS multiplier is around 2.00 and suddenly my bike jumps from 500W up to 1000W for a moment because it is sensing a non-existent spike in human effort.

I am going to try and pull my Bottom Bracket tonight and see if there is an orientation issue but I am pretty sure I did exactly what Grin recommended in the BB mounting video: https://www.youtube.com/watch?v=yg6cXP0oGQo

I did look at the hole in my BB shell this morning and it appeared that the BB had rotated a bit in the shell and could be pinching the wire at the edge of hole. I sure hope I didn't damage the wire!

For reference, here is my current CycleTruck with the Grin Tech kit installed.
Full Photo set here: https://www.flickr.com/photos/andy_squirrel/albums/72157695740591672

Nice setup Andrew, did you purchase the Sempu BB from Grin Tech? If not, and you purchased it from another supplier then it could well be a "1" wire unit. You'll need to change this parameter in the PAS Device setup screen. I run my Sempu at 14.5 v/nm., seems to be just the right sensitivity. Checkout this page for more technical details (if you hadn't had a chance to come upon it yet...)

http://www.ebikes.ca/learn/pedal-assist.html

http://www.ebikes.ca/shop/electric-bicycle-parts/torque-sensors.html


I also have a Kona Smoke, and am wondering what fork you got on yours. Did you make it yourself?
 
I believe I should have the newest Cycle Analyst firmware since i've purchased it within the past few months. When I turn it on I see version 3.1 across the splash screen. I have the USB-TTL cable in the mail now, probably going to hook it up to make sure everything looks good.


Sempu BB by Andrew Squirrel, on Flickr

I pulled out the Sempu BB over the weekend. I noticed that the BB had slipped in the shell and the wire is very close to being pinched with the edge of the hole. I removed it and reinserted so wire was centered again, went for a ride, came back and discovered it had slipped yet again even though I tightened it securely. I wonder if I need to offset it in the opposite direction so when it does slip then it centers the wire?
It also appears that the RTV has cracked at the location where the wire enters the BB. You can see it in the image below:


Sempu Torque Sensing Bottom Bracket by Andrew Squirrel, on Flickr

So at this point I'm not sure if I've injured the BB or if its just a flesh wound. I have not ridden the bike in rain yet, hesitant to get this area wet until I can add more RTV.

I am now getting really high Negative Human Watt numbers on my CycleAnalyst. I am also seeing negative numbers more often than before. Might try to contact Grin Tech to see if they have any ideas.

"Nice setup Andrew, did you purchase the Sempu BB from Grin Tech?"
Thanks, Yes, I purchased from Grin Tech with the eZee kit.

"Did you make it yourself?"
I put everything together myself but did not do any welding or major modifications.

Here is my spiel on the bike if you are interested:

"This CycleTruck is a modified 26" mtb retrofitted with a cargo bike kit designed locally by Cyclefab / Haulin Colin (http://www.haulincolin.com/racks.html). They have a rack & headtube extension that clamps on to an existing bike frame then the front fork and wheel are replaced by 20" version. The cargo bike is designed such that the large rack platform stays fixed in place while the handlebars and fork can turn independent from the load. The benefit to this is as you turn the bike the load does not dive into the corner and create instability in steering. In addition the front end geometry is designed with what is called "Low Trail Geometry" which is calculated to minimize wheel flop when handlebars are turned. To calculate trail you need to use an equation involving fork rake, head tube angle and wheel diameter (see: http://yojimg.net/bike/web_tools/trailcalc.php). There is a sweet spot that many front loading bikes use to improve steering over a stock road or mountain bike loaded similarly. The reason for choosing a smaller wheel in the front is to keep the load's center of gravity as low and stable as possible. Many cargo bikes will push the small front wheel out further and drop the cargo load even lower to the ground but unfortunately this makes the bike significantly longer and almost impossible to carry up stairs whereas mine can be brought inside an apartment or house if you don't have a garage to store it (I don't have a garage in my rental). Front wheel is a 20" Dia x 50mm Width, Rear is a 26" Dia x 60mm Width. Photos can be found here: https://www.flickr.com/photos/andy_squirrel/albums/72157695740591672"
 
Thank you for the information on the front fork. It's a fascinating setup.

As for your Sempu, hook it back up to the CA3 and take a screen shot of the PAS device screen under SETUP. It should roughly show 1.52v when no load is placed on the shaft. Next, rotate the shaft and monitor the pole and direction indicators to see that they are functioning correctly. This will verify the ground (black), power (red), pole (green), and direction(yellow) wires are ok. You might want to carefully, and gently wiggle (do not pull!) the wire assembly coming out of the Sempu to see if any of the above indicators start misbehaving. If so, you've either got a short or a broken wire inside the unit.

The last thing to test would be the torque wire. It's the white one. The easiest way is to re-install the Sempu, and hang a known weight off the right pedal. You'll need to place the pedal parallel to the ground. You should see the 1.52v increase in value. If not the white wire is either shorted or pulled.

If all wires are ok, then you have a setup parameter issue. The current firmware version is v3.12 which included changes to PAS.

By the way, you can take a pvc T, cut it in half and use it to shield the wire coming out of your bottom bracket. Strap the T using a hose clamp or similar to your downtube. Lastly, make sure you have some slack in that wire assembly going into the Sempu.
 
My Sempu gives a strange PAS-Signal, perhaps that's confusing the CA....

https://endless-sphere.com/forums/viewtopic.php?p=1421482#p1421482

regards
stancecoke
 
The grossy asymmetric Sempu waveform in combination with the high pole count has been found to cause certain CA 3.1+ PAS difficulties. This isn't directly related to the odd pulse timing your post illustrates, but it's another Sempu weirdness.

This issue has been addressed and will be released very shortly as 3.13b1. Stay tuned in the CA beta thread! :D

Anyhow - thanks for the post!
Interesting.
 
I don't know, if there is a mechanical problem with the plastic tooth disc, that is operating the two photoelectric sensors. Maybe it'll jam once a turn.
https://www.pedelecforum.de/forum/index.php?threads/integration-des-sempu-sensorlagers-in-den-fc.51648/page-3#post-929461

index.php
index.php


regards
stancecoke
 
Hey all,

I have installed a Sempu BB in my ebike this week and have run into a few troubles along the way first off the cadence was intermittently dropping to zero turned out the cable got pinched a little during the install. So after taking out the sempu pulling it apart and re soldering some new wires to the board I have completed the second install.

Went for the first ride yesterday only about 5k's and it performed perfectly. Smooth Power, No unusual fluctuations with the rpm or torque values and fairly accurate human power results. Loved it :D

Got the bike out today to go for a decent ride and its now sensing my input as negative human watts and very large values 300-700w for not too much pedal effort. Very disappointing to say the least.

Have been trying to troubleshoot but its got me stumped, Below is what I have tried so far

Removed BB inspected wires all okay
Dissassembled and reassembled BB all looks good
Checked Pedals where correctly oriented on the crank
Zeroed PasD 1.52v 0.0nm +- 0.1nm in Cycle Analyst
Pedaled around in the Pas settings screen reasonable values for torque were displayed around 30nm with light pedaling and no negative values
Checked rpm is being read correctly.
Updated cycle analyst to 3.12 then also tried 3.13b

Still no luck it just keeps giving negative human watts whenever I attempt to pedal.

If anyone has any advice it would be great.
 
I have a 2-wire Sempu with a 115mm spindle width. Mine initially gave wonky readings where it was displaying massive amounts of human power with the rear wheel lifted off the ground and spinning the crank gently by hand. I had to adjust the Nm/V parameter down from the 50.0 default to 7.5 Nm/V to get it to behave in a manner to where I could use it. I installed it according to the instructions in the video referenced in this topic, but it turns out mine is a recumbent trike in the form of a homemade velomobile.

As it is now, it kind of scales with my input, increasing watts the harder I pedal, but if I'm freewheeling, it shows hundreds of watts of human power. I set it to give me 3W from the motor for every 1W of human input, turning on at 100W. It feels like it turns on at less than 100W, and definitely does so while freewheeling. If I'm mashing up a steep hill with what is in reality 500W+ in a high gear, I get no human input reading and no assist, but if I switch to a lower gear on that same hill that requires less rider torque, I get immediate assist that is probably even more than appropriate for my pedal input.

I do want it to be functioning in a normal manner. So, rotating the crank 90 degrees relative to the indicator dot on the spindle may solve this issue and allow me to put the default setting back.
 
Andrew_Squirrel said:
I pulled out the Sempu BB over the weekend. I noticed that the BB had slipped in the shell and the wire is very close to being pinched with the edge of the hole. I removed it and reinserted so wire was centered again, went for a ride, came back and discovered it had slipped yet again even though I tightened it securely. I wonder if I need to offset it in the opposite direction so when it does slip then it centers the wire?

Same happened with my sempu T2. I also tried the offset idea, didn't work. It only needed a few km's to slip.
The solution which still works after 5 months cycling is to put an extra tiny screw through the frame and in the plastic locking ring (tap thread in the frame). This way it won't turn on its own. I made sure that it doesn't stick out of the locking ring so the sempu fits in normally.
And don't make the same mistake as I did, drilling through the frame when the locking ring is still in place. Now there is a metal burr between frame and ring, so I cannot screw the soft plastic ring out without ripping it open ...
Sempu 03.jpg
 
Back
Top